Seestar & Mak PixInsight Workflow

Processing guide · written 14 November 2026

Seestar & Mak PixInsight Workflow

A complete deep-sky workflow for PixInsight, written for your Seestar S50 and your SVBONY MK127 imaging rig. Each step explains why the process exists as well as which buttons to press, so you can adapt it to any target.

PixInsight 1.9.6 Lockhart · build 1718 (13 Nov 2026)WBPP 3.2.1 (30 Oct 2026)BlurXTerminator 2.6.11 · ML 5NoiseXTerminator 2.6.11 · ML 3.1StarXTerminator 2.6.11 · ML 11StarNet2Gaia DR3 + DR3/SP installed

These versions were read from your Mac's PixInsight install, and the RC Astro tools were checked to be running their newest models. Figures captioned Your install are screenshots taken from your own PixInsight on 14 November 2026, set to the values this guide recommends. The other images are from the reference documentation bundled with that install (© Pleiades Astrophoto S.L.), used here for personal study. Numbers marked starting point are typical values to adjust, not fixed rules.

Your two rigs

The single most important number in this guide is image scale: how many arcseconds of sky land on each pixel. (For a feel for how small an arcsecond is, see Appendix C.) It decides whether you drizzle, whether you bin, and how hard you can sharpen. The formula is:

scale (″/px) = 206.265 × pixel size (µm) ÷ focal length (mm)

Seestar S50MK127 + 0.65× reducer + ASI585MC
Optics50 mm triplet APO, 250 mm, f/5127 mm Maksutov-Cassegrain, ~1500 mm f/11.8 native → ~975 mm f/7.7 with reducer
SensorSony IMX462, 1920×1080, 2.9 µm, colour (OSC)Sony IMX585, 3840×2160, 2.9 µm, colour (OSC), uncooled
Image scale2.39″/px~0.61″/px (1.23″/px after 2×2 binning)
Field of view~1.28° × 0.72°~0.65° × 0.37°
Star FWHM~3 px measured (7–8″ on your M45 subs; ~2 px on your M31 stack) → adequately sampled. Seeing alone (2–3″) would give ~1–1.3 px, but the Seestar's optics and tracking add blur (see Appendix C)~3.5–5 px expected at 2–3″ seeing → oversampled (not yet measured)
MountAlt-az, so the field rotates during a sessionSky-Watcher GTi (equatorial), guided with UniGuide 32 + ASI120MM Mini in PHD2
What that meansDrizzle 2× only on sharp nights (FWHM under ~2 px in SubframeSelector) with enough subsDon't drizzle; consider 2×2 software binning
FiltersBuilt-in light-pollution filter (on/off per session)ZWO Duo-Band 2″ (H-α + O III), or none
Focal length to check. The ~975 mm focal length is nominal. With the reducer at its designed 55 mm back focus, plate solves typically land between 955 and 985 mm, varying with focus position. Use the focal length ImageSolver reports (see reducer spacing).

One-time setup

Do these once. Each takes a few minutes and prevents most of the errors people hit later.

  1. Update PixInsight. Run Resources › Updates › Check for Updates. WBPP 3.2.1 needs core build 1715 or later, so if WBPP complains about the core version, this fixes it.
  2. Point PixInsight at the Gaia databases. You already have ~/PixInsight-Data/gdr3-1.0.0 (astrometry) and gdr3sp-1.0.0 (spectra for SPCC). Open Gaia (Process › All Processes), click the wrench (Preferences), and add the .xpsd files for DR3 and DR3/SP separately. If SPCC later says "no catalog available", this is the cause.
  3. Finding processes. Every process named here is in Process › All Processes, or type its name into the Process Explorer search. Script menu paths are given in full.
  4. Turn on the auto-stretch display. Linear images look black. The ScreenTransferFunction (STF) stretches the screen view only without touching the data. Press Ctrl+A (⌘A on Mac) on any image. Use the chain icon to link or unlink RGB channels: unlinked hides colour casts, which is useful for judging gradients.
  5. Optional: MARS database for MultiscaleGradientCorrection. MGC is in your install but no .xmars files were found on your Mac. Without them, use GradientCorrection or GraXpert (both covered below). If you want MGC, download the MARS data from pixinsight.com and add it in MGC's database list.
ScreenTransferFunction toolbar with annotated buttons: auto-stretch, reset, link channels, track view
ScreenTransferFunction. The radioactive-symbol button (auto stretch) is the one you'll use constantly; the chain toggles linked vs unlinked channels.

Part A · Why it works

PixInsight is opinionated about order because each process assumes the data is in a particular state. Understand these five ideas and the workflow stops being a recipe.

Linear vs non-linear data

Straight off the stacker, your image is linear: a pixel's value is proportional to the number of photons it collected. Twice the light gives twice the value. Because faint nebulosity might be 0.1% of the brightness of a star core, almost everything sits in the bottom sliver of the histogram, so the image looks black.

Stretching applies a curve that lifts faint values far more than bright ones. After that the image is non-linear: values no longer track photon counts. You can't go back.

Why it matters. Gradient removal, colour calibration and deconvolution are all physics. They model light as it actually adds up (sky glow plus target, star profiles blurred by optics, flux through a filter). Those models only hold while the data is linear. That's why they all happen before the stretch.

Signal, noise and stacking

Every exposure contains signal (the target) plus random noise (photon shot noise, read noise, thermal noise). Signal adds up in proportion to the number of frames you stack, but random noise only grows with its square root. So:

SNR ∝ √N (N = number of subs, or total integration time)

Four times the integration doubles the SNR. One hour of Seestar 10 s subs is 360 frames; four hours is 1,440. That's also why the Seestar workflow leans on FastIntegration when sub counts reach the thousands.

Stacking does more than average. Pixel rejection compares each pixel across all frames and throws out statistical outliers: satellite trails, planes, cosmic rays and hot pixels that move between frames because of dithering or field rotation.

What calibration frames remove

Darks

Same exposure, gain and temperature as lights, lens capped. They record thermal signal, amp glow and hot pixels, which are then subtracted. Because the ASI585MC is uncooled, its dark current changes with sensor temperature, so darks need to be taken near the same temperature as your lights.

Flats

Evenly lit frames that record vignetting, dust donuts and sensor non-uniformity. These are multiplicative effects, so lights are divided by the normalised master flat. Flats must be shot without changing focus or camera rotation.

Dark flats (or bias)

Darks matched to the flat exposure. They remove the camera's offset from the flats before division; otherwise the flat correction is wrong. For CMOS cameras, dark flats are more reliable than bias frames.

Diagram: Light minus Master Dark, divided by Master Flat, gives Calibrated Light
Light calibration: subtract the master dark, then divide by the master flat.
Diagram: flat frame minus master dark-flat gives calibrated flat
Flats are calibrated first, with darks of their own exposure.

calibrated = (Light − MasterDark) ÷ normalise(MasterFlat − MasterDarkFlat)

Sampling, drizzle and binning

To record a star's shape you need its blur (the FWHM) to span roughly 2–3 pixels. Fewer and stars become blocky squares; more and you're spreading the same light over extra pixels, which adds read noise for no extra detail.

Seestar: few pixels per star on sharp nights

At 2.39″/px, seeing alone would put a star on about one pixel. Your measured subs are softer, about 3 px, which is already well sampled, so drizzle only pays off on nights when SubframeSelector shows stars under about 2 px. On those nights, drizzle integration helps. Each sub lands on the sky slightly shifted (the alt-az mount's tracking drift and field rotation give natural dithering), so drizzle "drops" each input pixel onto a finer output grid. With enough frames, the 2× output recovers real resolution. It needs many subs: plan for at least a few hundred.

MK127: too many pixels per star

At 0.61″/px, stars are 3.5–5 px wide even in good seeing. Drizzle would only make it worse. Instead you can bin 2×2 in software after stacking (IntegerResample, average). That gives 1.23″/px, halves the file dimensions and doubles the per-pixel SNR with no real loss of detail.

Undersampled star field integrated normally, showing blocky stars
Original undersampled image, before drizzle.
Same field after drizzle integration, showing round stars
A simulated ideal 2× reconstruction of the same field: what drizzle tries to approach with enough dithered subs. Both from the DrizzleIntegration docs.

Why the order matters

The diagram shows the order used in Parts B and C. Teal steps run on linear data and red steps after the stretch. The arrows are hard dependencies, not preferences.

Where does BXT go relative to SPCC? BlurXTerminator's manual says it wants linear data "ideally right after integration, channel combination, and perhaps colour calibration and gradient removal". A common modern order is: Correct Only pass before SPCC (rounder stars make the photometry more accurate), then the sharpening pass after. Both orders work, and the difference is small.

Part B · Seestar S50

The Seestar does its own live stacking, but its saved sub-frames contain far more potential. Processing them in PixInsight gets you better rejection, drizzle, proper colour calibration and real deconvolution.

Each step below shows a before/after from your own M45 subs from 31 October 2026, processed with exactly the settings given.

Your M45 master straight from stacking, red cast, linked auto-stretch
Start: the stacked master (step 3).
Your finished M45 image after the full workflow
Finish: after step 13.

Quick path: start from the Seestar's stacked FITS

If you only kept the Seestar's own stacked file (Stacked_…fit), you can still use most of this guide. That file is already debayered, rotation-corrected and stacked. Open it, then skip to step 4 (Crop). You'll miss drizzle and pixel rejection, but gradients, SPCC, BlurXTerminator and the rest work. To get SPCC working, plate solve it first with Script › Astrometry › ImageSolver (250 mm, 2.9 µm).

01

Collect the sub-frames

RAW
  1. In the Seestar app, enable the option to save every sub-frame before you start imaging. Its name varies between app versions; look for "save all frames" or similar in the stacking settings.
  2. After the session, copy the target's …_sub folder from the Seestar (it appears as a network or USB drive, under MyWorks) to your Mac. Keep one folder per target and night.
  3. Open one sub and check its FITS header (File › FITS Header, or the Image › FITS Header window). Look for:
    • BAYERPAT: should read GRBG. WBPP reads it automatically.
    • EXPTIME: usually 10 s (newer firmware allows longer).
    • Filter and LP-filter state, so you know which SPCC curves to use later.
Why bother with subs? The Seestar's live stack uses simple averaging with limited outlier rejection. In PixInsight, every frame is measured, weighted by quality and rejected per pixel, and you can drizzle.
Calibration on the Seestar. The Seestar captures its own darks, and community reports say the saved subs already have them applied. There are no flats. So in WBPP you'll load lights only. Don't add darks unless you captured raw frames some other way, such as through N.I.N.A.
02

Inspect and cull bad frames

Blink · SubframeSelectorLINEAR

Open Blink, add the subs, and press the auto-STF button so they're visible. Play through them. Delete or move frames with cloud, heavy wind blur, or a dew-fogged lens.

For a measured approach, run SubframeSelector with Subframe scale 2.39 and look at the FWHM, Eccentricity and Stars graphs. A frame with half the normal star count is almost always cloud.

SubframeSelector Measurements window: a table of 62 M45 subframes and a graph of FWHM against frame index, with star count as a dashed line and an FWHM histogram on the right
Your install: SubframeSelector measuring your 62 M45 Seestar subs from 31 October 2026 (30 s, subframe scale 2.39). The blue line is FWHM and the dashed line is the star count. The first 16 frames have soft stars (FWHM 4.4–6.4 px, which is 10–15″) and fewer detections; after that FWHM settles around 3 px (7–8″) and the star count roughly doubles. Those 16 were culled for the example images below, along with frame 23, which had only 67 stars (probably cloud).
SubframeSelector: every setting explained routine, scale, star detector, expressions
SettingWhat it doesBest practice
RoutineMeasure Subframes measures every frame; Output Subframes writes approved frames with a weight keyword; Star Detection Preview shows what the detector sees on the first frame.Measure first, inspect the graphs, then either delete bad frames yourself or let WBPP do the weighting. You rarely need Output Subframes when stacking with WBPP.
Subframe scaleImage scale in ″/px. Converts FWHM from pixels to arcseconds; only used when Scale unit is arcseconds.Seestar 2.39 (your solve measured 2.38). MK127 + 0.65×: ~0.61 until you plate solve it.
Camera gainElectrons per data number. Converts noise and SNR figures to electrons; only used when Data unit is electrons.Leave 1 and Data unit Normalized unless you want absolute noise in electrons.
Camera resolutionBit depth of the subs.16-bit for both cameras (12-bit ADC stored as 16-bit).
Site local midnightUTC hour of local midnight, used with DATE-OBS to group frames by night.Leave 24 (unknown) unless you compare several nights.
Structure layers / Noise layersWavelet layers for star detection. More structure layers detect bigger stars; noise layers smooth before detection.Defaults (5 / 0). Raise noise layers to 1–2 only if hot pixels are being counted as stars.
Hot pixel filter radiusMedian filter applied before detection so hot pixels are not found as stars.Default 1. Keep it on for the uncooled ASI585MC.
Sensitivity / Peak responseMinimum SNR and peakedness for a detection.Defaults. Lower sensitivity if very few stars are detected on short subs.
Maximum distortionRejects detections whose shape is far from round (0 round, 1 any shape).Default 0.6. Raise to ~0.8 if trailed or wind-shaken frames show zero stars and you want to measure them anyway.
PSF fitFunction fitted to each star for FWHM and eccentricity (Moffat4 default).Default. Max PSF fits 8000 is plenty.
Approval expressionJavaScript-like test; frames failing it are unapproved, e.g. FWHM < 3.6 && Stars > 120.Write limits from your own graphs, not generic numbers. For your M45 subs, FWHM < 4 && Stars > 100 (pixel units) rejects exactly the frames culled above.
Weighting expressionFormula for a per-frame weight.Leave blank when stacking with WBPP: WBPP computes its own PSF Signal Weight.

Source: SubframeSelector parameter tooltips in your PixInsight 1.9.6 install.

Why. WBPP weights frames automatically, but a sky full of thin cloud passes weighting and still adds gradients and colour noise. Throwing away the worst 5–10% usually improves the stack.
Don't over-cull. Satellite trails and planes are fine to keep, since pixel rejection removes them. Only cut frames whose whole field is degraded.
03

Stack with WeightedBatchPreprocessing

Script › Batch Processing › WeightedBatchPreprocessingLINEAR

WBPP automates calibration, debayering, measurement, registration, local normalisation and integration. For the Seestar, most of the work is choosing the right options.

SectionSettingWhy
LightsAdd Directory → the _sub folder. No bias, darks or flats.The subs are already dark-subtracted (see step 1).
Lights › DebayerCFA images ✔, Bayer/mosaic pattern Auto, method VNGAuto reads BAYERPAT=GRBG from the header. VNG preserves edges better than bilinear.
Lights › Cosmetic correctionOptional: a CosmeticCorrection template with Auto detect on (hot sigma ~3)Catches residual hot pixels that dark subtraction missed, before they're smeared by debayering.
Image RegistrationDefaults. Pixel interpolation Auto, or the new Cardinal Quintic B-Spline option in 3.2.1Registration also derotates each frame, which handles the Seestar's field rotation.
Local NormalizationEnabled (default)Evens out the sky brightness changes between frames as the target moves through different parts of your light dome.
Image IntegrationRejection AutoWith hundreds of frames, WBPP selects a suitable statistical rejection algorithm.
Drizzle configurationUsually off. On sharp nights: Enable ✔, Scale 2, Drop shrink 0.90, Function SquareRecovers resolution only when the subs are undersampled. Your measured Seestar stars are about 3 px FWHM, which is already well sampled, so leave drizzle off (Scale 1) unless SubframeSelector shows stars under about 2 px and you have ~200+ subs. Otherwise the drizzled image is noisier and patterned with no gain in detail.
Autocrop✔Field rotation leaves ragged, partially covered edges; autocrop trims to fully covered pixels.
Astrometric solution✔ (focal length 250, pixel size 2.9)Gives the master a plate solution, which SPCC needs. WBPP 3.2.1 solves against Gaia DR3.
WeightedBatchPreprocessing 3.2.1 main dialog on the Calibration tab, with one Light group of 45 frames (1080x1920, 30 s, IRCUT) and empty Bias, Dark and Flat tables, Grouping Keywords, Registration Reference Image and Output Directory panels, and buttons along the bottom for adding files
Your install: WBPP 3.2.1 with your 45 culled M45 subs loaded through + Directory (bottom left). They form one Light group: 1080×1920, 30 s, IRCUT. The warning icon only means there are no calibration frames, which is expected for Seestar subs. The tabs along the top hold the settings in the table above; set the output directory at the bottom right, then press Run.
WBPP Drizzle configuration panel with Enable checked, Scale, Drop shrink 0.90, Function Square
WBPP's Drizzle configuration panel. The image shows Scale 1; use 2 only on sharp Seestar nights (stars under ~2 px).
Bayer/mosaic pattern drop-down set to Auto
Bayer/mosaic pattern: leave on Auto so the FITS header's GRBG is used.
WBPP: the settings that matter rejection, weighting, LN, drizzle, registration
SettingWhat it doesBest practice
Debayer methodHow colour is reconstructed from the Bayer mosaic.VNG. Bilinear is softer; SuperPixel halves resolution (useful only for the oversampled Mak).
Rejection algorithm (Auto)Removes outlier pixels when integrating.Leave Auto. WBPP 3.2.1 picks Percentile clipping for ≤ 6 frames, Winsorized sigma for 7–14 and Generalized ESD for 15+. Your 35-frame M45 stack used ESD.
Sigma low / highClipping thresholds for sigma-type rejection.Defaults. If satellite trails survive, lower high slightly; if dark pixels or edges look ragged, the low value is too tight.
Large-scale rejectionRejects big bright or dark blobs (planes, clouds) as a unit, not just single pixels.Turn on high for Seestar sessions near airports or with passing cloud.
Subframe weightingPer-frame weight used during integration.Default PSF Signal Weight. It rewards sharp, high-signal frames without manual formulas.
Minimum weightFrames below this fraction of the best weight are excluded.Default. Lower it if WBPPLogAnalysis shows good frames excluded.
Local normalizationMatches each frame's background to a reference in a grid, so gradients that change through the night don't confuse rejection.On. It matters most for alt-az Seestar sessions and multi-night data.
Registration referenceFrame all others are aligned to.auto. For mosaic-mode subs, choose a frame from the panel you want, because frames from other panels may fail to register (10 of yours did).
Distortion correctionSpline-based registration that corrects field distortion.On for mosaics and the reducer-equipped Mak.
Drizzle scale / drop shrink / kernelOutput upsampling, size of each dropped pixel, and the shape of the drop.Seestar with stars under ~2 px FWHM and 200+ subs: scale 2, drop shrink 0.9, Square. Otherwise (including your typical ~3 px Seestar nights) and always on the Mak: drizzle off.
AutocropCrops the master to the area every frame covers.On for normal sessions. Off, or use the uncropped master, for mosaic-mode subs.
Astrometric solutionPlate solves the master for SPCC/SPFC/MGC.On, with focal length and pixel size filled in. If it fails, solve manually with ImageSolver (step 7).

Source: WBPP 3.2.1 scripts and tooltips in your install (rejection rules from BPP-FrameGroup.js).

Pick an output directory and press Run. You'll find the result in master/; with drizzle enabled, use the file whose name includes drizzle.

Thousands of subs? For very large sessions, try FastBatchPreprocessing (FBPP, in the same menu) or WBPP's fast integration option. Both use the FastIntegration engine, which trades some rigour for speed. Its documentation notes that with very large frame counts, rejection algorithms converge and the results become nearly equivalent.
FastIntegration process dialog
FastIntegration, built for aligning and integrating thousands of short exposures.
WBPP rejecting too many Seestar frames? Open the log with Script › Batch Processing › WBPPLogAnalysis to see why frames were excluded, then lower the Minimum weight setting.
WBPPLogAnalysis frame list showing per-frame status
WBPPLogAnalysis lists every frame and the reason it was excluded.
Seestar mosaic-mode subs. Your M45 subs were shot in mosaic mode, so their pointings wander across about 0.7° × 0.5°. When WBPP stacked them, 10 of the 45 frames failed registration because they barely overlapped the reference, and autocrop cut the master down to a 370-pixel-wide strip: the only area every frame covered. Use the master without _autocrop in its name, and crop it yourself in step 4. To use all of a mosaic's area, process it as a mosaic instead: see Seestar mosaics.
Your M45 master after stacking, shown with a linked auto-stretch: a strong dark-red cast over the whole frame, with the bright Pleiades stars and faint nebulosity around the right-hand star
Your M45 data: the WBPP master from 35 of your subs (17.5 minutes), shown with a linked auto-stretch (STF). The red cast is light pollution plus the sensor's colour response; SPCC removes it in step 7. Every "before/after" image below continues from this master.
04

Crop the edges

DynamicCropLINEAR

Apply STF auto-stretch and look at the borders. Even after autocrop, there are often a few dark or noisy pixel rows from rotation and drizzle. Open DynamicCrop, drag a rectangle inside the clean area, and apply (green check).

Why now. Gradient tools and SPCC sample the whole image. Black edges or stacking artifacts make the gradient model wrong and throw off the background measurement.
DynamicCrop process dialog
DynamicCrop. Draw the rectangle directly on the image, then apply.
05

Remove light-pollution gradients

GradientCorrection · GraXpert · MGCLINEAR

Sky glow from towns, the Moon, or low altitude adds an uneven brightness ramp on top of your target. It's additive: it adds to every pixel rather than multiplying, so it has to be subtracted.

Default choice: GradientCorrection

Open GradientCorrection (find it under Process › All Processes, or by typing its name in the Process Explorer search). Start with all defaults and apply. Enable Generate gradient model to see what was removed; it should be a smooth ramp with no trace of your target.

ParameterDefaultAdjust when
High threshold0.05The process's most important control. Raise it if the gradient remains; lower it if the target is being eaten (overcorrection).
Low threshold0.2Rarely needs changing.
Low tolerance0.5Lower it in small steps if dark areas are overcorrected.
Structure protectionOn (threshold 0.10, amount 0.50)Keep it on for large bright objects (a big galaxy or M42); it stops the model from absorbing them. See the M31 example.
GradientCorrection process interface
GradientCorrection's interface.
M101 before gradient correction, with strong uneven backgroundM101 after gradient correction, with even background
M101 from the GradientCorrection docs. Top: the original. Bottom: the final correction after tuning the parameters; defaults alone left a greenish bubble around the galaxy.
GradientCorrection: every setting explained thresholds, scale, protection, tuning order
SettingWhat it doesBest practice
High thresholdHow much relatively bright background structure is allowed into the gradient model. Default 0.05.The most important control. Raise it if a gradient remains; lower it if the target is being eaten. Change in small steps.
High toleranceHow strongly the bright structures selected above contribute.Default. Raise slightly to recover a mild under-correction.
Low thresholdHow much dark structure is included in the model. Default 0.2.Lower it if dark nebulae are being flattened.
Low toleranceHow strongly dark structures contribute. Default 0.5 is neutral.Lower below 0.5 only if dark areas are over-corrected and structure protection doesn't fix it; small steps.
ScaleSpatial scale of the multiscale analysis. Default 5.0.Higher for smooth gradients and big nebulae; lower for sharp, complex light-pollution gradients (at the risk of over-correcting small objects).
SmoothnessSmoothness of the model. Default 0.4.Raise together with Scale to protect faint large nebulosity.
Automatic convergenceRepeats the correction until the model stops changing (usually 3–6 iterations).Off by default; try it on stubborn gradients.
Simplified model / degreeRemoves a low-degree polynomial gradient first, then models the rest.Use for very bright, complex gradients. Never with "natural" gradients like the Milky Way along one edge.
Structure protection: threshold / amountBuilds a mask to stop bright large objects (galaxies, M42) from being subtracted.Enable when you see a dark halo around the target. Inspect the protection mask so it covers the object, then raise amount if the outskirts are still over-corrected.
Generate gradient modelOutputs the subtracted model as an image.Always check it: it should be smooth, with no trace of your target.

Tuning order from the docs: try the defaults; if the gradient remains, turn off protection and try the simplified model at degree 1; then raise High threshold until the background is flat; then re-enable protection and tune threshold and amount. Crop dim or noisy edges first: the algorithm is sensitive to sudden brightness changes at the borders.

Source: GradientCorrection reference documentation in your install.

Alternative: GraXpert

GraXpert is built into your install (Process › All Processes › GraXpert). It uses AI to model the background. It's often excellent on Seestar data with complex, multi-directional gradients (city glow plus the Moon).

GraXpert process dialog in PixInsight
The GraXpert process in PixInsight.

Most accurate, if you install MARS: MultiscaleGradientCorrection

MGC compares your image with professional survey data (the MARS database) to tell real faint nebulosity from sky glow. It needs a plate-solved image and a SpectrophotometricFluxCalibration run first. It's the best option for faint, extended targets, but MARS isn't on your Mac yet (see setup).

SpectrophotometricFluxCalibration dialog with Ideal QE curve and Sony IMX 462-662-585-UVIRCut R, G and B filters, Gaia DR3/SP catalog
Your install: SpectrophotometricFluxCalibration, run first. Use the same QE curve and filters as SPCC (step 7).
MultiscaleGradientCorrection dialog with an empty MARS Database list, reference image field, gradient scale 1024, structure separation 3 and scale factors
Your install: MultiscaleGradientCorrection. The MARS Database list at the top is empty until you add .xmars files.
M45 master before gradient correction, red cast, linked stretch
Before: the master, cropped 12 px on each edge.
M45 after GradientCorrection at defaults, nearly identical to before
After: GradientCorrection at defaults.
Your M45 data: very little changes here. This 0.7° Seestar field has only a mild gradient, and the colour cast is not a gradient (it covers the whole frame evenly), so it stays until SPCC.
GraXpert and MGC: key settings
SettingWhat it doesBest practice
GraXpert: CorrectionSubtraction or division of the background model.Subtraction for light pollution. Division only for very strong multiplicative effects such as uncorrected vignetting.
GraXpert: AI modelWhich neural network models the background.Reset selects the recommended model. Let downloads finish.
GraXpert: SmoothingExtra blur of the AI background model.0.0 in almost all cases.
GraXpert: Sample-free scale / thresholdNon-AI mode: size of structures treated as background, and where structure protection starts.Higher scale protects smaller structures; lower threshold protects fainter structures.
MGC: MARS databaseProfessional survey reference used to separate real signal from sky glow.Required. Run SPFC first so flux scale factors are correct.
MGC: Gradient scaleLargest structure size treated as gradient (default 1024 px).Lower for complex gradients; keep high to protect large faint nebulae.
MGC: Structure separationHow isolated large structures are represented in the model (default 3).Default; lower to make relative structure intensities more similar.
MGC: Scale factorsPer-channel flux scale from SPFC.Leave at the SPFC values; adjust only if the corrected background is visibly tinted.

Sources: GraXpert reference documentation and MultiscaleGradientCorrection tooltips in your install.

Classic mistake. Running gradient removal on a nebula that fills the frame. Every gradient tool needs some real background to measure. If the target fills the frame, be conservative and look hard at the gradient model.
06

BlurXTerminator: correction pass

BlurXTerminatorLINEAR

Check Correct Only and apply. This makes stars round across the field, fixing small amounts of coma, astigmatism, residual chromatic aberration, tracking elongation and 2× drizzle artifacts, without sharpening.

Why before SPCC. SPCC measures star brightness through photometry. Round, compact stars give cleaner measurements. And the manual notes that Correct Only also cleans up the "two-pixel pairing" that 2× drizzle can introduce.
BlurXTerminator 2.6.11 dialog: ML Version Latest (ML 5), Sharpen Stars 0.25, Adjust Star Halos 0.00, Automatic PSF on, Sharpen Nonstellar 0.50, Correct Only off
Your install: BlurXTerminator 2.6.11 on its newest model (ML 5), shown with step 8's sharpening values. For this correction pass, tick Correct Only; the sharpening sliders are then ignored.
1:1 crop of two bright Pleiades stars before BlurXTerminator, enlarged 2x
Before BXT.
The same crop after BlurXTerminator Correct Only, stars very slightly tighter
After Correct Only.
Your M45 data: a 260 × 260 px crop, enlarged 2× without smoothing. The change is small because your stars were already round; the correction pass matters most on data with coma, tilt or drizzle artifacts.
07

Colour calibration with SPCC

SpectrophotometricColorCalibrationLINEAR

SPCC finds stars in your image, looks up each one's measured spectrum in Gaia DR3/SP, predicts how bright it should appear through your exact filter and sensor curves, and scales R, G and B until your measurements agree. The result is colour that is physically right rather than a matter of taste.

SettingSeestar, LP filter OFFSeestar, LP filter ON
White referenceAverage Spiral Galaxy
QE curveStarting point: Ideal QE curve, the usual choice when the R/G/B curves describe whole camera channels, as these appear to. If the calibration graph looks poor, try Sony IMX585 instead (same sensor family).
Red filterSony IMX 462-662-585-UVIRCut RSeestar-LP-R
Green filterSony IMX 462-662-585-UVIRCut GSeestar-LP-G
Blue filterSony IMX 462-662-585-UVIRCut BSeestar-LP-B
Background neutralization✔. Optionally draw a preview on empty sky and set it as the background reference.
Generate graphs✔. Points should follow a straight line; a scattered cloud means the fit failed.

These curve names are exactly as they appear in your install's filter database.

SPCC: every setting explained references, filters, catalog, photometry
SettingWhat it doesBest practice
White referenceSpectrum that should come out white.Average Spiral Galaxy. Hot (blue) references push colour red; cool (red) references push it blue.
QE curveSensor quantum efficiency, multiplied with the filter curves.Ideal QE curve when the filter curves already describe the whole camera channel, as the Sony and Seestar curves do.
Red / Green / Blue filterTransmission curve of each colour channel.Seestar LP off: Sony IMX 462-662-585-UVIRCut. LP on: Seestar-LP. ASI585MC with no filter: the same Sony UVIRCut curves.
Narrowband filters modeReplaces curves with a centre wavelength and bandwidth per channel.Duo-Band data: H-α 656.3 nm and O III 500.7 nm, bandwidths from ZWO's spec sheet.
Catalog / limit magnitudeGaia DR3/SP spectra source and faintest star used.Local Gaia DR3/SP with automatic limit magnitude.
Background neutralization: lower / upper limitBackground pixels between these limits, in sigma from the median, set the neutral sky. Defaults −2.8 / +2.0.Defaults. Use a Region of Interest on empty sky if the target fills the frame.
Generate graphsPlots measured against catalog colour ratios for each star.Always on. A tight straight line is a good fit; a scattered cloud means a bad solve or wrong filters.
Saturation thresholdStars with pixels above this are excluded from photometry.Default (relative). Keeps clipped Pleiades-type stars out of the fit.
PSF growthSize of the flux-measurement region, in units of the width at one tenth of the peak. Default 1.25.The tooltip notes that raising it can improve accuracy on undersampled images. Your Seestar stars (~3 px) don't need it; try 1.5 only on unusually sharp nights (under ~2 px) if the graphs look scattered.
Minimum SNR / Max starsDetection limit and cap on measured stars.Defaults.

Source: SPCC/SPFC parameter tooltips in your install; Seestar curve names from filters.xspd.

"Image has no astrometric solution." Run Script › Astrometry › ImageSolver first (focal 250 mm, pixel 2.9 µm for the Seestar). WBPP's astrometric solution option avoids this step.
ImageSolver script dialog
ImageSolver: the plate solution SPCC, SPFC and MGC depend on.
SpectrophotometricColorCalibration dialog: white reference Average Spiral Galaxy, Ideal QE curve, Sony IMX 462-662-585-UVIRCut R, G and B filters, Generate graphs on, catalog Gaia DR3/SP, background neutralization on
Your install: SPCC set for the Seestar with the LP filter off. For LP-on sessions, switch the three filters to Seestar-LP-R/G/B.
M45 before SPCC: strong red cast across the frame
Before SPCC.
M45 after SPCC: neutral grey background, white and blue-white stars, blue nebulosity around Alcyone
After SPCC.
Your M45 data: plate solved first with ImageSolver against your local Gaia DR3 (it measured 251.6 mm and 2.38″/px), then SPCC with the IMX462 UV/IR-cut curves shown above. Both images use the same linked auto-stretch, so the difference is the calibration alone: the red cast is gone, the background is neutral, and the reflection nebula's real blue appears.
08

BlurXTerminator: sharpening pass

BlurXTerminatorLINEAR
ControlSeestar starting pointWhat it does
Sharpen Stars0.20–0.30Shrinks star diameter. Too high gives hard dots with dark rings.
Adjust Star Halos0.00Positive softens halos (keeps star colour); negative tightens them.
Automatic PSF✔Measures the blur from stars in each 512×512 tile. Seestar fields have stars everywhere, so this works well.
Sharpen Nonstellar0.40–0.60Sharpens nebulae and galaxies. 1.0 attempts full deconvolution; real detail recovered is limited by SNR.
Correct OnlyoffOff for this pass.
BlurXTerminator: every setting explained stellar, nonstellar, options, input rules
SettingWhat it doesBest practice
Sharpen StarsReduces star diameter (FWHM).0.20–0.30. If dark rings appear, lower it or raise Adjust Star Halos. High values on long-focal-length data leave flat patches around bright stars.
Adjust Star HalosBrightens (+) or tightens (−) star halos.0 for the Seestar; −0.05 to 0 for the Mak's obstruction halos. Raising halos while shrinking stars keeps the colour of stars with clipped cores.
Automatic PSFMeasures the blur from stars in each 512 × 512 tile.On when stars cover the whole field. Off for galaxy cores or star-poor areas, which would otherwise be over-sharpened.
PSF DiameterManual FWHM (pixels) for nonstellar sharpening.Set to the measured star FWHM (SubframeSelector or a PSF measurement). Maximum 8 px; data above that is 2× oversampled and should be binned.
Sharpen NonstellarFraction of the PSF to remove from nebulae and galaxies; 1.0 tries for a point PSF.0.4–0.6. Real gain is limited by SNR.
Correct OnlyMakes stars round (coma, tilt, astigmatism, CA, guiding, 2× drizzle pairing) without sharpening.First pass, before SPCC. RC Astro found SPCC fits more consistent after it.
Luminance OnlyDeconvolves only luminance.Off. Only for forcing colour retention; disables chromatic corrections.
ML VersionNeural-network model.Latest (ML 5 in your install).

Input rules from the manual: linear data only (a hard rule with AI4 and later), 32-bit float, no noise reduction before it, and no classical deconvolution before or after it. Remove stars after BXT. For mosaics like your M45 subs, the manual recommends Correct Only on each panel before assembly, then re-solving, then SPCC and sharpening on the assembled mosaic.

Source: BlurXTerminator reference documentation in your install (© RC Astro).

Why not use a classic Deconvolution process? Classical deconvolution assumes one blur shape for the whole image. BlurXTerminator measures it locally, so it can sharpen the centre and corners differently. The manual also warns against running classical deconvolution before or after it.
Two hard rules from the manual. No noise reduction before BXT: it destroys the faint detail BXT recovers. And no stretch: with AI4 and later, input must be linear.
1:1 crop of two bright stars after SPCC, before sharpening, enlarged 2x
Before sharpening.
The same crop after BlurXTerminator sharpening: smaller star cores and tighter faint stars
After: Sharpen Stars 0.25, Nonstellar 0.50.
Your M45 data: the same crop, enlarged 2×. Faint stars become smaller and crisper and bright cores tighten, without dark rings. At 0.25 the effect is deliberately modest; at 2.39″/px there's little real detail left to recover.
09

Separate the stars

StarXTerminatorLINEAR

Apply StarXTerminator with Generate Star Image ✔ and Unscreen Stars off. You get two images: a starless one and a stars-only one. Both stay linear.

Why. A stretch hard enough to show faint nebulosity bloats and saturates stars. Separating them lets you stretch the nebula aggressively and the stars gently, then combine them with a screen blend in step 13.
Unscreen is for stretched images only. RC Astro's usage notes say Unscreen should not be used on linear data; on linear data the star image is a simple subtraction (original − starless). Also don't press auto-STF on the stars-only image: StarXTerminator passes STF information along with it, and auto-stretch overwrites it.
StarXTerminator: every setting explained and how to recombine
SettingWhat it doesBest practice
Generate Star ImageCreates a stars-only image alongside the starless result.On whenever you plan to put stars back.
Unscreen StarsBuilds the star image so a screen blend reverses the removal exactly.Stretched (non-linear) images only. Off on linear data, where the star image is original − starless.
ML VersionNeural-network model.Latest (ML 11).
OverlapOverlap between processing tiles (default 0.20).Default. Raise if you see tile seams.

Usage notes from RC Astro: run it as early as possible, ideally on linear data after deconvolution, because ArcsinhStretch, GHS, MaskedStretch and HDR processing change star profiles enough that stars may not be recognised. Don't auto-STF the stars image. Stretch the starless and stars images separately, then combine with a screen blend (step 13).

Source: RC Astro StarXTerminator usage notes (rc-astro.com).

StarXTerminator 2.6.11 dialog: ML Version Latest (ML 11), Generate Star Image checked, Unscreen Stars unchecked
Your install: StarXTerminator 2.6.11 on its newest model (ML 11), set for linear data: Generate Star Image ticked, Unscreen Stars off.
M45 starless image, auto-stretched: grey background with blue nebulosity around where Alcyone was
Starless (shown auto-stretched).
M45 stars-only image on black after a gentle stretch
Stars only (after step 10's gentle stretch).
Your M45 data: M45 is a reflection nebula wrapped around very bright stars, so star removal leaves soft blotches where Alcyone, Atlas and Pleione were. They disappear again when the stars go back in at step 13.
Galaxy targets. For galaxies (M31, M51, M81/82), star removal often damages compact cores and spiral detail. Skipping it is fine: stretch the whole image with step 10's settings and move on.
10

Stretch to non-linear

MultiscaleAdaptiveStretch · HistogramTransformation · GHSTHE STRETCH

This is the point of no return. For the theory behind this step, see Appendix A (histograms) and Appendix B (stretching). Save a linear copy (…_linear.xisf) first.

Starless image: MultiscaleAdaptiveStretch (new in 1.9.x)

ParameterDefaultGuidance from the docs
Target Background0.150.10–0.20 looks natural. Lower gives a darker sky.
Aggressiveness0.7Black-point clipping. Higher is brighter but noisier.
Dynamic range compression0.4Higher softens bright cores (useful for M42).
Contrast Recovery—Turn it on. The docs call it "a necessary step in most images", especially with high dynamic range compression.
Color SaturationoptionalIncludes protection against background chroma noise.
M42 stretched with a conventional histogram stretch, cores blown out
M42 stretched with MultiscaleAdaptiveStretch, core detail preserved
M42, an HDR example from the MultiscaleAdaptiveStretch docs. Left: MAS with Color Saturation enabled, which can create colour artifacts on very bright cores. Right: MAS with Color Saturation disabled, followed by HDRMultiscaleTransform and CurvesTransformation. For high-dynamic-range targets, leave Color Saturation off and add colour later.
MultiscaleAdaptiveStretch dialog: target background 0.150, aggressiveness 0.70, dynamic range compression 0.40, Contrast Recovery on with scale separation 1024, Color Saturation on with amount 0.75 and boost 0.50
Your install: MultiscaleAdaptiveStretch at its defaults, which match the table above. For HDR targets, untick Color Saturation.
Starless M45 shown through the screen auto-stretch only
Before: linear, viewed through the STF.
Starless M45 after MultiscaleAdaptiveStretch: dark grey background with the blue nebula visible but grainy
After: MAS, target 0.22, aggressiveness 0.95, DRC 0.30.
Your M45 data: the stretch is now permanent. MAS at its defaults left this faint nebula almost invisible, so these settings are stronger; that's typical for a short integration of a faint target. The background is grainy because 17.5 minutes is not much signal. Step 11 deals with that.
Faint target, short integration? If MAS at its defaults leaves the nebula barely visible, raise Target Background to about 0.22 and Aggressiveness to 0.9–0.95, and lower Dynamic range compression to 0.3. That's what your M45 data needed.
MultiscaleAdaptiveStretch: every setting explained stretch, background ROI, contrast recovery, saturation
SettingWhat it doesBest practice
Target BackgroundBackground level after the stretch (0–0.5). Default 0.15.0.10–0.20 for most targets. Raise to 0.20–0.30 for faint targets or short integrations; 0.25–0.40 when the object fills the frame. Above ~0.25 looks washed out.
AggressivenessBlack-point clipping: how close to the sky peak the clip sits. Default 0.7.0.4–0.8 normally. Higher gives more contrast but a noisier background; your M45 needed 0.95.
Dynamic range compressionCompresses bright areas to keep star profiles and cores. Default 0.4.Lower (0–0.4) for gentle targets; raise (0.7+) to protect bright cores. 0 equals a plain MTF stretch.
Background ReferenceRegion used to measure the background.Use a preview on empty sky when the target fills the frame (M42, Veil).
Contrast RecoveryMultiscale recovery of fine contrast lost to compression.On for nebulae and galaxies. Don't judge it on a small preview: results depend on image size.
Scale separationBoundary between "fine" and "large" structures (8–1024 px).128–256 for most images; 512–1024 to also protect large nebulosity. The docs' M45 example raises it to protect bright stars.
IntensityStrength of contrast recovery. Default 1.0.Reduce if bright cores show artifacts.
Color Saturation: Amount / BoostSaturation curve in HSV; Boost shifts it toward faint colours.Amount 0.6–0.7 to start. Disable for HDR targets and add colour later with Curves.

Troubleshooting from the docs: background too grey → raise Aggressiveness or lower Target Background. Too noisy → lower Aggressiveness. Extended objects too dim → raise Target Background.

Source: MultiscaleAdaptiveStretch reference documentation in your install.

Stars image: a gentler stretch

Use ArcsinhStretch (it keeps star colour; see Appendix B) or HistogramTransformation with a moderate midtones move. Stop when the faint stars appear. Over-stretching stars produces big white blobs.

ArcsinhStretch and HistogramTransformation: key settings for the stars image
SettingWhat it doesBest practice
Arcsinh: Stretch factorLogarithmic slider 1–1000; noise is amplified by the same factor.Stars image: 10–50. For more than 1000, run it twice (three runs of 6.7 ≈ one run of 300).
Arcsinh: Black pointCoarse and fine sliders; values below are clipped.In the real-time preview, click Estimate Black Point (clips 2%), then fine-tune.
Arcsinh: Protect highlightsScales the image so no channel clips.On for stars; the result is a little darker.
Arcsinh: Use RGB working spaceLuminance weights for the stretch.Off (equal weights).
HT: Shadows / Midtones / HighlightsBlack point, MTF balance and white point.For the stars image, midtones only (your M45 used 0.04) and no clipping. Watch the clipped-pixel count when moving shadows.

Turn the STF off before using ArcsinhStretch's preview: an active STF distorts what you see.

Source: ArcsinhStretch and HistogramTransformation reference documentation in your install.

Manual alternative: HistogramTransformation from the STF

Apply STF auto-stretch, then drag the STF's triangle icon onto HistogramTransformation's bottom bar, and apply HT to the image. This makes the screen stretch permanent. It's quick but blunt; GeneralizedHyperbolicStretch gives finer control.

HistogramTransformation annotated interface
HistogramTransformation, annotated.
11

Noise reduction

NoiseXTerminatorNON-LINEAR

Apply NoiseXTerminator to the starless image. Starting point: Denoise 0.6–0.8, other options at defaults. Compare the result with the original by zooming to 1:1 on faint areas.

NoiseXTerminator 2.6.11 dialog: ML Version Latest (ML 3.1), Denoise 0.70, Iterations 2, color and frequency separation off
Your install: NoiseXTerminator 2.6.11 on its newest model (ML 3.1), at Denoise 0.70.
1:1 crop of the stretched starless nebula before noise reduction: heavy colour grain
Before NXT (1:1).
The same crop after NoiseXTerminator 0.70: smooth background with wispy nebula structure visible
After NXT 0.70 (1:1).
Your M45 data: a 420 × 420 px crop at full resolution. The colour grain is gone and the nebula's wispy structure, which was buried in it, is now readable. The smooth patch in the middle is where StarXTerminator removed a bright star.
NoiseXTerminator: every setting explained denoise, iterations, separation
SettingWhat it doesBest practice
DenoiseFraction of noise removed (0–1).0.6–0.85. RC Astro warns that 1.0 can look unnaturally smooth.
IterationsSuccessive-approximation passes (AI3 only).Default 2. More can recover detail in very noisy areas but may add artifacts; inspect carefully.
Intensity/Color SeparationSeparate sliders for brightness noise (Denoise) and colour noise (Denoise Color).Colour noise is more objectionable: set colour higher than intensity, e.g. 0.75 / 0.95.
Frequency SeparationSeparate sliders for fine (HF) and blotchy (LF) noise, split at the HF/LF scale.LF usually lower than HF, especially on dust and faint nebulae. Tune the scale with Denoise LF at 0 in the real-time preview.
Both separationsFour sliders: HF intensity, HF colour, LF intensity, LF colour.RC Astro's starting point: HF intensity 0.80–0.90, HF colour 0.90–1.00, LF intensity 0.50–0.70, LF colour 1.00.
DetailSharpening during denoising.AI2 only; no effect on the current model.
ML VersionModel.Latest (ML 3.1).

From the manual: works equally on linear and stretched data (it stretches internally). Never before deconvolution of any kind. Use the real-time preview: the right amount is subjective.

Source: RC Astro NoiseXTerminator 2 AI3 user manual (rc-astro.com).

Why now, not earlier. Running it after BlurXTerminator protects the fine detail BXT needs. Running it after the stretch lets you judge it on what you actually see. NXT also works on linear data, so applying a lighter pass (Denoise ~0.5) before the stretch is a valid choice; just never before BXT.
Plastic look? That's over-reduction. Some fine grain looks more natural than a smeared background. Seestar data with under 2 hours of integration benefits from more noise reduction; 6+ hours needs very little.
12

Contrast and colour

CurvesTransformation · LocalHistogramEqualization · RangeSelectionNON-LINEAR
  1. Make a mask. RangeSelection with Lightness on: lower limit just above the background (read the background median in Statistics and add about one noise σ), fuzziness 0.1–0.2, smoothness 3–5. Apply it with Mask › Select Mask. White areas of the mask are changed; black areas are protected. Mask › Show Mask toggles the red overlay; Mask › Remove Mask when done.
  2. Curves. Darken the sky with an unmasked RGB/K curve first. Then, with the mask active, a gentle S-curve on RGB/K adds contrast and raising the S (saturation) curve boosts colour on the target only.
  3. Local contrast. LocalHistogramEqualization (kernel radius 64–128, amount 0.2–0.4) brings out structure in bright nebulosity.
Curves, LocalHistogramEqualization and RangeSelection: key settings
SettingWhat it doesBest practice
Curves: RGB/KTone curve on all channels.Gentle S-curve. Pull the lowest point slightly to darken the sky (your M45: 0.20 → 0.13).
Curves: S (saturation)Saturation curve.Lift the middle a little (0.5 → 0.6–0.65). Use a mask so the background doesn't gain colour noise.
LHE: Kernel radiusNeighbourhood size for local histograms.32–128. Smaller is stronger but noisier and rings around stars.
LHE: Contrast limitMaximum slope of the local transfer curve.1.5–3. Higher approaches plain equalization and boosts noise.
LHE: AmountBlend with the original.0.2–0.4 for natural results. Test on previews larger than twice the kernel radius; protect stars with a mask.
RangeSelection: Lower / upper limitPixel range selected into the mask.Lower limit just above the background level.
RangeSelection: FuzzinessSoftens the limits.0.1–0.2 to avoid hard edges.
RangeSelection: SmoothnessGaussian blur of the mask.3–5 so the mask follows structures, not noise.
RangeSelection: Lightness / Screening / InvertBuild from L*, multiply by the image, or invert.Lightness on for colour images; invert to protect instead of select.

Sources: CurvesTransformation, LocalHistogramEqualization and RangeSelection reference documentation in your install.

CurvesTransformation annotated interface
CurvesTransformation, annotated.
RangeSelection process dialog
RangeSelection, for building luminance masks.
Why masks. Every contrast or saturation boost also amplifies background noise. A mask confines the change to where there's real signal.
Starless M45 after noise reduction and a global curve that darkens the background
1 · Global curve: background 0.20 → 0.13.
RangeSelection mask: white over the nebula around Alcyone and its fainter extensions, black over the empty sky
2 · RangeSelection mask.
Starless M45 after the masked curve: deeper, more saturated blue nebula with the background unchanged
3 · Curve applied through the mask.
Your M45 data: two curves. First, an unmasked RGB/K curve darkens the sky (0.20 → 0.13). Then RangeSelection builds a mask on the result: lower limit 0.167 (the background median plus 0.8 × its noise σ), upper limit 1, fuzziness 0.15, smoothness 4, Lightness on. White areas receive the change; black areas are protected. With the mask active, a second curve lifts contrast (RGB/K 0.30 → 0.38) and saturation (S 0.50 → 0.74) only on the nebula, so the background gains no colour noise. Faint horizontal bands on the left come from the uneven overlap of the mosaic panels, not from processing.
13

Put the stars back

PixelMathNON-LINEAR

Open PixelMath, leave Use a single RGB/K expression checked, and enter this expression, replacing the names with your image IDs:

~((~starless) * (~stars))

Under Destination, check Create new image and apply to either image.

PixelMath dialog with the RGB/K expression ~((~starless) * (~stars)) and Use a single RGB/K expression checked
Your install: PixelMath with the screen-blend expression. Expand Destination at the bottom to set Create new image.
Why screen, not add. This is a "screen" blend: 1 − (1−a)(1−b). Unlike simple addition, it can never exceed 1.0, so star cores don't clip where they overlap bright nebulosity. It is also the blend RC Astro recommends for putting StarXTerminator stars back after stretching the two images separately.
PixelMath: settings for recombining stars
SettingWhat it doesBest practice
Expression (RGB/K)Formula evaluated per pixel. ~x means 1 − x.~((~starless) * (~stars)): a screen blend.
Use a single RGB/K expressionApplies the same formula to all channels.On for recombination; off for channel mixing (HOO).
Rescale resultStretches the output to fill 0–1.Off: a screen blend already stays in range, and rescaling changes your tones.
TruncateClips values outside 0–1.On (default).
Destination: Create new imageWrites to a new window instead of overwriting.On, with a clear Image ID, so both inputs survive.
SymbolsDeclares variables for longer expressions.Not needed here.

Source: PixelMath parameter tooltips in your install.

Too many stars? Multiply first: ~((~starless) * (~(stars*0.8))).

Starless M45 after curves
Before: starless.
Final M45 image: stars recombined over the blue nebulosity around Alcyone on a dark background
After: stars screened back in.
Your M45 data: the finished image after steps 2–13, from 35 × 30 s Seestar subs. The blotches left by star removal are covered by the stars again, and star colours are natural because the stars were stretched separately and gently.
14

Save and export

FINAL
  • Save the master as XISF (32-bit float). That's your archive; everything else is derived from it.
  • For sharing: Image › Sample Format › 16-bit (on a duplicate), then save as TIFF for further editing or JPEG (quality 90+) for the web. Make sure the colour profile is sRGB for web use.
  • Keep the linear master from step 10. If you learn a better stretch in six months, you start there.

Seestar mosaics

A mosaic covers more sky than one frame by combining frames aimed at different points. The Seestar makes them in two ways: its Mosaic framing mode, which moves between positions during a session, and longer sessions where the alt-az mount and field rotation make the pointing drift. Either way, the subs don't share one field, and a normal WBPP run quietly throws much of the data away.

Signs you have mosaic data

  • File names start Light_mosaic_.
  • The RA/DEC keywords differ by more than a few arcminutes between subs.
  • WBPP reports "registration failed" for some frames, and autocrop leaves a narrow strip.

Your M45 session

The 45 good subs' pointings spread over about 0.7° × 0.5°. Plain WBPP registered 35 of them and autocrop cut the master to 370 px wide. Processed as a mosaic, all 45 contribute and the field grows from about 0.70° × 1.25° to 1.31° × 1.27°.

For planning a mosaic (coverage, time per panel, targets) see Mosaics with the Seestar in Appendix C.

Route A: put every sub on one canvas (recommended for Seestar subs)

This is what produced the images below. Each sub is plate solved, reprojected onto a common canvas that covers the whole mosaic, and integrated with empty areas ignored.

  1. Calibrate and debayer only. Run WBPP on the culled subs as in step 3, but with Image Registration and Image Integration off. You only need the debayered folder.
  2. Plate solve every sub. Script › Astrometry › ImageSolver, with the subs added as target files: focal length 250 mm, pixel size 2.9 µm, distortion correction on, local Gaia DR3. All 45 of yours solved.
  3. Reproject. Script › Astrometry › MosaicByCoordinates: add the solved subs, leave the geometry automatic and press Calculate. Yours gave a 2160 × 2586 px canvas at 2.377″/px. Run it to write one reprojected image per sub.
  4. Integrate. ImageIntegration on the reprojected images with Clip low range on (range low 0) so the zero-filled areas are ignored, rejection Generalized ESD, normalization Additive with scaling, weights PSF Signal Weight.
  5. Crop to good coverage. The canvas edges are ragged and covered by only a few subs. Crop to the area most subs cover (step 4), then continue with steps 5–14 on the cropped mosaic. Plate solve the result again before SPCC.
The full mosaic canvas from your 45 M45 subs, auto-stretched: the whole Pleiades cluster with ragged black edges where no sub reached
The integrated canvas, before cropping. The black notches are areas no sub covered.
The cropped mosaic master, auto-stretched, with a red cast
Cropped to 2016 × 1980 px, before any processing.
Your M45 data: compare with the single-panel master in step 3, which covered only the left part of this field. Both use a linked auto-stretch.
The mosaic after gradient correction, BXT Correct Only and SPCC: neutral background and blue nebulosity around the brightest stars
After GradientCorrection, BXT Correct Only and SPCC.
RangeSelection mask of the mosaic: white over the nebulae around Merope, Maia, Electra and Alcyone
The step 12 RangeSelection mask, made with the same rule (median + 0.8 σ).
Finished M45 mosaic: the Pleiades cluster with blue reflection nebulosity around Merope, Maia, Electra and Alcyone on a dark background
Your M45 data, finished: 45 × 30 s subs (22.5 minutes) processed as a mosaic, then through steps 4–13 with exactly the settings used for the single-panel example. North is up. The brightest nebula is around Merope (lower centre); Maia is upper right, Electra right, Alcyone centre, and Atlas and Pleione are the pair on the left.
Disk space. Every reprojected sub has the full canvas size. Your 45 subs made 2.8 GB of reprojected files; a 1,000-sub night would need about 60 GB. For long sessions, stack the subs in groups of nearby pointings first (Route B), or delete the reprojected files once the integration is saved.
Uneven depth. The centre of a drifting mosaic gets many more subs than its edges, so noise varies across the frame, and faint banding can appear where coverage changes. Crop generously, and judge noise reduction at the edges as well as the centre.

Route B: stack each panel, then merge

Use this when the panels are distinct: planned multi-panel mosaics, panels shot on different nights, or very long sessions where Route A's disk use gets out of hand. It follows the mosaic workflow in BlurXTerminator's manual and the GradientMergeMosaic documentation. I ran it on the same 45 M45 subs, first with autocrop on each panel and then with autocrop off. The recipe below is the version that worked best.

  1. Group the subs into panels. Read each sub's RA/DEC keywords and group neighbouring pointings. Keep at least 8 subs per panel, or pixel rejection has too little to work with. The Seestar may revisit a position later in the night, so don't group by time alone. Your session visited about 15 positions with 2–4 subs each; grouping by position gave 4 panels of 8, 14, 8 and 15 subs.
  2. Stack each panel separately in WBPP with normal registration and local normalization, and autocrop off. Autocrop throws away each panel's outer strip, and in a mosaic those strips are the overlaps.
  3. On each panel master: crop 8 px from the edges, then BXT Correct Only. The corners of neighbouring panels have different aberrations, so they're corrected before the panels meet. Leave gradient correction until after the merge for nebula-rich targets: each panel is small, and the nebula can fill much of it.
  4. Plate solve each corrected panel with distortion correction and local Gaia.
  5. MosaicByCoordinates on the panel masters. List the panel you want to set the orientation first.
  6. GradientMergeMosaic on the reprojected panels: Average, shrink radius 10 (the maximum), feather radius 30, black point 0. Tick Generate Mask to see how the panels overlap.
  7. Crop to the fully covered rectangle, then GradientCorrection, plate solve, SPCC, BXT sharpening, and steps 9–14.
Uncropped panels need a bigger shrink and feather. With autocrop off and GradientMergeMosaic at its defaults (shrink 1, feather 10), your mosaic showed a green and purple blotch above Alcyone, left of Maia, and an exaggerated dark patch beside Alcyone. The panels' outer strips are covered by only a few subs, so they're noisy and slightly off in colour, and the defaults blended them straight in. Shrink 10 and feather 30 discard more of each border and blend over a wider band, which removed most of it. Moving gradient correction from each panel to the merged mosaic made no difference to these artifacts, but it's still the safer order for a nebula that fills each panel.
Contrast-boosted crop of the merged starless mosaic with a green blotch and a purple patch near a panel boundary
Shrink 1, feather 10 (defaults).
The same crop with shrink 10 and feather 30: the green blotch is mostly gone
Shrink 10, feather 30.
Your M45 data: the starless image after the global curve, with contrast boosted so the artifacts are easy to see. Both use uncropped panels.
GradientMergeMosaic result from four autocropped M45 panels, auto-stretched, on a stepped outline
The first attempt, with autocropped panels (auto-stretched, before SPCC).
GradientMergeMosaic mask: four overlapping grey rectangles of different brightness showing each panel's footprint
GradientMergeMosaic's mask: each grey level is a different combination of panels.
Your M45 data: with autocrop on, the four panels merged onto a 1729 × 2245 px canvas whose largest fully covered rectangle was only 1646 × 1441 px. With autocrop off, the canvas grew to 2008 × 2491 px and the usable rectangle to 1972 × 1620 px.
Finished Route B M45 mosaic: the Pleiades with blue nebulosity around Merope, Maia, Electra and Alcyone, nearly as wide as Route A
Your M45 data, Route B finished: uncropped panels, shrink 10, feather 30, and the same steps 9–13 settings as Route A. It's nearly as wide as Route A but a little shorter.
1:1 crop of Merope's nebula from Route A
Route A, 1:1 around Merope.
1:1 crop of Merope's nebula from Route B, nearly identical to Route A
Route B, the same area at 1:1.
Your M45 data: at full resolution the two routes are practically identical in detail and noise. The differences are coverage, disk space and effort.
Route A: one canvasRoute B, autocrop onRoute B, autocrop off
Subs used454545
Usable field (your M45)1990 × 1930 px (1.31° × 1.27°)1646 × 1441 px (1.09° × 0.95°)1972 × 1620 px (1.30° × 1.07°)
Reprojected files2.8 GB (one per sub)179 MB231 MB
Seams and artifactsNone: one integrationNone visible with defaultsNeeds shrink 10 and feather 30
Extra workPlate solve every subGroup subs, one WBPP run and plate solve per panel
Best forSeestar sessions whose pointing drifts; up to a few hundred subsDistinct panels, multi-night panels, very long sessions
MosaicByCoordinates, ImageIntegration and GradientMergeMosaic: key settings for mosaics
SettingWhat it doesBest practice
MBC: Input imagesSolved images to reproject. The first one sets the mosaic's rotation.Every image needs a solution computed with distortion correction and the local Gaia XPSD databases; online catalogs can't constrain the corners, where tiles overlap.
MBC: Projection / Center / Resolution / Rotation / DimensionsGeometry of the common canvas. Each is automatic unless you tick it.Leave all automatic and press Calculate to see the canvas size before running. Resolution defaults to the finest input, so never mix in one long-focal-length frame by accident: the canvas grows with the square of the scale ratio.
MBC: Pixel interpolation / ClampingResampling method and ringing suppression.Auto (Lanczos-3 for unchanged scale) and clamping 0.3.
MBC: Output directory / suffixWhere the reprojected images go (suffix _ra).A scratch folder with plenty of space: every output has the full canvas size.
II: Range rejection, clip low (range low 0)Excludes pixels at or below the low limit.Essential. Areas a frame doesn't cover are filled with zeros, and the docs warn that zeros treated as data cause dark seams and depressed backgrounds.
II: RejectionOutlier rejection.Generalized ESD for 15+ frames per pixel; Winsorized sigma for fewer.
II: NormalizationMatches frame levels before combining.Additive with scaling.
II: WeightsPer-frame weighting.PSF Signal Weight.
GMM: Type of combinationOverlay (later images hide earlier ones) or Average (averages overlapping pixels).Average for panels of similar depth; it also improves SNR in the overlaps.
GMM: Shrink radiusPixels trimmed from each panel's border (0–10).Default 1 for autocropped panels; 10 for uncropped panels, whose borders are noisy.
GMM: Feather radiusWidth of the blend across seams.Default 10; 30 for uncropped panels or when stars or colour blotches appear at the seams.
GMM: Black pointValues at or below this count as empty.0 for reprojected tiles.

Sources: MosaicByCoordinates and GradientMergeMosaic reference documentation in your install; ImageIntegration parameter tooltips.

Quick path and planning. The Seestar's own Stacked_…mosaic…fit is already assembled; for a quick result, start it at step 4 like any Seestar stack. To plan a mosaic for the MK127 rig, Script › Astrometry › MosaicPlanner lays out tiles on a reference image for a given optical train.

Part C · MK127 + ASI585MC

The processing after integration is the same as Part B. What changes is everything around it: you need your own calibration frames, the data is oversampled, and with the ZWO Duo-Band you're doing narrowband.

Capture habits that make processing easier

Build a calibration library

FrameHowHow manyReuse?
DarksSame gain, offset and exposure as lights. Cap the scope (or camera) in a dark place. Shoot sets at a few temperatures (for example 5 °C, 15 °C, 25 °C).30–50 per setYes, for months, if gain, offset and temperature match.
FlatsSame night, focus and rotation as lights. Use a light panel or T-shirt over the dew shield, with N.I.N.A.'s flat wizard targeting around 40–50% of the histogram.20–30No. Redo whenever focus, rotation or dust changes. The Mak + reducer vignetting makes flats essential.
Dark flatsSame exposure and gain as the flats, capped.20–30Yes, if flat exposure is unchanged.

WBPP with full calibration

Load everything at once with Add Directory. WBPP sorts frames by type, exposure, filter and binning from their FITS headers, and matches darks to lights automatically.

SectionSettingWhy
DarksOptimization off if exposure and temperature match; on if using a dark from a different temperatureOptimisation scales the dark to fit the light's thermal signal, partly compensating for temperature mismatch.
FlatsDark flats are matched to flats by exposure automaticallyRemoves the offset from the flats before they're used.
Lights › DebayerCFA ✔, pattern Auto, VNGN.I.N.A. writes BAYERPAT to the header.
Cosmetic correctionAuto-detect template, as in Part BUncooled sensors have more hot pixels.
DrizzleOffThe data is already oversampled.
Autocrop✔Removes dither edges.
Astrometric solution✔ (focal ~975, pixel 2.9; use your measured focal length once known)Needed for SPCC.
Calibrated frame with hot and cold pixels and a defective column
Same frame after CosmeticCorrection with defects removed
CosmeticCorrection before and after, on a crop of a single 600 s frame (from the docs).
LocalNormalization process interface
LocalNormalization, which WBPP runs for you. It matters more across multiple nights, when sky conditions vary between sessions.

Fix the oversampling: 2×2 binning

C1

Downsample the master

IntegerResampleOPTIONAL

Right after integration: IntegerResample, Downsample, factor 2, mode Average. The image goes from 3840×2160 at 0.61″/px to 1920×1080 at 1.23″/px.

Why. Averaging 4 pixels into one halves the noise per pixel (√4 = 2). You lose no real detail, because seeing already blurred anything finer. Every later step runs four times faster. BlurXTerminator's manual says images with FWHM above 8 px are over two times oversampled and can safely be halved; the Mak is close to that on poor nights.

Imaging without the reducer? See Reducer or native?: at native focal length, 3×3 binning gives about the same 1.2″/px.

When to skip it: on nights of excellent seeing (FWHM under 2″, which you can check in SubframeSelector), on small planetary nebulae where you want maximum scale, or when BlurXTerminator's PSF readout shows stars around 3 px or less.

IntegerResample process interface
IntegerResample.

Linear stage: same steps, different settings

Follow Part B steps 4–9. The differences are:

StepMK127 difference
GradientsThe 0.65° field is small, so gradients are usually a simple linear ramp. GradientCorrection at defaults is normally enough. Watch for overcorrection around galaxies; turn on structure protection.
BXTSharpen Stars 0.20–0.30. The Mak's large central obstruction spreads light into bright halos around stars: try Adjust Star Halos −0.05 to 0.00. Sharpen Nonstellar 0.5–0.7 suits small galaxies and planetaries. If a galaxy core has few stars around it, turn off Automatic PSF and set PSF Diameter to the measured FWHM in pixels.
SPCC (no filter or UV/IR cut)QE Ideal QE curve; filters Sony IMX 462-662-585-UVIRCut R/G/B; white reference Average Spiral Galaxy.
SPCC (Duo-Band)Your database has no ZWO Duo-Band entry. Either use SPCC's narrowband mode (H-α 656.3 nm and O III 500.7 nm, with bandwidths from ZWO's spec sheet), or skip SPCC and go to the HOO path below, which assigns colour by palette.

Worked example: a galaxy (M31, Seestar data)

About this data. These images use your Seestar M31 mosaic (902 × 10 s, 2.5 hours, 21 August 2026), because you haven't imaged with the MK127 rig yet. The Seestar had already calibrated and stacked it, so the calibration and Duo-Band parts of Part C can't be shown with it. What it does show is the galaxy-specific advice above, which applies to either rig: structure protection, the SPCC curves both cameras share, BXT's PSF setting, when binning helps, and stretching a galaxy without removing the stars.

Gradient correction: keep structure protection on

M31 fills much of the frame, which is exactly the case GradientCorrection's structure protection exists for. In your install it's on by default (threshold 0.10, amount 0.50).

Gradient model without structure protection: a bright diagonal band in the shape of the galaxy runs through the model
Gradient model, protection off.
Gradient model with structure protection: smooth background without the galaxy shape
Gradient model, protection on (default).
Your M31 data (Seestar): the models GradientCorrection subtracted, auto-stretched hard so their shape is visible. Without protection, the model contains a bright band in the shape of the galaxy: it would have subtracted part of M31's outer disc. With protection on, the model is a smooth background. Always check Generate gradient model on large targets.

SPCC with the shared Sony curves

The Seestar's IMX462 and your ASI585MC share one set of curves in the filter database, Sony IMX 462-662-585-UVIRCut R/G/B, so this is the setting you'll use for unfiltered Mak data too. The solve measured 251.7 mm and 2.38″/px.

M31 Seestar stack before processing, auto-stretched, with a warm cast
Before: the Seestar stack, cropped 16 px.
M31 after gradient correction, BXT Correct Only and SPCC: neutral background
After GradientCorrection, BXT Correct Only and SPCC.
Your M31 data (Seestar): both shown with the same linked auto-stretch. The warm cast is removed and the background neutralised; the galaxy keeps its yellow core.

BlurXTerminator: automatic or manual PSF

Part C suggests a manual PSF Diameter when a galaxy core has few stars around it. SubframeSelector measured this stack's star FWHM at 2.0 px, so the manual run used PSF Diameter 2.0 with Sharpen Stars 0.25 and Sharpen Nonstellar 0.50.

2x enlarged crop of M31's dust lanes and stars before BlurXTerminator sharpening
Before sharpening (2× enlarged).
The same crop after BlurXTerminator with a manual PSF of 2.0 px: slightly tighter stars and crisper dust lanes
After: manual PSF 2.0 px.
Your M31 data (Seestar): the dust lanes beside the core. The gain is modest at 2.38″/px. The automatic PSF gave a result indistinguishable from the manual one here, because stars cover the whole field, including around the core. Manual mode matters on long-focal-length data, where a galaxy can fill a 512 × 512 tile with few stars.

Binning: only when the data is oversampled

2x enlarged crop of a star field at full resolution
Full resolution (2× enlarged).
The same area after 2x2 binning, shown 4x enlarged to match: blockier stars
After 2×2 IntegerResample (4× enlarged to match).
Your M31 data (Seestar), as a counter-example: stars here are 2.0 px wide, so binning makes them blocky and loses real detail. Background noise fell by only 15%, not the 50% that averaging four independent pixels would give: registration and stacking had already correlated neighbouring pixels. On the Mak, with stars 3.5–5 px wide, binning loses no detail; check its noise gain on your own data the same way, with Statistics before and after.

Stretch: no star removal, and protect the core

Step 9's note applies: for galaxies, skip StarXTerminator and stretch the whole image. M31's core is far brighter than its disc, so raise MultiscaleAdaptiveStretch's Dynamic range compression.

M31 core after MAS at defaults: a small bright concentrated core
MAS defaults (DRC 0.40).
M31 core after MAS with dynamic range compression 0.60: the core's brightness spreads more gradually into the disc
DRC 0.60, Color Saturation 0.60.
Your M31 data (Seestar): the core at half size. With more compression the core rolls off gradually into the disc instead of forming a hard bright blob, and the dust lanes next to it keep their contrast. I also tried DRC 0.70 with saturation off, as the docs suggest for extreme HDR targets like M42; on M31 it left the core grey and flat, so a middle setting worked better.

Finish: noise reduction and a masked curve

NoiseXTerminator at 0.60, a gentle global curve to darken the sky (0.15 → 0.11), then a RangeSelection mask (lower limit = background median + 1 σ, fuzziness 0.15, smoothness 6, Lightness on) and a curve through it that adds contrast and saturation (S 0.50 → 0.72) to the galaxy only.

RangeSelection mask: the galaxy, its companions and the stars in white on black
The mask. Without star removal, stars are selected too; that's fine for colour.
Finished M31: yellow core, dust lanes, bluish outer disc, companions M32 and M110, on a dark background
Finished.
Your M31 data (Seestar), finished: GradientCorrection with protection → BXT Correct Only → SPCC → BXT (manual PSF 2.0 px) → MAS (DRC 0.60) → NXT 0.60 → curves through the mask. M32 is the compact companion below the core, and M110 the elongated one to the right.

ZWO Duo-Band → HOO palette

With a dual-band filter on a colour sensor, the red pixels record mostly H-α and the green and blue pixels record mostly O III (green and blue both respond near 500 nm). Extracting them gives two narrowband images you can combine into the classic HOO palette: H-α red, O III teal.

  1. After gradient correction and BlurXTerminator (still linear), run Script › Suite Astrocitas › DBXtract (installed with your PixInsight). It produces separate H-α and O III images. Script › SetiAstro › NBExtractPI is an alternative.
  2. Balance the channels. O III is almost always much fainter. Use NarrowbandNormalization or LinearFit to bring the two to a comparable level before combining.
  3. Combine. In PixelMath with Use a single RGB/K expression unchecked, set R = Ha, G = OIII, B = OIII, and create a new RGB image. For a richer blend, try G = 0.4*Ha + 0.6*OIII.
  4. Continue at Part B step 9 (star removal) and carry on as normal.
DBXtract script dialog
DBXtract: dual-band extraction for colour cameras.
Why extract at all? Treated as plain RGB, Duo-Band data looks red-magenta with weak teal. Extracting H-α and O III separately lets you stretch each to its own signal level, which reveals the O III structure that the red channel would otherwise swamp.

For SHO, pseudo-SHO, H-α blends and filter choice, see Appendix D.

Star colour in HOO. Narrowband stars come out pink or white. A common fix: take the stars from a short unfiltered RGB session (no Duo-Band), or reduce star saturation before recombining.

Cheat sheet

Printable: three-page version (page 1 processing, narrowband and planets, page 2 sky, angles, time and mosaics, page 3 capture, gear and data) · PDF

#StepProcessSeestar S50MK127 + 585MC
1CullBlink / SubframeSelectorDrop cloud and dew framesSame; check FWHM for focus drift
2StackWBPPLights only · drizzle only if FWHM < ~2 px · autocrop · solveDarks + flats + dark flats · no drizzle · solve
3BinIntegerResample—2× average (optional)
4CropDynamicCropRotation edgesDither edges
5GradientGradientCorrection / GraXpertDefaults; GraXpert if complexDefaults; structure protection on galaxies
6CorrectBXT Correct Only✔✔
7ColourSPCCSeestar-LP or IMX462 curves · Ideal QEIMX585 curves · or Duo-Band → HOO
8SharpenBXTStars 0.25 · halos 0 · nonstellar 0.5Stars 0.25 · halos −0.05 · nonstellar 0.6
9Stars outStarXTerminatorStar image ✔ · Unscreen off on linear data (skip for galaxies)
10StretchMAS / ArcsinhStretchStarless: MAS 0.12–0.15 background · Stars: gentle arcsinh
11DenoiseNoiseXTerminator0.7–0.80.5–0.7 (more integration)
12EnhanceCurves / LHE + maskGentle S-curve · saturation · LHE 0.2–0.4
13RecombinePixelMath~((~starless)*(~stars))
14ExportSave AsXISF master · 16-bit TIFF · sRGB JPEG

Histograms at a glance

Background in Appendix A.

ItemRuleWhat to look for
Value range0.0 (black) … 1.0 (white)Any format; masters are 32-bit float (~7 significant digits)
Main peakSky background = medianLinear stack: ~0.001–0.005. After stretch: aim 0.10–0.15
Peak widthNoise = MADNarrower is smoother; read both in Statistics
Right tailNebula just right of the peak, stars out to 1.0Faint signal sits only a few MADs above the median
Spike at 0Clipped shadowsLost data; keep to a tiny fraction of a percent
Spike at 1Clipped highlightsA few star cores are normal; many means over-stretch or overexposure
R, G, B peaks apartColour castNormal before SPCC; should line up after background neutralisation
Linear-only toolsUse values as physicsGradient removal, SPCC, BXT, rejection
Stretched-only toolsUse tonal positionCurves, LHE, saturation, RangeSelection masks
STF auto-stretchblack = median − 2.8 × MAD · background → 0.25Display only; data stays linear
STF linked / unlinkedLinked shows true colour; unlinked hides castsJudge colour linked, structure either way

Stretching at a glance

Background in Appendix B; settings in step 10.

ItemRuleNote
Midtones functionMTF(x) = (m − 1)x ÷ ((2m − 1)x − m)Input m → 0.5 output; m = 0.5 changes nothing; smaller m = harder
First stretchm ≈ 0.005Sky 0.002 → 0.285 (×143); star 0.9 → 0.999 (×1.1)
NoiseGain at the background applies to noise tooSNR (integration time) sets how hard you can stretch
Black pointClips everything below it, permanentlyWatch clipped-pixel counts in HT
ColourPer-channel curves push bright stars to whiteArcsinhStretch keeps colour ratios
StarlessMAS: background 0.15 · aggressiveness 0.7 · DRC 0.4 · contrast recovery onFaint/short data: 0.22 / 0.9–0.95 / 0.3. Galaxy cores: DRC 0.6
StarsArcsinhStretch 10–50, protect highlights; or HT midtones onlyStop when faint stars appear
HTBlack, midtones, white once; matches the STFQuick stretches and small tweaks
GHSContrast placed at a symmetry pointStaged, precise stretches
StrategySave linear first · starless and stars separately · two moderate stretches beat one extremeCheck at 1:1
Double-stretch trapReset the STF after a permanent stretchBlown-out look = STF still on

Arcseconds at a glance

Background and derivations are in Appendix C.

RuleYour numbers
1° = 60′ = 3600″1″ = 4.85 µrad · eye resolves ~60″
scale = 206.265 × px(µm) ÷ f(mm)Seestar 2.39″ · MK127 0.61″ · binned 1.23″ · guider 6.4″
FOV = scale × pixels ÷ 60Seestar 76′ × 43′ · MK127 39′ × 22′
FWHM ≈ √(seeing² + diffraction² + guiding² + …)The largest term dominates
Airy FWHM ≈ 1.03 λ/DSeestar 2.3″ · MK127 0.92″ (550 nm)
FWHM(px) = FWHM(″) ÷ scaleSeestar measured ~3 px (7–8″) · MK127 ~3.5–5 px at 2–3″ seeing
ideal scale ≈ FWHM ÷ 2…3< 2 px: drizzle (Seestar, sharp nights) · > 3 px: bin 2×2 (MK127)
FWHM ≈ 2 × HFRN.I.N.A. HFR to FWHM
guiding blur ≈ 2.35 × RMSKeep RMS ≤ scale ÷ 2: ~0.3″ unbinned, ~0.6″ binned on the Mak
sky motion = 15.04″/s × cos δCrosses an MK127 pixel in 0.04 s, a Seestar pixel in 0.16 s
1 mas = 0.001″Gaia parallax and proper motion units

Coordinates and units at a glance

Details and worked examples are in Appendix C. A printable three-page version of this whole cheat sheet (Letter, landscape) is also available as a PDF.

ItemRuleExample / note
Angle units1° = 60′ = 3600″41° 30′ = 41.5°, not 41.30°
DMS → decimald + m/60 + s/360051° 28′ 40.1″ = 51.47781°
Decimal → DMSRemainder × 60 → minutes; remainder × 60 → seconds0.47781° × 60 = 28.67′ → 0.67′ × 60 = 40.1″
SignsN and E positive; S and W negativeThe minus applies to the whole value: −0° 30′ = −0.5°
RA (time) → angle1ʰ = 15° · 1ᵐ = 15′ · 1ˢ = 15″M31 RA 00ʰ 42ᵐ 44.3ˢ = 10.6846°
RA on the sky1ˢ of RA = 15″ × cos(Dec)At Dec +41°: 11.3″
DecDegrees, + north / − south of the celestial equatorM31 Dec +41° 16′ 09″ = +41.2692°
FITS headersRA/DEC decimal degrees; OBJCTRA/OBJCTDEC sexagesimal text00 42 44.3, +41 16 09
Ground distance1″ latitude ≈ 31 m · 1′ ≈ 1.85 km · 1° ≈ 111 kmLongitude × cos(latitude)
Site precision2–3 decimal places (≈1 km–110 m)Enough for GoTo, flips and polar alignment
Polar alignmentPolar axis elevation = site latitudeGTi wedge set to your latitude
True vs magnetic northtrue = magnetic + declination (E +, W −)Look up magnetic declination (NOAA/NCEI WMM); take compass bearings away from steel
UTC and time zonesUTC = local − offset21:00 at UTC−5 = 02:00 UTC next day · FITS DATE-OBS is UTC · DST shifts the offset 1 h
Sidereal day23 h 56 min 4 sStars rise ~4 min earlier each night, ~2 h a month
Local sidereal timeLST = GMST + longitude ÷ 15 · HA = LST − RAHA = 0: transit, highest, meridian flip · negative = rising
Clock error → pointing1 s = 15″ · 1 min = 15′ · 1 h = 15°Check PC and SynScan time, zone and DST before a session
Maximum altitude90° − |latitude − Dec|Reached at transit (HA = 0)
Airmass≈ 1 ÷ cos(zenith angle)45° = 1.41 · 30° = 2.0 · 20° = 2.9 · stay above ~30°
Seeing vs altitudeFWHM × airmass0.6At 30°: stars ~50% wider
Refraction≈ 58″ × tan(zenith angle)45° = 1′ · 30° = 1.7′ · horizon ≈ 34′ · objects appear higher
Dispersion≈ 1.5% of refraction (450–650 nm)30° = 1.6″ (2.5 Mak px) · blue up, red down · ADC for planets
Field rotation (alt-az)15.04 × cos lat × cos az ÷ cos alt °/hSeestar: avoid within ~10° of the zenith
Sky brightnessmag/arcsec² (bigger = darker)1 mag = 2.512× · 5 mag = 100× · Bortle 8 ≈ 18.0 vs Bortle 2 ≈ 21.9 (36×)
Bortle → SQM1: 22.0 · 3: 21.7–21.9 · 4: 20.5–21.7 · 5: 19.5–20.5 · 6: 18.9–19.5 · 7: 18.4–18.9 · 8–9: < 18.4Approximate; Bortle is visual
Integration vs skytime ∝ sky brightnessEach 1 mag brighter ≈ 2.5× the time (faint broadband targets)
FiltersDuo-Band / Seestar LP: emission nebulae in any skyGalaxies, clusters, M45, dust: filter off, dark site, no Moon
Lunar cycle29.53 days · rises ~50 min later each dayBroadband window: last quarter → new → first quarter, Moon below the horizon
Moon distance≥ 60° broadband · ≥ 30° narrowbandMoonlight hurts O III more than H-α
DarknessSun > 18° below the horizonCivil 0–6° · nautical 6–12° · astronomical 12–18° · none above ~48.6° latitude at midsummer
Polar error → driftdrift ≈ error × 7.29×10⁻⁵ /s1° ≈ 0.26″/s · aim ≤ 5′ for the Mak · Polaris is ~37′ off the pole
Where the site goesN.I.N.A. (Options › General › Astrometry), GS Server; Seestar uses the phone's locationKeep N.I.N.A. and the mount in agreement; PixInsight only needs RA/Dec

Narrowband and filters at a glance

Details, palettes and PixelMath walkthroughs are in Appendix D.

ItemRuleExample / note
LinesH-α 656.3 · [O III] 500.7 · [S II] 671.6/673.1 · H-β 486.1 nmS II ≈ 5–20% of H-α · [N II] 658.3 sits inside H-α filters wider than ~3 nm
No filterGalaxies, clusters, reflection and dark nebulaeA narrowband filter removes their starlight
Duo-Band / Seestar LPH II regions, planetaries, SNRs, Wolf-RayetAny sky; Moon ≥ 30° away · O III suffers first
Broadband LPGalaxies under old sodium/mercury lightLittle help against LEDs · colour cast → SPCC
UV/IR cut · IR-passPlanet colour · planets in poor seeingASI585MC window already blocks IR for deep sky
HOOR=Ha · G=OIII · B=OIIIRicher: G = 0.4*Ha + 0.6*OIII
Pseudo-SHOR=Ha · G=0.6*Ha+0.4*OIII · B=OIIIGold H-α, blue O III without S II
SHOR=SII · G=Ha · B=OIIIThen SCNR green 0.6–1.0 (Average Neutral) → hue curves
H-α into RGBHa − k(R − med R), k 0.2–0.5Add w = 0.5–1 to R, after SPCC · or NBRGBCombination
WorkflowExtract (DBXtract) → gradients per channel → stretch each channel separatelyNo SPCC for SHO · NXT weak channels · RGB stars
IntegrationH-α : O III : S II ≈ 1 : 2 : 2Colour camera: H-α and S II on ¼ of pixels
Fast opticsPassband shifts blue: f/2 ≈ 5–9 nm, f/2.8 ≈ halfUse 6–7 nm or high-speed filters on the Samyang 135
SunFull-aperture solar film onlyNever an eyepiece "sun filter"

Mosaics at a glance

Planning details per rig are in Appendix C (Seestar, MK127, ASI585MC + 135, R8); processing in Seestar mosaics.

ItemRuleExample / note
Coverageframe × (N − (N−1) × overlap)Overlap 15% (ASI585MC, Seestar, Mak) · 20% (R8)
Seestar0.72° × 1.27° · 2×2 = 1.33° × 2.34° · 3×3 = 1.94° × 3.42°Each panel gets 1/N of the session · above ~2×2 use the 135
MK12739′ × 22′ · 2×2 = 73′ × 41′ · portrait 3×2 = 60′ × 73′~4 h/panel binned · N.I.N.A. Center every panel
ASI585MC + 1354.7° × 2.7° · 2×2 = 8.7° × 4.9° · 3×3 = 12.8° × 7.2°Heart & Soul 2×1 · Cygnus 3×3 · Orion ~4×4
R8 + 135 / 40015.2° × 10.2° · 2×2 = 27° × 18° / 5.2° × 3.4° · 2×2 = 9.3° × 6.2°Broadband: dust, Milky Way, galaxies · flats per lens
CaptureSame rotation, sub length, gain/ISO and total time per panelShoot western panels first · narrowband panels OK under Moon
Route ASolve every sub → MosaicByCoordinates → ImageIntegration (clip low)Seestar app mosaics and drift · disk = subs × canvas
Route BWBPP per panel (autocrop off) → crop · BXT correct → solve → MosaicByCoordinates → GradientMergeMosaicShrink 10 · feather 30 (50–100 on big mosaics) · Mak: bin each panel first
GradientsPer panel if open sky; after merge if nebula fills the panelsMGC + MARS best on the finished mosaic
SeamsFix gradients first, then depth, then featherDifferent seeing → BXT correct + cull by FWHM
Moon / landscapesNo stars to solve / foregroundMoon: AutoStakkert! → ICE/Hugin · Milky Way: Sequator/SLS → PTGui/Hugin

Capture and gear at a glance

Details in Appendix E.

ItemRuleExample / note
Sub lengthsub ≥ 10 × RN² ÷ sky e⁻/px/sMak 30–60 s / Duo 180–300 s · 135 f/2 15–30 s / Duo 60–120 s · Seestar LP on: 20–30 s
ASI585MC gain~252 (HCG on), default offsetOne gain/offset for every rig → one dark library
Focus tolerance±1.34 × N² µm135 f/2 ±5 µm · Mak f/7.7 ±80 µm · refocus each 1–2 °C
Polar alignmentTPPA, PHD2 static, SharpCapMak < 5′ · 135 < 10′
PHD23.75 µm, 120 mm → 6.4″/px · 1–2 s · Guiding AssistantRMS: Mak ≤ 0.3″ (0.6″ binned) · 135 rigs ≤ 2–4″
DitherIn guide pixelsMak 1–2 · 135 rigs 3–5 · every 1–3 subs
DewKeep optics 2–3 °C above ambientMak shield + strip · lens and guide strips · Seestar heater
Power (C300)~245 Wh usableMak rig ~8–10 h · 135 rig ~9–12 h
MK127Cool 45–60 min · 55 mm reducer back focusSolves 955–985 mm · collimate with 3 push-pull pairs
Reducer or nativeReducer + 2×2 ≈ native + 3×3 per pixelNative for planets, Moon, doubles, tiny planetaries
CalibrationDarks per exposure in 5 °C bins · flats + dark flats every sessionR8: dither + flats · Seestar: lights only
DataKeep lights, masters, project, TIFFDelete WBPP temp · 3-2-1 backup
CometsStar-register → CometAlignment → two stacksSub motion < 1 px
ScienceLinear, calibrated data onlyAAVSO · ETD/ExoClock · IOTA · MPC
ExportsRGB → resize → JPEG 90–95AnnotateImage for labelled versions

Troubleshooting

SymptomLikely causeFix
Strong green cast after stackingNormal for OSC sensors: there are two green pixels per Bayer cellSPCC fixes it. If SPCC failed, check the plate solve, then try again.
SPCC: no catalog / no stars matchedGaia DR3/SP not configured, or a wrong focal length or pixel size in the solveSee setup step 2. Re-solve with 250 mm (Seestar) or your measured Mak focal length.
Diagonal streaks ("walking noise")No dithering; hot pixels and pattern noise line up across framesDither in N.I.N.A. For the Seestar, more subs over a longer span help.
Dark rings around stars after BXTSharpen Stars too high, or halos too negativeLower Sharpen Stars, raise Adjust Star Halos.
Target faded or a dark hole around a galaxyGradient overcorrectionLower GradientCorrection's high threshold, or turn on structure protection.
Drizzled image looks noisy or cross-hatchedDrizzle used on data that isn't undersampled (stars ~3 px), or too few subsUse Scale 1. Drizzle only when SubframeSelector shows stars under ~2 px and you have 200+ subs. BXT Correct Only reduces mild drizzle artifacts.
Blotchy, plastic backgroundNoise reduction too strongLower NXT Denoise; accept some grain.
Dust donuts survive calibrationFlats shot after focus or rotation changed, or missing dark flatsReshoot flats without touching the optical train.
Mak stars bloated in some subsFocus drift (mirror moves as it cools)Cull by FWHM in SubframeSelector; autofocus every 1–2 °C in N.I.N.A.

Appendix A · Histograms and your image data

Read this appendix first; Appendix B builds on it. A histogram is the most honest view of your data: it shows what the numbers actually are, regardless of how the screen happens to display them.

What a histogram is

A histogram counts pixels by brightness. The horizontal axis is pixel value; the vertical axis is how many pixels have that value. PixInsight normalises every image to the range 0.0 (black) to 1.0 (white), whatever the file format:

FormatDistinct levelsWhere you meet it
8-bit integer256JPEG and your screen
12-bit (IMX462 / IMX585 readout)4,096Raw sensor ADC; stored in 16-bit FITS
16-bit integer65,536Subs from N.I.N.A. and the Seestar; TIFF exports
32-bit float24-bit precision (about 7 significant digits) at any brightnessEvery WBPP master and all processing

Stacking is why the 32-bit float format matters. Averaging hundreds of 16-bit subs produces values between the original integer levels. That extra precision is real information about faint signal, and only floating point keeps it.

Reading an astro histogram

A deep-sky histogram has a characteristic shape, and each part means something physical:

FeatureWhat it isWhat it tells you
The main peakThe sky background: most pixels in a deep-sky image are "empty" skyIts position is the sky brightness. In PixInsight it's measured as the median.
Width of the peakRandom variation of the background: noiseA narrow peak means a smooth background. Measured as MAD (median absolute deviation), a noise estimate that ignores stars.
Long tail to the rightNebulosity, galaxies, then starsThe faint target sits just right of the peak. Stars stretch far out to 1.0.
Spike at 0.0Clipped shadowsData crushed to pure black and lost. Normal only in cropped-away edges.
Spike at 1.0Clipped highlightsSaturated star cores. A few are unavoidable; many means overexposed subs or an over-hard stretch.
Separated R, G, B peaksColour castBefore colour calibration this is normal (OSC green, light-pollution orange). After SPCC with background neutralisation, the peaks should line up.

The linear histogram problem

On linear data, almost the whole image lives in a tiny sliver at the far left. A typical sky background on a stacked master sits around 0.001–0.005. Faint nebulosity might be 0.0005 above that. A bright star core can be 0.9 or more.

Simulated data: 60,000 pixels with a sky background at 0.002, faint nebulosity and a few stars. Both panels are drawn to scale with 128 bins. On the linear side, about 97% of all pixels fall in the very first bin (0 to 0.0078).

That has three practical consequences:

  1. You can't see it. A screen shows 256 levels. If the useful data spans 0.000 to 0.005, it maps to only the first one or two of those levels: black.
  2. You can't place controls on it. A curve point, a mask threshold, or a histogram slider would need to sit at something like 0.0023 versus 0.0027. In a histogram widget a few hundred pixels wide, that difference is less than one pixel of mouse travel.
  3. Histogram-shaping tools misbehave on it. More on that below.

Why some work needs stretched data and some needs linear data

Tools that use your image's statistics fall into two groups, and the difference explains the whole workflow order.

LINEAR Statistics as physics

Gradient correction, SPCC, BlurXTerminator, and pixel rejection during stacking calculate with the actual values: medians, noise estimates, star fluxes. They rely on the values being proportional to light. They don't care that the histogram is squashed, because they read numbers, not a picture.

A stretch would break them. SPCC compares star fluxes with Gaia predictions; after a curve, a star twice as bright is no longer twice the value, so the photometry is meaningless.

NON-LINEAR Histogram as picture

Curves, LocalHistogramEqualization, saturation, and RangeSelection masks change or select data by where it sits in the histogram. They assume the tonal range is spread out so that "shadows", "midtones" and "highlights" each contain meaningful structure.

On linear data, all of the nebula, the sky and its noise share the same few levels in the shadows. These tools can't tell them apart, so they either do nothing visible or amplify noise along with signal.

Some concrete examples:

ToolWhat it does with the histogramOn linear data
LocalHistogramEqualizationRedistributes values in each neighbourhood so they fill the range more evenly, boosting local contrastEach neighbourhood is mostly noise around the sky level, so it equalises the noise: a grainy, blotchy mess.
RangeSelection masksSelects pixels between a lower and upper valueNebula and sky differ by ~0.0005, which is within the noise. The threshold can't separate them cleanly, so the mask is speckled.
CurvesTransformationRemaps values through a curve you drawAll the interesting data lies under the first sliver of the curve. You can't place points accurately there.
Saturation boostsPush colour away from greyFaint areas have little colour signal relative to noise, so the boost mostly amplifies colour noise.
Key point. "Work that needs histograms" really means work that depends on the distribution of tones: tools that decide what to change by where a pixel falls between dark and bright. Those need the tonal range opened up first, which is exactly what a stretch does. Work that depends on the values themselves (physics) must happen before the stretch.

The STF: a stretch for your eyes only

The ScreenTransferFunction resolves the conflict. It applies a stretch to what is displayed while the underlying data stays linear. That's why you can see a linear image while running gradient correction and SPCC on it.

STF auto-stretch (the button in the toolbar) does this:

  1. Measures the median and MAD of each channel.
  2. Places the black point a little below the background: by default, 2.8 MADs below the median. That clips almost nothing real, but removes the empty range to the left of the peak.
  3. Picks a midtones value that moves the background to 0.25 on screen.

Those are the same three moves you'll make in a permanent stretch: black point, midtones, and sometimes white point. This is why dragging the STF's settings onto HistogramTransformation reproduces exactly what you see.

Don't trust the STF for colour or for final contrast. A linked STF shows the real colour balance (and casts). An unlinked STF stretches each channel separately, which hides casts. It's useful for judging structure, but misleading for colour. And because auto-STF is aggressive, it makes noise look worse than a careful final stretch would.

Histograms in PixInsight

Annotated diagram of the GeneralizedHyperbolicStretch graph: pre- and post-transform histograms, transformation curve, identity line, zoom controls and symmetry point marker
GeneralizedHyperbolicStretch's graph, annotated in its documentation. The outlined histogram is the image before transformation; the filled one is the estimated result; the red curve is the transformation itself.

Appendix B · Stretching in depth

Step 10 told you how to stretch. This appendix explains what a stretch is mathematically, why it's needed, what it costs, and how to choose between PixInsight's stretch tools.

Why stretch at all

Your camera and your eyes respond to light in very different ways:

A stretch is a non-linear compression of dynamic range. It gives the faint end of the data a much bigger share of the output levels and squeezes the bright end into what's left. It's the astronomical equivalent of what your eye does naturally.

The midtones transfer function

PixInsight's basic stretch, used by HistogramTransformation and the STF, is the midtones transfer function (MTF). For a midtones balance m between 0 and 1:

MTF(x) = (m − 1)·x ÷ ((2m − 1)·x − m)

Its key property: an input value equal to m comes out as exactly 0.5. Set m to your sky background, and the background is lifted to mid-grey. Smaller m means a harder stretch. At m = 0.5 the curve is a straight line and nothing changes.

Curves calculated from the formula above and drawn to scale. The red curve (m = 0.005) is typical of a first stretch on a linear stack.

A worked example

Take three values from a typical linear stack and apply an MTF with m = 0.005:

PixelLinear valueAfter MTF (m = 0.005)Change
Sky background0.0020.285×143
Faint nebula0.0040.444×111
Bright star0.9000.999×1.1

The nebula was 0.002 above the sky; now it's 0.159 above it, so the contrast between them grew about 80×. Meanwhile, the star barely moved. That asymmetry, huge gain in the shadows and almost none in the highlights, is the whole point of a stretch.

What a stretch costs: noise and clipping

Noise is stretched too

The slope of the curve at any point is its gain there. At the background, the red curve is extremely steep, so every bit of random noise around the sky level is amplified by the same factor as the faint signal. The stretch can't tell them apart.

This is why SNR decides how hard you can stretch. With little integration, the faint nebula is barely above the noise; stretching enough to see it also shows a grainy background. With more hours, the signal grows faster than the noise (√N, see Part A), so the same stretch reveals the nebula with a smoother background. No processing replaces integration time.

Clipping is permanent

Moving the black point right sets every value below it to zero. A little clipping removes the empty range below the sky peak and adds contrast. Too much destroys the faintest signal, which sits just above the background. In HistogramTransformation, watch the clipped-pixel counts: keep shadow clipping to a tiny fraction of a percent.

Moving the white point left saturates everything above it. That's rarely useful for deep-sky images; star cores are already near 1.0.

Colour can shift

Stretching R, G and B separately with the same curve changes the ratios between them. A star that is slightly orange in linear data (R a bit higher than B) gets pushed toward white, because its high values are all compressed toward 1.0. That's why heavily stretched stars look white and why some tools stretch only brightness.

PixInsight's stretch tools compared

ToolHow it worksStrengthsUse it for
HistogramTransformation (HT)Black point, midtones (MTF) and white point, applied onceSimple, predictable, matches the STF exactlyQuick stretches; reproducing an STF view; small final tweaks
ArcsinhStretchInverse hyperbolic sine curve applied to brightness, with the same factor applied to R, G and BPreserves colour ratios, so stars keep their colourStars-only images; colourful targets
GeneralizedHyperbolicStretch (GHS)A family of curves: you choose where contrast goes (symmetry point), how much (stretch factor), and how concentrated (local intensity), with optional shadow and highlight protectionPrecise control; can add contrast exactly at the nebula's brightness levelCareful, staged stretches when you want full control
MultiscaleAdaptiveStretch (MAS)Automatic stretch to a target background, with dynamic range compression and multiscale contrast recoveryReproducible; protects small bright structures and star profiles; good results in one stepThe default starless stretch in this guide; HDR targets like M42

Practical stretching strategy

  1. Save the linear image first. A stretch can't be undone, and you'll want to try again later.
  2. Stretch starless and stars separately (Part B, step 9). The nebula usually needs a hard stretch; stars need a gentle one, or they bloat into white discs.
  3. Aim the background at 0.10–0.15. The STF's 0.25 looks bright and washed out when printed or shown on a phone. A darker background leaves room for contrast.
  4. Prefer two moderate stretches to one extreme one when using HT or GHS. Check the histogram between them and stop when the peak is clear of the left edge with no spike at zero.
  5. Check at 1:1 zoom. A stretch that looks fine when zoomed out often shows background grain or hard-edged faint structures up close.
  6. Judge with the STF off. After a permanent stretch, reset the STF (or you'll be seeing a stretch on top of a stretch).
The double-stretch trap. If your image suddenly looks blown out right after a stretch, the STF is probably still active. Press the STF reset button to see the real result.

Appendix C · Arcseconds: seeing the unit behind every number

Image scale, FWHM, seeing, guiding error and field of view are all measured in arcseconds. This appendix gives you a feel for how small one is, then follows the unit through every stage from the atmosphere to the pixel, using your two rigs.

What an arcsecond is

A full circle is 360 degrees. Each degree is 60 arcminutes (′), and each arcminute is 60 arcseconds (″). So one arcsecond is 1/3600 of a degree, or 1/1,296,000 of a full circle.

A circle is also 2π radians, so one radian is 1,296,000 ÷ 2π = 206,265″. That's the constant in the image-scale formula. For small angles, the angle in arcseconds is simply size ÷ distance × 206,265:

angle (″) ≈ 206,265 × size ÷ distance

The image-scale formula in Your two rigs is the same equation: the "size" is one pixel (µm) and the "distance" is the focal length (mm). The ÷1000 for the unit change turns 206,265 into 206.265.

How small is one arcsecond?

Each object below covers one arcsecond when seen from the stated distance (from the formula above):

ObjectSizeDistance for 1″
A human hair~70 µm14 m: across a tennis court
A pencil, end on7 mm1.4 km
A US dime17.9 mm3.7 km
A US quarter24.3 mm5.0 km
A golf ball42.7 mm8.8 km

Your unaided eye resolves about 1′ (60″). One Seestar pixel is 1/25 of that, and one MK127 pixel is 1/100 of it.

The diagram below zooms in by 60× twice, from a degree down to an arcsecond. Everything is drawn to scale.

Drawn to scale. Betelgeuse is one of the largest stars in apparent size, and its disc is still 1/14 of an MK127 pixel. Every other star is smaller. The stars in your images are discs only because of blur, which is the subject of the rest of this appendix.

Angular sizes of familiar targets

ObjectAngular sizeIn Seestar pixels (2.39″)In MK127 pixels (0.61″)
Andromeda Galaxy, M31~190′ × 60′far bigger than the field (mosaic)far bigger than the field
Pleiades, M45~110′~2,760 (mosaic)~10,800 (mosaic)
Full Moon~31′ (1,865″)~780~3,040 (wider than the 22′ short side)
Orion Nebula, M42~85′ × 60′larger than the fieldlarger than the field
Dumbbell Nebula, M27~8′ × 6′~200 × 150~790 × 590
Whirlpool Galaxy, M51~11′ × 7′~280 × 180~1,080 × 690
Ring Nebula, M57~1.4′ × 1′~35 × 25~140 × 100
Jupiter (near opposition)~45″~19~73
Saturn's globe~18″~8~29
Albireo, separation~34″~14~55
Epsilon Lyrae ("Double Double"), each pair~2.3″~1 (unresolved)~4

Your fields of view are 76′ × 43′ (Seestar, 1920 × 1080 px at 2.39″) and 39′ × 22′ (MK127 + 0.65× and ASI585MC, 3840 × 2160 px at 0.61″). The Mak's field is narrower than the full Moon. Field of view is simply image scale × pixel count, ÷ 60 for arcminutes.

Where a star's size comes from

A star is a point, far smaller than any pixel. The disc you see is the sum of several separate blurs, each measured in arcseconds. They add roughly in quadrature (square, add, square-root), so the largest one dominates:

FWHM ≈ √(seeing² + diffraction² + guiding² + focus/optics²)

Blur sourceWhat it isSeestar S50 (50 mm)MK127 (127 mm)
SeeingTurbulence in the atmosphere; the same for both scopes on a given nighttypically 2–4″typically 2–4″
DiffractionLight passing through a finite aperture. FWHM of the Airy disc ≈ 1.03 λ/D (λ = 550 nm). Dawes limit ≈ 116 ÷ D(mm)2.3″ (Dawes 2.3″)0.92″ (Dawes 0.91″)
Guiding / trackingMount error during the exposure. A guiding RMS of σ adds about 2.35 σ to the FWHMalt-az tracking and field rotation; short subs keep it smallPHD2 total RMS 0.8″ ≈ 1.9″ of FWHM
Focus, optics, windDefocus, temperature drift, aberrations, vibrationvariesvaries; refocus as the Mak cools
Pixel samplingBlur from the pixel's own width, small once FWHM ≥ 2 pxsignificant at 2.39″negligible at 0.61″

So on a 2.5″ night:

No pixel scale beats the blur. Smaller pixels can't show detail finer than the FWHM. Once the star is 2–3 px wide, more pixels just spread the same light thinner, which is the reason for binning the Mak.

Pixels per star: sampling

Here is the same patch of sky, 14.4″ square, on each of your sensors, with a star of the stated FWHM centred on a pixel. Each square is one pixel, and its shade is the light that pixel collects. The dashed red circle is the FWHM.

Calculated by integrating a Gaussian star over each pixel and drawn to scale, with a square-root display stretch. Each panel is normalised to its own brightest pixel, so it shows shape, not brightness.

The usual rule of thumb is 2 to 3 pixels across the FWHM, so the ideal image scale is the FWHM ÷ 3 to the FWHM ÷ 2:

Star FWHM on the nightIdeal image scaleSeestar 2.39″MK127 0.61″MK127 binned 1.23″
1.5″ (excellent)0.5–0.75″/pxundersampledidealundersampled
2.5″ (good)0.83–1.25″/pxundersampledoversampledideal
4″ (average)1.33–2.0″/pxslightly undersampledoversampledslightly oversampled
7″ (Seestar measured)2.3–3.5″/pxidealoversampledoversampled

This is how to decide for each session. Measure the FWHM in SubframeSelector (below), then check the table. It's also why drizzle helps the Seestar only when its stars are genuinely small, on the best nights with many subs.

Arcseconds per second: how fast the sky moves

The sky turns 360° in one sidereal day (23 h 56 min), which is 15.04″ per second of time at the celestial equator. Further from the equator it's slower, by cos(declination). Without tracking, a star near the equator would cross:

A 120 s sub on the Mak needs the mount to cancel 1,800″ of sky motion with an error well under 1″, about 1 part in 2,000. That's what PHD2 and the GTi are doing.

Coordinates: degrees, minutes, seconds and decimals

The same 60-based system that gives arcminutes and arcseconds is used for positions on Earth (latitude and longitude) and in the sky (declination). Each value can be written in three ways, and your software uses all of them.

NotationExample (Royal Observatory, Greenwich)Where you meet it
Degrees, minutes, seconds (DMS)51° 28′ 40.1″ N, 0° 00′ 05.3″ WStar charts, catalogues, mount handsets, Dec in PixInsight dialogs
Degrees and decimal minutes (DDM)51° 28.668′ N, 0° 00.088′ WGPS units, some phone apps
Decimal degrees (DD)51.47781, −0.00147N.I.N.A. site settings, FITS headers, web maps, scripts

Converting

decimal = degrees + minutes ÷ 60 + seconds ÷ 3600

Going the other way: the whole number is the degrees; multiply the remainder by 60 for minutes, and multiply that remainder by 60 for seconds. For 51.47781°: 0.47781 × 60 = 28.6686′, then 0.6686 × 60 = 40.1″, giving 51° 28′ 40.1″.

Signs and hemispheres

41.30° is not 41° 30′. 41° 30′ is 41.5°. Typing minutes into a decimal field, or the reverse, is the commonest coordinate mistake. For latitude it moves your site by up to 22 km; for a declination it can put a small target outside the Mak's 22′ field.

Right ascension is measured in time

Declination uses degrees, but right ascension (RA) is written in hours, minutes and seconds of time, because the sky turns once a day: 24 h of RA = 360°. So:

RA unitEquals (at the celestial equator)
1 hour (1h)15°
1 minute of time (1m)15′
1 second of time (1s)15″

Away from the equator, lines of RA converge toward the pole, so 1s of RA spans 15″ × cos(Dec) of actual sky. At M31's declination (+41°) that's 11.3″; near Polaris it's almost nothing. Don't confuse a second of time in RA with an arcsecond.

Worked example, M31: RA 00h 42m 44.3s, Dec +41° 16′ 09″.

FITS headers often carry both forms: RA and DEC in decimal degrees, and OBJCTRA / OBJCTDEC as text such as 00 42 44.3 and +41 16 09. PixInsight's ImageSolver takes RA in hours, minutes and seconds and Dec in degrees, minutes and seconds, or fills them from the header or a Search by object name.

Arcseconds on the ground

One degree of latitude is about 111 km, so one arcminute is about 1.85 km (a nautical mile) and one arcsecond about 31 m. Longitude spans shrink by cos(latitude): at 45° they're about 0.71 of those values.

Decimal placesLatitude stepDistance (north–south)
1 (0.1°)6′~11 km
2 (0.01°)36″~1.1 km
3 (0.001°)3.6″~110 m
4 (0.0001°)0.36″~11 m

For astrophotography, two or three decimal places are plenty. Your site's position only needs to be good to about a kilometre: an error of 1 km shifts where the mount thinks the pole is by about 30″, well under what plate-solved GoTo and PHD2 correct.

Where your site coordinates are used

True north and magnetic north

An equatorial mount like the GTi has to point its polar axis at the celestial pole, which lies directly above true north. A compass doesn't point there.

True (geographic) northMagnetic north
Defined byEarth's rotation axis: the direction along a line of longitude toward the North PoleEarth's magnetic field, which a compass needle lines up with
Relation to the skyThe celestial pole is above it, at an altitude equal to your latitudeNone. It's an accident of the planet's molten core
StabilityFixedDrifts: the magnetic pole moves tens of km a year, and the local offset changes by up to a few tenths of a degree a year
Who uses itMaps, GPS, plate solving, N.I.N.A., your mount's alignment modelCompasses, including a phone compass set to "magnetic"

The angle between them at your location is the magnetic declination (also called variation). It has nothing to do with the declination of a star. It's measured east (positive) or west (negative) of true north, and across populated areas it ranges from about 0° to 20° or more either way.

true bearing = magnetic bearing + magnetic declination (east positive)

Example: if your declination is 10° W (−10°), a compass reading of 0° is actually 350° true, so true north is 10° to the right of where the needle points. Look up your current value with the NOAA/NCEI magnetic field calculator (World Magnetic Model). It changes over time, so check it every few years.

Steel throws compasses off. The mount, tripod legs, counterweights, a car or rebar in a concrete pad can deflect the needle by several degrees. Take the bearing a few metres away from the rig, pick a landmark in that direction, and aim the mount at the landmark. A phone compass with "true north" enabled applies declination for you, but it's just as easily disturbed by metal and needs calibrating.

Why it matters: polar alignment error

A misaligned polar axis makes stars drift in declination and the field slowly rotate. The worst-case drift is about the polar error × the sidereal rate (7.29 × 10⁻⁵ rad/s):

Polar alignment errorTypical causeWorst-case driftOver a 120 s sub
10° or moreCompass, magnetic declination ignored~2.6″/s~315″: unusable
2°Compass, declination applied, metal nearby~0.5″/s~63″
0.6° (37′)Aimed at Polaris itself, not the pole~0.16″/s~19″
5′Polar scope, or a quick plate-solved alignment~0.02″/s~2.6″
1′Plate-solved alignment refined (N.I.N.A. TPPA, SharpCap)~0.004″/s~0.5″

PHD2 can correct drift, but every correction is a chance for error, and field rotation can't be guided out at all. For the Mak at 0.61″/px, aim for 5′ or better. So use the compass, with declination applied, only to get the tripod roughly facing north. Set the altitude to your latitude, then refine with the polar scope or a plate-solve routine, which measures the true pole directly from the stars.

Polaris isn't the pole. In 2026 Polaris is about 0.6° (37′) from the true celestial pole, and slowly getting closer until around 2100. Polar-scope reticles and apps show where to place it on its small circle around the pole for the current time.

The Seestar in its normal alt-az mode doesn't need north at all: it plate solves and tracks in altitude and azimuth. North only matters if you use it on a wedge in equatorial mode, where the same rules apply.

Time zones and sidereal time

Telescopes run on two clocks. Civil time (UTC and your time zone) says when something happened. Sidereal time says which part of the sky is overhead. Mixing them up is behind many GoTo misses and confusing file timestamps.

UTC, time zones and daylight saving

WhereWhat it usesWatch for
FITS DATE-OBSUTC, ISO 8601, e.g. 2026-10-31T03:12:45: the start of the exposureNot your local time. Your 31 October M45 subs may show 31 Oct or 1 Nov depending on your offset
SubframeSelector Site local midnightThe UTC hour of your local midnightUTC−5 → 5. UTC+1 → 23. Only needed to group frames by night
N.I.N.A., GS Server, PHD2The mini PC's clock and time zoneSet the PC's time zone correctly and let Windows sync the time. The software converts to UTC itself
GTi (SynScan app)Date, time, time zone and DST you enter or take from the phoneA wrong DST setting is a one-hour error
SeestarYour phone's time and locationUsually automatic

Sidereal time: the sky's clock

Earth turns once relative to the stars in 23 h 56 min 4 s, a sidereal day. That's about 3 min 56 s shorter than the 24 h solar day, because Earth also moves along its orbit each day. As a result:

Local sidereal time (LST) is the right ascension currently on your meridian, the north–south line through the zenith. It's measured in the same hours as RA, and it ties RA directly to the clock:

hour angle (HA) = LST − RA

Hour angleWhere the target isWhat it means for imaging
Negative (e.g. −3 h)East of the meridian, risingTransit in 3 hours
0On the meridian (transit)Highest in the sky: least air, best seeing and least light pollution. A German equatorial mount like the GTi does its meridian flip here
PositiveWest of the meridian, settingGetting lower; plan to finish before it reaches the murk

LST comes from UTC and longitude. Greenwich sidereal time (GMST) is a formula of the UTC date and time, then:

LST = GMST + longitude ÷ 15 (longitude in degrees, east positive; hours)

Every 15° of longitude is one hour of sidereal time, the same 15°-per-hour rule as RA. You'll rarely compute it yourself: N.I.N.A. shows LST and each target's transit time in the Sky Atlas and Framing Assistant, and the mount computes it from its clock and site.

Worked example, M31 (RA 00h 42.7m) on 15 November 2026: GMST at 00:00 UTC is 3h 36m. M31 transits when LST = 0h 43m:

Sites near the middle of their time zone see a target transit at roughly the same local clock time. Sites near a zone edge are off by up to about 30 minutes, and DST adds an hour.

Clock errors are pointing errors

The mount converts RA to a direction using LST, so a clock error turns directly into a pointing error in RA:

Clock errorPointing error (at the celestial equator)Effect
1 s15″Negligible: plate solving corrects it
1 min15′Small target near the edge of the Mak's 22′ field
1 h (DST or time-zone mistake)15°GoTo lands in completely the wrong place; alignment stars aren't where expected
12 h (AM/PM mix-up)180°The mount points at the opposite side of the sky

Plate-solved GoTo in N.I.N.A. hides small errors, but a wrong time zone or DST setting breaks meridian-flip timing and altitude limits, and can send the mount below the horizon. Before a session, check that the mini PC and the SynScan app agree on time, time zone and DST, and sync the PC's clock while it still has internet. Offline in the field, a PC clock can drift by seconds a day.

Altitude, airmass and atmospheric refraction

Every photon you collect has crossed the atmosphere, and the lower the target, the more air it crossed. Altitude affects brightness, sharpness, colour and even where the target appears to be.

Altitude and how high a target gets

Altitude is the angle above the horizon (0° horizon, 90° zenith); azimuth is the compass direction, measured from true north through east. The zenith angle is 90° − altitude. A target reaches its highest altitude when it transits (hour angle 0, see above):

maximum altitude = 90° − |latitude − declination|

At latitude 40° N, M31 (Dec +41.3°) transits 88.7° up, almost overhead, while M42 (Dec −5.4°) only reaches 44.6°. A target whose declination is below latitude − 90° never rises.

Airmass: how much atmosphere

Airmass is the thickness of air along the line of sight, relative to looking straight up. Away from the horizon it's about 1 ÷ cos(zenith angle). The table uses the more accurate Kasten–Young formula, and shows what the extra air does:

AltitudeAirmassSeeing (FWHM × airmass0.6)Extinction, V band (~0.2 mag per airmass)RefractionColour dispersion, 450–650 nm
90°1.00× 1.000.20 mag00
60°1.15× 1.090.23 mag34″0.5″
45°1.41× 1.230.28 mag1.0′0.9″
30°2.0× 1.510.40 mag1.7′1.6″
20°2.9× 1.900.58 mag2.7′2.5″
10°5.6× 2.811.1 mag5.4′4.9″
0° (horizon)~38—~7.6 mag~34′—

Airmass from Kasten–Young; refraction from Bennett's formula at standard temperature and pressure; dispersion from the refractive index of air (Edlén). Extinction varies with site and haze; 0.2 mag/airmass is typical of a clear, dark site, and it's larger in blue than in red.

So a target at 30° is seen through twice the air of one overhead. Its stars are about 50% wider for the same seeing (2.5″ becomes ~3.8″), it's 0.2 magnitude dimmer, and it usually sits in the brighter sky glow near the horizon. That's why the 30° rule of thumb exists: below about 30° altitude, the data degrades fast.

Atmospheric refraction

Air bends light downward toward the denser layers, so every object appears higher than it really is. At moderate altitudes the lift is about 58″ × tan(zenith angle). At the horizon it reaches about 34′, more than the Moon's diameter: when the Sun appears to touch the horizon, it's geometrically already below it. Refraction changes by about 1% for every 3 °C of temperature or 10 hPa of pressure.

Atmospheric dispersion: refraction by colour

Blue light is bent more than red, so at low altitude each star is smeared into a tiny vertical spectrum, blue on top and red underneath. Between 450 and 650 nm the spread is about 1.5% of the refraction: 0.9″ at 45°, 1.6″ at 30°, 2.5″ at 20°.

Alt-az mounts and the zenith (Seestar)

An alt-az mount like the Seestar's has to rotate the field as it tracks. The rotation rate is:

field rotation (°/h) ≈ 15.04 × cos(latitude) × cos(azimuth) ÷ cos(altitude)

At latitude 40°, due south, that's 13° per hour at 30° altitude, 23°/h at 60°, 66°/h at 80° and 130°/h at 85°. Near the zenith the mount can't keep up, and short subs smear at the corners. Plan Seestar sessions so the target isn't within about 10° of the zenith; for the Seestar, the best part of the sky is roughly 30°–75° altitude. The GTi, being equatorial, has no field rotation and is happy overhead (except around the meridian flip).

Using altitude in practice

Light pollution, sky brightness and the Bortle scale

Sky brightness is measured in magnitudes per square arcsecond (mag/arcsec²): the brightness of one 1″ × 1″ patch of sky, on the same logarithmic scale as star magnitudes. Bigger numbers mean darker skies. A Sky Quality Meter (SQM) reads it directly, and light-pollution maps estimate it from satellite data.

The Bortle scale

John Bortle's 9-class scale (2001) describes the sky by what you can see with your eyes. The SQM and naked-eye limiting magnitude (NELM) values below are the commonly quoted approximate equivalents; Bortle is a visual description, so the boundaries are fuzzy.

ClassDescriptionSQM (mag/arcsec²)NELMWhat you see
1Excellent dark site21.99–22.07.6–8.0Zodiacal light, gegenschein and airglow; the Milky Way casts shadows
2Truly dark21.89–21.997.1–7.5Milky Way highly structured; M33 easy naked-eye
3Rural21.69–21.896.6–7.0Some glow on the horizon; Milky Way still complex
4Rural/suburban transition20.49–21.696.1–6.5Light domes over towns; Milky Way visible but washed out near the horizon
5Suburban19.50–20.495.6–6.0Milky Way faint overhead, gone near the horizon; clouds lit from below
6Bright suburban18.94–19.505.1–5.5Milky Way only at the zenith, if at all; sky grey-white
7Suburban/urban transition18.38–18.944.6–5.0Whole sky grey; M31 and M44 barely visible
8Citybelow ~18.384.1–4.5Sky orange or white; only bright constellations
9Inner citybelow ~18.04.0 or lessOnly the Moon, planets and a few bright stars

Bortle, J. E. (2001), "Introducing the Bortle Dark-Sky Scale", Sky & Telescope. SQM and NELM ranges are the commonly used approximate mapping.

Sky per pixel: your two rigs

Each pixel sees (image scale)² square arcseconds of sky, so the sky's brightness per pixel depends on the image scale. The rate it fills a pixel per second depends on the focal ratio:

Seestar (2.39″/px, f/5)MK127 (0.61″/px, f/7.7)MK127 binned (1.23″/px)
Sky area per pixel5.7 arcsec²0.38 arcsec²1.5 arcsec²
Per-pixel sky magnitudeSQM − 1.9SQM + 1.1SQM − 0.4
Sky signal per pixel per second (relative)2.4×1×4× (after binning)

The faster Seestar fills each pixel with sky glow about 2.4 times as fast as the Mak (that's (7.7 ÷ 5)²). So in the same sky, the Seestar's 10–30 s subs are already sky-limited, while the Mak needs longer subs to swamp its read noise.

What light pollution costs

On a faint target, the dominant noise is the shot noise of the sky glow itself. Noise grows with the square root of the sky brightness, so to reach the same SNR, the required integration time grows in direct proportion to the sky brightness:

time needed ∝ sky brightness → each 1 mag/arcsec² brighter ≈ 2.5× the integration

Moving fromToSky differenceIntegration for the same SNR (broadband, faint target)
Bortle 8 (18.0)Bortle 6 (19.2)1.2 mag~3× less
Bortle 8 (18.0)Bortle 4 (21.0)3.0 mag~16× less
Bortle 6 (19.2)Bortle 2 (21.9)2.7 mag~12× less

One hour from a Bortle 4 site does roughly the work of two nights in a Bortle 8 city on a faint galaxy. Light pollution also adds the gradients you remove in step 5 and the colour cast SPCC removes in step 7, and it makes noise reduction work harder.

Filters and target choice

Measuring your own sky

Moon phases and planning a session

The Moon is the brightest light-pollution source most imagers deal with, and it moves on a predictable schedule. Planning around it, together with darkness, altitude and weather, decides more about a night's data than any processing step.

The lunar cycle

The Moon goes from new to new in a synodic month of 29.53 days. It rises on average about 50 minutes later each day (anywhere from about 30 to 70 minutes, depending on season and latitude). Because its phase depends on its angle from the Sun, each phase has typical hours in the sky:

PhaseAge (days)IlluminatedRises / sets (approx.)Dark sky
New00%Up with the Sun, in daylightAll night
Waxing crescent3–510–30%Sets a few hours after sunsetFrom late evening
First quarter7.450%Rises ~noon, sets ~midnightAfter midnight
Waxing gibbous10–1370–95%Sets in the early morningOnly before dawn
Full14.8100%Rises at sunset, sets at sunriseNone
Waning gibbous16–1995–70%Rises in mid-eveningEarly evening only
Last quarter22.150%Rises ~midnight, sets ~noonBefore midnight
Waning crescent25–2730–10%Rises a few hours before dawnMost of the night

So the useful window for broadband imaging runs from about the last quarter, through new Moon, to about the first quarter: roughly two weeks a month, shrinking at each end as the Moon is up for part of the night. Mean phase dates for the coming weeks (each within about a day; check an almanac or N.I.N.A. for exact times):

New MoonFirst quarterFull MoonLast quarter
9 Nov 202617 Nov24 Nov2 Dec
9 Dec 202616 Dec24 Dec31 Dec

Your 31 October M45 subs were shot about 20 days into the cycle, with a waning gibbous Moon roughly three-quarters lit. That didn't matter much for the bright Pleiades stars, but moonlight raises the background and its noise, which is part of why the faint nebulosity needed such a hard stretch (step 10).

How much the Moon matters

Target typeMoon phase that worksNotes
Galaxies, reflection nebulae (M45), dust, faint clustersWithin ~5 days of new Moon, or with the Moon below the horizonBroadband, no filter: the Moon is the biggest single factor
Emission nebulae with the Duo-Band or Seestar LP filterAny phase; best away from fullKeep 30°+ away; favour H-α-rich targets when the Moon is bright
Bright star clusters, double starsAnyMoonlight mostly affects the background
The Moon itselfCrescent to gibbousDetail is best along the terminator; a full Moon is flat. The Mak and its 2.25× Barlow suit it
PlanetsAnySeeing and altitude matter far more; use the ADC

Darkness: twilight

The sky isn't fully dark at sunset. Twilight is defined by how far the Sun is below the horizon:

TwilightSun below horizonFor imaging
Civil0–6°Bright; set up, polar align roughly, focus on a bright star
Nautical6–12°Plate solve, start guiding, take flats if you shoot them at dusk; narrowband can start
Astronomical12–18°Broadband sky still slightly bright, especially toward the sunset
Astronomical darknessmore than 18°Full darkness: the broadband window

Near the summer solstice, latitudes above about 48.6° never reach astronomical darkness (the Sun gets no more than 90° − latitude − 23.4° below the horizon at midnight). Narrowband still works then.

Planning a session, step by step

  1. Pick the dark window. Astronomical darkness, minus the hours the Moon is up (for broadband). N.I.N.A.'s Sky Atlas shows twilight and Moon times for your site.
  2. Pick a target that's high in that window. Use its transit time (LST = RA, see time) and keep it above 30° (altitude), and at least 60° from a bright Moon. Targets transit about 2 hours earlier each month, so a target that transits around midnight is in season.
  3. Check the forecast. Cloud, plus transparency (haze) and seeing forecasts, such as Clear Outside, Astrospheric or Meteoblue's astronomy seeing. Transparency matters most for faint broadband targets, seeing most for the Mak at 0.61″/px.
  4. Choose the filter and sub length. Duo-Band when the Moon is up or the sky is bright; broadband in the dark window. Seestar 10–30 s; Mak 30–60 s unfiltered, 120–300 s with the Duo-Band.
  5. Plan the meridian flip. On the GTi, note when the target transits and let N.I.N.A. flip a few minutes after.
  6. Plan focus and calibration. Refocus every 1–2 °C as the Mak cools. Take flats without disturbing the optical train, and make sure your dark library matches the night's gain, exposure and temperature.
  7. Spread long projects. A faint target needs several dark windows. Shoot it on consecutive moonless nights, or across two new-Moon periods, and stack everything in WBPP, with local normalisation handling the different skies.
Automating it. N.I.N.A.'s Advanced Sequencer can wait for twilight, start when the target rises above an altitude, and stop at dawn. The Target Scheduler plugin adds Moon avoidance and picks the best target for the night from a list.

Planets with the MK127

Planets are tiny in arcseconds, but bright, which makes them the MK127's strongest targets. Its 127 mm aperture resolves about 0.9″, finer than most nights' seeing, and its long native focal length is made for them. Planetary imaging works differently from deep sky: you record short videos and keep the sharpest frames, so the gear setup and software change too.

Setting up the Mak for planets

ConfigurationFocal length / ratioImage scale (ASI585MC)Use it for
0.65× reducer (deep-sky setup)~975 mm, f/7.70.61″/pxNot for planets: too few pixels across the disc
Native, no reducer~1500 mm, f/11.80.40″/pxPoor seeing; the Moon
Baader Q-Barlow at 1.3×~1950 mm, f/15.40.31″/pxThe usual choice: close to the f/14.5 that suits 2.9 µm pixels
Baader Q-Barlow at 2.25×~3375 mm, f/26.60.18″/pxOnly excellent seeing (stars under ~1″); otherwise dimmer with no extra detail

Barlow factors are nominal and change with spacing; measure the real scale from a capture of a double star or of Jupiter's known diameter. The f/14.5 figure is the common lucky-imaging rule of f-ratio ≈ 5 × pixel size in µm.

Capture and processing

StageTypical settingsWhy
Capture softwareSharpCap or FireCapture (free), writing SER video. N.I.N.A. is built for deep-sky subs, not high-rate videoHigh frame rates from a small ROI
Exposure / gain3–15 ms, high gain; histogram peak around 50–70%Short enough to freeze the seeing
FormatRAW8 (bright planets) or RAW16, ROI about 640 × 480Debayer later; small ROI = high frame rate
Video lengthJupiter ≤ ~3 min; Saturn and Mars up to ~5 min; Venus and Moon unlimitedJupiter rotates in 9 h 56 min: at 45″, features move 0.9″ (the Mak's resolution) in about 3.5 minutes
StackingAutoStakkert!: best 5–20% of frames, multiple alignment pointsLucky imaging: keep only the sharpest moments
SharpeningRegiStax wavelets, or PixInsight MultiscaleLinearTransform / deconvolutionPlanets need much stronger sharpening than deep sky
Derotation (optional)WinJUPOSCombines several Jupiter videos into one sharper, lower-noise image

PixInsight's deep-sky workflow (WBPP, SPCC, BXT and so on) doesn't apply. BlurXTerminator isn't designed for planetary images; use wavelets instead.

What the MK127 shows, and when

Sizes, dates and visibility below were calculated for this guide from standard orbital elements (JPL approximate ephemeris), so dates are good to about a day. Declination tells you how high the planet gets: for northern-hemisphere sites, positive declinations are best.

PlanetBest windowSizeOn the ASI585MC at 0.31″/pxWhat 127 mm shows
JupiterOpposition ~11 Feb 2027 (Dec +15°): up all night. Morning sky now; evening from about March37″ now; 45″ at opposition~145 pxBelts and zones, the Great Red Spot, festoons, the four Galilean moons and their shadows crossing the disc
MarsOpposition ~20 Feb 2027 (Dec +16°). A distant (aphelic) opposition7″ now; 13.8″ at opposition; above 10″ from about early January to late March~45 pxPolar cap, dark albedo markings (Syrtis Major and others), clouds. Small this time: the next big opposition is 2033
SaturnEvening sky now, setting earlier each week; behind the Sun around March 2027; opposition ~18 Oct 2027 (Dec +7°)Globe 19″ now, 20″ at opposition; rings about 2.3× widerGlobe ~65 pxRings, the globe's shadow on them, Titan and several smaller moons. The rings are still fairly narrow after their 2025 edge-on crossing; the Cassini Division (~0.7″) is at the limit
VenusMorning sky now. Greatest western elongation ~4 Jan 2027 (47°); then a long, low morning apparition through spring50″ thin crescent now; 25″ half-lit in January; 10–17″ gibbous by spring~80 px at elongationPhases only; cloud markings need a UV filter. Image in twilight or daylight, when it's higher and the glare is lower
MercuryMorning elongation ~21 Nov 2026 (20° W); evening elongations around 3 Feb and 28 May 20275–8″~20 pxIts phase. Always low: find it in twilight and use the ADC
UranusOpposition ~26 Nov 2026 (Dec +21°): well placed all winter3.8″~12 pxA small blue-green disc; the brightest moons (Titania, Oberon) with long exposures
NeptuneEvening sky now; next opposition ~29 Sep 2027 (Dec +1°)2.3″~8 pxA tiny blue disc, and Triton beside it
This winter in short. Jupiter and Mars both reach opposition in February 2027, high in the sky for northern observers, and Uranus is at its best now. Saturn is an early-evening target until about January. For Jupiter's Great Red Spot, check a transit-time predictor (Sky & Telescope publishes one), because it's only facing Earth for a couple of hours per 10-hour rotation.

Seeing decides everything. Planets don't need a dark sky, so Moon phase and light pollution hardly matter. A steady atmosphere does: watch the Meteoblue or Astrospheric seeing forecast, image when the planet is highest (near transit), and avoid nights after a cold front or under the jet stream. The Seestar can also image planets in its planet mode, but at 2.39″/px Jupiter is only about 19 pixels across.

Double stars with the MK127

Double stars are the purest arcsecond test there is: two points of light a known distance apart. They need no dark sky and little integration, and the MK127 is well suited to them. Its long focal length and 127 mm aperture split pairs close to 1″. Many pairs are true binaries orbiting each other, so their separations and angles slowly change.

How close the MK127 can split

LimitFormulaMK127 (550 nm)What it means
Rayleigh1.22 λ/D ≈ 138 ÷ D(mm)1.09″Each star's peak sits on the other's first dark ring: a clear dip between them
Dawes116 ÷ D(mm)0.91″Empirical limit for two equal ~6th-magnitude stars: elongated with a slight notch
Sparrow≈ 0.95 λ/D0.85″The dip vanishes; the pair is just an elongated blob. Software can still measure it
SeeingStar FWHM on the nightUsually 1.5–3″Usually the real limit. Close pairs need steady nights, short exposures, or both

Ring and limit values were calculated for this guide from the diffraction pattern of a 127 mm aperture at 550 nm. Blue light splits slightly closer and red slightly wider.

The central obstruction in a Maksutov changes the pattern. Calculated for an obstruction of about a third of the aperture, which is typical of 127 mm Maks:

A faint companion 1–2″ from a bright star sits right on that ring and is easy to miss. Unequal pairs are therefore harder in a Mak than equal pairs at the same separation.

Sampling: which setup for which pair

SetupScalePixels across a 1″ pairComfortable for pairs wider than
MK127 + Q-Barlow 1.3× (planetary setup)0.31″/px3.2~1″: the close pairs
MK127 native0.40″/px2.5~1.3″
MK127 + 0.65× (deep-sky setup)0.61″/px1.6~2″
Seestar S502.39″/px0.4~8″ (its 50 mm aperture resolves ~2.3″, and stars are 7–8″ FWHM)

Use the planetary setup for anything under ~3″. For it to split cleanly, a pair should be at least about 3 pixels apart and wider than the night's FWHM.

Capturing

Measuring separation and position angle

Double stars are described by two numbers:

Measuring them is a real, publishable kind of amateur science; the Washington Double Star Catalog (WDS) accepts amateur measures. In PixInsight:

  1. Plate solve the stacked image with ImageSolver. This gives the exact scale and the orientation of north.
  2. Run DynamicPSF, click both stars, and read their centroid X and Y.
  3. Calculate the separation: ρ = √(Δx² + Δy²) × scale. Compare it with the catalogue value as a check on your scale.
  4. Calculate the PA. In an image with north up and east left:
    • the companion's north offset is −Δy and its east offset is −Δx, because image y increases downward;
    • θ = atan2(east, north), plus 360° if negative, then corrected by the solved rotation angle.

Without a plate solve, use the drift method. Switch off tracking while recording: stars drift west, so the trail direction is PA 270°.

Pairs to try

Separations and magnitudes below are approximate catalogue values. Orbiting pairs (marked ⟳) change by a little each year; check Stelle Doppie (the WDS-based online catalogue) for current figures. The "Season" column assumes northern mid-latitudes, around 9–10 pm local time.

PairSeparationMagnitudesSeasonWhy it's worth it
Albireo (β Cyg)34.7″3.1 / 5.1Summer–autumn; low in the west nowGold and blue: the classic colour pair. Seestar-friendly
Almach (γ And)9.6″2.3 / 5.0Autumn–winter: overhead nowOrange and blue. Seestar can just split it
Mesarthim (γ Ari)7.5″4.5 / 4.6Autumn–winter: nowMatched white twins
Achird (η Cas)13″3.5 / 7.4 ⟳All year (circumpolar); best in autumnYellow and reddish
Polaris (α UMi)18″2.0 / 9.1All yearFaint companion close to the bright star; ties in with polar alignment
ε Ari~1.4″5.2 / 5.5Autumn–winter: nowAn equal pair near the limit: a good seeing test
ζ Aqr~2.4″4.3 / 4.5 ⟳Autumn; early evening nowEqual pair, slowly widening
Rigel (β Ori)9.5″0.1 / 6.8Winter: from late evening nowThe test of glare: a 6.7-magnitude difference
Trapezium (θ¹ Ori)~9–21″ (four bright stars)5–8; E and F 10–11WinterStars E and F sit within ~4″ of the bright stars, inside the Orion Nebula. A transparency and seeing test
Alnitak (ζ Ori)~2.3″1.9 / 3.7WinterBright, close and unequal: a good test of the Mak's first ring
52 Ori~1.0″6.0 / 6.1WinterEqual pair at the Dawes limit: the MK127's ceiling
β Mon7.1″ and 2.8″4.6 / 5.0 / 5.4WinterA fine triple
Castor (α Gem)~5.5″1.9 / 3.0 ⟳Winter–springBright white pair
ι Cnc30″4.0 / 6.6Winter–springA "winter Albireo": gold and blue
Sirius (α CMa)~11″−1.5 / 8.4 ⟳Winter, at transitThe Pup, a white dwarf, is near its widest. It's only visible on the steadiest nights, in short frames with the primary saturated
Algieba (γ Leo)4.7″2.4 / 3.6 ⟳SpringGolden pair
Mizar (ζ UMa)14.4″2.2 / 3.9All year; best in springWith Alcor 12′ away. Seestar-friendly
Izar (ε Boo)2.9″2.6 / 4.8Spring–summerOrange and blue-green, close and unequal
Double Double (ε Lyr)~2.2″ and ~2.4″, the pairs 208″ apart5.0–6.1SummerTwo close pairs in one field: fits the 0.61″/px setup
Make it a check on your gear.
  • Measured scale: image a wide pair with a well-known separation, such as Mizar, Almach or Albireo, and divide by its separation in pixels. That calibrates the Q-Barlow's real factor (see planets).
  • Optics and seeing: splitting ε Ari or 52 Ori tells you your collimation and that night's seeing are both good, before you spend the night on Jupiter.

Deep-sky targets for the Seestar

The Seestar's strength is its wide field: 43′ × 76′ at 2.39″/px, about 1.4 × 2.5 full Moons. Target choice comes down to three numbers:

All three are just arcminutes and hours of RA.

Size: what fits

Target outlines drawn to scale from catalogue sizes and orientations; real nebulae have soft edges that extend further in deep images.

Target sizeOn the Seestar (2.39″/px, ~25 px per arcminute)What to do
Larger than ~70′Overflows the frameMosaic mode, or image the brightest part. See Seestar mosaics for processing
20–70′500–1,750 px: fills the frameThe Seestar's sweet spot: large nebulae, big galaxies, clusters
8–20′200–500 pxGood: medium galaxies, small nebulae. Crop in processing
3–8′75–200 pxRecognisable but soft. With stars 7–8″ FWHM (about 3 px), you get shape, not detail
Under ~3′Under 75 pxSmall planetary nebulae and galaxies: a better job for the MK127 at 0.61″/px

LP filter: on or off

The Seestar's built-in LP filter is a dual-band filter: it passes only narrow bands around H-α and O III.

Use the matching SPCC filter curves for each (step 7).

Season: when it's up in the evening

A target is highest when its RA equals the local sidereal time (see sidereal time). At about 9 pm standard time, mid-month, the sidereal time is roughly:

MonthJanFebMarAprMayJunJulAugSepOctNovDec
Transiting RA4ʰ6ʰ8ʰ10ʰ12ʰ14ʰ16ʰ18ʰ20ʰ22ʰ0ʰ2ʰ

The sidereal time advances 2ʰ per month, or 1ʰ per hour of the night. Under daylight saving time, add 1ʰ to the clock time. For a northern site at latitude φ, a target's maximum altitude is 90° − |φ − Dec|. Below about 30° (Dec south of about φ − 60°) seeing, extinction and light pollution all get worse (see altitude).

Targets by season

Sizes are approximate catalogue values for the visible extent. "Best" is the month the target transits around 9 pm; it's well placed for about two months either side. Rows are for northern mid-latitudes.

TargetTypeSize · Seestar pxFitLPBestNotes
Autumn–winter: now
M31 Andromeda GalaxyGalaxy178′ × 63′ · 4,470 px longMosaic (2–3 panels)OffNovOne frame holds the core with M32 and M110. Bright: 1–2 h gives a good result
M33 Triangulum GalaxyGalaxy73′ × 45′TightOffDecLow surface brightness: 3 h or more, dark moonless sky
NGC 869 / 884 Double ClusterOpen clustersPair spans ~60′FitsOffDecBoth clusters in one frame. 30–60 min; great under moonlight
NGC 281 PacmanEmission35′ × 30′FitsOnNovCircumpolar from mid-northern sites
NGC 7635 BubbleEmission15′ × 8′ · ~380 pxFitsOnNovShares the field with M52 if framed between them
IC 1805 HeartEmission~2.5°MosaicOnDecIts core cluster (Melotte 15) region fits one frame
NGC 891Galaxy13′ × 3′ · ~330 × 75 pxFits (small)OffDecEdge-on with a dust lane; a test of seeing and focus
M45 PleiadesCluster + reflection~110′MosaicOffJanYour Part B target. Reflection nebulosity is blue starlight, so the LP filter would remove it
NGC 1499 CaliforniaEmission145′ × 40′Mosaic (2 panels)OnJanFaint, nearly pure H-α: 3 h+
M1 CrabSupernova remnant6′ × 4′ · ~150 pxSmallOnFebShape only; filaments need the MK127
M42 / M43 Orion NebulaEmission85′ × 60′Core fits; 2 panels for allOn (or off for natural colour)FebVery bright core: the Seestar's 10 s subs help. 30 min already looks good
IC 434 / NGC 2024 Horsehead & FlameEmission + darkBoth within ~50′FitsOnFebFrame with Alnitak just off-field; it causes glare. 2 h+
M78Reflection8′ × 6′FitsOffFebDarker sky needed; dust lanes around it
NGC 2237 RosetteEmission~80′1×2 mosaic or tightOnFebStrong O III and H-α: dual-band shines here
NGC 2264 Cone & Christmas TreeEmission + cluster~60′ × 40′FitsOnFebFox Fur detail needs 3 h+
M35 (+ NGC 2158)Open clusters28′FitsOffFebA rich cluster with a distant compact one beside it
NGC 2359 Thor's HelmetEmission16′ × 8′FitsOnMarDec −13°: low from northern sites; strong O III
Spring: galaxy season
M81 + M82Galaxies27′ and 11′, 38′ apartFitsOffAprCircumpolar: possible all year
M97 + M108Planetary + galaxy3′ and 9′, 48′ apartFitsOffAprAn odd but appealing pair
Leo Triplet (M65, M66, NGC 3628)GalaxiesGroup ~45′FitsOffAprA Seestar classic
Markarian's ChainGalaxies~90′Mostly fits; 2 panels for allOffMayA dozen galaxies in one field
M51 WhirlpoolGalaxy11′ × 7′ · ~275 pxFitsOffJunSpiral arms visible; best as a close crop
M101 PinwheelGalaxy29′FitsOffJunLow surface brightness: 3 h+
M104 SombreroGalaxy9′ × 4′SmallOffMayDec −12°: catch it at transit
Summer
M13Globular cluster20′FitsOffJulBright; 30–60 min. Short subs keep the core unsaturated
M8 Lagoon (+ M20 Trifid, 1.4° north)Emission90′ × 40′Mostly fits; mosaic for bothOnAugDec −24°: low from mid-northern sites
M16 Eagle / M17 SwanEmission35′ / ~40′FitsOnAugDec −14° / −16°
M27 DumbbellPlanetary8′ × 6′SmallOnSepThe Seestar's best planetary nebula: bright with strong O III
M57 RingPlanetary1.4′ · ~35 pxTinyOnAugA ring, but just; one for the MK127
NGC 6888 CrescentEmission18′ × 13′FitsOnSepH-α shell with an O III halo in long integrations
NGC 6960 / 6992 VeilSupernova remnant~3° loopMosaic, or one arc per frameOnSepEach arc fits one frame; fine filaments in H-α and O III
NGC 7000 North AmericaEmission120′ × 100′Mosaic, or the "Gulf" / Cygnus WallOnSepBright H-α; the Wall region fits one frame
IC 1396 Elephant's TrunkEmission~170′Mosaic, or the Trunk aloneOnOctCircumpolar from mid-northern sites
NGC 7293 HelixPlanetary25′FitsOnOctLarge but faint; Dec −21°, low from the north

"Best" months were calculated from each target's RA with the sidereal-time rule above. Integration times are rough guides for a Bortle 5–6 sky and 10 s subs.

Tonight's short list (mid-November, evening).
  • Moonless window: M31, M33, the Double Cluster and NGC 891.
  • With the LP filter, under any Moon: the Pacman, Bubble and Heart nebulae.
  • Later in the night: M42, the Rosette and the Horsehead rise.
For broadband targets, use the Moon section to pick dates. Check each target against the sky-brightness integration table before committing a whole night.

Deep-sky targets for the MK127

The MK127 complements the Seestar. Its field is only 39′ × 22′ at 0.61″/px, about a quarter of the Seestar's area. In return it puts about 4 times as many pixels across each arcminute, and it collects 6.5 times as much light in total. Its targets are the small ones the Seestar renders as smudges:

Size: what fits

Each frame is the MK127 + 0.65× + ASI585MC field, 39′ × 22′, north up; shaded = inside, dashed = outside. Outlines are drawn to scale from catalogue sizes. Pixel counts are at the native 0.61″/px (~98 px per arcminute; ~49 when binned 2×2).

Target sizeOn the MK127 (0.61″/px)What to do
Larger than ~35′Overflows the frameA Seestar target, or image one part of it: a pillar, a wall, an arc
15–35′1,500–3,400 pxFills the frame. Rotate the camera to line up long targets
3–15′300–1,500 pxThe sweet spot: most galaxies, the larger planetary nebulae, globular clusters
1–3′100–300 pxSmall planetary nebulae and galaxy groups. Seeing sets the detail
Under 1′Under 100 pxBright tiny planetaries: use short subs, or the planetary setup at 0.31″/px (see planets)

Light per pixel: why binning matters

The signal each pixel gets depends on (aperture × pixel scale)². Both cameras have 2.9 µm pixels, so this is the same as comparing f-ratios.

SetupScaleLight per pixel vs SeestarUse
Seestar S50, 50 mm f/52.39″/px1×—
MK127 native, 127 mm f/7.70.61″/px0.42×Bright, small targets: planetary nebulae, globular cores, M42's core
MK127 binned 2×21.23″/px1.7×Faint galaxies and nebulae (see 2×2 binning)

Two consequences:

Filters and subs

Targets by season

"Best" is the month the target transits around 9 pm (see the RA-by-month table in Seestar targets). Sizes are approximate catalogue values. Southern declinations are low from northern mid-latitudes. Duo = ZWO Duo-Band; none = no filter or UV/IR cut only.

TargetTypeSize · MK127 pxFilterBestNotes
Autumn–winter: now
NGC 7331 + Stephan's QuintetGalaxy + group10′ × 4′ · ~1,000 px; Quintet ~3′, 30′ awayNoneOctBoth fit corner to corner (the diagonal is 45′). The Quintet's galaxies are 13–14 mag: 4 h+, binned
NGC 7662 Blue SnowballPlanetary~0.5′ · ~50 pxDuo or noneNovBright and tiny: short subs, or planetary setup
NGC 7635 BubbleEmission15′ × 8′; bubble ~3′DuoNovThe bubble shell finally shows structure; strong O III
NGC 891Galaxy13′ × 3′ · ~1,300 × 250 pxNoneDecEdge-on with its dust lane resolved. Rotate the camera to lay it along the frame
M74Galaxy10′ · ~1,000 pxNoneDecFace-on spiral with very low surface brightness: binned, 5 h+, darkest sky
M77Galaxy7′ × 6′NoneDecBright core, faint outer ring
M76 Little DumbbellPlanetary2.7′ × 1.8′ · ~270 pxDuoDecFaint outer loops in long O III
M33 core / NGC 604Galaxy / H II regionCore region fills the frame; NGC 604 ~1.5′None + DuoDecNGC 604 is a star-forming region 2.7 million light-years away, resolved as a nebula
M1 CrabSupernova remnant6′ × 4′ · ~600 pxDuoFebFilaments show. A Mak classic
M42 core and TrapeziumEmissionCore ~20′: fills the frameDuo or noneFebExpose for the core with 5–15 s subs and the rest with longer subs, then blend (HDRComposition)
Horsehead (Barnard 33)Dark on emission~8′ × 6′ head, in IC 434DuoFebFrame the head and the edge of IC 434; Alnitak just out of field. 4 h+
M78Reflection8′ × 6′NoneFebDust lanes around a blue glow
NGC 2261 Hubble's Variable NebulaReflection~2′ × 1′ · ~200 pxNoneFebChanges shape over weeks as shadows move across it: image it monthly
Cone NebulaEmission + dark~10′ regionDuoFebClose-up of the Seestar's NGC 2264 field
NGC 2359 Thor's HelmetEmission16′ × 8′DuoMarFills the frame; strong O III. Dec −13°
NGC 2392 EskimoPlanetary~0.8′ · ~80 pxDuo or noneMarBright and tiny: short subs
Spring: galaxy season
M81Galaxy27′ × 14′NoneAprRotate the camera so its long axis runs along the frame; spiral arms resolved. Circumpolar
M82Galaxy11′ × 5′None + DuoAprAdd H-α for its red outflow filaments
NGC 2903Galaxy12′ × 6′NoneAprBright barred spiral, overlooked
M65 + M66Galaxies~10′ each, 20′ apartNoneAprTwo of the Leo Triplet (NGC 3628 needs the Seestar or 2 panels)
M97 OwlPlanetary3.4′ · ~330 pxDuoAprThe "eyes" need good seeing and O III depth
M106Galaxy19′ × 8′None + DuoMayAnomalous red arms in H-α
M64 Black EyeGalaxy10′ × 5′NoneMayDark dust band over a bright core
M63 SunflowerGalaxy12′ × 7′NoneMayFlocculent arms resolve into knots
M104 SombreroGalaxy9′ × 4′ · ~880 pxNoneMayDust lane crisp at this scale. Dec −12°: image at transit
NGC 4565 NeedleGalaxy16′ × 3′NoneMayEdge-on, laid along the frame's long side
M51 WhirlpoolGalaxy pair11′ × 7′ · ~1,100 pxNone + DuoJunThe showpiece: arms, dust, the bridge to NGC 5195, pink H II knots
M101 PinwheelGalaxy29′ × 27′NoneJunTaller than the frame: crop or 2 panels. Very faint arms: bin, 6 h+
M3, M13Globular clusters18′, 20′NoneJun, JulResolved nearly to the core. 30–60 s subs; 1–2 h. M13's companion galaxy NGC 6207 is 28′ away and fits in the frame
Summer
NGC 6543 Cat's EyePlanetary~20″ core · ~35 pxDuo or noneAugVery bright core in short subs; faint outer halo (~5′) in long Duo-Band
M57 RingPlanetary1.4′ × 1′ · ~140 pxDuoAugThe ring's structure, central star and faint outer halo. Galaxy IC 1296 sits 4′ away
M16 Pillars of CreationEmissionPillars ~4′ in a 7′ coreDuoAugDec −14°: low, so pick steady nights
M17 Swan, M20 TrifidEmission~11′ bright core; 28′Duo (M20: none for the blue half)AugLow from northern sites; M20 just fits diagonally
M27 DumbbellPlanetary8′ × 6′ · ~800 pxDuoSepThe "apple core" and its faint outer shell
NGC 6888 CrescentEmission18′ × 13′DuoSepFills the frame; braided H-α filaments
Pickering's Triangle (Veil)Supernova remnant~30′ sectionDuoSepFine filaments at full resolution
Cygnus Wall (NGC 7000)EmissionA 39′ stripDuoSepThe ridge of the North America Nebula
NGC 7009 Saturn NebulaPlanetary~0.5′ with its "rings"Duo or noneSepBright and tiny: short subs
NGC 7023 IrisReflection18′NoneSepBlue petals with brown dust: dark sky, 5 h+
Elephant's Trunk (vdB 142)Emission + darkTrunk ~20′DuoOctClose-up of the Seestar's IC 1396 field

"Best" months were calculated from each target's RA with the sidereal-time rule. Integration times are rough guides for a Bortle 5–6 sky with the GTi guided.

Tonight's short list (mid-November).
  • Moonless window: NGC 891, NGC 7331 and M77, binned.
  • Under any Moon: the Bubble, M76 and NGC 7662 with the Duo-Band.
  • After midnight: M1, then M42's core.

Deep-sky targets for the Canon R8

The R8 fills the gap above the Seestar: fields from 5° to 100° across, which neither telescope can reach. It has a full-frame 6000 × 4000 sensor with large 5.98 µm pixels. With your lenses it covers constellations, the Milky Way, and big nebula complexes that are mosaics on the Seestar. On the GTi with the Samyang 135 mm f/2, it collects light per unit of sky about six times faster than the Seestar's f/5.

Lenses, scale and field

Lens on the R8ScaleFieldUntracked limit (NPF)Best for
Laowa 15 mm f/2 (at f/2.8)82″/px100° × 77°~17 s at f/2 (~19 s at f/2.8)Milky Way landscapes, meteor showers, aurora, zodiacal light
RF 24–50 mm (24 mm f/4.5)51″/px74° × 53°~14 sConstellations, Milky Way panoramas
RF 24–50 mm (50 mm f/6.3)25″/px40° × 27°~8 sOrion with Barnard's Loop, Cygnus, Taurus with the Hyades and Pleiades
Samyang 135 mm f/2 (at f/2.8)9.1″/px15.2° × 10.2°~2 s: track itThe deep-sky workhorse: nebula complexes, M31, Orion's belt and sword
RF 100–400 mm (400 mm f/8)3.1″/px5.2° × 3.4°~1 s: track and guideAll of M31, the Veil, California, Rosette with the Cone
RF-S 55–210 mm (210 mm f/6.3, crop mode)5.9″/px6.1° × 4.1°~2 sWorks, but slow and only 9 MP: the 135 or 400 are better

Scale = 206.265 × 5.98 µm ÷ focal length; field from the 36 × 24 mm sensor. The untracked limit uses the simplified NPF rule, t ≈ (35 N + 30 p) ÷ f, for pinpoint stars. The old "500 rule" allows about twice as long, but with visible trails at 6 µm pixels.

All three rigs together give a zoom ladder: the 135 mm frame holds about 170 Seestar frames, and the Seestar frame holds about 4 MK127 frames. A typical plan:

  1. Frame a region with the R8.
  2. Pick a nebula from it for the Seestar.
  3. Pick a detail from that for the Mak.

What's different about the R8

Processing differences in PixInsight

Targets by season

"Best" is the month the target is high around 9 pm. Sizes are approximate. The "LP filter" column applies to the Samyang 135 with the drop-in adapter.

TargetSizeLensLP filterBestNotes
Autumn–winter: now
M31 with M32, M1103° × 1°135 mm or 400 mmOffNovThe 400 mm frames it like the Seestar's 3-panel mosaic in one shot
M31 + M33 region15° apart50 mmOffNovTwo Local Group galaxies in one frame
Double Cluster + Heart & Soul~7° span135 mmOnDecClusters and nebulae together; the nebulae are faint on a stock camera
M45 + California~13° apart135 mm (diagonal)OffJanBlue reflection nebula and red emission in one frame; dark sky needed. In very dark skies, faint dust (Taurus molecular clouds) surrounds both
Taurus: Hyades, Pleiades, Aldebaran~20°50 mmOffJanStar colour and two open clusters
Orion's belt and sword~8°135 mmOff or onFebM42, Running Man, Horsehead, Flame and M78 in one frame
Orion with Barnard's Loop~25°24–50 mmOffFebThe loop is faint pure H-α: hard on a stock R8; several hours, dark sky
Rosette + Cone (NGC 2264)~5° apart135 mmOnFebBoth nebulae plus open clusters
Winter Milky Way (Cassiopeia to Orion)~90°15 mmOffDec–FebFainter than the summer core, but rich in clusters
Geminids meteor showerWhole sky15 mmOffPeak night 13–14 Dec 2026Moon is a thin evening crescent (new 9 Dec): an excellent year. Hundreds of 15–20 s frames; keep those with meteors and composite
Spring
Virgo Cluster with Markarian's Chain~8°135 mmOffMayDozens of galaxies as small smudges; crop for Markarian's Chain
Leo Triplet + M95/M96/M105~8° span135 mmOffAprTwo galaxy groups in one field
M81/M82 region~1° pair400 mmOffAprThe dust around them (integrated flux nebula) needs a very dark site
Rho Ophiuchi complex~5°135 mmOffJunColourful clouds around Antares. Dec −24°: low from the north, so catch it at transit
Summer
Milky Way core~30°15–24 mmOffJun–Aug (spring mornings)The classic landscape target. Low from the north: find a dark southern horizon
Lagoon + Trifid (M8, M20)1.5° apart135–400 mmOnAugLow (Dec −24°); M20's blue half shows well on a stock camera
Cygnus: Sadr region, Butterfly, Crescent~10°135 mmOnSepRich H-α region, still impressive on a stock camera with long integration
North America + Pelican~3°135–400 mmOnSepThe brightest big H-α target for an unmodified camera
Veil complex~3° loop400 mmOnSepThe whole loop in one frame
IC 1396 Elephant's Trunk region~3°135–400 mmOnOctSingle frame instead of a Seestar mosaic

Field widths and separations are approximate. Geminid peak date from its usual 13–14 December maximum; Moon phase from the Moon section.

Tonight's short list (mid-November).
  • On the GTi with the 135 mm: M31, then the Double Cluster, then the Pleiades with the California Nebula.
  • Later: Orion's belt and sword.
  • On a tripod with the 15 mm: the winter Milky Way, and then the Geminids on 13–14 December.

The R8 on the MK127? It's possible, but it would need an RF T-ring, which isn't in your kit. With the 0.65× reducer it would give 1.27″/px over 2.1° × 1.4°, close to the ASI585MC's binned scale over a larger field. The ASI585MC is still the better match there. It's lighter and records H-α fully, and N.I.N.A. controls it more easily.

Targets for the Samyang 135 on the ASI585MC

The ZWO EOS-T2 adapter lets the ASI585MC use the Samyang 135 mm f/2, an EF-mount lens; the Rokinon 135 is the same optical design. This gives a fourth rig, between the R8 and the Seestar:

Big emission nebulae that are mosaics on the Seestar and faint on the R8 fit one frame here.

How it compares

RigScaleFieldLight per pixel vs SeestarH-α
Seestar S502.39″/px0.7° × 1.3°1×Full
ASI585MC + Samyang 135 f/24.43″/px4.7° × 2.7° (14 Seestar fields)6.3× (3.2× at f/2.8)Full
R8 + Samyang 135 f/2.89.1″/px15.2° × 10.2°13.6×, but its larger pixels each cover 4.3× more skyWeak (stock filter)
R8 + RF 100–400 at 400 mm f/83.1″/px5.2° × 3.4°1.7×; per area of sky, 8× slower than the 135 at f/2.8Weak

Scale = 206.265 × 2.9 µm ÷ 135 mm; field from 3840 × 2160 px. Light per pixel ∝ (aperture × scale)², which for equal pixel sizes reduces to (f-ratio)⁻².

Frames are 4.73° × 2.66°, north up; outlines drawn to scale from catalogue sizes and positions. Red = emission nebula, teal = galaxy.

Setting it up

Processing

Use the MK127 workflow (Part C) with three changes:

SPCC uses the same IMX585 curves, with the Duo-Band settings for HOO.

Targets by season

Targets chosen to fill the 4.7° × 2.7° frame. "Best" is the month the target transits around 9 pm (see the RA-by-month table). Duo = ZWO Duo-Band; none = no filter. Rotate the camera to fit long targets.

TargetSize · pxFilterBestNotes
Autumn–winter: now
M31 with M32, M1103° × 1° · ~2,500 px longNoneNovThe whole galaxy and its halo in one frame. 2–3 h at f/2 goes deep
Bubble + M52 + Cave Nebula~1.5° groupDuoNovA cluster and two nebulae in one field
M3373′ × 45′ · ~990 pxNoneDecSmall in the frame, but faint outer arms show quickly at f/2
Heart (IC 1805) or Soul (IC 1848)~2.5° each, ~2.7° apartDuoDecOne per frame; both in a 2-panel mosaic (N.I.N.A. framing assistant)
California (NGC 1499)2.4° × 0.7°DuoJanPure H-α, the target the stock R8 struggles with. Fits diagonally
M45 Pleiades~2° with nebulosityNoneJanReflection nebulosity plus the surrounding dust in a dark sky
Flaming Star + Tadpoles (IC 405, IC 410)~1.6° apartDuoFebBoth in Auriga, plus the Spider and Fly (IC 417) nearby
M42 to the Horsehead~3.7° spanDuo or noneFebCamera rotated to portrait: M42, Running Man, Horsehead, Flame in one frame. Blend short and long subs for M42's core
Jellyfish + Monkey Head (IC 443, NGC 2174)~2.6° apartDuoFebA supernova remnant and an H II region
Rosette1.3° · ~1,100 pxDuoFebFits with its surrounding faint nebulosity
Cone + Fox Fur (NGC 2264)~1° × 1.5°DuoFebWith the dusty surrounding region
Seagull (IC 2177)~2.5°DuoMarWings spread across the frame. Dec −10°
Spring
M81 + M82 + NGC 3077~1° groupNoneAprSmall, but room for the faint surrounding dust in a dark sky
Markarian's Chain~1.5° · ~1,200 pxNoneMayWhole chain plus M87 and many fainter galaxies
Rho Ophiuchi with Antares~4°NoneJunColourful dust and reflection nebulae. Dec −24°: at transit, from a dark southern horizon
Summer
Lagoon + Trifid (M8, M20)~1.4° apartDuo (none for M20's blue)AugBoth fit comfortably
Eagle + Swan (M16, M17)~2.4° apartDuoAugRotate the camera to fit both
Sadr region (IC 1318, Butterfly)~3–4°DuoSepFills the frame with H-α and dark lanes, with the Crescent at its edge
North America + Pelican~3.3° spanDuoSepBoth in one frame; HOO shows the Cygnus Wall's O III edge
Veil complex~3° loopDuoSepTight: rotate the camera or accept trimmed edges. The best O III target here
Elephant's Trunk (IC 1396)~3°DuoOctThe whole region with the Trunk inside

Sizes and separations are approximate catalogue values; "Best" months from each target's RA. Integration times assume a Bortle 5–6 sky.

Tonight's short list (mid-November).
  • Moonless window: M31.
  • Under any Moon, with the Duo-Band: the Heart, then the Soul.
  • After midnight: the California Nebula, then the Flaming Star and Tadpoles.
Under moonlight, the Duo-Band at f/2–2.8 still gives deep H-α in a couple of hours.

Which rig for which target:

The ranges overlap, so the Seestar and the 135 can share a target, the 135 for the wide context and the Seestar for the core.

Mosaics with the Samyang 135

The ASI585MC + 135 mm frame is 4.7° × 2.7°. The biggest structures in the sky are 5–20° across. A mosaic tiles several frames into one image: it keeps the 4.43″/px detail and full H-α of the ASI585MC, at the cost of a panel's worth of integration per tile. The R8 + 135 covers 15° in one frame, but it records H-α only weakly. For emission nebulae, an ASI585MC mosaic goes much deeper.

Size and cost

Panels (landscape)Coverage at 15% overlapFinal sizeRGB float fileAt 2 h per panel
2 × 18.7° × 2.7°7,100 × 2,160 px0.2 GB4 h
2 × 28.7° × 4.9°7,100 × 4,000 px0.3 GB8 h
3 × 212.8° × 4.9°10,400 × 4,000 px0.5 GB12 h
3 × 312.8° × 7.2°10,400 × 5,800 px0.7 GB18 h
4 × 416.8° × 9.4°13,600 × 7,700 px1.3 GB32 h

Coverage = frame × (panels − (panels − 1) × overlap). Turn the camera 90° for portrait panels (2.7° × 4.7°) when the target runs north–south.

Positions from catalogue coordinates, to scale. Shaded strips are the overlaps, where neighbouring panels are matched.

Mosaic targets

TargetPanelsFilterBestNotes
Heart & Soul2 × 1DuoDecThe easiest first mosaic: two panels side by side. Now
M31 with its outer halo2 × 1NoneNovRoom around M31 for its faint outer disc and the gradients around it
Flaming Star, Tadpoles, M36–M382 × 2DuoFebAuriga's nebulae and clusters together
Rosette, Cone and Christmas Tree2 × 2 portraitDuoFebRosette to the Cone is ~5°, mostly north–south
California + Pleiades2 × 4 portraitNoneJan~13° north–south. Under a very dark sky the Taurus dust connects them. Long and faint: a season's project
Orion: belt, sword and Barnard's Loop~4 × 4 portraitDuoFebThe loop is ~10–15° across. The classic wide-field mosaic; a multi-night project
Seagull + Thor's Helmet2 × 2DuoMarAbout 5° apart in Canis Major. Low from the north
Virgo cluster3 × 2NoneMayMarkarian's Chain to M49 and M61: hundreds of galaxies
Rho Ophiuchi, Antares and M42 × 2 portraitNoneJunColourful dust. Low from the north: shoot at transit
Lagoon, Trifid, M16, M172 × 3 portraitDuoAugFour showpieces along the Milky Way. Low
Cygnus: NGC 7000 to the Crescent3 × 3DuoSepThe diagram above: H-α across the whole field
Veil complex and surroundings2 × 2DuoSepThe whole loop with its faint outer filaments
IC 1396 + Flying Bat (Sh2-129)2 × 2DuoOctThe Bat holds the Squid (Ou4), a very faint O III target: 20 h+

Planning and capture

  1. Frame it in N.I.N.A.'s Framing Assistant.
    • Set focal length 135 mm and pixel size 2.9 µm. Load a sky survey image of the target.
    • Set the panel count and 15–20% overlap, and drag the grid onto the target.
    • Set the camera rotation to match the grid.
    • Send the panels to the Advanced Sequencer. N.I.N.A. creates one target per panel, named "Target Panel 1", "Panel 2" and so on.
  2. Keep the rotation fixed. Every panel and every night must use the same camera angle, or the overlaps won't line up cleanly. Use the plate-solved rotation in N.I.N.A. to check before each session.
  3. Equal integration per panel. Uneven depth shows as panels with different noise. Shoot each panel to the same total, even if one has less nebula.
  4. Order: each night, start with the panels that set first, on the western side. Alternatively, shoot one panel per night under similar conditions, which helps match gradients.
  5. Narrowband helps: with the Duo-Band, panels can be shot under moonlight and on different nights with similar results (see Moon). Broadband mosaics need all their panels shot under similar dark skies.
  6. Calibration: one flat set works for all panels if the camera, lens and focus haven't moved. Take new flats each session if they have.

Processing in PixInsight

This is Route B of Seestar mosaics, tested there on your M45 panels, with the 135 mm values:

  1. Stack each panel separately in WBPP: group by the N.I.N.A. target name; autocrop off (the edges are the overlaps); astrometric solution with focal length 135, pixel size 2.9.
  2. On each panel master: crop ~10 px of edge, then BXT Correct Only.
  3. Plate solve each panel with ImageSolver, distortion correction on. Even small lens distortion misaligns stars in a 15% overlap without it.
  4. Gradients:
    • Panels with open sky: GradientCorrection on each, so they start level.
    • Panels full of nebula (Cygnus, Orion): leave gradients until after the merge; per-panel correction would remove real nebulosity.
    • If you've installed the MARS database, MultiscaleGradientCorrection on the finished mosaic is the most reliable at keeping large faint structures (see step 5).
  5. MosaicByCoordinates on all panels; list the centre panel first.
  6. GradientMergeMosaic: Average, shrink radius 10, feather radius 30 (or larger, ~50–100, at these image sizes), Generate Mask on.
  7. Crop to the fully covered area. Then continue on the whole mosaic:
    • gradient correction;
    • plate solve again;
    • SPCC (or DBXtract → HOO for Duo-Band data);
    • BXT sharpening, star removal, stretch and the rest of the workflow.
Seams. A visible line or brightness step between panels has three usual causes:
  • gradients that differ between panels: correct them before the merge, or use MGC afterwards;
  • different integration depth: equalise the panels;
  • too narrow a feather: raise it.
Fix gradients first. Feathering only hides a mismatch; it doesn't remove it.
Disk and memory.
  • Each reprojected panel is full mosaic size: a 3 × 3 mosaic writes nine 0.7 GB files.
  • BlurXTerminator and StarXTerminator on 10,000+ px images take minutes per run on the MeLE mini PC.
  • Test settings on a preview, and keep 20 GB free for a big mosaic.

Your first one: the Heart and Soul, 2 × 1 with the Duo-Band, starting this month. Shoot about 2 h per panel and process with the steps above. The single overlap makes seams easy to check and fix before you try a 3 × 3.

Mosaics with the Canon R8

R8 mosaics come in two kinds:

The R8's large sensor means fewer panels for a given area than the ASI585MC needs: one R8 + 135 frame covers as much sky as about 12 ASI585MC + 135 frames.

Coverage

Lens2 × 12 × 23 × 2Final size (2 × 2)
400 mm (5.2° × 3.4° frame)9.3° × 3.4°9.3° × 6.2°13.4° × 6.2°10,800 × 7,200 px, ~0.9 GB as 32-bit RGB. 3 × 2: 15,600 × 7,200, ~1.4 GB
135 mm (15.2° × 10.2°)27.3° × 10.2°27.3° × 18.3°39.5° × 18.3°
50 mm (39.6° × 27.0°)71° × 27°71° × 49°103° × 49°

At 20% overlap: more than the ASI585MC's 15%, because full-frame lenses vignette and distort more toward the corners. Turn the camera to portrait for north–south targets.

Star and nebula positions from catalogue coordinates; Barnard's Loop drawn approximately.

Deep-sky mosaic targets

For a stock R8, broadband targets work best: dust, star fields, galaxies and reflection nebulae. Emission targets work too, but they need much longer integration than on the ASI585MC.

TargetLens · panelsBestNotes
Orion constellation with Barnard's Loop and the λ Ori ring135 mm · 2 × 2 portraitFebThe diagram above. The loop is faint on a stock camera: 2–3 h per panel from a dark site
Taurus dark clouds with the Pleiades, Hyades and California135 mm · 2 × 2JanDust is broadband, the R8's strength. Needs Bortle 3 or darker
Cassiopeia to Perseus: Heart, Soul, Double Cluster135 mm · 2 × 1DecClusters and Milky Way star fields. Now
M31 + M33135 mm · 2 × 1, rotated ~45°Nov15° apart along a diagonal: turn the camera so the long side of the mosaic runs from M31 to M33. Now
Cygnus Milky Way (Deneb to Albireo)135 mm · 2 × 1SepDense star clouds and dark lanes (the Cygnus Rift)
Rho Ophiuchi to the Pipe Nebula135 mm · 2 × 1Jun–JulColourful dust, then dark nebulae against the Milky Way core. Low: dark southern horizon
Sagittarius–Scorpius core50 mm · 2 × 1, or 135 mm · 3 × 2JulLarge-scale Milky Way detail
M31 with surrounding field400 mm · 2 × 1NovM31 with room for its faint outer disc, at 3.1″/px
Orion's belt and sword400 mm · 2 × 2 portraitFebM42, Running Man, Horsehead, Flame, M78
Markarian's Chain to M49400 mm · 2 × 2MayThe heart of the Virgo cluster

Capturing a deep-sky mosaic

Processing a deep-sky mosaic

Same as ASI585MC + 135 mosaics (Route B of Seestar mosaics), with these R8 differences:

  1. WBPP per panel from the CR3 files: RGGB, autocrop off, astrometric solution with focal length 135 (or 400) and pixel size 5.98. At 50 mm and shorter, turn on Distortion correction in registration.
  2. ImageSolver with distortion correction on each panel master. This is essential at 135 mm and wider, where lens distortion over a 15° field would otherwise misplace stars by several pixels in the overlaps.
  3. Gradients: correct each panel before merging when there's open sky. In Milky Way and dust fields there's almost no true background. Merge first, then use MultiscaleGradientCorrection with MARS if installed, or a cautious GradientCorrection on the whole mosaic.
  4. MosaicByCoordinates, then GradientMergeMosaic (Average, shrink 10, feather 50–100).
  5. Crop, re-solve, SPCC with a DSLR or Ideal QE curve, then BXT, star removal and stretch as usual.
Projection. MosaicByCoordinates uses a flat (gnomonic) projection, which stretches the edges increasingly beyond about 30–40° from the centre. A 135 mm 2 × 2 (27° × 18°) is fine. For wider 50 mm mosaics or whole-sky panoramas, use panorama software with a cylindrical or stereographic projection instead.

Milky Way panoramas (15 mm, 24 mm)

PanoramaLens and layoutFrames
Milky Way arch, horizon to horizon (180°)15 mm, portrait, one row3–4 at 30% overlap
Arch with more detail24 mm, portrait, two rows5 per row, 10 total
Single section (summer core, winter Milky Way)24–50 mm, 2 × 2 or 3 × 24–6

First R8 mosaic: M31 + M33 (135 mm, 2 × 1, rotated) this month. It's broadband, so the stock sensor is no handicap, and the two galaxies make it easy to check alignment across the seam.

Mosaics with the MK127

The MK127's 39′ × 22′ frame is smaller than many showpiece galaxies and nebulae: M33, M42, the Leo Triplet, the Veil's arcs, and the Moon itself. A mosaic of 2–6 panels covers them at 0.61″/px (1.23″/px binned), about four times finer than the Seestar. The catch is time. The Mak is slow per pixel (see MK127 targets), and each panel needs a full integration. Plan Mak mosaics as multi-night or whole-season projects.

Coverage

PanelsLandscape (39′ × 22′ frames)Portrait (22′ × 39′ frames)Final size, binned 2×2At 4 h per panel
2 × 173′ × 22′41′ × 39′~3,550 × 1,080 px8 h
2 × 273′ × 41′41′ × 73′~3,550 × 2,000 px16 h
3 × 2106′ × 41′60′ × 73′~5,200 × 2,000 px24 h
3 × 3106′ × 60′60′ × 106′~5,200 × 2,900 px36 h

At 15% overlap. Unbinned sizes are twice as large in each direction. "4 h per panel" is a typical binned galaxy depth for a Bortle 5–6 sky; bright nebulae need less.

M33 drawn to scale from its catalogue size (73′ × 45′, position angle 23°); NGC 604 from its catalogue position.

Mosaic targets

TargetPanelsFilterBestNotes
Autumn–winter: now
M333 × 2 portraitNone + DuoDecThe diagram above. H-α from the Duo-Band makes NGC 604 and the other H II regions pop. A season's project
M31 core with M32 and NGC 2062 × 2NoneNovThe inner disc, dust lanes and star cloud at full detail. The whole galaxy would be ~15 panels: a Seestar or 135 mm job
Double Cluster2 × 2NoneDecBoth clusters at 0.61″/px with resolved star colours. Short subs; bright
Heart core (Melotte 15, Fish Head)2 × 2DuoDecThe central pillars and cluster in detail
Pleiades core3 × 2NoneJanThe brightest stars and the reflection nebulosity's texture. Bright stars mean short subs and BXT halo care
M42, M43 and the Running Man3 × 3Duo or noneFebThe full nebula complex. Blend short subs for M42's core (HDRComposition)
Horsehead + Flame2 × 2DuoFebBoth with the edge of IC 434 between them. Keep Alnitak's glare in an overlap, not a panel centre
Cone, Fox Fur and Christmas Tree2 × 2 portraitDuoFebThe NGC 2264 region from the Cone to S Mon
Spring
M81 + M821 × 2 portraitNone (+ Duo for M82)AprTwo panels, one galaxy each, 38′ apart north–south
Leo Triplet2 × 2 portraitNoneAprAll three galaxies, including NGC 3628's tidal plume, at full scale
M1012 × 2None + DuoJunIts 29′ disc with outer H II regions. Very faint arms: bin, 6 h+ per panel
Markarian's Chain3 × 2, rotated along the chainNoneMayA dozen galaxies at 1.23″/px binned
Summer
M16 Eagle2 × 2DuoAugThe whole nebula with the Pillars at full scale. Low
M8 Lagoon3 × 2DuoAug90′ × 40′ fits. Low: a steady night at transit
M20 Trifid2 × 2Duo (none for the blue part)AugEmission and reflection halves together
Western Veil (Witch's Broom) / Eastern Veil1 × 2 portrait eachDuoSepEach arc runs north–south. HOO filaments at 0.61″/px
Cygnus Wall3 × 1DuoSepThe ridge of the North America Nebula in one strip
Pelican core2 × 2DuoSepThe "beak" and its ionisation front
Any month
The Moon2 × 2 at 0.61″/px; ~3 × 3 at native 0.40″/pxNone (or IR-pass)Any, ideally near the quartersVideo, not subs: see below

Capturing a Mak mosaic

Processing a Mak mosaic

Route B again (Seestar mosaics, 135 mm mosaics), with the Mak's calibration and binning:

  1. WBPP per panel with darks, flats and dark flats (see WBPP with calibration); autocrop off; astrometric solution with focal length 975 and pixel size 2.9.
  2. Bin each panel 2×2 (IntegerResample, Average), as in Fix the oversampling, unless the seeing was excellent. This halves the noise, quarters the file sizes, and makes BXT and the merge much faster.
  3. Crop ~8 px of edge from each panel, then BXT Correct Only. The Mak's corners differ between panels, and correction before merging keeps star shapes matched across seams.
  4. Plate solve each panel (ImageSolver, distortion correction on).
  5. Gradients per panel for galaxies and clusters, where there's open sky around the target. For nebulae that fill every panel (M42, Lagoon), merge first and correct once.
  6. MosaicByCoordinates, then GradientMergeMosaic: Average, shrink 10, feather 30.
  7. Crop to the fully covered area. Then, on the whole mosaic:
    • gradient correction;
    • re-solve;
    • SPCC (or DBXtract → HOO for Duo-Band data);
    • BXT sharpen, star removal, stretch.
Seams from different seeing. Panels shot on different nights can have different star sizes, for example 2″ and 3″ FWHM, and the seam shows even when brightness matches. Two remedies:
  • BXT Correct Only plus a modest BXT sharpen on the merged mosaic evens them out;
  • cull more aggressively by FWHM, so every panel stacks only similar nights.

Moon mosaics

The full Moon is 29–34′ across: taller than the reducer's 22′ frame and twice the native frame. Lunar mosaics follow the planetary method, not the deep-sky one:

  1. Capture each tile as a short SER video in SharpCap or FireCapture: about 1–2 ms exposures, low gain, 500–2,000 frames. Overlap the tiles by 20–30%, moving along the terminator first, because it changes fastest.
  2. Stack each tile in AutoStakkert! (best 10–25%, many alignment points), then sharpen with wavelets.
  3. Stitch in Microsoft ICE (no longer developed, but still works), Hugin or PTGui. The Moon has no stars to plate-solve, so PixInsight's MosaicByCoordinates can't be used.
  4. Finish with curves. For colour, a mild saturation boost brings out mineral colour differences.

First Mak mosaic: M81 + M82 in spring (1 × 2 portrait: two panels, one galaxy each). Before that, the Moon near first quarter (16 Dec) is a quick 2 × 2 practice run for framing and overlap.

Mosaics with the Seestar

Part B's Seestar mosaics covers processing (Route A and Route B, tested on your M45 data). This section covers planning:

Coverage and cost

PanelsCoverage (15% overlap)Final sizeEach panel's share of a 3 h session
10.72° × 1.27°1,080 × 1,920 px3 h
2 × 11.33° × 1.27°~2,000 × 1,920 px90 min
2 × 21.33° × 2.34°~2,000 × 3,550 px45 min
3 × 21.94° × 2.34°~2,900 × 3,550 px30 min
3 × 31.94° × 3.42°~2,900 × 5,200 px20 min

The Seestar splits one session's time across all positions, so every panel gets 1/N of the integration. A 2 × 2 needs four times the session length for the same depth as a single frame. Its large 2.39″/px pixels and f/5 speed make that affordable for bright targets, but faint ones quickly need several nights.

Setting up a mosaic

Which route in PixInsight

DataRouteWhy
One night, app mosaic mode, up to a few hundred subsA: plate solve and reproject every subUses every sub regardless of where it landed. Your M45 result
Long or multi-night sessions (1,000+ subs)B: group by position, stack each group, mergeRoute A would write tens of GB of reprojected files
A single target that drifted (no mosaic mode)ARecovers the subs that plain WBPP rejects

Seestar mosaic targets

TargetPanelsLP filterBestAlternative rig
Autumn–winter: now
M31 whole3 × 3 (2 × 2 for most of the disc)OffNovOne frame on the ASI585MC + 135 or the R8 + 400
M33 with its outer arms2 × 1OffDec—
Heart or Soul core2 × 2 each (the full nebulae need ~4 × 2)OnDecBoth in a 2 × 1 on the 135
M45 with surrounding dust2 × 2OffJanYour M45 session was effectively a 2 × 1
California3 × 2OnJanOne 135 frame (ASI585MC)
M42 + Running Man2 × 2On or offFeb—
Rosette2 × 1 (tight) or 2 × 2OnFeb—
Horsehead, Flame and M422 × 4OnFebOne portrait 135 frame
Spring–summer
Markarian's Chain3 × 1OffMay—
Lagoon + Trifid3 × 2OnAugOne 135 frame
M16 + M172 × 3OnAugOne 135 frame, rotated
North America + Pelican5 × 2OnSepOne 135 frame
Veil complex5 × 3OnSepOne 135 frame (tight)
IC 13965 × 3OnOctOne 135 frame

Panel counts are for 15% overlap, worked out from each target's catalogue size and position. They assume the frame's long side runs roughly north–south; the actual orientation depends on the alt-az mount, so confirm the grid in the app's framing view.

When to mosaic with the Seestar, and when not:
  • Up to about 2 × 2: the Seestar is a good choice. It's quick to set up, and bright targets like M42, the Rosette, M45 and M31's core reach decent depth in one or two nights.
  • 3 × 3 and larger (North America, the Veil, IC 1396 need 10–15 panels): each panel gets so little time that the ASI585MC + Samyang 135 usually does better in one frame. It has 2.7 times the field width, 6 times the light per pixel at f/2, and a single gradient to correct (see Samyang 135 targets). Keep big Seestar mosaics for nights when the 135 rig isn't set up, or for a multi-night project.

Next Seestar mosaic: M31 as a 2 × 2 this month, around the new Moon of 9 December, with LP off and every frame saved. Process it with Route A, exactly as for your M45 session.

Arcseconds in your software

WhereWhat it showsConverting
ImageSolver / ASTAPMeasured image scale (″/px) and focal length. Your Seestar solved at 2.38″/pxUse the solved value everywhere else; plate solve the MK127 once to replace the nominal 0.61″
SubframeSelectorFWHM and eccentricity, in pixels or ″Enter the Subframe scale (″/px) and set Scale unit to arcseconds to read FWHM in ″ directly
N.I.N.A.HFR (half-flux radius), in pixelsFor a Gaussian star, FWHM ≈ 2 × HFR. Multiply by the image scale for ″
PHD2Guiding RMS (RA, Dec, total) in ″ and in guide-camera pixelsGuide scale: 206.265 × 3.75 µm ÷ 120 mm ≈ 6.4″/px (ASI120MM Mini on UniGuide 32). A common target is total RMS ≤ half the imaging scale: ~0.3″ unbinned, ~0.6″ binned on the Mak; much looser on the Seestar
BlurXTerminatorPSF diameter, in pixels× image scale = FWHM in ″. On the Mak, a PSF above ~8 px means the image is more than 2× oversampled
SPCC, Gaia, cataloguesStar positions; parallax and proper motion in milliarcseconds (mas)1 mas = 0.001″. Gaia DR3 measures positions to well under 1 mas for bright stars, a thousandth of an MK127 pixel
206,265 one more time. A parsec is the distance at which 1 AU (the Earth–Sun distance) subtends exactly 1″ of parallax, so 1 parsec = 206,265 AU ≈ 3.26 light-years. It's the same small-angle formula as image scale, at a cosmic distance.

Quick reference

1° = 60′ = 3600″1″ = 4.85 µrad
scale = 206.265 × px(µm) ÷ f(mm)Seestar 2.39″ · MK127 0.61″ · binned 1.23″ · guider 6.4″
FOV = scale × pixels ÷ 60Seestar 76′ × 43′ · MK127 39′ × 22′
Airy FWHM ≈ 1.03 λ/DSeestar 2.3″ · MK127 0.92″ (550 nm)
ideal scale ≈ FWHM ÷ 2…32–3 px across a star
FWHM ≈ 2 × HFRN.I.N.A. HFR to FWHM
guiding blur ≈ 2.35 × RMSkeep RMS ≤ scale ÷ 2
sky motion = 15.04″/s × cos δunguided drift rate
decimal = d + m/60 + s/360041° 30′ = 41.5°, not 41.30°
1ʰ = 15° · 1ᵐ = 15′ · 1ˢ = 15″RA time to angle (× cos δ on the sky)
1″ of latitude ≈ 31 m0.001° ≈ 110 m
true = magnetic + declinationpolar error 1° ≈ 0.26″/s drift
airmass ≈ sec z · R ≈ 58″ tan zstay above 30° (airmass 2)
time ∝ sky (mag/arcsec²)1 mag brighter ≈ 2.5× integration
synodic month 29.53 dMoon ≥ 60° away · Sun > 18° down
planetary f-ratio ≈ 5 × px(µm)MK127 no reducer + 1.3× = f/15.4, 0.31″/px
Dawes ≈ 116 ÷ D(mm)MK127 0.91″ · Rayleigh 1.09″ · Seestar 2.3″
size(px) = size(′) × 60 ÷ scaleSeestar ~25 px/′: fits 43′ × 76′ · LP on for emission, off for galaxies/clusters
light/px ∝ (D × scale)²MK127 vs Seestar: 0.42× native · 1.7× binned · ~98 px/′ in a 39′ × 22′ field
NPF t ≈ (35 N + 30 p) ÷ fR8 (5.98 µm): 15 mm ~17 s · 135 mm ~2 s → track · 135 mm = 9.1″/px, 15.2° × 10.2°
585MC + 135 mm: 4.43″/px4.7° × 2.7° · f/2 = 6.3× Seestar light per pixel · full H-α
HA = LST − RAsidereal day 23 h 56 m · 1 s clock = 15″

Appendix D · Narrowband, SHO and astro filters

Broadband imaging records all the light in a colour band. Narrowband imaging records only a few nanometres around the specific wavelengths that glowing gas emits. That changes what you can image, where, and when, and it changes how you process the result. This appendix covers:

The emission lines

LineWavelengthColourComes fromTypical strength
H-α656.3 nmDeep redHydrogen ionised by hot stars, recombiningThe strongest line in nearly every emission nebula
[O III]500.7 and 495.9 nmTealDoubly ionised oxygen; needs very hot stars or shocksBright in planetary nebulae, supernova remnants and Wolf-Rayet bubbles. Weak or absent in many H II regions
[S II]671.6 and 673.1 nmDeep redSingly ionised sulphur, at ionisation fronts and in shocksUsually the weakest: often 5–20% of H-α. Stronger in supernova remnants
H-β486.1 nmBlue-greenHydrogen, like H-αAbout a third of H-α. Mainly a visual filter (Horsehead, California)
[N II]654.8 and 658.3 nmDeep redIonised nitrogenSits beside H-α: any H-α filter wider than ~3 nm includes it

Square brackets mark "forbidden" lines: transitions that happen only in the near-vacuum of nebulae, which is why they don't appear in ordinary light sources.

Why narrowband beats light pollution and the Moon. Skyglow and moonlight are spread across the whole spectrum. Emission nebulae put most of their light into a few lines. A 7 nm filter passes about 2–3% of the visible spectrum, removing about 97% of the sky while keeping nearly all the nebula's light in that line. A 3 nm filter rejects even more, but it needs longer subs and accurate focus.

Filter types and when to use them

FilterPassesUse it forIn your kit
Single narrowband (H-α, O III, S II; 3–7 nm)One lineMono cameras: one stack per line, combined as SHO. On a colour camera only some pixels see each line, so it's slow—
Dual-band (H-α + O III)Two linesEmission and planetary nebulae on colour cameras, under any sky and most Moon phases. Gives HOOZWO Duo-Band 2″; the Seestar's built-in LP filter is also dual-band
Dual-band (S II + O III)Two linesAdds the S II channel to Duo-Band data for full SHO on a colour camera— (e.g. Askar ColorMagic D2)
Tri-band / quad-bandH-α, O III, H-β (and S II), wider bandsColour camera on emission nebulae with more natural colour than a dual-band. Less contrast in bright skies—
Broadband light-pollution (L-Pro, IDAS LPS-D types)Visible light minus the old sodium and mercury linesGalaxies and clusters under moderate, older-style light pollution. They do little against white LEDs and add a colour cast that SPCC must removeCanon drop-in "astro LP" (EF lenses on the R8)
UV/IR cut~400–700 nmPlanetary colour imaging; cameras without their own IR block; refractors that focus IR badlyOptolong UV/IR Cut 1.25″. The ASI585MC's window already blocks IR for deep sky
IR-pass (685, 742, 807 nm)Near-infrared onlyPlanets and the Moon in poor seeing (longer wavelengths are steadier). Monochrome result—
Methane (889 nm)A methane absorption bandCloud altitudes on Jupiter and Saturn. Mono, long exposures—
UV (~350 nm)Near-ultravioletVenus cloud markings (see planets)—
Solar film (full aperture)~1/100,000 of sunlightSunspots in white light. Always over the front of the telescope or lens. Never use an eyepiece-end "sun filter". H-α solar detail needs a dedicated solar telescope, not a deep-sky H-α filter—
Comet (Swan band C₂)~510 nm bandComet comas—
L, R, G, BBroad colour bandsMono cameras: colour from three filtered stacks plus luminance—

Which filter for which target

Target typeFilterWhy
Galaxies, star clusters, reflection and dark nebulaeNone (or UV/IR cut)Their light is broadband starlight; a narrowband filter would remove almost all of it. Add H-α only for galaxies' star-forming regions
H II regions (Heart, Rosette, NGC 7000, California)Duo-Band → HOO; add S II for SHOStrong H-α; O III varies by region
Planetary nebulae (M27, M57, M97)Duo-BandVery strong O III: HOO shows it best
Supernova remnants (Veil, Crab, Jellyfish)Duo-Band, plus S II if you have itAll three lines strong; SHO separates the shock structure
Wolf-Rayet bubbles (Thor's Helmet, Crescent)Duo-BandO III shells around H-α
CometsNone, or Swan bandGas glows at C₂ and CN; dust and tail are broadband
PlanetsUV/IR cut; IR-pass in bad seeingSee planets

Palettes: mapping lines to colours

PaletteMappingTypical lookWhen to use
HOOR = H-α · G = O III · B = O IIIRed and teal, close to natural colourDuo-Band and Seestar LP-on data. The default for colour cameras
SHO ("Hubble palette")R = S II · G = H-α · B = O IIIGold and blue after processingShows where S II, H-α and O III differ: shock fronts, ionisation layers, pillars
Bicolour "pseudo-SHO"R = H-α · G = blend of H-α and O III · B = O IIIGold and blue, without real S IIDuo-Band data with an SHO-like look
HSO, OSH and other permutationsLines to other channelsVariesAesthetic choice; the structure is the same
H-α + RGBH-α blended into the red channel of a broadband imageNatural colour, with pink H II regions boostedGalaxies (M33, M51, M82, M106) and broadband nebula images
None of these, except H-α + RGB, is true colour.
  • SPCC calibrates broadband colour against star catalogues, so it doesn't apply to SHO.
  • For HOO, SPCC's narrowband mode (see step 7) can balance the H-α and O III signals.
  • For SHO, the colour balance is set by eye. There's no wrong answer, only consistent processing.

How narrowband processing differs

The linear and non-linear stages are the same as in Part B. What changes is that you work on separate line images and combine them yourself.

StageBroadband (Parts B, C)Narrowband
ChannelsOne RGB imageOne image per line. Colour cameras: extract with DBXtract (see Duo-Band → HOO). Mono: one stack per filter
GradientsOnce on RGBPer channel. O III gradients (moonlight) usually differ from H-α's
Colour calibrationSPCCNone for SHO; SPCC narrowband mode for HOO
NoiseNXT on the stretched imageWeak channels (O III, S II) need more: NXT per channel before combining, or stronger after
StretchOne stretchEach channel separately, to a similar background and brightness. Otherwise H-α dominates: SHO comes out green, HOO red
Combine—PixelMath or ChannelCombination, choosing the palette
Colour workCurves and saturationSCNR (removes green), hue curves, selective saturation. Much more work
StarsNatural colourMagenta, small, or haloed. Replace them with RGB stars
IntegrationOne totalWeak lines need more: a common split is about 1 : 2 : 2 of the time for H-α : O III : S II

PixelMath walkthroughs

All four start from linear, gradient-corrected, BXT-corrected images, with stars removed by StarXTerminator. Keep the star images for step 4 of the walkthroughs. Name the line images Ha, OIII and SII. In PixelMath, leave Use a single RGB/K expression unchecked and create a new RGB image.

1 · HOO from your Duo-Band or Seestar LP data

  1. Run DBXtract on the linear master to get Ha and OIII.
  2. Stretch each one separately, aiming for the same background level (~0.10–0.15) in both. Use MAS or the GHS script, and push O III harder.
  3. Combine. Raising G's share of H-α warms the teal toward white in bright regions.
    • Basic: R = Ha · G = OIII · B = OIII.
    • Richer: G = 0.4*Ha + 0.6*OIII.
  4. Add stars back from the original colour image's stars with the screen blend ~((~starless)*(~stars)) (step 13).

2 · Pseudo-SHO from Duo-Band data (no S II)

  1. Extract and stretch as above.
  2. Combine: R = Ha · G = 0.6*Ha + 0.4*OIII · B = OIII. H-α regions come out gold (strong red, medium green); O III regions come out blue.
  3. Adjust the 0.6 to move the gold between orange and yellow, then fine-tune with hue curves.

3 · True SHO (with S II data)

  1. Stretch all three lines separately to similar backgrounds. S II usually needs the hardest stretch and the most noise reduction.
  2. Combine: R = SII · G = Ha · B = OIII. The result is mostly green, because H-α is the strongest line.
  3. SCNR: Color to remove Green, Protection method Average Neutral, Amount 0.6–1.0. Green H-α regions shift to gold.
  4. Curves › Hue: separate the gold, cyan and blue. ColorSaturation: boost blue and gold selectively.
  5. Many imagers use a "dynamic" palette instead of a fixed mapping. The Foraxx formula is the best-known example. It blends H-α into green only where O III is weak, which avoids the green cast with less SCNR. Search for the current formula; it's a single PixelMath expression per channel.
  6. Use stars from an unfiltered RGB session or a short colour run. Narrowband stars turn magenta after SCNR.

4 · H-α into a broadband galaxy image

  1. Register the H-α image to the RGB image (StarAlignment).
  2. Remove the starlight from H-α (continuum subtraction), still linear: Ha − k*(R − med(R)), with k about 0.2–0.5, raised until stars and the galaxy's smooth disc disappear and only H II knots remain.
  3. Add the result to red: R = R + w*Ha_cs, with w about 0.5–1. The NBRGBCombination script does the same with a preview.
  4. Run SPCC on the RGB image before adding H-α, then continue as normal.
Common narrowband mistakes.
  • Linked stretch: one channel swamps the others.
  • Too much SCNR: the image turns grey-gold.
  • Saturating noisy channels: colour blotches in the background.
  • Mismatched gradients between channels: colour bands across the field.
  • Stars: combining narrowband stars, then fighting magenta halos.

SHO on a colour camera

Notes for your rigs

RigNarrowband optionsWatch for
Seestar S50LP filter on = dual-band H-α + O III → HOO via DBXtractSmall sensor and short subs: O III needs hours. Switch LP off for galaxies and clusters
MK127 + ASI585MCDuo-Band; 120–300 s subsAt f/7.7 the filter works at its design band. Halos around very bright stars from reflections between the filter and the sensor
ASI585MC + Samyang 135Duo-Band; 60–180 s subsSee the f-ratio note below
Canon R8Drop-in LP filter with the EF Samyang onlyThe stock R8 records only a fraction of H-α and S II: narrowband work belongs on the ASI585MC
Fast lenses shift narrowband filters. Light arriving at an angle sees a filter's passband shifted toward blue. At f/2, rays at the edge of the light cone are about 14° off-axis, which moves the H-α passband about 5–9 nm bluer, depending on the filter's coating. At f/2.8 the shift is about half that, and at f/7.7 it's under 1 nm. A 3 nm filter at f/2 can lose much of its line. Use 6–7 nm filters, buy "high-speed" (pre-shifted) versions, or stop down to f/2.8. Calculated with the standard interference-filter formula λ = λ₀ √(1 − (sin θ ÷ n)²), for effective indices of 1.45 to 2.05.
Halos and rings. Bright stars such as Alnitak and Merak can grow rings through dual-band filters. These come from reflections between the filter, the sensor window and the lens. Tilting the filter slightly, or moving it further from the sensor, changes them. In processing, replacing stars with RGB stars hides them.

Appendix E · At the telescope: capture, gear and data

Parts B and C start from data already on disk. This appendix covers what happens before that:

Sub-exposure length and gain

A sub is long enough once the sky background's own noise swamps the camera's read noise. Beyond that point, longer subs gain almost nothing per hour; they just risk more lost time to satellites, wind and guiding errors. Shorter than that, read noise adds to every sub and you need more total time for the same result.

minimum sub ≈ 10 × (read noise)² ÷ (sky electrons per pixel per second)

At this length read noise adds only about 5% to the total noise (√1.1 ≈ 1.05). Sky flux depends on aperture, pixel scale, filter bandwidth and sky brightness, so each rig and filter has its own answer.

Rig and filterBortle 3 (21.5)Bortle 4–5 (20.5)Bortle 6 (19.5)Bortle 7–8 (18.5)
MK127 + 585MC, no filter (0.61″/px)~135 s~55 s~20 s~10 s
MK127 + 585MC, Duo-Band~15 min*~6 min*~140 s~60 s
Samyang 135 f/2 + 585MC, no filter~7 s~3 s~1 s< 1 s
Samyang 135 f/2 + 585MC, Duo-Band~45 s~20 s~7 s~3 s
R8 + Samyang 135 f/2.8, no filter~12 s~5 s~2 s~1 s
Seestar S50, LP off~3 min*~70 s*~30 s~10 s
Seestar S50, LP on~20 min*~8 min*~3 min*~75 s*

Calculated for this guide from a simple photon-budget model. Inputs:

Treat the results as ±2× guides. Measure your own sky level on a real sub to refine them: median ADU minus offset, times the camera's e⁻/ADU at your gain, divided by the exposure time. * marks lengths beyond what's practical (guiding, satellites) or allowed (the Seestar's sub choices).

Focus

Focus tolerance is set by the critical focus zone: ±2.44 λ N², about ±1.34 N² µm at 550 nm. That's how far the sensor can sit from perfect focus before stars grow measurably.

Setupf-ratioCritical focus zoneIn practice
Samyang 135 at f/2f/2±5 µmThe hardest to focus of all your rigs: a hair's turn of the ring. Use a Bahtinov mask and tape the ring
Samyang 135 at f/2.8f/2.8±10 µmTwice as forgiving
Seestar S50f/5±34 µmAutofocus built in. Re-run it after big temperature drops
MK127 + 0.65× reducerf/7.7±80 µmForgiving at the sensor, but the Mak moves its primary mirror to focus, so the knob is sensitive and shifts the image
MK127 native / planetaryf/11.8–f/15±190–300 µmEasy; seeing dominates

GTi, polar alignment and PHD2

  1. Mount mode: long deep-sky subs need the GTi in equatorial mode: on a wedge tilted to your latitude, with the counterweight bar fitted. Balance slightly east-heavy in RA so the gears stay engaged.
  2. Driver: run GS Server (GSS) and connect N.I.N.A. and PHD2 to its ASCOM driver. Enter the same site coordinates in GSS and N.I.N.A. (see coordinates).
  3. Polar alignment: rough-align with a compass and latitude scale (see true vs magnetic north), then refine with a camera.
    • Free: N.I.N.A.'s Three Point Polar Alignment (TPPA) plugin, or PHD2's Static Polar Alignment.
    • Paid: SharpCap Pro's polar alignment.
    Aim for under 5′ of error for the Mak, under 10′ for the 135.
  4. PHD2 profile (New Profile Wizard): camera ASI120MM Mini, pixel size 3.75 µm; focal length 120 mm (UniGuide 32); mount ASCOM GS Server. This gives the 6.4″/px guide scale. Let the wizard compute the calibration step size.
  5. Calibrate near declination 0° and within an hour of the meridian, on a reasonably bright star. Recalibrate only when the guide scope or camera moves; PHD2 adjusts the calibration for declination and pier side automatically.
  6. Guiding Assistant: run it for 2–3 minutes with guiding off. Accept its suggested minimum-move values and Dec backlash compensation. It also measures polar alignment error and seeing.
  7. Starting settings:
    • exposure 1–2 s, multi-star guiding on;
    • RA: Hysteresis algorithm, aggression ~70%, minimum move ~0.15–0.2 px;
    • Dec: Resist Switch, Auto direction.
    Lower aggression if the RA graph oscillates; raise it if errors persist.
  8. Targets: total RMS ≤ half the imaging scale, which is ~0.3″ on the Mak (0.6″ if binning). The 135 rigs are much looser: ~2″ for the ASI585MC and ~4″ for the R8. A GTi typically manages 0.6–1.2″ on a calm night, so the Mak is usually binned anyway.
  9. Dithering (from N.I.N.A.), with dither amounts in guide pixels. One guide pixel is 10 Mak pixels, 1.4 ASI585MC + 135 pixels, or 0.7 R8 + 135 pixels. That means:
    • 1–2 px for the Mak;
    • 3–5 px for the 135 rigs.
    Settle at 0.5–1 px for 10 s.
  10. Meridian flip: in N.I.N.A. set "minutes after meridian" to 5–10, re-centre after the flip with a plate solve, and let PHD2 flip its calibration. Check that nothing (cables, the guide scope) can hit the tripod.

Dew and power

Dew forms when a surface cools below the dew point. Optics radiate heat to the clear sky and can sit 2–5 °C below the air temperature, so dew arrives long before the air itself reaches the dew point.

Power budget for the Anker SOLIX C300: 288 Wh, of which about 85% (~245 Wh) is usable after conversion losses.

LoadTypical draw
GTi tracking (12 V)~6–10 W (more while slewing)
MeLE Quieter 4C running N.I.N.A., PHD2~8–15 W
ASI585MC + ASI120MM Mini (USB)~2 W
Mak dew strip at 50%~4–5 W
Lens or guide-scope strip at 50%~1.5–3 W each
Seestar charging over USB-C~10–20 W while charging
Setup on the C300TotalRuntime
MK127 rig (mount, PC, cameras, Mak and guide-scope dew)~25–30 W~8–10 h
Samyang 135 + ASI585MC rig~20–28 W~9–12 h
R8 + 135 on the GTi, no PC (handset or app)~10–15 W (plus R8 batteries)~16–24 h
Any rig + charging the Seestar+10–20 W for 2–3 h1–3 h less

Typical figures for these device classes, not measurements of your units. A USB-C or DC inline power meter on each cable gives your real numbers in one evening. Winter nights cut battery capacity and raise dew-heater demand; plan for the low end.

Remote operation with the MeLE

The MeLE Quieter 4C is fanless, small and runs on the mount. With N.I.N.A., PHD2 and GS Server on it, the whole rig can run from indoors: you connect from a laptop, tablet or phone, start the sequence, and check in occasionally. The goal is a setup that survives a dropped connection: the MeLE does the work, and the remote screen is only a window onto it.

Connecting to it

OptionCostGood forNotes
Windows Remote Desktop (RDP)Free with Windows ProHome network; the smoothest on a laptopThe MeLE must run Windows Pro to host RDP (check Settings › System › About). Works without a monitor attached. Clients: Windows App on Mac and iPad, and Windows
NoMachineFreeHome network, any client OSWorks on Windows Home. It mirrors the real display, so fit an HDMI dummy plug to give the MeLE a proper screen resolution when headless
RustDesk, AnyDesk, TeamViewerFree or freemiumFrom anywhere via the internetRelay through the vendor's servers; needs internet at the scope. Fine for checking in; laggy for focusing
Tailscale + RDP or NoMachineFree tierSecure access from outside homeA private VPN between your devices. Never expose RDP directly to the internet

Networking

Preparing Windows on the MeLE

  1. Never sleep: Settings › System › Power: screen and sleep to Never when plugged in. In the advanced power options, set USB selective suspend to Disabled, so cameras don't disconnect.
  2. Windows Update: set active hours across the night, or pause updates before an imaging run. A forced restart at 2 am ends the session.
  3. Automatic sign-in, and startup apps (GS Server; optionally N.I.N.A.), so a power blip or reboot brings the system back without a keyboard.
  4. Time: keep Windows' automatic time sync on (see time). In the field without internet, a USB GPS receiver with a time-sync utility keeps the clock accurate. A 1 s clock error is 15″ of pointing error.
  5. Drivers: install the ZWO ASCOM and native drivers, ASCOM Platform and GS Server, and test every device indoors in daylight before the first remote night.
  6. USB: ASI585MC to a USB 3 port directly. Guide camera, mount cable and any focuser through a powered USB hub. Strain-relieve every cable at the mount, so a slew or meridian flip can't pull a plug.
  7. Storage: save images to the MeLE's internal SSD, never across the network during capture. Check free space before each night (see data for sizes).

A night run remotely

  1. At the scope, at dusk:
    • Set up, power on, and put the dew heaters on.
    • Remove the lens cap.
    • Rough polar alignment.
    • Check the remote connection works from indoors before leaving the scope.
  2. Indoors, remotely:
    • Run the polar alignment refinement (TPPA).
    • Focus with the HFR readout; with a motor focuser, autofocus.
    • Start PHD2 guiding and check the graph.
    • Load the sequence.
  3. Let N.I.N.A.'s Advanced Sequencer run the night. A typical template:
    • wait for astronomical darkness;
    • slew, centre and plate solve the target;
    • start guiding;
    • loop exposures until the target reaches the altitude limit or dawn, with dithering every N frames and refocus on a temperature change;
    • handle the meridian flip;
    • at the end: stop guiding, park the mount, and optionally shut down the PC.
  4. Notifications: the N.I.N.A. Ground Station plugin sends messages (Pushover, Telegram, email and others) on errors, sequence end, or failed guiding. You don't need to watch the screen.
  5. Check in occasionally: look at HFR, star count and the guiding RMS in N.I.N.A.'s image history. Rising HFR means dew or focus; falling star count means cloud. Don't run focusing over a laggy internet relay; use the local network.
  6. Battery: plan the sequence to end before the C300's estimated runtime (see power). As the last sequence item, an External Script running shutdown /s /t 60 shuts the MeLE down cleanly after parking.
  7. Morning: copy the night's data to the main computer over the network, or with a sync tool such as Syncthing, then back it up. Cap the scope and bring it in.
Unattended safety.
  • Rain: without a cloud or rain sensor, nothing stops the sequence if rain arrives. Only leave the rig unattended on nights with a confident clear forecast (see planning tools), and keep a waterproof cover nearby. N.I.N.A. supports ASCOM Safety Monitor devices (cloud and rain sensors) that can stop and park the mount automatically, if you add one later.
  • Collisions: check that the camera and guide scope clear the tripod legs at every position, and through the meridian flip, before leaving it. A slew into a tripod leg can strip a GTi's gears.
Processing on the MeLE?
  • It can run PixInsight, but a fanless mini PC is slow for WBPP and the XTerminators.
  • Keep it as the capture computer and process on your main machine.
  • The exception is a quick check stack of the night's subs, to confirm focus and framing.

Keeping the MK127 in shape

Reducer or native?

The 0.65× reducer isn't required. For deep sky it buys field and convenience, not depth at the resolution you'll actually keep.

With 0.65× reducerNative
Focal length, ratio~975 mm, f/7.7 (solves 955–985 mm)~1500 mm, f/11.8
Image scale (ASI585MC)0.61″/px0.40″/px
Field39′ × 22′25.6′ × 14.4′ (2.3× less sky)
Stars at 2–3″ seeing3.5–5 px FWHM5–7.5 px FWHM (heavily oversampled)
Usual binning2×2 → 1.23″/px3×3 → 1.20″/px
Signal per binned pixelThe same: it depends only on aperture (127 mm either way) and the final scale
Read noise per binned pixel2 × RN (4 pixels summed)3 × RN (9 pixels summed): negligible once subs are sky-limited
Sky-limited sub, unfiltered, Bortle 5~55 s~2 min (each native pixel gets ~2.3× less sky)
Sky-limited sub, Duo-Band, Bortle 5~6 min~14 min: impractical, so accept some read noise
Pointing and plate solvingMore forgiving of GTi GoTo errorTighter; use plate-solve centring
Things that can go wrongNeeds exactly 55 mm spacing; corners soften if spacing or tilt is offSimpler optical train; flat field across the small sensor

Sub lengths from the photon-budget model in sub-exposure length; ±2×.

Use the reducer forGo native for
  • Most deep sky: galaxies over ~5′, nebulae, globular clusters.
  • Targets that need the extra field: M81, M101, M13 with NGC 6207, NGC 7331 with Stephan's Quintet.
  • Mosaics: fewer panels.
  • Duo-Band work: practical sub lengths.
  • Planets and the Moon, with the Q-Barlow (see planets).
  • Double stars (see double stars).
  • Small bright planetary nebulae: M57, NGC 7662, the Cat's Eye, the Eskimo.
  • Small galaxies under ~5′ on nights of excellent seeing, unbinned.
Rule of thumb: keep the reducer on for deep sky. Take it off for planets, the Moon, doubles and tiny planetary nebulae. If you go native for deep sky, bin 3×3 (IntegerResample) and expect longer subs.

A calibration library for uncooled cameras

The ASI585MC has no cooler, so its sensor follows the air temperature. Dark current roughly doubles every ~6 °C, so darks must match the lights' exposure, gain, offset and temperature.

FrameHow manyWhen to retakeNotes
Darks20–30 per setEvery 6–12 months, or after a firmware or driver changeOne set per exposure length, in temperature bins of ~5 °C (e.g. 0, 5, 10, 15 °C). Shoot them cap-on on cloudy nights outdoors, or in a fridge or cool garage, to hit the bins
Flats20–30Every session, or whenever the camera, focus, filter or aperture changesDust moves; flats must match the optical train exactly. An LED panel or tablet with an even white screen. Aim for a histogram peak at ~40–50%
Dark flats20–30With the flatsSame exposure as the flats, cap on. Used instead of bias frames for flats
Bias50RarelyOptional with dark flats; WBPP can use them for dark scaling

Data management

SourcePer subA typical night
ASI585MC, 16-bit FITS (3840 × 2160)~16.6 MB4 h of 60 s subs ≈ 240 files ≈ 4 GB
Seestar, 16-bit FITS (1080 × 1920)~4 MB3 h of 10 s subs ≈ 1,080 files ≈ 4.5 GB
Canon R8 CR3 (24 MP)~25–35 MB2 h of 60 s subs ≈ 120 files ≈ 3.5 GB
WBPP intermediates3–5× the raw lightsDelete after integration

Comets, meteors and sky events

Finding comets:

Processing in PixInsight:

  1. Calibrate and register on the stars in WBPP as usual.
  2. CometAlignment on the registered frames. Mark the comet's nucleus in the first and last frames; it interpolates the motion between them.
  3. Make two integrations:
    • stars, from the star-registered frames, with rejection removing the comet;
    • the comet, from the comet-aligned frames, with rejection removing the trailed stars.
  4. Remove stars from the comet stack (StarXTerminator), process both stacks, and combine them with the screen blend from step 13.

Keep subs short enough that the comet moves less than one pixel per sub: motion (″/min) × sub length (min) < image scale.

Meteor showers, with Moon conditions (calculated from the mean lunar cycle, ±1 day):

ShowerPeakMoonVerdict
Geminids 202613–14 Dec 2026New 9 Dec: thin crescent, sets earlyExcellent
Quadrantids 20273–4 Jan 2027Waning crescent; new ~7 JanGood (short peak)
Lyrids 202722 Apr 2027Full ~20 AprPoor
Eta Aquariids 20276 May 2027New ~5 MayGood (best from the south)
Perseids 202712–13 Aug 2027First quarter ~10 Aug; sets around midnightGood after midnight
Orionids 202721 Oct 2027Full ~15 Oct; waning gibbousFair
Leonids 202717 Nov 2027Full ~13 NovPoor
Geminids 202714 Dec 2027Full ~13 DecPoor

Shoot meteor showers with the R8 and the 15 mm (see R8 panoramas), as hundreds of 15–20 s frames. Keep the frames with meteors, align them on the stars (or on the radiant for a radiant composite), and blend them with Lighten mode.

Eclipses: the total solar eclipse of 2 August 2027 crosses southern Spain, North Africa and Egypt, with up to about 6 minutes of totality near Luxor. Only during totality is it safe to look or shoot without a solar filter. Check lunar eclipse dates for your location on timeanddate.com.

Science with your gear

ProjectWhat you measureBest rigSoftware and where results go
Variable starsBrightness changes of stars (Miras, eclipsing binaries, novae) relative to comparison starsASI585MC + 135 or Mak; slightly defocused stars, unsaturatedASTAP's photometry tool or AstroImageJ → AAVSO (its DSLR/CMOS guides explain using the green channel as V)
Exoplanet transitsA 0.5–2% dip as a hot Jupiter crosses its starMak, binned, defocused; 2–4 h continuous, guidedAstroImageJ or HOPS → Exoplanet Transit Database, ExoClock
Asteroid occultationsThe exact moment a star blinks out behind an asteroidMak + ASI585MC, videoOccult Watcher for predictions; IOTA for reporting. Needs accurate timing (GPS or a tightly synced clock); rolling-shutter cameras like the ASI585MC need timing corrections
Asteroid and comet astrometryPrecise positions to refine orbitsMak or 135ASTAP or Astrometrica → Minor Planet Center (needs an observatory code)
Double-star measuresSeparation and position angleMak, planetary setupPixInsight (see double stars) → WDS submissions
Supernova checksNew stars in galaxiesMak or SeestarCompare your images with references; report via the Transient Name Server after confirmation

For photometry, don't stretch, sharpen, denoise or run BXT. Measure calibrated, linear subs or stacks: everything in Parts B–C after step 6 changes brightness relationships.

Planning tools

ToolBest for
Telescopius (web)Target lists by season, framing with your exact camera and lens, altitude charts
Stellarium (desktop, free)Planetarium with sensor-field overlays (Oculars plugin), comets, planets, eclipses
N.I.N.A. Sky Atlas and Framing AssistantTarget search with altitude and Moon distance; framing and mosaics sent straight to the sequence
Clear Outside (app/web)Hourly cloud by layer (low, mid, high), dew point, an overall "go/no-go"
Astrospheric (app/web)Cloud, transparency and seeing forecasts (North America); smoke
Meteoblue astronomy seeing (web)Seeing in arcseconds and jet-stream speed: the planetary and double-star forecast
Windy (app/web)Cloud and wind models side by side, to see whether forecasts agree
Light pollution map (lightpollutionmap.info)Bortle and SQM estimates for nearby dark sites
SkySafari (app)In-field finder, and it can control the GTi over Wi-Fi for visual use
Seestar app planningScheduling Seestar targets with start times and mosaic options
Heavens-Above (web)ISS and satellite passes (to avoid or to capture)

Capture troubleshooting

SymptomLikely causeFix
Stars elongated, all in the same directionGuiding error, wind, poor balance, cable dragCheck the PHD2 graph for the axis; rebalance east-heavy; tie cables; shield from wind
Elongation varies around the frameTilt or wrong reducer spacingAberrationInspector; see Mak spacing
Periodic RA streaks every few minutesMount periodic errorGuide (shorter exposure, a bit more aggression); PEC in GSS if available
Dec guiding slow to respond, or overshootsDec backlashGuiding Assistant's backlash measurement; Dec "North" or "South" only, with a slight deliberate polar misalignment
Stars swelling over the nightFocus drift (temperature) or dewRefocus each 1–2 °C; dew heater; cull by FWHM
Star count falling, background risingThin cloud, haze or dewCheck the sky and optics; cull with SubframeSelector (stars, median)
Diagonal correlated "rain" streaks after stackingWalking noise (no dithering)Dither every 1–3 subs
Straight lines through some subsSatellites and planesPixel rejection (Generalized ESD) needs ≥ 15–20 subs; keep the subs
Dark or bright rings after flatsFlats from a different session or focus; dust movedFresh flats every session without touching the train
Overcorrected (bright) vignettingNo dark flats or bias; light leakUse dark flats; cover gaps in adapters
Halos around bright starsDew, filter reflections, chromatic errorDew heater; tilt the filter; stop the lens down; replace stars in processing
Plate solve failsWrong focal length or pixel size; too short an exposure; cloudsCheck the values (955–985 mm, 2.9 µm for the Mak); 5–10 s solve exposures; give a search radius
Meridian flip fails or loses the targetFlip settings; mirror shift on the MakRe-centre with a plate solve after the flip; a pause before the flip
Seestar stars trailed or rotated near the zenithAlt-az field rotationAvoid targets above ~80° altitude; EQ mode

Exporting and sharing

  1. Keep a master: the final XISF plus a 16-bit TIFF. Every export starts from it, never from a JPEG.
  2. Colour profile: convert to sRGB with ICCProfileTransformation before saving web images. Phones, browsers and social sites assume sRGB, and an untagged wide-gamut image looks dull or oversaturated.
  3. Size: use Resample to fit the destination.
    • AstroBin: full resolution.
    • Forums: 2,000–3,000 px on the long side.
    • Instagram: 1,080 px wide, cropped to 4:5 for portrait posts.
    After a big reduction, add a very gentle sharpening (UnsharpMask, small radius, low amount).
  4. Banding: smooth gradients (dark sky, galaxy halos) can show steps in 8-bit JPEGs. Save at quality 90–95, avoid extreme curves in the last step, and check the export at 100% before posting. A faint noise floor helps hide banding, so don't denoise the background to perfect smoothness.
  5. Annotations: Script › Render › AnnotateImage on the plate-solved image labels catalogue objects and draws a coordinate grid. Save it as a separate version.
  6. Details to post with an image:
    • target;
    • rig (telescope or lens, camera, mount);
    • filter;
    • sub length × count and total integration;
    • site Bortle and Moon;
    • processing summary.
    AstroBin fills much of this in from your equipment list.
  7. Check on a phone before posting: most people will see it there. Images processed on a bright monitor often look too dark or too saturated on phones.

About the screenshots

Every image in this guide is a real PixInsight image. Dialogs captioned Your install were captured from your PixInsight 1.9.6 (build 1718) on 14 November 2026, with each process set to this guide's values. They cover SubframeSelector, the WBPP main dialog, SPCC, SPFC, MGC, BlurXTerminator, StarXTerminator, MultiscaleAdaptiveStretch, NoiseXTerminator and PixelMath. The SubframeSelector graphs are measurements of your own M45 subs. Images captioned Your M45 data come from processing those subs through Part B: 16 soft frames and 1 cloudy frame culled, 35 of the remaining 45 stacked in WBPP (the other 10 were from a non-overlapping mosaic panel), then every step run with the settings in this guide. Linear stages are shown with a linked auto-stretch so they're visible; non-linear stages are shown as they are. Images captioned as the M45 mosaic come from all 45 culled subs, plate solved, reprojected with MosaicByCoordinates and integrated (Route A in Seestar mosaics); the Route B images come from the same subs grouped into 4 panels, stacked without autocrop, merged with GradientMergeMosaic (shrink 10, feather 30) and processed with the same settings. The remaining images come from the bundled documentation.

Still to do:

Sources