Calibration frames remove sensor artifacts from your images: bias frames (near-zero-length exposures) capture read noise, dark frames (same exposure and temperature as your lights) capture thermal noise, and flat frames (even illumination shots) correct vignetting and dust. Combine each type into a master frame, then subtract darks and divide by flats to calibrate your stacked light frames, per Celestron's guide.
The four frame types, and what each one corrects
Every raw frame your camera produces carries noise and optical artifacts that have nothing to do with the sky — some baked in by the sensor's own electronics, some introduced by dust and vignetting in your optical train. Calibration frames measure those artifacts on their own, with no starlight in the way, so your stacking software can subtract or divide them back out of your actual data. There are four types, and each one isolates a different piece of the problem Celestron, “The Ultimate Guide to Calibration Frames”.
| Frame type | What it captures | Typical exposure & settings | What it corrects |
|---|---|---|---|
| Bias | Read noise + fixed pattern noise | Near-zero length (under 1s), sensor covered | The read-noise floor present in every other frame type |
| Dark | Thermal noise (dark current) + hot/cold pixels | Same length & sensor temperature as your lights | Thermal noise and hot/cold pixels in the stack |
| Flat | Vignetting + dust shadows | Even illumination, unchanged optical train & focus | Vignetting and dust doughnuts in the light frames |
| Flat-dark | Thermal noise at the flat's own (short) exposure | Matched to the flat frame's length & temperature | Calibrates the flat frames themselves before use |
Bias frames
A bias frame is the shortest exposure your camera can take — under a second, sensor covered or lens capped so no light reaches it at all — and it isolates read noise: the noise a sensor's electronics add just by reading out a frame, plus any fixed pattern noise baked into the chip itself Celestron, “The Ultimate Guide to Calibration Frames”. Because the exposure is so short, there isn't time for meaningful dark current (thermal noise) to accumulate, which is why bias frames are the one calibration type that's unaffected by cooling Sky & Telescope, Richard Wright — a bias frame shot warm and one shot cold look essentially the same. That property is also what puts bias frames at the center of the debate below: if a temperature- and exposure-matched dark frame already contains read noise as part of matching your lights, what is a separate bias frame actually adding?
Dark frames
A dark frame is exposed for the same length of time, at the same sensor temperature, as your light frames — with the optical path still blocked, so nothing but sensor noise reaches the pixels Celestron, “The Ultimate Guide to Calibration Frames”. It captures thermal noise (dark current, the electrons a sensor generates on its own as it warms) along with any hot or cold pixels that would otherwise show up as fixed bright or dark specks across every light frame in the stack. Temperature matching isn't optional here: dark current roughly halves for every 6–7°C the sensor is cooled Sky & Telescope, Richard Wright, so a dark shot even a few degrees off your lights' actual sensor temperature will under- or over-subtract, leaving residue or dark blotches in the stack rather than cancelling cleanly.
Flat frames
A flat frame is an even, uniformly lit exposure — a twilight sky, an electroluminescent panel, or a light box — shot through the exact same optical train, at the same focus and camera rotation, as your lights Celestron, “The Ultimate Guide to Calibration Frames”. Because every pixel should be receiving identical illumination, any variation the flat records — a dimmer corner from vignetting, a dark ring or doughnut from a speck of dust on a filter or sensor window — is optical, not astronomical. Software turns those variations into a per-pixel correction factor and divides your lights by it, which is why flats have to be reshot any time the imaging train's optics or dust pattern change, not just when the target changes.
Flat-dark frames
A flat-dark (sometimes called a dark-flat) frame is a dark frame matched to the flat's own exposure length and temperature rather than the light frame's, since a flat exposure is often a fraction of a second to a few seconds — far shorter than a typical light sub Celestron, “The Ultimate Guide to Calibration Frames”, corroborated by Practical Astrophotography's calibration-frame guide Practical Astrophotography. Its job is narrow: calibrate the flat frames themselves, removing the same read-noise-and-thermal-noise floor from the flat that a regular dark removes from a light, before that flat gets turned into a correction factor. Some imagers use a flat-dark here instead of a bias frame for reasons that show up again in the debate below.
Do you need bias frames if you already shoot matched darks?
This is the one place in calibration-frame procedure with a real, ongoing disagreement — not a settled answer we're smoothing over for the sake of a tidy article. Two things are true and agreed on by both sides: a bias frame isolates read noise on its own, and a dark frame that's properly exposure- and temperature-matched to your lights already contains that same read-noise signal, because reading out any frame — bias, dark, or light — adds it Sky & Telescope, Richard Wright. Where imagers split is what follows from that shared fact.
Position A holds that a separate bias frame is redundant once your darks are properly matched in both exposure length and sensor temperature to your lights — the bias signal is already embedded in the dark frame, so subtracting the dark handles both jobs in one step Sky & Telescope, Richard Wright. Position B holds that bias frames still earn their place: they let you scale a single dark library to light frames shot at a different exposure length without reshooting darks from scratch, and some sensors — the ZWO ASI1600 is the example that comes up repeatedly — show banding artifacts at the very short exposures a bias frame requires, which pushes those imagers toward flat-darks in place of a bias everywhere one would otherwise be used Cloudy Nights. What both agree on: the underlying physics above; the disagreement is entirely about whether isolating that signal separately is worth the extra library once your darks are already matched.
This gets argued out across multiple active Cloudy Nights threads without a converged answer Cloudy Nights, and it isn't Dew & Dark's place to declare a winner where the community hasn't reached one. If your darks are reliably matched to your lights in both exposure length and regulated sensor temperature, Position A says you can drop bias frames from your workflow. If you shoot varying exposure lengths, or your sensor is one of the ones known to band at bias-length exposures, Position B's flat-dark- based workflow is the safer default — and the flat-dark frame from the section above already gives you the tool that side relies on.
How many frames you need, and how to build the masters
The count that's actually well-sourced here is flats: a workable range is roughly 20–50 individual frames combined into one master flat, shot per filter you use, since dust position and vignetting can differ slightly between filters in the same imaging train BBC Sky at Night Magazine; AstroBackyard.
Bias and dark libraries don't have an equivalent published number. Despite “how many dark frames do I need” being one of the most commonly searched questions about calibration frames, no manufacturer guide or comparably authoritative source in our research states a canonical bias or dark count — so treat any specific figure you see quoted online for bias or dark frames as a practitioner's habit, not a sourced spec. What is true, and is the entire reason to combine frames into a master in the first place, is that averaging more sub-frames together reduces the noise the master itself contributes to your calibration; a master built from more frames is a cleaner correction than one built from fewer.
Your capture or stacking software combines each stack of same-type frames — bias, dark, flat, or flat-dark — into one master, typically by averaging or median-combining the set, which also rejects one-off outliers like a cosmic-ray hit on a single sub. Because bias and dark frames don't depend on the sky, a matched-temperature library can be reshot occasionally and reused across many imaging sessions rather than captured fresh every night. Flats are the exception, since they're tied to the current state of your optical train — see the retake section below.
The calibration order, worked through
Once your masters exist, calibrating a light frame is a fixed two-step operation, not a checklist you can reorder: subtract, then divide Celestron, “The Ultimate Guide to Calibration Frames”.
Building the masters comes first. The bias subs combine into a master bias, if your workflow uses one. The dark subs — matched in exposure length and sensor temperature to your lights — combine into a master dark. The flat subs combine into a raw master flat, which then gets calibrated in its own right: subtract a master flat-dark (or master bias, depending on which side of the debate above your workflow follows) from the raw master flat, then normalize the result so its average pixel value works out to 1 rather than some fraction of the sensor's full well. That normalization step is what keeps flat correction from darkening your whole image instead of just evening it out.
Calibrating each light frame is then two operations, always in this order: subtract the master dark, then divide by the normalized master flat Celestron, “The Ultimate Guide to Calibration Frames”. Subtraction removes the additive noise — read noise, thermal noise, hot and cold pixels — that the dark frame captured. Division removes the multiplicative pattern — vignetting, dust shadows — that the flat frame captured, and it has to happen second: dividing a frame that still has thermal noise and hot pixels baked in would scale that noise unevenly across the image instead of correcting it cleanly. Only once every light in the set has been through both steps does the stack get combined into your final integration.
When to retake your calibration frames
Three things reliably invalidate an existing calibration set, and none of them are about how long ago you shot it.
Sensor temperature drift is the one that matters for darks. As our camera-buying guide's cooling section covers, a regulated setpoint is what actually matters — ZWO's own ASI2600MM/MC Pro manual puts dark current at roughly 0.0022e⁻/s/pixel at a 0°C setpoint, about 0.7e⁻ accumulated over a 300-second sub ZWO ASI2600MM/MC Pro Manual — and that figure only holds at the specific temperature it was measured at. Change your setpoint, or image on an uncooled camera on a night with a different ambient temperature than your existing darks were shot at, and dark subtraction stops cancelling cleanly.
A gain or offset change is the second trigger, and it isn't confined to bias and dark libraries. Changing either setting shifts the read-noise and pixel-value-floor relationship your existing bias and dark frames were built around, so treat any settings change as a reason to reshoot those rather than reuse an older library. The same logic reaches your flat calibration too: whichever frame calibrates the raw flat before normalizing — a flat-dark or a bias — was shot at the old gain or offset, so rebuild that calibrating frame and its master flat alongside your bias and darks rather than assuming a settings change only touches the sensor-noise frames. See our gain and offset guide for how a gain or offset change invalidates your existing calibration set.
Flats are the odd one out for what invalidates the raw flat's own optical correction — that part isn't about the sensor at all, it's a snapshot of your optical train's current dust and illumination pattern. Refocusing between sessions doesn't invalidate a flat as long as nothing physically moved, but rotating the camera in its adapter, swapping or cleaning a filter, or reseating any component in the train does, since each of those can shift exactly the dust shadows and illumination falloff the flat exists to correct. What does depend on sensor settings, per the gain and offset trigger above, is the flat-dark or bias that calibrates the flat itself.
Elongated or seagull-shaped stars in the corners of a stack is not a symptom flats fix. Flats correct uneven illumination and dust shadows — they have no effect on the shape of individual stars. If that's what you're chasing, it's very likely a back-focus or tilt problem instead, and reshooting flats won't touch it. Our tilt-vs-back-focus diagnostic is where to go next, not a new set of calibration frames.
A cooled camera with a regulated, fan-forced setpoint — the ZWO ASI2600MC Pro, the camera this hub keeps coming back to — turns hitting the same sensor temperature every session into a settings choice rather than something you have to chase with ambient conditions; check current price at Agena. An uncooled camera can still be calibrated well — you just have to shoot darks the same night as your lights, since ambient temperature is what's actually being matched.
FAQ
Do I need bias frames if I already shoot matched darks?
It's genuinely unresolved in the community, not a settled house recommendation. One position says no — a dark frame properly matched in exposure and sensor temperature to your lights already contains the bias signal, so a separate bias frame is redundant. The other says bias frames still earn their keep for scaling a dark library across exposure lengths, and that some sensors band at bias-length exposures, which pushes those imagers toward flat-darks instead Sky & Telescope, Richard Wright; Cloudy Nights. Pick based on your workflow: matched darks every session favors dropping bias; varying exposure lengths favors keeping it.
How many dark or bias frames do I need?
No manufacturer guide or comparably authoritative source we found states a canonical count for bias or dark frames — unlike flats (below), this isn't a sourced number. More sub-frames in a master reduce the noise that master contributes to your calibration, so more is generally better, but treat any specific figure you see quoted online as a practitioner's habit rather than a spec.
How many flat frames do I need for a master flat?
Roughly 20–50 individual flats combined into one master, shot per filter you use, since dust position and vignetting can differ slightly by filter BBC Sky at Night Magazine; AstroBackyard.
Does a dark flat (flat-dark) replace a bias frame?
For calibrating your flat frames themselves, yes — that's a flat-dark's defined job, matched to the flat's own short exposure and temperature. Whether a dark-flat-based workflow can replace bias frames everywhere else in your calibration is the unresolved debate above, not a settled swap Cloudy Nights.