How to shoot the frames
Everything colorcal tells you is derived from the photographs you feed it. If the camera moved the goalposts between shots, the numbers that come out the far end are describing the camera, not the wall. These nine rules are the whole job. None of them need color science, and each one says why it matters so you can judge when to bend it.
Skim this
The checklist
Tap any line for the reasoning behind it.
- 01Shoot straight on. Panels change color when you look at them from an angle.
- 02Lock exposure. Auto exposure re-scales every frame and breaks the comparison.
- 03Lock white balance. Auto white balance silently removes the error you are measuring.
- 04Match shutter to the wall's refresh. Otherwise you get bands across the frame.
- 05No LUT, no creative look. A baked look cannot be separated from the wall's error afterwards.
- 06Export 16-bit if you can. More levels means a finer correction, especially near black.
- 07Fill the frame, keep the whole chart in it. Big patches average out noise. Cropped edges lose the chart.
- 08Shoot a black frame too. It sets black level, which is half the correction.
- 09Focus reasonably sharply. A soft chart smears patches into each other.
The one-sentence version
Put the camera in full manual, point it square at the middle of the chart, and change nothing at all between the shots except which pattern is on the wall.
The reasoning
Nine rules, and why
01 Shoot straight on
Put the camera on axis: lens centered on the middle of the chart, at the same height, square to the wall rather than off to one side. Back up far enough that the wall stays flat in frame instead of wrapping away from you. If the wall is curved, shoot the flattest section you can find and treat that as your reference.
LED panels are directional. The light they put out changes with the angle you view them from, and the three color channels do not fall off at the same rate, so an off-axis view is both dimmer and differently tinted. That is a property of the panel, not a fault.
Shoot at an angle and you measure the wall's off-axis behavior, then bake that angle into the correction. The wall ends up tuned for the one place you happened to stand, and looks wrong from everywhere else, including the seats and the camera that is actually shooting the show.
02 Lock exposure
Full manual. Fix the aperture, the shutter, and the ISO, and do not touch any of them again until every frame is shot. Set the exposure so the brightest white patch sits comfortably below clipping. If your camera has a histogram or highlight warning, use it: the white patch should be bright but not pinned against the right-hand end.
Auto exposure re-meters every single frame. Point it at the black frame and it opens up, point it at a bright chart and it stops down. colorcal works by comparing frames to each other and to the known chart, so if the camera quietly rescales between shots, every comparison is measuring the camera's decisions rather than the wall.
Clipping matters for the opposite reason. Once a channel hits the top of its range it stops recording differences, so a clipped white patch looks perfectly neutral even on a wall with a strong cast. Leave headroom and the cast stays visible in the data.
03 Lock white balance
Set a manual white balance and leave it. A fixed Kelvin value near daylight, around 5600K to 6500K, is a sensible default. Do not use auto, and do not white-balance off the wall itself between shots.
White balance is the camera's own color correction: it decides which part of the scene is meant to be neutral and cancels the cast it sees. That is exactly the job you are asking colorcal to do. Leave it on auto and the camera fixes the wall's cast in the file, the measurement comes back clean, and you learn nothing about a wall that is visibly green to everyone in the room.
The exact value you choose matters far less than keeping it constant. Any fixed white balance leaves the wall's error intact and measurable. Avoid extreme settings only because they push one channel toward clipping and another toward the noise floor, which costs you accuracy at both ends.
04 Match shutter to the wall's refresh rate
Pick a shutter speed that lines up with how fast the wall refreshes. A slower shutter, 1/50 or 1/60 of a second rather than 1/500, is the easiest fix because it averages several refresh cycles together. Shoot a test frame and look at it at full size before you commit: if you see horizontal bands of differing brightness, change the shutter and shoot again until they are gone.
An LED wall is not lit continuously. Each panel flickers on and off very fast, and controls its brightness by how much of each cycle it spends lit. Meanwhile most camera sensors do not capture the whole image at one instant, they read it out row by row from top to bottom.
Put those two together and different rows of the picture catch different amounts of the flicker, which shows up as bands rolling across the frame. Those bands are a brightness error of several percent that varies by position, so a patch near a dark band measures dimmer than the same patch would near a bright one. A longer exposure spans enough cycles that the flicker averages out.
05 No LUT, no creative look
Turn off any look, film emulation, or picture profile that changes color rather than just recording it. Shoot the camera's standard or neutral mode. A LUT on your monitor is fine, because it is only affecting what you see. A LUT baked into the recorded file is not.
A LUT, short for lookup table, is a fixed recipe that maps every input color to a different output color. That is the same kind of thing colorcal is trying to work out for your wall. If the file already has one applied, the wall's error and the LUT's intent are added together into a single result, and nothing downstream can tell you which part came from which.
The practical symptom is a correction that fights the grade. You end up adjusting the processor to undo a look somebody chose on purpose, and the wall drifts further from correct with every pass.
06 Export 16-bit if you can
A 16-bit TIFF or PNG is ideal. Regular 8-bit files still work, so do not stop the job over it, but take the 16-bit option when your camera or your export dialog offers it. Avoid heavy compression, and turn off aggressive noise reduction and sharpening if you have the choice.
Bit depth is how many distinct brightness levels a file can hold per color channel. An 8-bit file has 256 of them. A 16-bit file has tens of thousands. In the bright half of the picture 256 is plenty, but the dark patches all crowd into the bottom handful of those levels, so the difference between a good black and a slightly green black may be a couple of steps, sitting in the same range as the sensor's own noise.
More levels means the low end stays measurable, which is where most of the visible improvement on an uncorrected wall actually comes from. Compression and noise reduction cost you the same thing from the other direction: they smooth and invent detail across flat patches, which is precisely the region being measured.
07 Fill the frame, keep the whole chart in it
Get close enough that the chart occupies most of the picture, but leave a margin so the outer edge of the chart is fully inside the frame with a little wall visible around it. Nothing should be cropped, and nothing should be in front of the wall casting a shadow or a reflection onto it.
Each patch is measured by averaging the pixels inside it. More pixels means the sensor's random noise cancels out and the average lands closer to the true value. A chart that fills the frame can give you thousands of pixels per patch, and a chart shot from across the room might give you a few dozen.
The margin matters because the tool has to find the chart before it can read it, and it does that by locating the outer boundary. Clip a corner and the geometry no longer resolves, so every patch is sampled slightly off-center or the frame is rejected outright.
08 Shoot a black frame too
Put full black on the wall and shoot it with the identical settings you used for every other frame. Same exposure, same white balance, same distance, same position. Do not skip it because the picture looks like nothing. Keep the room lighting the same as it was for the other shots.
A wall showing black is not emitting nothing. There is a residual glow from the panels, an offset the processor is holding, ambient light in the room bouncing off the surface, and flare inside your own lens. All of that adds a floor underneath every other measurement, and it is rarely neutral, which is why uncorrected walls so often have colored shadows and a milky, low-contrast look.
Once colorcal knows the floor it can subtract it, and the numbers for everything above it get better at the same time. Black level is genuinely half the correction. It is also the single change most likely to make somebody walking past say the wall looks better, because contrast is what the eye notices first.
09 Focus reasonably sharply
Focus manually on the chart and check it at full magnification before you shoot the set. It does not need to be critically sharp, it needs to be clearly in focus. If you see a rippling interference pattern across the picture, back the focus off very slightly or change your distance to the wall until it goes away.
A soft image spreads each patch into its neighbours. The tool samples the middle of every patch expecting one flat color, and instead gets a blend of that patch and whatever is next to it, which pulls every reading toward the average of the chart and flattens exactly the differences you are trying to measure.
The ripple is called moire, and it happens when the grid of LEDs on the wall and the grid of pixels on your sensor land close enough in size to interfere with each other. It creates false color patterns that are not on the wall at all. A touch of defocus, or a small change in distance, breaks the interference without meaningfully softening the patches.
Worth knowing
A few things that trip people up
- Shoot the whole set in one sitting. Walls drift as they warm up, and the room's ambient light changes as the day does. Frames shot an hour apart are describing two slightly different walls.
- Let the wall warm up first. Give it ten or fifteen minutes of content before you measure. LED output shifts noticeably in the first few minutes from cold.
- Do not move the camera between frames. A tripod makes this easy and also lets you use the slower shutter from rule 04 without smearing.
- Send the pattern at full scale, unscaled. If the processor is stretching or resampling the chart, the patch edges soften and the values shift. Match the pattern's resolution to the wall where you can.
- A phone can work. It needs an app that gives you manual exposure, manual white balance, and manual focus, and it needs to save something better than a heavily processed JPEG. Many do.
Next
You have the technique. Now the patterns.
Six frames is the whole shoot: the neutral hex chart at four brightness levels, the grayscale wedge, and one black frame. They are all free downloads from testpattern.app.