Hi Jim , i was playing around with some raws in ACR this morning and noticed the same happening with my raw files from the sony a74 the more i pushed the exposure slider the image was not clipping only pushing the lighter tone together up against the wall. sony does exactly the same with the microphone gain controll.
April 10, 2023
57
-
-
Yes I notice that as well with ACR, but only up to a point and it varies with images.
On just about every image, once I push the "exposure" slider more than about 1.5 stops to the right the clipping becomes obvious, especially if the highlight clipping indicator is turned on.
-
how do you turn that on ? i clicked on the little box in the top right and hi lighted it but nothing happened.
-
figured it out it turns on the red and blue blinkies.
-
The thing is that the Rawdigger suggest "the same" on RAW files.
-
- By bobn2 on April 24, 2023, 6:47 p.m..
-
The Canon EOS 10D is non-linear (compressed response) at ISO 100 near raw saturation. I did a split image a long time ago with a lighting gradient on a wall where I exposed ISO 100 manually 2x as much as ISO 200, demosaiced each, and the tones are exactly the same for the darkest parts of the gradient and the image, but as the tones got brighter closer to the bulb, they split with a visible seam, with the ISO 100 side darker. That's with just simple demosaic, no WB or any other conversion, with linear raw color space mapped linearly to sRGB. The OOC JPEGs and default conversion parameters kept this non-linear range hidden. The camera had a bad reputation for discolored highlights when pulling brightness down after ETTR at the time, as converters did not correct the non-linearity.
-
It is certainly less problematic with a monochrome sensor.
-
I was thinking of more contemporary cameras. Canon 10D was launched in 2003.
-
Well, how many cameras have been tested for linearity? Just because my example is from a camera designed over 20 years ago, there is no reason to assume that it was the last one.
-
When I test cameras, I always do a linearity check by examining the photon transfer curve, although I'm usually looking for gross departures from linear performance.
-
A few relatively recent Panasonic cameras are intentionally that way at the lower ISO settings.
-
I have a question.
I put my camera on a tripod to photograph an architectural (static) subject. I set the lens to F11 to get everything in the scene in focus using DoF. I set the camera to base ISO 64ISO. I do not care about shutter speed. It might be 1/20 second or it might be 30 seconds depending on the intensity of light falling on my subject.
For my two extreme shutter speeds, is using base ISO 64 the best solution to reduce the noise level in my picture to a minimum? Remembering that shutter speed is of no interest to me.
-
Valid point. I assume that non-linearity at raw saturation would cause problems with ETTR and would have been reported.
-
I your case, it is close to the best solution. The best solution is to have as large an exposure as you can, subject to pictorial requirements. Since you don't have an exposure time constraint your constraint is the largest exposure that your camera will accept, which is likely a bit more than nominal exposure for ISO 64. That is, you could expose to the right. DxO mark suggests a 'measured ISO' (misleading term) of 46, which suggests you could get a bigger exposure, especially if your scene had no specular highlights. Iliah might have some comments about whether or not you're going into non-linearity at that exposure.
-
Only if I know which camera is it ;)
-
Nikon D850 - Nikon Z7
-
Thanks for the interesting and useful reply.
-
With those cameras, I wouldn't worry about loss of linearity before clipping even at base ISO. I've tested both. But Iliah is the expert on this subject.
-
At ISO 64 setting both Nikon D850 and Z7 with recent stock firmware are linear (I never check any lossy pseudo-raw modes as I don't use them).
-
What is the best way to determine whether a camera's sensor is being used in the linear range only and thus ETTR will not degrade IQ?
-
Here's what I do. At base ISO, plot rms noise versus average signal level for the upper few stops of the dynamic range.
How to do that?
You can photograph a gradient, take two shots, add the two raws together giving you image A and subtracting one from the other giving you image B. Pick one of the raw planes. Measure the average level of image A for a bunch of crops. Divide that by two to get average signal level. Using the same crops, measure the standard deviation of image B. Divide by 1.414 to get rms noise level. Plot the two against each other, one point for each crop, on log log paper.
-
A simple way: clipping point in electrons multiplied by ISO setting is close to constant for the range of non-extended ISO settings .
