I've seen it mentioned before that underexposing a raw shot by shooting at base ISO, and then boosting the rendered brightness afterwards using the "Exposure" boost in a raw developer, instead of capturing an appropriate exposure at high ISO, results in different colours.

(The advantage of shooting this way, with a camera with a low-noise ADC, has been covered here: https://www.pentaxforums.com/forums/pentax-k-5/135603-isoless-sensor.html)

So I decided to investigate the validity of this assertion.



Here is the test setup:

A camera with a low-noise ADC system: Pentax K-5.
Set to raw.
Set to M mode.
Set AE to centre-weighted.
Set ISO to ISO1600
Set shutter speed & f-stop to centre metering.
Shot and checked histogram. Ended up using +1.7 EV comp to get a bit below clipping (checked raw histogram in RawDigger).
Took a shot.
Dropped ISO to ISO100 and took another shot.
Opened both raw files in RawTherapee (important). I had set the default processing to "Neutral".
Bought up ISO100 with +4 "Exposure Compensation".
Compared the 2 images. Saw that the ISO100 was a bit different in level and in colour to the ISO1600. Why?
Noticed that in the camera I was using Auto WB, and RawTherapee was set to "Camera Setting" for the WB.
Comparing the two WBs:

ISO1600
Temperature: 4979
Tint: 0.941

ISO100
Temperature: 3698
Tint: 0.836

Next, I change the WB in RT, for the ISO100 file, to the same as what's in the ISO1600. Now the colours were very similar, but the ISO100 was brighter. So I dropped the Exposure Compensation of the ISO100 file from +4 to +3.60, so that the luminance peak on the histograms matched.

LESSON: Don't use AWB for this test, or you'll be lead astray, because the the camera's WB will change with ISO setting, and this will affect both brightnesss and colour.

So I set the camera's WB to Cloudy (I was shooting indoors using just natural illumination and it was a cloudy day) and tried again. (It doesn't particular matter what WB you use for this test, as long as it's fixed). Now, the WB as shown in RT was the same and the images looked close. There was a minor difference between the top and the bottom and I needed 4.05-4.10 to match the brightness.

LESSON: Try to use artificial illumination so it's more consistent then my variable cloud-cover mediated illumination. (I did this test a number of times to try getting ISO100 & ISO1600 pairs that were close together in time.)

So the matter of the different appearance of images at different ISO has been resolved to my satisfaction. (Note: I did no other image adjustment. Normally, if this was not a test, I would have adjusted the contrast and sharpness.) But what about the noise difference caused by using the base ISO instead of ISO1600?

The exposure was the same: same scene luminance (as well as I could with natural illumination), same shutter speed & same f-stop. But the ISO100 image was sent to the ADC at 1/16th of the level of the ISO1600 shot, so it was "under-brightened" in the stored image file.

I shot a part of the side of a cardboard box from a wine cask. This part looked white & black, with a grey separation region. What I was looking for was any change in the level and character of the noise in the dark region.

Ev Comp in the camera for both images was +0.7EV. The grey separation line between white & black is 5.5mm high. (I used a 100mm macro lens.)



This is a comparison of 100% crops. The lighter spots in the dark region are imperfections in the cardboard surface. You can see there is a slight increase in noise in the boosted ISO100 shot, but there's not much in it, and it would be hard to notice a difference without a side-by-side comparison.

Original files available, upon request, if you want to check the Exif, or do your own raw conversion.

Dan.

[fs999]

Thank you for the comparison !

The ISO 100 shot of the cardboard is not exactly the same place on the ISO 1600.
It would be interesting to see exactly the same dotted pattern, if there is a difference or not.

[dosdan]

Thank you for the comparison !

The ISO 100 shot of the cardboard is not exactly the same place on the ISO 1600.
It would be interesting to see exactly the same dotted pattern, if there is a difference or not.

That's true, but they are close enough. The two dot imperfections on the left side image are just to the right of centre on right image.

Any sharpness difference is probably an indication of Pentax's AF performance repeatability, rather then anything else.

An insight: the PGA in the K-5 has a 24 dB (4 stops, 16x) adjustment range of analogue gain. Since I've enabled Expanded Sensitivity, the base ISO is ISO80, not ISO100. This means that the highest ISO before non-defeatable NR kicks in is now ISO1250, not ISO1600. So the ISO1600 version will have raw NR applied. I didn't think about this when I did the test, and just realised it recently. Therefore, I think I'll repeat this test comparison using ISO80 & ISO1250 as the two comparison sensitivities.

But I will not be surprised if the sharpness difference is still caused by Pentax Phase-AF variation. Perhaps it will be better if I do this test in LV. I'll think further on this.

Dan.