[AstroDave]

Inspired by this recent thread (Pixel depth on electronic shutter - PentaxForums.com) wondering about whether or not electronic shutter (ES) exposures used the full dynamic range of a camera (i.e. the maximum number of bits), I’ve made some relevant measurements with my K-1 and K-3 III cameras. I had already done a quick check on my K-3 III and showed that all 14 bits are used for electronic shutter on that camera. But here I report in a bit more depth (no pun intended ).

As a by product, I have also checked whether or not ES exposures are the “same” as mechanical shutter exposures, as well as whether or not .DNG and .PEF files are the “same.” By same, I mean whether or not the exposure results are equivalent, in terms of pixel values for the same exposure parameters (lighting, shutter speed, f-stop, and aperture setting).

I can now say for certain that for my K-1, electronic shutter exposures also use all 14 bits. I upgraded the firmware to the latest version (2.20), and now have an electronic shutter option for the K-1 when it is used in live view. Using Raw Digger software (highly recommended, and I can’t really find any other software to get the RAW pixel values, unless one is ambitious enough to take on the DCRAW software!), I can get the actual value of individual R,G,B pixels. Typically, I extract the average pixel value within a rectangle placed over the area(s) of interest in an image - typically gray scale patches similar to those in a Gretag-MacBeth color checker card, or even just the sample color(s) on a paint sample card from your favorite paint store.

I did not deliberately over expose any of the K-1 images, so the highest pixel count I measured was not at saturation, but was at about 15300 - close enough to the absolute possible maximum pixel value of 16383/4 for 14 bits.

So, at least for the Pentax K-1 and K-3 III, electronic shutter uses all the available bits (14) for images, same as mechanical shutter exposures.

I also took and compared exposures with the same exposure settings in both electronic and mechanical shutter modes, as well as a few duplicate exposures with the K-1 set for either .PEF RAW files or .DNG RAW files. Adjusting the camera for these settings (pushing various buttons) as well as shutter-button pressing caused minor variations in the pointing of the camera, leading to minor variations in the lighting of the sampled target areas (I know for sure that my lighting was not completely uniform). I also can not be sure that the voltage to my light sources was invariant at some very low level, which might well cause changes in the light output of the LED bulbs. Hence, I do not expect that all the shots actually have exactly the same exposure, but they should be very similar.

And, indeed they are - variations between exposures are at the 1-2 % level or less for the two cameras. I saw similar variations even when all I did was take two exposures in a row with no changes in any camera setting.

Hence, I am pretty confident that no matter how you take a particular picture with your choice of electronic shutter or mechanical shutter, and the camera set for either native Pentax RAW format (.PEF files) or the alternative .DNG option, you will get the “same” picture.

One last test gave the expected result as well, but it’s always good to verify assumptions: I used the Adobe negative converter program to turn a couple of my PEF files into DNG files. In this case, the RAW Digger program reports exactly the same pixel values for the two versions of an image file.

Bottom line: Use your Pentax DSLR cameras with whatever shutter type and RAW file type you prefer - the results don’t depend on that (although, as I hope we are all aware, ES is not so good when there is motion in the field of view!!).

And, finally, I checked the linearity of the K-3 III when using electronic shutter. Here’s the result



The linearity is very good (a power of 1.000 for the x-value would be “perfect” and the exponents are within a per cent of that) and the quality of the fits is also just about perfect (R-square values of 1 to 4 decimal places). See the reference to my earlier efforts in the next paragraph for a bit more explanation of what I did here.

The linearity is just as good as I found for the camera when using mechanical shutter and similar to results reported previously for my K-3 and K-1 ( How Linear Are Some Pentax Camera Sensors? - PentaxForums.com - those measurements were done with mechanical shutter). The camera is linear over a range of pixel values for a factor of around 3000 in exposure value, or about 11 and a half bits. At the lowest data values (below about 3), noise starts to dominate, at least for short exposures.

[Pentax Syntax]

Thank you. Very convincing post and I was contemplating trying some tests myself before your hard work.

[Medex]

I did not deliberately over expose any of the K-1 images, so the highest pixel count I measured was not at saturation, but was at about 15300 - close enough to the absolute possible maximum pixel value of 16383/4 for 14 bits.

Thank you for nice post.
I can't reach 16383 value with my K-1 color sensor, 16316 only.
Same in case with monochrome picture shot with debayered sensor (16316), but preprocessed RAW (in Monochrome2DNG converter, demosaicing skipped) delivers higher - 16386 value. BTW, Mono2DNG converter somehow adds "new" pixels - 7392x4950 instead of 7360x4912 (shown in Lightroom, but RAW Digger shows 7392x4950 pix even for color pictures)
What does it means?

[ProfessorBuzz]

Great work, AstroDave. I've suspected what you've confirmed - thanks for making the effort. The Full Well Capacity of the sensors aligns pretty well with the 14-bits - I just wish I understood the saturation a bit better, and the read noise.

[AstroDave]

I can't reach 16383 value with my K-1 color sensor, 16316 only.

Yep - that's the maximum I see (reported in my original linearity measurements and for the K-3 as well: How Linear Are Some Pentax Camera Sensors? - PentaxForums.com) I think they leave a bit of headroom for the A/D conversion. Also allows a bit of room at the top for post-processing progams - you don't want to risk flipping all the bits and finding a maximum value of -1 !!

I get 16318 for the K-3 III.

but preprocessed RAW (in Monochrome2DNG converter, demosaicing skipped) delivers higher - 16386 value. BTW, Mono2DNG converter somehow adds "new" pixels - 7392x4950 instead of 7360x4912 (shown in Lightroom, but RAW Digger shows 7392x4950 pix even for color pictures)

Interesting! I don't know anything about that software. Generally, there are various guard bands and zero level reference pixels around the area deemed "active" for a camera sensor. Maybe some of these get "promoted."

16386 is cheating!? 2^14 = 16384 Do they go to 16 bit output (which I think is OK/legitimate in TIFF)?

[Medex]

Interesting! I don't know anything about that software. Generally, there are various guard bands and zero level reference pixels around the area deemed "active" for a camera sensor. Maybe some of these get "promoted."

16386 is cheating!? 2^14 = 16384 Do they go to 16 bit output (which I think is OK/legitimate in TIFF)?

Zero level reference pixels... Maybe... Image in RawDigger has some black stripes on the edges - 5 to 12 pixel width. Probably they are counted somehow.
16386 value is from my monochrome picture. Maybe RawDigger dors some calculation error.

[ProfessorBuzz]

Zero level reference pixels... Maybe... Image in RawDigger has some black stripes on the edges - 5 to 12 pixel width. Probably they are counted somehow.

Many of the Sony (and other makers) CMOS APS sensors have reference pixels in an "overscan" area. These are usually outside the active pixel array, and are used for things such as black-level measurement. Some software and camera firmware can pass along the overscan pixels in addition to the active pixel area.

[Medex]

Many of the Sony (and other makers) CMOS APS sensors have reference pixels in an "overscan" area. These are usually outside the active pixel array, and are used for things such as black-level measurement. Some software and camera firmware can pass along the overscan pixels in addition to the active pixel area.

Seems you are right.
Selected red area 3x82 pix shows values from 0 to 3, it means completely or almost black