If you set "1/60" shutter speed, what exposure time do you expect?

I think this is maybe more about naming things than logic or physics.

I have no idea. I also have no idea how (decimal) 1055 would relate to "1/66sec".

I don't see how that follows.

It doesn't make much sense to add 'exposure time' to 'readout duration', unless referring to an electronic second curtain, where 'readout duration' is, in effect, the shutter travel time AKA rolling shutter.

You also wrote:

This is what I thought you meant by "integration time".

Which is (shutter travel time) + (exposure time)

Which we can deduce from the X-sync speed of your camera and your chosen shutter speed ("1/6000") is something like 5ms, or 1/200 s.

Definitely not anything nearly as large as 1/66 s.

Which is completely normal for electronic or mechanical focal plane shutters (linear or rotary).

John Sheehy's idea can possibly give a time comparable with the "1/66sec" (15ms) that you refer to.

John Sheehy's sequence is:

1. SD9 sensor is reset.
2. SD9 sensor released from reset. (Start clock).
3. Start first shutter curtain
4. 1/6000 s later, start second shutter curtain.
5. (SD9 shutter travel time ~= 4.8ms) later, second curtain has covered sensor.
6. Start reading SD9 sensor
7. Finish Reading SD9 sensor. (Stop clock).

Anyway, whatever it is that Sigma decided to call "Integration time", it's a weird thing specific to Sigma. "Integration time" would make more sense as a synonym for exposure time.

I think all this is an exercise in engineering archeology.
Some Sigma engineer wanted a a quantity that interested them in the EXIF, and slightly randomly called it "integration time".
Trying to find some established but uncommon meaning for this "integration time" phrase may not be fruitful.

It may have nothing to do with exposure at all. Some Pentax cameras have an "AF integration time" tag.

Putting: exif "integration time"
into a search engine, apart from a lot of astrophotography references where it means total exposure time, I found this:
exiftool.org/forum/index.php?topic=10134.0
In post #4 there's EXIF for a picture at "1/1000", where "Integration Time" is "1/1020".

That would be exactly the sum of exposure time and rolling shutter time. If that's the case, then we should be able to get the rolling shutter time from the EXIF by subtraction of actual exposure time from Integration time. I just looked at some Canon images and they don't have that field.

If my original guess is correct, then a camera that offers full electronic shutter would have the same value for integration time as actual exposure time.

For EFCS, it is impossible to have a consistent "pixel active time" down the frame, because the readout roll is much slower than the activation roll.

I take it that nobody here knows how EXIF's IntegrationTime, tag no, 0x0423 is measured or calculated?

Found this:

dictionary.university/Integration%20Time

Neither the tag number nor the tag name 'IntegrationTime' are unique to Sigma, John
They are also used by:

MRC FEI12: 419
Kodak IFD: 0x0423

Both tagged 'IntegrationTime'.

Now looking for a definitive Kodak description because of the same tag number.

Found this list repeating the above tag.

Found this note:

Electronic shutter is not used. Integration time equals frame time

Found this diagram:

Ignore the EXIF - FastStone Viewer strikes again 😟

Both in here, www.kodak.com/global/plugins/acrobat/en/digital/ccd/products/interline/KAI-11000CMLongSpec.pdf

All I need to know is what is meant by 'frame time' because my camera does not shoot video!

I suppose that the SD9 only offers full-mechanical shutter, and no EFCS? If so, all you can do to reverse-engineer the EXIF field is just graph exposure time against "integration time" and see what signal is there. If the camera had other shutter modes, then it would be easier to figure out. I have cameras here with e-shutter, EFCS, and full-mechanical, but my cameras do not have the field, so I can't do it.

Correct. Manual says:

Vertical-travel metal focal plane shutter, electronically controlled through entire speed range

Not sure that I understand that, John, but the IntegrationTime varied little for 1/6000, 1/3000 and 1/2000 sec.

It's just a way to be thorough, to see if there are surprises, although I wouldn't expect any. It would take another shutter mode to compare to, to get a clear answer. In any event, 1/66s is too long for it to be for 1/6000 shutter speed and the rolling shutter. 15.15ms minus 0.17ms is 14.98ms, much longer than the likely rolling shutter speed of the SD9. What is the max flash synch?

No

Yes

1/180 sec

As to the rolling shutter, it is probably inherited from the previous film camera SA-9 for what that's worth and about which I know nothing.

Further research today is leading me to believe that IntegrationTime is the period during which the sensor or parts thereof are being iluminated.

The difference between the smallest IntegationTime 1/66 sec and the sync-speed 1/180 sec remains an anomaly yet to be explained.

Just now looked at that. They're discussing the DP1 Merrill compact which has a leaf shutter so not the kind of shutter we're discussing.

1/180s = 5.45ms
1/6000s = 0.17ms
1.180s + 1/6000s = 5.45ms + 0.17ms = 5.62ms
1/66 = 15.15ms

You seem sold on your original idea, but it does not seem possible.

This indicates that whatever definitions you may find of "Integration Time" in your searches, it is not what the SD9's EXIF is reporting in that field.

The "active pixel time" explanation is the only plausible one so far.

Thank you for quantifying the anomaly.

Sorry, I seem to have missed that explanation ... Where was it?

Thank for your continuing interest!

If the time axis is linear, then that's enough for me to draw the conclusions that I drew.

In my first reply to you. The pixels are activated to take on charge a while before the first mechanical curtain opens, and gets read out and stops taking charge a bit after the second mechanical curtain has closed on each row of pixels.

We can't know from your number "1/66s" whether that is per-pixel or from the activation of the first row until the readout of the last row, but it seems to have to include time where the mechanical curtains are covering the pixels.

Ah, by talking about pixels by row, I got confused because, on the SD9, there is no alignment of the mechanical curtains with pixels as far as I know.

I doubt that it is 1/66 sec "per pixel" and I don't know whether the curtains' travel is somehow synchronized to the sensor's row-by-row line scan. If the shutter is inherited from the SA-9, then such synchronization is most unlikely.

[edit]Maybe when the sensor is covered after exposure, there is a long read time just as you imply.

I'm looking at a paper for that sensor, as we speak. [/edit]

We're talking still shot (SS) mode and not rolling shutter (RS) mode as far as I can tell.

The paper is about the sensor - RS mode can not be selected or activated automatically on my camera.

Which is a great pity because the only mentions of the subject at hand are in the RS mode.