IQ reduction with higher shutter speeds?

[smeggy]

I have done basic diffraction testing. A modified white LED (lens portion removed), with a thin black cover with a 0.6mm hole drilled into it, 4m away from the 7D with a 17-40mm L lens on it, helps ensure the imager sees a point source of light so making any effect of diffraction obvious. I varied the current though the LED to compensate the change of exposure time. Obviously, the ISO and aperture remained unchanged.

Results: nada!

I had to greatly overexpose the image (very high ISO) in order to see the effect of diffraction; only then could I see the tell-tale, sideways-Saturn-looking vertical spikes around the point (a few pixels) of light. So the diffraction effect is present, but it is certainly isn’t dominant. So I think my theory is “Busted”.
So why my previous second result? I think I owe folks an apology. I suspect my second test was subject to a slight amount of focus error (even with LiveView, go figure!), hence it affected my interpretation of the result. As most folks here should know, stopping down can sharpen the out-of-focus portions of an image <kicks self>
It has been a really crap day here in London (really thick fog all day), so I couldn’t confirm that result (or do any other IQ testing).

I suspect the IS doesn’t make any significant difference if the exposure time is so short. I can’t see how the IS mechanics could accelerate within such a short amount of time; this is regardless of sample rate, but if the sample rate is really is up to 1000 Hz, then the exposure time is much shorter than the sample interval. So I reckon if IS was the problem, any negative contribution should be proportionally worse with longer exposures as the processing still wouldn't be updated in that time.

mreco99,
How did you take images with different exposures, yet keeping all else unchanged? Were the exposures compensated in other ways, or were the resulting images of varying brightness? Due to the non-linearity of the photon to bitcount conversion (the ‘gamma’ if you wish), it is not impossible that the contrast of the images was different; perhaps this could be interpreted as a loss of IQ/sharpness.

 

[Edwin Herdman]

You'd have to figure out, in other words, how far apart the aperture blades of a given camera actually get.

I just went one better: I measured them

Nice!

The differences in spacing between the 1000D and the 5D are interesting - and in the direction I'd have expected. Although I'd like to know what the 1/4000 sec distance is for the 5D (3.4mm? Am I discounting geometry here?)

I suspect the IS doesn’t make any significant difference if the exposure time is so short. I can’t see how the IS mechanics could accelerate within such a short amount of time; this is regardless of sample rate, but if the sample rate is really is up to 1000 Hz, then the exposure time is much shorter than the sample interval. So I reckon if IS was the problem, any negative contribution should be proportionally worse with longer exposures as the processing still wouldn't be updated in that time.

Agreed. I'm having trouble trying to visualize a situation where the varied frequencies of the shutter and the IS update would destructively interfere with each other. We also know that the longer the shutter speed gets, the better the effect of IS versus an unstabilized image (for handheld, non-tripod use, of course).

 

[epsiloneri]

Is it possible to have any loss in image quality soley by using the highest shutter speed (ie 1:6400) assuming ISO and everything else was equal?

"Impossible" is a strong word, but I tested exposures 1/2000 & 1/8000 seconds with EOS 7D and EF 70-200/2.8 IS II @ 200/2.8 and EF 400/2.8L IS @ 400/2.8, both handheld, shooting a test chart at 3 meter and 6 meter distance, and noticed no difference in IQ (as expected really). The exposure level was made equal by moving the studio light (distance difference a factor of 2 gives illumination difference of 4).

I also tested the claim that IS could deteriorate IQ at high shutter speeds by switching the IS on/off, and noticed no difference in IQ under these conditions. I made sure there was no AF error variability by reacquiring AF multiple times in each setting.

I would like to try with a smaller aperture as well, but as even my bright studio lights are not sufficient for these high shutter speeds, I will have to wait 5 months for some decent daylight to test it.

 

[smeggy]

The differences in spacing between the 1000D and the 5D are interesting - and in the direction I'd have expected. Although I'd like to know what the 1/4000 sec distance is for the 5D (3.4mm? Am I discounting geometry here?)

What you say makes sense. Given that speed of the shutter blades never change for a given part of the imager (not affected by exposure time), the gap between the blades should be in proportion to the exposure time (when both are simultaneously in motion).

However, I realised another important factor: the acceleration of the shutter blades. Being as the shutter is a simple spring-loaded mechanism, I strongly suspect the blades don’t immediately get to speed and then hold that speed; the speed of the blades should constantly increase as they fall. So at the start of the exposure (the gap at the top of the imager) I think the blades should a closer together than at the end (gap at the bottom of the imager). Again I don’t have a quantitive feel, but I reckon this factor won’t be of any significance. I will try to measure this later.

 

[mreco99]

So to sum up (so far) we havent actually got any hard evidance conclusions?

 

[epsiloneri]

When the shutter speed of the camera is (say) 1/100 of a second when IS is active, the lens can have moved the image back and forward from a single point 10 times during the exposure and it will appear solid. On the other hand, if the shutter is set to 1/2000 of a second, you'll capture the lens moving the image between two points.

Thanks for the explanation, Dilbert... but it still doesn't make sense to me. Let's say that for a slow exposure IS moves between positions A and B sufficiently many times that, effectively, the image is superposition of IS at A and B. On the other hand, with a very fast exposure, you may catch the IS while moving from A to B. This is what you are saying, right? But then why would the second case be less sharp than the first case? Why is moving between A and B less sharp than showing A and B simultaneously?

To think of it in a different way... our eyes are able to respond at around 10 to 12hz. If you look at a CRT that has a frequency of 50 or 60hz, you see a solid image. If you take a photo of that CRT using a camera set to (say) 1/100 of a second, you get a less than solid picture on the CRT.

The reason for this is that a CRT draws only part of the screen at the same time, while other parts of the screen quickly fade to black until updated in the next cycle. The equivalent would be to shoot a scene that is alternating to black at high frequency. Then yes, it could make sense to average over a couple of cycles, instead of catching it while in the black phase. But that is not what happens with IS.

You should instead compare to an LCD screen, which shows the same picture until updated (also at 60 Hz). If you expose for short enough time you may catch the LCD during an update and get part of the screen showing a frame and parts of it showing the next frame. With a longer exposure you may get both or more frames. There is no big difference for the IQ in this case.

And why isn't this trivial to test on an IS lens?

Because it's not trivial to come up with a lighting source that illuminates your subject such that you can shoot it at 1/2000 or 1/1000 at with the camera/lens set for good IQ (ISO 100, f/8, etc.)

I understand that, because it requires bright conditions. But if it's hard to come up with even a test case where IS is a problem for short exposures, then it probably isn't a problem. And if you, ever, get sufficient light that you can use ISO 100 with f/8 and a very short exposure, then it should be trivial to switch between IS on and off, with everything else being the same, to confirm the potential IQ degradation introduced by IS.