A closer look at Shake Reduction on the K10D

You may remember that I threatened to do a bit more of a careful investigation into the allegation that shake reduction on the K10D causes shake when the camera is still.

Well I have, and the results are in.

For those that just want the bottom line:

- If you want to eliminate shake on a tripod, there is one way, and one way only, and that is to use a remote release, (edit: or the 2s delay, as has been kindly pointed out).

- If you insist on releasing the shutter with your finger for a long (several seconds) exposure, you should turn SR off – not, however, for the reason you might think.

- If you release the shutter with your finger, SR does not make the shake either worse or better.


For those that want a bit more detail:

Basic set-up:

- K10D on the tripod
- Manual focus lens (50mm A F1.7 or 100mm A macro F4)
- Static subject, centre focus point, centre of green hexagon confirmation range.
- ISO 100

The following were varied:

- SR on or off
- Exposure times
- Lens (50mm or 100mm)
- SR focal length setting

When the 50mm lens was on the camera, the full frame looked like this:




And with the 100mm, it looked like this:




When there was shake, it was very obvious, as was the degree of shake. No shake looked like the following 100% crop:




When there was shake, it would be some variant of this:





I’m not going to describe everything here, but here’s the gist of the main experiments.

Experiment 1: Vary the exposure, with and without SR.

This involved taking pictures at a series of shutter speeds from 1/20s to 2s, at 50mm and 100mm, with SR on and off. The SR focal length was set correctly, according to the lens.

Results:

Shake was visible in most of the shots when the exposure was longer than 1/5s, with SR both on and off. It varied from one shot to the next, and it was not possible to conclude whether it was, on average, better or worse in either case.

Experiment 2: Intentionally set a greater SR focal length to amplify any effects.

The hypothesis that SR somehow ‘creates’ shake when none exists, as has been alleged, was further probed by intentionally setting the SR focal length to a length (300-500mm) much greater than the actual focal length of the lens (50-100mm). The idea of this was that SR has much more work to do at longer focal lengths, so any effect would be amplified.

Results:

With the exception of one shot, the results with SR on and off were not noticeably different – shake was present to a variable extent in both situations. The one shot that was different, though, the SR went into overdrive, and shook the sensor far more than was necessary, resulting in a completely blurred image. I’ve reported this effect before, with hand-held shots. However, I don’t think this effect is helpful in answering the main question, and since mostly the results were not obviously worse, I abandoned this line of investigation.



It was clear by this time, that the amount of shake varies from shot to shot. So the final experiment was to keep the exposure time fixed, and take series of shots with SR on and off, to see if the average amount of shake was obviously different in each case.

Experiment 3: 10 shots at 1s, SR on and then 10 shots at 1s, SR off. Repeat for 50mm and 100mm.


Results:

Again, shake was visible in most of the shots, with SR both on and off. It varied from one shot to the next, and it was not possible to conclude whether it was, on average, better or worse in either case.

Conclusions

A more rigorous investigation would be to measure the amount of shake in each shot, calculate means and standard deviations, and test whether the results with SR on and off were statistically different. However, life is too short for this, and in any case, in both cases, finger-induced shake is very noticeable, and with SR on and off, is plenty bad enough to ruin any shot you wanted really sharp.

I did, however, repeat experiments 1 and 3, and the results were much the same.

So based on these series of tests, judging the shots visually, my conclusion is this:

It is not possible to conclude that SR causes any more shake than is present without it. When you release the shutter with your finger with no delay, this causes shake. It seems more likely to me that the amount of shake is below the SR’s ability to detect it, and all that happens is that the shake is not eliminated.

So why do I say that you should switch off SR for several-second, finger-triggered exposures?

Well, twice I found that with SR on (and confirmed on with the hand icon), and a 2.5s exposure set, a) the mirror flipped up, but the shutter did not open, and then b) the camera would not switch off. I had to remove the battery to get it to switch off. After a heart-stopping few moments, replacing the battery and switching back on restored normal function. This did not happen with SR off under the same conditions. You’ll forgive me if I don’t repeat this test too often, though.

Bottom line: based on these tests, I'm afraid I don't believe Don's or anyone else's assertion that SR causes shake.

On the tripod, I can't see any difference in the amount of shake between SR on and off. Occasionally I got a shake-free shot with SR off, and a shaky one with SR on (and vice versa), but over lots of shots, it was clear that this is just the normal variability of finger-induced shake.

I shall be leaving SR switched on, since any time I want shake-free tripod shots, I'll be using the remote, and this switches SR off anyway.

What's wrong with the 2 second timer?

Thanks anyway for your post. When I have time, I will re-read it several times to try and digest the details!

If you want to eliminate shake on a tripod, there are two ways, and two ways only, and that is to use a remote release, or the 2s delay.

I meant of course, remote as in not touching the camera.

I've done a quick check with the same setup, and indeed, I can see no shake using a finger but with the 2s delay.

Excellent!

Incidentally, I couldn't see any difference between using the remote release and no delay, or the 2s delay, or the 3s delay. This doesn't mean there isn't any though - I'd expect mirror-induced shake to die down more quickly than finger-induced shake, so it would probably only be relevant with exposures shorter than a second, which is what I was looking at.

my observation was, I was getting sharper images handheld, with the 1.4 and wide angles like the 16-45, with long exposures. I stand by that observation.
my hypothesis was that the camera was inducing shake with the sr on and the camera stable. I accept your results. (well done!).

now to complete your study: (I for one don't have time to thouroghly test this one).

to prove or disprove my original observation, you may want to repeat your tests with the lenses I used (I already pointed out the effect didn't happen wth telephoto/longer focal lengths (thank you for confirming that with the 100mm)) and no tripod.

I'd suggest firing off a 1000 or more shots handheld with sr on and another with sr off, using the wides to standard lenses, after firing off 1000 images with sr off at increasingly long shutter speeds to determine the longest shutter speeds you can handhold.
then analyze your data to determine where the sr helped, by how much, and if the sr hindered and by how much.

Happy Shooting!

Well, the 1.4 you refer to is the 50mm 1.4 I assume, and I did most of the testing with a 50mm lens.

I don't have a 16-45, but I do have a MF 28, the 10-20 in the picture, and of course the 18-55. I can repeat the test quickly now, so what exposures would you like?

Well yes, but you haven't given any evidence that you've eliminated any of the things that account for your observation that some of your SR-off shots were sharper than some of your SR-on ones. Hand-held shooting is notoriously unreliable for this, and if your assertion is that it is the stillness of the camera that causes the shake, basing your conclusion on hand-held shots is invalid, since you haven't shown that you can hold the camera steadily enough.

Well, I've done some trials, not particularly rigorous ones, and SR improves my hand-held shots by about 2-3 stops. But this is not relevant to the question of whether SR induces shake when the camera is still.

Sometimes I get shake-free shots without it, and sometimes I get shakier shots with it. But this proves nothing.

It's just as possible to account for this by the natural shake variation from one shot to another, and requires less in the way of extraordinary assertions to explain the results.

Unless you do a repeatable experiment that properly eliminates causes of the effect you're describing other than the one you believe in, subjective experience is virtually worthless.

Believing something despite a) evidence to the contrary, and b) not having reliable evidence for it, is not rational.

tell that to the eggheads like Einstien and Edison who believed in the existance of radiation, before they devised ways to see it.

questioning your methods does not make one irrational.

I'm suggesting you need to do a statistical analysis.

first finding out where your limits on handheld abilities lay, then working out where the sr is affecting your averages. Ideally you'd want to repeat the test with more than one shooter (100's would be best) to call it "evidence"

I made an observation, (so I did see something, I've spent years practicing handholding in lowlight to make myself much better than average when I walk into a dim church for client, and I did see a drop in my averages when I started using the k10, but not the ds2, with the same lenses) and adapted my shooting for it, and resolved the issue for me. I'm just trying to help you get it sorted for yourself, so if your convinced...I'm good.

Don, you're not listening.

Questioning methods is fine. Believing something despite contrary evidence, and despite absence of evidence for it, is what I'm saying is irrational.

Don, you're still not listening.

Whether SR improves hand-held shots, and by how much, is not relevant to whether SR makes tripod shots worse.


Well it shouldn't have resolved the issue if that's all the evidence you had.

You haven't eliminated other possible causes - like for a start, the fact that the K10D isn't either the same size or weight, so you don't know that your technique with it is the same.

Well, you're very kind.

But the point is, you're asserting that SR causes shake when the camera is steady.

And in support of this, you cite your hand-held technique.

This implies that your HH technique is equivalent in steadiness to a tripod, which is not credible.

At the very least, you should test with a tripod, you should use a small aperture, to eliminate the possibility of confusing blur due to lack of DOF with shake, and you should do enough shots in each case to get a fair impression of whether any effect still present is within normal variability.

Claiming that noticing something with one camera, and something else with another camera, and attributing the whole effect to one feature of one camera is bad enough.

But when, at the same time, you simply ignore the things that come with the territory of hand-held shooting, then it's just bad science to continue to claim what you're claiming.

But if that's what you want to do, I'm good. I'll happily try at shorter focal lengths than 50mm, though. Watch this space.

Ok, done. Same as before, 1s exposures with the 18-55, at 26mm and 18mm, SR on and off for each.

Also some control shots with the 3s delay to establish a baseline 'no shake' situation.

I won't bore you with the details, but the results were the same.

At 26mm, and even more so at 18mm, the amount of shake was less than at 50mm and more (for reasons that should hopefully be obvious), but pretty much the same with and without SR in each case.

Unless someone can see a glaring hole in the method, I can see no other explanation than simply that SR does not affect the amount of shake when the camera is on the tripod, and a finger is used for immediate release, if you'll forgive the double entendre there.

By the way... I never said I was convinced. All I'm claiming is that I've seen no evidence, despite trying my best to design an experiment that tests for what I'm looking for.

But there may well be something I hadn't thought of, and a different experiment would reveal it.

I also understand your contention, that if, hypothetically, you took 1000 hand held, long exposure shots with, and then without SR, and got a statistically significant reduction in shake with (edit: I mean without, of course) SR, and there were no other variables, then it would suggest that SR was causing shake.

However, the devil is in the detail, and it would be exceedingly difficult to ensure that there were no other variables, taking pictures hand-held.

Hence the need for an experiment that is designed to eliminate those other variables.

Chris,

A very interesting experiment.

When you refer to the use of a remote, do you mean both the IR remote and the wired remote or just the IR remote :

I don't have a wired one, so just the IR remote.

With and without the 3s delay, and also without the remote, but with the 2s delay.

nice experiment Chris, without wearing out a K10D and boring yourself to death when you should be out taking photographs I can't see a way to improve it.

The only thing I would add to this discussion is that for exceptionally long focal lengths it is important to use both a remote release (either the wired or IR one is fine) AND the 2/3s delay. Finger shake doesn't die down fast enough at 1000mm+.

Don said "my hypothesis was that the camera was inducing shake with the sr on and the camera stable. I accept your results. (well done!). "

So I did listen.

Don also said "ok here's one for all you low light shooters:
if using sr with a tripod causes blur in the images...
what happens if you have good technique, and steady hands....no tripod...will sr induce blur, or improve your shots?
I've been experimenting. With shots in low light, 1/2 sec exp at f2 iso 100, I'm getting sharper images without sr....
any steady handed shooters out there care to try also, and let me know how it works out for you?
I'd like to rule out any problems with my cameras before drawing conclusions."

So I do see that your experiment resolved one issue, but your methods did not replicate the original observations circumstances.

so you did not prove me wrong on my observation, except the hypothosis that a stable camera on a tripod (my hands are not as stable as a tripod, or I wouldn't spend money on three of them) would get shake induced.

that I could see a difference, grab the ds2 and get better results, switch off sr on the k10, and get better results, and grab Gerri's k10 and repeat the same results, proves either that my case study is too small, or both Gerri's k10 and mine have the same glitch, or I have some kind of hand shake that only happens when when I pick up a k10 with sr on...don't forget I also shoot video, shaky hands would manifest there, real good.

if your tests disprove the common belief that sr will induce shake on a tripod mounted camera, I accept that.

I asked if anybody had similar experiance, and thank you for ruling out that variable.
now if nobody else has had similar observations then it would be a safe assumption that what I observed was nothing more than a coincidence.

But so far you have only proven the first part, of a two part question to be based on a misconception...the second half cannot be answered by your test methods...

my method of polling a larger sample will produce more accurate results than your method of using a tripod, to find out if a steady handed shooter without tripod is getting adverse results.

so if we can get a couple dozen shooters that consider themselves steady handed to fire off a couple hundred shots each (as I suggested before, under similar circumstances as described...) and let us know if thier averages improve or decrease or stay the same, we'll know if the original obsevation was correct.