E-510 IS at Higher Shutter Speeds?

[cstirlingbartholomew]

I am wondering if anyone has tested E-510 IS at shutter speeds above 100th of a second. For shooting with long teles it is important to know if IS E-510 produces a sharper image at 125th, 250th, and so forth. A 250th is a slow shutter speed for lenses 150mm and beyond. It would be really worth knowing if the 40-150mm kit lens zoomed all the way out would gain a stop or two worth of sharpness with the shutter speed set at 125th or 250th. All the tests I have seen so far have been about shutter speeds under 100th.

Thank You,

Clay

[windsprite]

Good question, Clay. I don't have time to process all the images now, but I just tried the 40-150 at maximum aperture and zoom, at around 1/200" and ISO 100, taking three images without IS followed by three with IS.

Looking at them on the LCD, the non-IS images look as clear or clearer overall than the IS images.

I tried the same setup holding the camera with only one hand, and I saw little difference there as well.

It's getting dark right now, so I can't get enough shutter speed without going higher on the ISO, but I'd like to try this again in better light. From today's tests, I would conclude that for myself, at 150mm in good light I would be better off turning the IS off, which is okay because I never felt a need for it anyway at that focal length and shutter speed. I'm curious now about shutter speeds like say 1/250" with focal lengths of 283mm (50-200 with TC-14).

Since getting the 510 and doing various IS tests, I've come to the conclusion that whether you use IS or not, it's best to take multiple shots when possible, because there can be a LOT of variation in clarity from image to image under exactly the same shooting conditions.

[E B]

Sure. I haven't been doing any tests per se but all of my recent animal posts were done with the E-510 with IS turned on and the shutter speed at 1/125 or faster and mostly using longer focal lengths. EXIF data is attached to the image in all cases if you have the software to read it. Seems to me to work fine. Each image is about 300K for easy download:

http://img.photobucket.com/albums/v1...707140243b.jpg

http://img.photobucket.com/albums/v1...707140238b.jpg

http://img.photobucket.com/albums/v1.../P7180382b.jpg

http://img.photobucket.com/albums/v1.../P7180389b.jpg

http://img.photobucket.com/albums/v1.../P7160355b.jpg

http://img.photobucket.com/albums/v1.../P7160360b.jpg

http://img.photobucket.com/albums/v1.../P7070102b.jpg

http://img.photobucket.com/albums/v1.../P7070103b.jpg

[cstirlingbartholomew]

Thank you Julie and E.B.,

I looked at all the images and the appear very sharp. I also looked at some of the images in the reviews shot with the 40-150mm kit lens at 150mm between 100th and 200th and they all looked very good as well.

What I have not seen is anyone who has done a series of 10 to 20 exposure side by side IS on and IS off tests using 150mm or longer at exposures between 100th and 250th with a subject that has high contrast edges which would allow us to evaluate the camera shake issue. This test would need to be set up in such a manner that camera shake would be sufficient to have an impact on the image sharpness with IS turned off at these short exposure times.

This is a fairly complex problem actually. The tests which choose the sharpest image with IS off in the sampling group and compare them with the the sharpest image with IS on group do not really address the problem. You really need to evaluate the entire range of the sample from the most unsharp to the sharpest image and see if the IS on and off groups are significantly different across the entire range.

Thanks again for responding,

Clay

[jebir]

Hi Clay,

I don't see the problem really!

No matter what the exposure time you use, the camera will shake in the same fashion. The camera will sense that and start following the wobbling projected image with the sensor movement. If it helps at long exposure times, it must also help for shorter ones up to the limit when there would not have been any camera shake visible anyway.

Maybe I am missing something (then please explain what) but to me it seems like you are looking for a problem that shouldn't exist.

Cheers, Jens.

[mckennma]

I thought IS had a limit around 1/125" to work properly.

[cstirlingbartholomew]

Jens wrote:


"No matter what the exposure time you use, the camera will shake in the same fashion. The camera will sense that and start following the wobbling projected image with the sensor movement. If it helps at long exposure times, it must also help for shorter ones up to the limit when there would not have been any camera shake visible anyway."

Thank your for asking this question, it is a valid one. What I am trying to determine is the limits of the IS mechanism in terms of response time. I would assume that there is a time interval X which is the minimum required for the IS engine to detect motion and respond to it with a correction. This time interval may be so short that the whole question is moot but I don't know that as a fact. Julie's post in response to your question indicated that up around 100th of a second the IS didn't appear to make a difference. Perhaps this was a result of Julie's steady hand and her familiarity with sharpshooting technique. In a less steady hand (mine due to old age) an exposure of 100th to 250th at her test focal length would produce some degrading of the image without IS.



Clay

[mckennma]

I usually change the ISO from 100 to 200 once I get between 1/6" - 1/30". Above 1/100", I can usually hold it steady enough on my E-500. Slower than 1/6" usually means ISO 400-1600.

[windsprite]

Thank your for asking this question, it is a valid one. What I am trying to determine is the limits of the IS mechanism in terms of response time. I would assume that there is a time interval X which is the minimum required for the IS engine to detect motion and respond to it with a correction. This time interval may be so short that the whole question is moot but I don't know that as a fact. Julie's post in response to your question indicated that up around 100th of a second the IS didn't appear to make a difference. Perhaps this was a result of Julie's steady hand and her familiarity with sharpshooting technique. In a less steady hand (mine due to old age) an exposure of 100th to 250th at her test focal length would produce some degrading of the image without IS.

I don't know if I have a particularly steady hand, Clay; it may be that my judgement of "sharp enough" is more lenient than other people's. It's a subjective thing. Anyway, I tried to address that by shooting a group one-handed.

Either way, as I say there was a lot of variation in sharpness between individual images in each group, and shooting some writing on an outdoor sign I didn't see an advantage to IS at the shutter speeds I used, either in considering individual images (choosing the sharpest of the group) or taken as a whole. To me that is counter-intuitive, and I want to give the test another try (I don't know if I can manage a series of 10 images each like you suggested above--that's a lot of time in PS to do 100% crops, etc.). But it could be as you say; the IS needs time to kick in and is only effective under a certain speed.

If that's the case, you could always stop down or use an ND filter.

Wink wink!

[cstirlingbartholomew]

Thanks Julie,

I gave this some more analysis while I was walking this afternoon and found a big gaping hole in my logic. I was assuming that the IS engine does sampling for correction of hand shake during the exposure. This is not something that can be assumed. Sampling might take place just prior to exposure and in that case it might be independent of exposure time. In other words if the IS engine looks for image movement just prior to the shutter opening then the sampling period could be a fixed value unrelated to shutter speed. If this is how things are done then you should get image stabilization at all shutter speeds even though it will probably be less obvious as the shutter speed increases.

Clay

[windsprite]

Clay, I don't have any shots to post right now, but I tried something different: purposely jiggling my hands severely while taking the shot. I obtained an AF lock before jiggling and kept the shutter half pressed, and shot at 150mm up to 1/400" and then 283mm up to 1/800".

At these speeds, it was fairly obvious that the IS was working. Beyond that, it's harder to tell at these focal lengths because the shutter speed stopped the motion pretty well. Unfortunately I don't have anything longer than 283 mm to work with, but my guess is that the results would be the same. At any rate, it seems like IS would definitely help someone who has very bad shaking in their hands. Moderate shaking, I don't know.

Personally, I would increase the ISO and maybe fire off a few shots with and without IS if I wanted to be sure of getting the shot. High ISO really is very clean on these new cameras--almost unnoticable noise up to 400 unless you crop severely (and that's in JPEG with the noise filter off)--so that is something else to consider; it's at least as important as the IS, I think.

[windsprite]

Thanks Julie,

I gave this some more analysis while I was walking this afternoon and found a big gaping hole in my logic. I was assuming that the IS engine does sampling for correction of hand shake during the exposure. This is not something that can be assumed. Sampling might take place just prior to exposure and in that case it might be independent of exposure time. In other words if the IS engine looks for image movement just prior to the shutter opening then the sampling period could be a fixed value unrelated to shutter speed. If this is how things are done then you should get image stabilization at all shutter speeds even though it will probably be less obvious as the shutter speed increases.

Yes, I get the sense that's how it works.

When you use live view and press the IS button to see what the stabilized image looks like, it takes a second or two for the image to "calm down" completely. So you might need to give the IS a second or two to kick in when shooting through the viewfinder as well, I don't know.

[cstirlingbartholomew]

Julie,

Thanks for all your work on this. I am sure you will take photos with more confidence now having pushed the limits of your hardware. I cannot image needing to test anything longer than 283mm. That is certainly long enough.

The reason I am being so inquisitive is that the Mega OIS engine in my Panasonic FZ20 was hailed when it came out as being on the leading edge of IS technology (ca. late 2004). Working with it extensively for over a year I found the OIS very spooky. It was a confidence destroying experience working with it. If your platform was too steady, bracing against a fence or a light post, you could get really bad images. I always used Mode-2 which limited OIS to the time of exposure. You didn't see a stabilized image in the viewfinder while you were composing the shot. Mode-2 was supposed to be more reliable than Mode-1 which was continuous OIS.

After months of working with it I got the system more or less under control so that I could rely on bringing home good images but there was always a shadow of doubt about what the OIS system was going to do in a given situation.

While there is really no reason to assume that Oly's IS engine will be as difficult to live with as the completely different OIS technology in my FZ20, I still think it will be worth the trouble to test the Oly's IS in challenging real world shooting situations before depending on it when something important is a stake.

Thanks again for the help with this.

Clay

[windsprite]

No thanks necessary, Clay. I'm naturally curious and like to test things, although I don't care to get too scientific about it. I honestly don't know to what extent I'll find IS useful in real life. Mostly low-light snapshots, I guess. I think Olympus has implemented the function very well, but it's not perfect; it doesn't work every single time. Many times I have gotten sharper results without it than with. To use it to your best advantage, I guess it takes a lot of experimentation like you did with your FZ20. There is only one page dedicated to IS in the 510 manual. I think Olympus could offer a little more guidance for getting good results.

[Stuttaton]

Any mechanical system has inertia, which works like a low pass filter. It means that really quick movements are "filtered out" by the inertia. The question here is which cutoff frequency this low pass filter effect has. This depends on the mass of the moving parts, among other things.
If it's going to do anything useful at 1/125 second exposures, the cutoff frequency needs to be considerably higher than 125 Hz. I doubt this is possible with the masses involved and the driving power available.
You could compare it with a bass speaker cone, except this has to be able to move in two directions, whereas the speaker cone moves in just one direction.

[jebir]

Hej Jörgen,

I don't see how the exposure time can be related to the cutoff frequency at the high end. If the camera shakes in a certain fashion (typically 1-10 Hz) the system must apply the exact same corrections to the sensor no matter the exposure time. EDIT: OK, for transients like mirror slap you do have a point since they die out fast and will not affect longer exposures as much as short ones.

If the system is incapable of correcting for high frequencies (say 100 Hz as you suggest), then: Would that not just be as detrimental for the image sharpness at long exposure times as short ones? I think the difference would be that at long exposures, there will be a pretty consistent shot-to-shot smear of the image, while at short exposures there will be a more stochastic variation in the shot-to-shot sharpness (the exposure each time sampling different relations between the un-synced image and sensor movements). EDIT: This would not apply for mirror slap as that is a fast dying transient.

FWIW, when the E-510 was announced, I read somewhere an interview with an Olympus product manager (or maybe it was a presentation) where he was very proud of the fact that Olympus had been able to solve the problems of contolling the sensor movement down to 7 Hz (which was implied to be a critical frequency for this application). I don't understand the details but it seems plausible that a camera shake caused by the human body should be in the range 1-10 Hz.

Cheers, Jens.

[Stuttaton]

Hej Jörgen,

I don't see how the exposure time can be related to the cutoff frequency at the high end. If the camera shakes in a certain fashion (typically 1-10 Hz) the system must apply the exact same corrections to the sensor no matter the exposure time.

Hi Jens. I suppose you are right. But for the IS to be OF ANY REAL USE at higher shutter speeds (like 1/125 etc) it needs to move quickly enough to do something useful in that amount of time.
The body movements in the range 1-10 Hz will already be taken care of by the relatively fast shutter speed, won't they?
Of course, a long lens will amplify these movements and if the IS tracks all the time, I suppose it could have some effect even at higher shutter speeds.

[jebir]

Hi Jörgen,

But for the IS to be OF ANY REAL USE at higher shutter speeds (like 1/125 etc) it needs to move quickly enough to do something useful in that amount of time.

aha, I see what you mean:
[loud thinking]
If you by "move quickly enough" mean that the sensor has to accelerate to the target velocity after the shutter button is pressed, then I agree that there could be a shutter lag.
In the other case, that the sensor is constantly on the move, the sensor should for the most part be a bit behind its target movement (probably more behind for higher frequencies), and a fast exposure time could sample that lag. Then one would see an improvement in sharpness in those cases when the sensor is moving in the right direction but a deteoriation if it happens to be moving in the wrong direction. Hence that could lead to a scatter in the image quality. Maybe that is what Julie is seeing when she reports about very varying results.
[/loud thinking]

After all, any IS system must have some limitations, but it must not make things worse if the user try to operate it outside its optimized parameters. I'm sure Olympus' engineers already have been there and implemented some cutoff for the IS function.

The body movements in the range 1-10 Hz will already be taken care of by the relatively fast shutter speed, won't they?
Of course, a long lens will amplify these movements and if the IS tracks all the time, I suppose it could have some effect even at higher shutter speeds.

Yes, that is more or less the way I see it and I don't know if there are any appreciable amount of high frequency components in the typical vibrations of a handheld dSLR. Oly could have the camera switch off the IS as the shooting conditions pass a certain limit, such as the 1/2xfocal length-rule.

Cheers, Jens.

[E B]

I think that having the IS degrade image quality when the camera is tripod mounted argues for the position that the IS is operational 100% of the time that the camera, and IS, are turned on and, probably, at all shutter speeds though a higher shutter speed could ameliorate deterioration of image quality due to IS being on while tripod mounted, just as it ameliorates deterioration of image quality due to hand movement when the camera is hand held with no IS function.

[Stuttaton]

In the other case, that the sensor is constantly on the move, the sensor should for the most part be a bit behind its target movement (probably more behind for higher frequencies

Yes, this is exactly what I mean. There will be an increasing phase shift between the subject movement and the sensor movement when the rate of movement nears the cutoff frequency. At a certain rate of movement they will be 180 degrees out of phase.
But note that the above is valid only for movement back and forth. For linear movement (within the timeframe the shutter is open), there is a lag, that is increasing if the sensor can't keep up with the speed. So in that case the IS will partially reduce motion blur. If the lag stays the same within the time frame there should be no blur but if it increases or decreases, there will be some blur.
Am I thinking right?

/Jörgen

[jebir]

Am I thinking right?

don't know... there is a lag between my thoughts and the discussion so my inner vision of the problem is getting blurred for most of the time but occasionally, it becomes crystal clear...

I think I'm reaching my cut-off or I may be in need of refilling the batteries...

Cheers! Jens.

[cstirlingbartholomew]

Yes, this is exactly what I mean. There will be an increasing phase shift between the subject movement and the sensor movement when the rate of movement nears the cutoff frequency. At a certain rate of movement they will be 180 degrees out of phase.
But note that the above is valid only for movement back and forth. For linear movement (within the timeframe the shutter is open), there is a lag, that is increasing if the sensor can't keep up with the speed. So in that case the IS will partially reduce motion blur. If the lag stays the same within the time frame there should be no blur but if it increases or decreases, there will be some blur.
Am I thinking right?

/Jörgen

I think this could explain a certain level of randomness in the results. Assuming that hand shake is a wave form of some sort, and also assuming that the IS system is always correcting for a situation which is now history, in the past, depending on the timing of the shutter release you might get good stabilization if the release point is on a flat part of the curve so that there is no change in slope between the IS sample and the capture period. On the other hand if the capture period falls toward the end of a cycle then the IS sample period might be rather different than the capture period.

If the hand shake was a more or less regular sign wave then the IS engine could predict what was going to happen during the capture period by looking at the last few cycles of the history and calculating the slope of the curve at capture time before releasing the shutter.

Clay

[cstirlingbartholomew]

This afternoon I ran some high shutter speed tests using the Panasonic Mega OIS system which is the same technology used in the lens that ships with the Leica/Panasonic 4/3 bodies.

The Mega OIS delivers significant image enhancement at 1/200th and 1/320th. I was shooting at 432mm (equiv) and introducing exaggerated hand jitter so that the image was dancing around in the viewfinder. I did more than ten exposures of each test set up. In the 1/200th series tests the worst stabilized image of the 10 was about equal to the best unstabilized image and both of them were freak shots, not even close to the others in the group. In the 1/320th tests most of the unstabilized images were very fuzzy and the best was quite soft. Half or more of the stabilized images were quite sharp with only one fuzzy one about equal again to the best unstabilized image.

When I tested at 1/500th of a second I changed from exaggerated jitter to a pan with a continuos linear motion past the target timing the exposure so that the target was captured in the exposure. This produced a different set of results. There was no perceptible improvement with Mega OIS turned on. The range of image quality was essentially identical for both sets of exposures. Most of the images were quite soft.

I should go do some 1/500th tests using jitter rather than pan because most IS systems treat jitter and pan differently. It is hard to jitter violently enough to cause image softness at 1/500th. At a some point the g forces introduced by the hand shaking will begin to interfere with the movement of the OIS engine and at that point things will get complicated.

Anyway, I am satisfied that Mega OIS really does significant things up to 1/320th of a second. If the E-510 in body IS engine performs as well as the results I found in my tests I would be a good thing to have. But lens based IS and sensor based IS are two entirely different worlds.

Clay