IS On or Off & More On Image Stabilization

BirdPhotographers.net

Help Support BirdPhotographers.net:

I'm attaching a some plots of an accelerometer attached to a camera to illustrate the vibrations due to framing during imaging. In the attached plot, I did two tests. My accelerometer measures acceleration in units of g (acceleration due to gravity = 1) from which velocity and position as a function of time can be derived. My accelerometer samples at 320 Hz. These data have been high pass filtered, meaning they show higher frequency vibrations only, and not movement due to panning.

First I tightly held the Canon 1D Mark IV with 500 mm f/4 L IS attached tightly in my arms and held tightly to my chest. With one hand on the camera and pushing the camera tightly to my chest and a finger on the shutter button, I fired two short sequences at 10 frames per second. This test illustrates my ability to dampen high frequency vibrations caused by mirror movement and is shown as the red curve in the figure. Mirror slap causes vibration up to about 25 microns (about 5 pixels) in 1/320 second.

The second test was a bird in flight test. I ran 8 tests and this sequence is the best (lowest vibrations). In this test, I held the 500 with one hand with my elbow tucked to my side. With my second hand, I held the camera tightly to my cheek and firmly in my hand with my index finger on the shutter button. I panned and during panning, I fired two sequences at 10 frames per second (blue line in the plot). The fact that the two sequences have timing where firing frames occurs at nearly the same time is a coincidence.

The BIF results show vibrations of up to almost 80 microns (the 1D mark IV has 5.7 micron pixels) in 1/320 second, or over 13 pixels. The peaks are caused by mirror slap and there is usually a bounce of the mirror also recorded in a second peak. Then the camera settles down a little when the shutter opens, but during the actual imaging period, vibrations of over 30 microns occur in 1/320 second (over 5 pixels).

The vibrations recorded are limited by the accelerometer and the vibrations at 1/1,000 second (kilohertz) are likely higher. But even assuming these numbers are the maximum, and that the vibrations scale linearly, a 5+ pixel shift in 1/320 second translates to about 1.7 pixel shifts at 1/1,000 second exposures, and 0.8 pixel shift at 1/2,000 second exposures.

Image stabilization does a great job of producing sharp images. This data says that the IS is removing the effects of these vibrations at least for exposure times as fast as 1/2,000 second and that one should use IS up to at least 1/2,000 second.

A side note. I found an interesting effect on hand holding. When hand holding and pointing in one direction at a static subject, I produced more vibrations, but when I started panning, those body-induced vibrations smoothed out. So I would say hand held panning works better than static hand held imaging.

Roger
 

Attachments

  • plot.bif.sensor.a.tgif.jpg
    plot.bif.sensor.a.tgif.jpg
    181.7 KB
Very interesting, Roger! I am not sure how much this really tells us about its effect on producing images in the real world though. The real world just has too many variables to ever really know. I guess thats one of the things that make photography so fascinating.

I found your comment about IS being effective up to 1/2000 to be interesting. Perhaps I'm not seeing a difference until 1/1250 because the difference is so small at 1/1600 or 1/2000 that its not noticible.
 
Very interesting, Roger! I am not sure how much this really tells us about its effect on producing images in the real world though. The real world just has too many variables to ever really know. I guess thats one of the things that make photography so fascinating.

I found your comment about IS being effective up to 1/2000 to be interesting. Perhaps I'm not seeing a difference until 1/1250 because the difference is so small at 1/1600 or 1/2000 that its not noticible.

Hi Jim,

Certainly, the differences will be smaller at 1/2000 second versus 1/1000 second, but I bet if you did a BIF session with 1/2000 second exposures, 100 exposures with IS on and 100 with IS off, you would have a higher percentage of critically sharp images in the IS on set. The difficulty in such a test is BIF photrography is so variable depending on what the birds are doing, so "real world" test can be biased unless the sample set is very large. Controlled tests, like the one I showed can tell the real story. But, if you have "vice grip" like hands, you may be able to dampen the vibrations better than I can, so would see less effect. So my test is strictly valid for me. I did the hold tight to the chest test to find the probable lower bound of what is possible, to show there was still a significant amount of induced vibrations. Note: 50 microns is really small: 0.002 inch, and one pixel would be a movement of 0.00022 inch. I doubt anyone can hold a 1DIV camera to within 0.00022 inch with the mirror swinging up and down at 10 frames per second. It is actually amazing how well IS works, and to impressively small tolerances!

Roger
 
I guess for me the bottom line is that either way it doesn't affect my SOP which is to leave IS on mode 2 all the time when shooting BIF. If I always have it on when shooting bif, it doesn't make any difference to me if it becomes effective at 1/1250 or 1/2000, I'm covered either way. :S3:
 
Very good Roger. There is nothing like applying some good science to a question like this. As you were measuring acceleration at the camera body, I assume the actual movement of the projected image on the sensor depends on the focal length of the lens would it not?

Do you think the movement amplitudes you were seeing are large enough to affect the ability of the AF system to do it's job?
 
Very good Roger. There is nothing like applying some good science to a question like this. As you were measuring acceleration at the camera body, I assume the actual movement of the projected image on the sensor depends on the focal length of the lens would it not?

Hi John,

Thanks, John.
I computed movement of the entire camera, so in linear units of camera movement, it is independent of focal length. But the angular movement that the IS system needs to compensate for vibration is dependent on focal length. Interestingly, the longer the focal length, the less the in-lens IS angular position needs to move.


Do you think the movement amplitudes you were seeing are large enough to affect the ability of the AF system to do it's job?

I think it could be a negative factor if one were not using IS. With IS on, the stabilization helps both the image and the AF system (as you know). The AF system is taking its measurements when the mirror is down, which is in between the mirror down position and the mirror up at the start of the next frame, and that time interval seems to be very short. That period is also when the maximum vibrations are occurring, so IS on certainly helps stabilize the image for AF measurement and tracking.

Over the next few weeks/months I'll test tripods, bean bags, and long lens technique and mirror lock-up.

Roger
 
I am not sure if this methoid is valid. The readings from accelerometer seem extremely noisy and the peaks are just spurious spikes, I am not sure it is possible to enumerate such readings and draw conclusions.


An image shift of 80 microns! at 1/320 sec will make the image as soft as a drunk man's dream :S3: I rountinly make tack sharp images with lenses that don't even have IS (like the 200mm f/2) at 1/320 and down to 1/100 sec easily. I am sure most poeple can do this. I think the best rule of thumb for determining the speed form which IS becomes effective is the 1/EFL rule that is recommended by the manufactruers. Maybe a bit more consevertive if the rig is too heavy or if someone has weak/shaky hands. If you can't make sharp images say at 1/1000sec with the 500mm wiht IS OFF most likely it is the physical limitation of the photographer themesleves.


Anyways, not much more information to add I guess it is best to stick with whatever works for you.
 
Last edited:
Last note from Chuck Westfall:

As far as official Canon documentation is concerned, "tripod sensing IS" is an undocumented feature of the 1st generation IS super telephoto lenses. The closest reference I've seen in print appears in EF Lens Work III. The current version (Edition #13, released in March, 2011) states the following:

Tripod- and monopod-compatible Image Stabilizer

"The EF300mm f/2.8L IS II USM and other IS lenses automatically prevent accidental activation of image stabilization when used on a tripod. This eliminates the need for having to manually turn the Image Stabilizer off. With the EF200mm f/2L IS USM, EF800mm f/5.6L IS USM and various other EF lenses, image stabilization reduces minor blurring caused by camera shake and other factors when shooting with a tripod. And when a monopod is used with any lens in the IS series, image stabilization is identical to that achieved during hand-held photography."

The group of "various other EF lenses" includes the 1st generation IS super telephotos as well as the EF70-200mm f/2.8L IS II USM and the EF70-300mm f/4-5.6L IS USM, even though they are not explicitly mentioned.

Chuck
 
I am not sure if this methoid is valid. The readings from accelerometer seem extremely noisy and the peaks are just spurious spikes, I am not sure it is possible to enumerate such readings and draw conclusions.

An image shift of 80 microns! at 1/320 sec will make the image as soft as a drunk man's dream :S3: I rountinly make tack sharp images with lenses that don't even have IS (like the 200mm f/2) at 1/320 and down to 1/100 sec easily. I am sure most poeple can do this. I think the best rule of thumb for determining the speed form which IS becomes effective is the 1/EFL rule that is recommended by the manufactruers. Maybe a bit more consevertive if the rig is too heavy or if someone has weak/shaky hands. If you can't make sharp images say at 1/1000sec with the 500mm wiht IS OFF most likely it is the physical limitation of the photographer themesleves.

Anyways, not much more information to add I guess it is best to stick with whatever works for you.

Arash,

First, the 80 micron spike is the mirror slap, not the blur during exposure. During exposure, the vibration is in the 20 to 30 micron range.

One can not extend the data to other lenses without more data. The major movement of the camera is caused by the torque of the mirror swinging up and down. Attach a small lens and fire a few frames and this torque is easily felt on the 1D4. But on a big lens, the mass is large so the system has too much inertia so translates into significant movement because the center of mass is so far from the camera. This translates to a rotation of only about 1.3 arc-minutes. Try holding a laser pointer to that accuracy--it is difficult for most people. The camera on a smaller lens can rotate more with less translation of the sensor, and with the shorter focal length, less impact on image quality.

Note too that the 1D4 has the highest torque and mirror slap in the Canon line due to the 10 frames per second. Other cameras should have lower vibrations.

Regarding noise, I plotted the red line to show the accelerometer has inherently low noise compared to that in the BIF test. When the camera is not firing, the vibration data shows peaks of only 10 to 15 microns, but the second plot shows that those peaks are much lower frequency, and indicates I would be getting sharp images (movements well under 1 pixel) at exposure times as long as 1/100 second if there were no vibrations from the mirror.

Finally, regarding the 1/focal length rule, it was derived in film days. Today's digital cameras have higher focal plane resolution so the rule needs modification, like 1/(x*focal length), where x is a larger number for cameras with smaller pixels. When pixel size is about 8 microns, x should be about 1, and for the 7D with 4.3 micron pixels, it would be closer to 2. The 1D4, x ~ 1.4, or easier to remember, 1.5.

Roger
 
Last edited:
great to bring some real scientific measurements into the discussion Roger, excellent work. I look forward to your new data. Do you have a plot of the above tests without IS on? This seems like an excellent way to assess the effectiveness of IS rather than the subjective tests normally seen (i.e. number of sharp-rated shots out of 10/20 etc. at lengthening SS).
 
great to bring some real scientific measurements into the discussion Roger, ...This seems like an excellent way to assess the effectiveness of IS rather than the subjective tests normally seen (i.e. number of sharp-rated shots out of 10/20 etc. at lengthening SS).

As true as it may sound, does it really matter to you or anybody else for all practical purposes if you or anybody else cannot see the impact of its effectiveness in the raw files if not the final images?
 
Desmond, my point is more that many people trot out these/their subjective measurements as fact, including 'selling' particular ways of holding the camera, without any data to back up their claims. Roger's data really highlights the error introduced JUST by mirror slap, and he can start to bring the 1/focal length rule into the digital age with real data. People have sensibly started to include a 'crop factor' mod in their 1/mm calc, but Roger's posts are the first time I've seen clear evidence for a conversion based on actual pixel sizes and his accelerometer measurements.

As in many things, the variation in the general population means that the ultimate results will vary for some people, the same way some people over/undersaturate their photos because of differences in colour perception. However, without real measured data to start from I find people's claims that one way is better than another pretty meaningless.

Now we need to find Roger some funding for a panning robot to remove the human factor (him) from his data!
 
Desmond, my point is more that many people trot out these/their subjective measurements as fact, including 'selling' particular ways of holding the camera, without any data to back up their claims. Roger's data really highlights the error introduced JUST by mirror slap, and he can start to bring the 1/focal length rule into the digital age with real data. People have sensibly started to include a 'crop factor' mod in their 1/mm calc, but Roger's posts are the first time I've seen clear evidence for a conversion based on actual pixel sizes and his accelerometer measurements.

As in many things, the variation in the general population means that the ultimate results will vary for some people, the same way some people over/undersaturate their photos because of differences in colour perception. However, without real measured data to start from I find people's claims that one way is better than another pretty meaningless.

Now we need to find Roger some funding for a panning robot to remove the human factor (him) from his data!

Hi Geoff,

I agree with Desmond. The reason is the accelerometer data shows what the camera body is doing, but it does not tell the effect on the image with IS on because the IS is compensating for the camera movement. The only way to evaluate the effectiveness of the IS is to evaluate the images. I posted the data because one person made a statement that IS only helped to stabilize the image for the AF system and another said it was effective only at shutter speeds of about 1/1000 second and slower. The data do show that vibrations effect exposures at 1/2000, though the effect is getting smaller as shutter speed gets shorter.

I do agree that such data could improve imaging methods, and could help diagnose a problem an individual may be having if they were tested. As I work with the data more, I am beginning to understant more of the effects. The data I showed in the above plots was a summary, but single axis data I think is more informative. I can see clearly the start and stop of the mirror movement in the z-axis data (up-down). I'll post it once I get a little further along in the analysis.

Another example is long lens technique with a big telephoto on a tripod. My preliminary data points to another method that is slightly more effective (or at least similar in effectiveness) and more comfortable for me. I'll post that when I get more data analyzed.

Accelerometer data can also provide a quantitative measure of the effectiveness of tripods and tripod heads.

Roger
 
Lens jiggle helped IS to work

I was shooting some telephoto landscape pictures on a very windy day with a 100-400 zoom at 400mm, on a tripod, from about 1/4 to 1/15th second. The wind was sometimes causing the lens to vibrate enough to keep from getting a sharp image. I was hoping the IS would help to stabilize the vibration of the lens, but it seemed to make the image wander and jump around too much whenever the wind slowed down a little, as I was watching through live view. It seems that there wasn’t quite enough movement for the IS to work well. I got several shots that were not very sharp.


Then I got an idea to try something. Just before tripping the shutter with a remote cord with IS on, (in live view mode), I jiggled the lens a little, tapping it with my finger, and then immediately fired the shutter. Amazingly that shot was tack sharp, much sharper than any others that were made using IS or not.
I’m not sure if this fits into this discussion, but maybe … there has to be some movement for the IS to work, and the wind was not quite enough, so I created some more and the IS worked great.


The old IS on the 100-400 is different, I know.

Just wondering if anybody else has tried this. I only tried it once, so it may be a fluke, but it worked perfectly that time.
 
I meant 1/EFL (Equivalent Focal length) so that somewhat takes into account the pixel size.

I also agree your data does shows qualitative trends but I wouldn't extract numbers from it as is, it also depends how you attach the accelerometer and how you handhold your gear when you are shooting.

I think the bottom line for me is that I am not seeing blur by mirror vibration at 1/1000sec with the 500 f/4 when I handhold the gear, because I have made many sharp shots consistently (with the 1D4 and at 10 fps and IS OFF) but I understand for different handholding techniques the threshold might be different.

In any case the point of this discussion was the IS (mode 2) is beneficial for handheld flight photography which we all seem to agree on :S3: The only time I turn it off is on tripod and for slower shutter speeds as discussed above.

thanks for comments
 
Last edited:
I meant 1/EFL (Equivalent Focal length) so that somewhat takes into account the pixel size.

Arash,
Actually it (Equivalent Focal length) only propogates the crop factor myth. For example, apply it to a Canon 5D Mark II (full frame) and a 30D (1.6x crop). Both have the same pixel size so if you were using a 500 mm lens on both cameras, your formula would give different answers for the same pixel size, the same number of pixels on a subject, and where the exposure time rule should be the same.

Roger
 
Arash,
Actually it (Equivalent Focal length) only propogates the crop factor myth. For example, apply it to a Canon 5D Mark II (full frame) and a 30D (1.6x crop). Both have the same pixel size so if you were using a 500 mm lens on both cameras, your formula would give different answers for the same pixel size, the same number of pixels on a subject, and where the exposure time rule should be the same.

Roger

I agree pixel size has to be taken into account not the crop factor.
 
I was shooting some telephoto landscape pictures on a very windy day with a 100-400 zoom at 400mm, on a tripod, from about 1/4 to 1/15th second. The wind was sometimes causing the lens to vibrate enough to keep from getting a sharp image. I was hoping the IS would help to stabilize the vibration of the lens, but it seemed to make the image wander and jump around too much whenever the wind slowed down a little, as I was watching through live view. It seems that there wasn’t quite enough movement for the IS to work well. I got several shots that were not very sharp.


Then I got an idea to try something. Just before tripping the shutter with a remote cord with IS on, (in live view mode), I jiggled the lens a little, tapping it with my finger, and then immediately fired the shutter. Amazingly that shot was tack sharp, much sharper than any others that were made using IS or not.
I’m not sure if this fits into this discussion, but maybe … there has to be some movement for the IS to work, and the wind was not quite enough, so I created some more and the IS worked great.


The old IS on the 100-400 is different, I know.

Just wondering if anybody else has tried this. I only tried it once, so it may be a fluke, but it worked perfectly that time.

When shooting landscapes I normally switch to manual focus. It eliminates the possibility of the AF choosing something not planned on.
 
Artie,

Looks like I did explain well, what I meant was that in my experience the IS system did not seem to perform correctly when on tripod and delivered soft images, I can post some examples later.

I have not seen Canon advertise this feature in their manuals, white papers or other publications, in fact as I mentioned CPS recommended IS be set to OFF on tripod. That's why I asked.

I agree IS improves sharpness when handheld, it also helps with keeping the bird centered while tracking. I always use mode 2.

Please let us know if he can provide point us to a Canon document that explains how this feature is supposed to work.
hello everyone how are you, we wish you a new year full of goodness and blessings.
I agree with you. It happened to me when I turned on the image stabilizer in tripod mode, and when focusing, the focus point jumped and the images were distorted, and when I turned off the image stabilizer, the focus was normal.
 

Latest posts

Back
Top