Jump to content

Image Stabilization test


henrik.ploug

Recommended Posts

Digital Photography Review has added a new IS-test to their already extensive

tests. Here they have tested the Sony A100:

 

http://www.dpreview.com/reviews/sonydslra100/page14.asp

 

The test is good, because you get a more precise indication of how many stops

the image stabilization will give you.

 

Do anybody know, where I can find a reliable IS-test of the new Canon 70-200mm

f4 IS - or other new Canon IS-lenses?

Link to comment
Share on other sites

I find the DP Review tests intersting -- they test the Sony Alpha with a Minolta 50mm 1.4 lens.

 

At 1/10 of a second they got very few usable shots. However at 1/20 most of their shots came out very well. Seems like P&S quality of the in-body IS.

 

To me this is no big deal, I would expect Canon's IS lenses to do much better -- in fact I've have seen

excellent photos taken at over 100mm with IS at shutter speeds 1/15 or longer (even down to 1/6).

Link to comment
Share on other sites

First, I must confess I haven't read the article yet but it seems to me that MOST of the success or failure of IS is tied directly to one's hand-holding technique (or lack thereof).

 

In short, if you have Parkinson's Disease, chances are IS won't do you much good and you're better off with a tripod. So trying to quantify its usefulness, to me, is a gesture in futility because your mileage will definitely vary depending on how steady you are.

 

I'm personally a firm believer in IS though. And now there's the raging debate over IS being implemented in the body (Sony/Minolta) versus in the lens (Canon's IS lens lineup for example)

 

I've begun seeing Canon advertisements that address this with Canon claiming IS is better implemented in the lens, their rationale being that the longer the lens, the more precise the movements must be, which makes sense to me, I suppose. Although I admittedly have not studied the matter.

 

Just for grins here's my "test". This shot was taken with a Canon 20D and the 70-200mm f/2.8L IS lens, zoomed all the way (200mm or with the "crop factor" of the 20D 320mm some might say) at ISO 1600 with a shutter speed of 1/20th of a second, hand-held. Without Image Stabilization, I'm quite certain it would look like colorful goo: http://www.photo.net/photo/5304968

 

BTW, that shot's a jpeg, straight out of the camera - no sharpening was applied.

 

Best wishes . . .

Link to comment
Share on other sites

I used to badly want an in body IS before , but after seeing some test, I think it is not as effective as in lens IS, Though still useful on a cheap lens like the kitlens and Its better than no IS at all IMO.<p>

 

I posted a test before of Canon IS in this <a href="http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00IttQ" >link</a><p>

 

The IN lens IS works very nicely, But I hope the in body IS will improve and that Canon would offer such body, I want to IS my prime lens too.

Link to comment
Share on other sites

I wouldn't base much on current testing methods. They are way too uncontrolled.

 

The effectiveness of IS depends not only on the system used (lens/body, version 1/version2) but on the user. It almost certainly depends on both the frequency and amplitude of the movement as well as whether it's rotation or translation and which axis the movement is around or along.

 

There are so many variables that scientific testing (which this wasn't) will be difficult and it may not always accurately predict how well it will work FOR YOU on any given day and in any given situation.

Link to comment
Share on other sites

I wouldn't base much on current testing methods. They are way too uncontrolled.

 

The effectiveness of IS depends not only on the system used (lens/body, version 1/version2) but on the user. It almost certainly depends on both the frequency and amplitude of the movement as well as whether it's rotation or translation and which axis the movement is around or along.

 

There are so many variables that scientific testing (which this wasn't) will be difficult and it may not always accurately predict how well it will work FOR YOU on any given day and in any given situation.

 

This test in isolation means little. You need to compare results of the same test done on different systems.

Link to comment
Share on other sites

I consistently take excellent photos with the Canon 70-300mm f4-5.6 IS lens at 300mm and speeds as low as 1/30 sec.

 

And at 1/15 sec. at 300mm, I'd estimate that I have about a 20-25% fallout rate (ie. 1 in 4 or 5 photos is unacceptable).

 

I don't know if this provides any good information to you... but I know that I'm impressed with the lens.

 

//Chuck

Link to comment
Share on other sites

The reason why I asked this question is because I want to buy a new DSLR for my brothers birthday.

 

He wants the Sony A100 because of the built in IS. But I think he would be better of with the new Canon 400D, because of its ISO-performance.

 

See these (large) pictures for comparison:

 

Sony A100 - ISO 400:

 

http://gallery.photographyreview.com/showphoto.php?photo=40291&size=big&cat=518

 

Canon 400D - ISO 1600:

 

http://gallery.photographyreview.com/showphoto.php?photo=52779&size=big&cat=518

 

Sony A100 - ISO 1600:

 

http://gallery.photographyreview.com/showphoto.php?photo=40290&size=big&cat =518

 

As you can see, the 400D's performance at ISO 1600 is almost identical to the A100's performance at ISO 400.

 

My conclusion is, that this alone will make up for the two stops he would gain with the A100's built in IS.

 

But to be sure to convince him, I would like to be able to show him a reliable IS-test, which document, that IS in the lens really is better than IS in the camera.

Link to comment
Share on other sites

I have the Sony A100 and I can tell you that the IS is highly effective and the picture quality is excellent (at low ISO). Some claim that lens based IS is more effective especially at long focal lenghts (a claim borne out of thin air-me thinks). I've used the A100 at 300mm at it works like a charm. One the negative-it is plagued by noise at ISO greater than 400 and it has poor flash implimentation (slow pre-flash). The best choice (though expensive) would be a good ISO performer like a Canon with lens based IS.
Link to comment
Share on other sites

One factor that is not properly highlighted in DPReview's test is the subject distance. This is highly significant, since it's clear that at the 1.2m distance they used you are entering the area where image magnification makes a difference to the capability of stabilisation, and is a factor not accounted for in Lester's tests. As has been discussed before, Nikon acknowledge that for magnifications of more than 1:30 (about 3m subject distance), image stabilisation becomes progressively less effective with their VR macro lens, and at 1:1 they actually recommend turning it off altogether. This is because the relative importance of angular shake (pitch/yaw) that IS/VR is designed to cope with becomes progressively swamped by other vibrations that the system can't cope with. This is entirely consistent with being able to achieve maybe up to 3 stops of stabilisation on the Sony Alpha at reasonable subject distances without suggesting that DPReview's findings are incorrect. IIRC, KM's claims for anti shake at short subject distances were in the order of 1 stop improvement.
Link to comment
Share on other sites

"One factor that is not properly highlighted in DPReview's test is the subject distance. This is highly significant, since it's clear that at the 1.2m distance they used you are entering the area where image magnification makes a difference to the capability of stabilisation, and is a factor not accounted for in Lester's tests. As has been discussed before, Nikon acknowledge that for magnifications of more than 1:30 (about 3m subject distance), image stabilisation becomes progressively less effective with their VR macro lens, and at 1:1 they actually recommend turning it off altogether. "

 

Hi Mark

 

Yes this is an important point thanks for bringing it up. I have not done the maths modelling to check for the onset point of this effect, and I suspect getting reliable data on human handhold vibration spectra (without spending cash) is not to easy.

 

I had not heard of Nikon's 1:30 magnification claim - I expect they know what they are doing though.

 

Up to and including 300mm my tests are done at 1:27 or less, so only just in the Nikon limit. Above 300mm the test magnification is less than 1:13.

 

However, these are quite practical usage magnifications for general photography and much wildlife photography and so hopefully of practical interest. Most of the results seem to be more or less in line with Canon's estimates, at least for the 300mm lens.

 

At 24mm the characteristics of softness verses shutter speed are quite different to longer focal lengths with IS on or Off. Something that is probably to do with handholding vibration spectrum I suspect. If this trend scales in a linear way for magnification I have no idea.

 

Hopefully the error between these tests (which are not expected to be exact) and usage at lower magnification is small enough that are still a helpful guide.

 

I agree very much with the macro use issue.

 

However I have noticed that using the 300mm with tubes, the IS does still seems to have some positive effective between magnifications of 0.25 and 0.6X (operational observation rather than test). Testing at this magnification using non-subjective sharpness criteria is however difficult due to test target edge resolution.

Link to comment
Share on other sites

I just shot 3 picts in low light with a 30D and a 17-85mm IS all of the same scene outdoor scene. 1 was at 5.6 1/10 and 2 more at 5.6 and 1/8.

 

I'm 55 and I used to have a steady hand ( used to ). I shot three hoping to get one good one. They all came out rather sharp except for some depth of field issues ( 5.6).

I went out and shot another pict the next day with a tripod. It really didn't look any better.

 

That concludes my IS test.

Link to comment
Share on other sites

Thanks a lot for all your answers!

 

Apparently it is rather difficult to meassure the effectiveness of IS. I was looking for something in the line of Lesters test.

 

And even though there could be a problem with the test - which Mark point out - I think my brother will find the results convincing.

 

I also think, that he will agree, that the Canon 400D is better for indoor photography without flash than the Sony A100.

Link to comment
Share on other sites

My experience with IS is that it is a bit of a lottery. Assuming I can pretty much nail an acceptably sharp shot a 1/f shutter speed then with IS 1 stop extra is virtually gareenteed to be acceptably sharp. 2 stops works about 85 per cent of the time. 3 stops about 60 per cent of the time and 4 stops about 30 per cent of the time.

 

Perhaps the most frustrating thing is that you can never be sure whether the shot worked until you review it on properly on computer. It also slows down shooting a bit as I find you have to have depress the shutter and wait about half a second for the IS to settle down.

Link to comment
Share on other sites

Yes if you look at my test results any handheld shot will have some statistic variability about sharpness, seen by the random distribution of sharpness around the trend line.

 

This is why the 1/f rule is a little suspect, I tend to use 1/2.c.f if I can where c is the format crop factor.

 

Another variablity with IS is the startup time, press the shutter too fast after the IS has started and it may not be fully effective, give it more time and it might improve.

 

Typically 0.5-1 second of IS run is needed, some more resent versions improve with 4 seconds of lead time. It is quite hard to keep this consistant even in a test.

Link to comment
Share on other sites

I like the objectivity of Lester's methodology for assessing the degree of blur. I'm sure that his observation that 24mm seems to be more effectively stabilised than 105mm on the 24-105 is simply a manifestation of the magnification issue combined with the wider field of view angle.

 

If we ignore vibrations along the lens axis (these really only affect focus and field of view at macro distances), and "roll" rotations about the lens axis (e.g. due to poor technique in jabbing the shutter), neither of which can be corrected by IS/VR or sensor anti shake (although Pentax do claim that their SR mechanism can help with roll), then you can basically resolve other movements into a combination of angular shake or lateral/vertical shake (if necessary with a suitable change of vector coordinates/origin and a little trignometry).

 

If you consider a lateral shake with an amplitude of 1mm, then the field of view of the image moves by 1mm (it's a parallel shift). With a macro lens at 1:1 that 1mm is a significant proportion of the sensor/film width - and it will be 1mm on the sensor/film, independent of focal length. Use a 500mm lens at a subject distance of 50m, and you have subject framing of about 3560x2380mm on a FF body, so the 1mm shift in framing equates to barely a pixel on the image. You will get about the same subject framing with a 24mm lens at a distance of 2.4m. The effect of lateral shifts depends solely on the image magnification, athough the interaction between focal length and subject distance for a given magnification can give a misleading impression.

 

If you now consider an angular shake which displaces the image by a similar amount on the sensor as 1mm of lateral shake at 1:1 then it entails an adjustment of 100mm at the subject with either example, equivalent to about 2 degrees with the 24mm lens, and only just over 0.1 degrees at 500mm. The effect of angular shake for a given angle is actually independent of subject distance, and solely depends of the effective focal length of the lens.

 

From the above is should be reasonably easy to see that

 

AS/SR will be more effective at close subject distances than IS/VR (although at very high magnification the mechanism will no longer be able to cope);

 

IS/VR may still be useful with longer focal lengths at close focus;

 

IS/VR has a clear advantage with more distant subjects and telephoto lenses in terms of the vibration amplitude it can neutralise.

 

The capabilities of the systems also depend on the frequencies they can handle - something which is beyond simple analysis.

 

When I tried AS (on a KM 5D), I noted that the viewfinder feedback (a bar of up to 4 lights indicating the amount of shake being compensated for) did promote better technique in holding the camera and lens and releasing the shutter. You do not get that feedback using an IS lens, as poor technique is soon masked by the stabilised viewfinder image. However, poor technique will clearly influence the shutter speed at which an acceptable image can be shot.

Link to comment
Share on other sites

"I like the objectivity of Lester's methodology for assessing the degree of blur."

 

Thanks Mark.

 

"I'm sure that his observation that 24mm seems to be more effectively stabilised than 105mm on the 24-105 is simply a manifestation of the magnification issue combined with the wider field of view angle."

 

Well in fact what I observed was at 24mm was the sharpness was proportional to shutter speed, the IS being on changed the gradient of this line.

 

The big difference at longer lengths (105mm and above) was the sharpness undergoes a step function going from very soft to more or less full sharpness with a small change is shutter speed. The IS being turned on lowers the transition shutter speed.

 

So for longer focal lengths there was a simple ~2 stop advantage for IS but at shorter focal lengths there is a situation of reducing advantage where the system gave a 3 stop advantage for very soft images but only about 0.5 stop for a very high sharpness criteria.

 

I suspect this is due to a combination of vibration spectra of the human hand and mirror slap and the closed loop frequency response of the IS system.

 

However as I am not planning to write a Phd thesis on this I don't plan to look into futher, rather I plan to take photos!

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...