fM: Comparison of anti-shake technologies of today
Ilya Zakharevich
During transfer in Frankfurt airport, I found an abandoned fotoMagazine
2007/01; there was the first on-equal-terms comparison of different
anti-shake technologies.
Well, the pecking order of this mag looks quite low (mostly naked gals
and naked lenses as illustrations ;-). And the picture quality measurement
(both for lenses they test, and for anti-shake) is incomprehensible: it
is measured in PERCENTS (?!) going from 0 to about 140. However, as far
as the researchers were unbiased (unfortunately, no way for me to check,
my German is too rusty now), such an investigation still carries some
information. Here are their raw findings (with\without anti-shake):
1/640 1/320 1/160 1/80
Canon EOS 400D, EF 2.8/70-200 L IS USM 100\60 90\50 50\0 10\0
Nikon D80, AF-S 3.5-5.6/18-200 G ED 100\90 100\90 90\40 50\30
Pentax K100D, SMS DA 4,0-5.6/50-200 ED 100\80 70\60 50\40 50\30
Sony Alpha 100, 2.8/70-200 G 100\90 100\60 90\30 50\20
1/200 1/100 1/50 1/25
Panasonic Lumix L1, Leica 2.8-3.5/14-50 ASPH 100\90 80\60 50\10 40\0
[I can make no comment on why their no-anti-shake results are so different
for different lenses...] The methodology was: for every shutter
speed, take 5 shots of the standard digital-test-chart with
anti-shake, and 5 without antishake; assign "quality" (in percents) to
each batch of 5. (I did not find the focal distance they used;
judging by results, most probably it was 200mm [equivalent of 300mm
for FF].)
(My) summary: The gain of "switching on anti-shake" (for this
particular person!) was approximately:
Sony Alpha A100: 2 steps
Canon: 1.5 steps
Nicon: 1 step
Pentax & Panasonic: 0.5 steps (better on the lower quality shots)
(comparing the shutter speeds where the quality is about 50-70%).
Overall "picture quality" vs shutter speed with anti-shake on:
winners are Nikon and Sony (identical curves), Canon is 1 step behind,
Pentax is 1.2 steps behind (Panasonic is not a player at this focal length).
[Note that Sony manages to get the same quality as Nicon with a much
brighter lens.]
Hope this helps,
Ilya
P.S. To somebody with better German, and access to pp.52-59 of the mag:
I think you would be able to make a better summary;
could you contribute?
Bill Tuthill
> ... Here are their raw findings (with\without anti-shake):
> 1/640 1/320 1/160 1/80
> Canon EOS 400D, EF 2.8/70-200 L IS USM 100\60 90\50 50\0 10\0
> Nikon D80, AF-S 3.5-5.6/18-200 G ED 100\90 100\90 90\40 50\30
> Pentax K100D, SMS DA 4,0-5.6/50-200 ED 100\80 70\60 50\40 50\30
> Sony Alpha 100, 2.8/70-200 G 100\90 100\60 90\30 50\20
> 1/200 1/100 1/50 1/25
> Panasonic Lumix L1, Leica 2.8-3.5/14-50 ASPH 100\90 80\60 50\10 40\0
I don't quite understand the raw data. Does this mean that with Canon 400D,
only 10 of 100 shots with IS came out sharp, and none without IS?
If that is the proper interpretation, it seems that Nikon outperforms Canon,
which is not what you say below.
> (My) summary: The gain of switching on anti-shake (for this person) approx:
> Sony Alpha A100: 2 steps
> Canon: 1.5 steps
> Nikon: 1 step
> Pentax & Panasonic: 0.5 steps (better on the lower quality shots)
This is contrary to most tests, which show better performance from in-lens
anti-shake (IS or VR) than from in-body AS.
Ilya Zakharevich
Bill Tuthill
<ccre...@yahoo.com>], who wrote in article <45ac...@news.meer.net>:
> Ilya Zakharevich <nospam...@ilyaz.org> wrote:
> > ... Here are their raw findings (with\without anti-shake):
> > 1/640 1/320 1/160 1/80
> > Canon EOS 400D, EF 2.8/70-200 L IS USM 100\60 90\50 50\0 10\0
> > Nikon D80, AF-S 3.5-5.6/18-200 G ED 100\90 100\90 90\40 50\30
> > Pentax K100D, SMS DA 4,0-5.6/50-200 ED 100\80 70\60 50\40 50\30
> > Sony Alpha 100, 2.8/70-200 G 100\90 100\60 90\30 50\20
> > 1/200 1/100 1/50 1/25
> > Panasonic Lumix L1, Leica 2.8-3.5/14-50 ASPH 100\90 80\60 50\10 40\0
>
> I don't quite understand the raw data. Does this mean that with Canon 400D,
> only 10 of 100 shots with IS came out sharp, and none without IS?
> If that is the proper interpretation, it seems that Nikon outperforms Canon,
> which is not what you say below.
My german is not good enough to find the methodology in this (long)
article. They estimate lenses usings similar axes, and there the
"the sharpness" is measured in percents (going from 0 to 140 ;-).
Thus I "expect" they take an average "sharpness" of 5 shots they made.
> > (My) summary: The gain of switching on anti-shake (for this person) approx:
> > Sony Alpha A100: 2 steps
> > Canon: 1.5 steps
> > Nikon: 1 step
> > Pentax & Panasonic: 0.5 steps (better on the lower quality shots)
> This is contrary to most tests, which show better performance from in-lens
> anti-shake (IS or VR) than from in-body AS.
Well, *I* did not see any other on-equal-terms comparison. Do you
have an URL?
Thanks,
Ilya
Ilya Zakharevich
Bill Tuthill
<ccre...@yahoo.com>], who wrote in article <45ac...@news.meer.net>:
> > ... Here are their raw findings (with\without anti-shake):
> > 1/640 1/320 1/160 1/80
> > Canon EOS 400D, EF 2.8/70-200 L IS USM 100\60 90\50 50\0 10\0
> > Nikon D80, AF-S 3.5-5.6/18-200 G ED 100\90 100\90 90\40 50\30
> > Pentax K100D, SMS DA 4,0-5.6/50-200 ED 100\80 70\60 50\40 50\30
> > Sony Alpha 100, 2.8/70-200 G 100\90 100\60 90\30 50\20
> > 1/200 1/100 1/50 1/25
> > Panasonic Lumix L1, Leica 2.8-3.5/14-50 ASPH 100\90 80\60 50\10 40\0
>
> I don't quite understand the raw data. Does this mean that with Canon 400D,
> only 10 of 100 shots with IS came out sharp, and none without IS?
Ex of my interpretation:
Canon 400D, 1/320: with anti-shake on, the "average sharpness" of
images is 90%; without anti-shake, the average sharpness is 50%.
> If that is the proper interpretation, it seems that Nikon outperforms Canon,
> which is not what you say below.
Nikon outperformes Canon in the *improvement* given by anti-shake:
e.g., sharpness of 50% is
for Canon: at 1/600 without, at 1/320 with (less than 1 step difference)
for Nikon: at 1/200 without, at 1/80 with (more than 1 step difference)
(I averaged several improvements at a few values of sharpness to get these:).
> > (My) summary: The gain of switching on anti-shake (for this person) approx:
> > Sony Alpha A100: 2 steps
> > Canon: 1.5 steps
> > Nikon: 1 step
> > Pentax & Panasonic: 0.5 steps (better on the lower quality shots)
Hope this helps,
Ilya
Bill Tuthill
>
> My german is not good enough to find the methodology in this (long)
> article. They estimate lenses usings similar axes, and there the
> "the sharpness" is measured in percents (going from 0 to 140 ;-).
> Thus I "expect" they take an average "sharpness" of 5 shots they made.
Perhaps the editors were too busy looking at the pictures of nekkid women
to notice that 140 is not a valid percentage in this test.
> ... Here are their raw findings (with\without anti-shake):
> 1/640 1/320 1/160 1/80
> Canon EOS 400D, EF 2.8/70-200 L IS USM 100\60 90\50 50\0 10\0
> Nikon D80, AF-S 3.5-5.6/18-200 G ED 100\90 100\90 90\40 50\30
> Pentax K100D, SMS DA 4,0-5.6/50-200 ED 100\80 70\60 50\40 50\30
> Sony Alpha 100, 2.8/70-200 G 100\90 100\60 90\30 50\20
Note that none of the numbers "with" anti-shake is greater than 100.
Thus I assume they took 100 pictures in all cases. Thus, the fact that
the Nikon D80 and Sony Alpha 100 were able to get 100 sharp pictures
at 1/320 second might just indicate better handholding ergonomics.
>> This is contrary to most tests, which show better performance from in-lens
>> anti-shake (IS or VR) than from in-body AS.
>
> Well, *I* did not see other on-equal-terms comparison. Do you have URL?
Yes, here are a few, admittedly Minolta and not Sony:
http://www.luminous-landscape.com/reviews/cameras/minolta-7d.shtml
http://www.aliasimages.com/KM7D_AS_Test.htm
http://www.epinions.com/content_174057361028
Ilya Zakharevich
Bill Tuthill
> Ilya Zakharevich <nospam...@ilyaz.org> wrote:
> >
> > My german is not good enough to find the methodology in this (long)
> > article. They estimate lenses usings similar axes, and there the
> > "the sharpness" is measured in percents (going from 0 to 140 ;-).
> > Thus I "expect" they take an average "sharpness" of 5 shots they made.
>
> Perhaps the editors were too busy looking at the pictures of nekkid women
> to notice that 140 is not a valid percentage in this test.
I have no idea why do you think so. Anyway, it is not relevant, since
140 does not appear in the results of *this* test (but "sharpness"
above 100% appears in tests of especially high-quality Leica and Zeiss
lenses).
> > ... Here are their raw findings (with\without anti-shake):
> > 1/640 1/320 1/160 1/80
> > Canon EOS 400D, EF 2.8/70-200 L IS USM 100\60 90\50 50\0 10\0
> > Nikon D80, AF-S 3.5-5.6/18-200 G ED 100\90 100\90 90\40 50\30
> > Pentax K100D, SMS DA 4,0-5.6/50-200 ED 100\80 70\60 50\40 50\30
> > Sony Alpha 100, 2.8/70-200 G 100\90 100\60 90\30 50\20
>
> Note that none of the numbers "with" anti-shake is greater than 100.
> Thus I assume they took 100 pictures in all cases. Thus, the fact that
> the Nikon D80 and Sony Alpha 100 were able to get 100 sharp pictures
> at 1/320 second might just indicate better handholding ergonomics.
Definitely - and also relate to weights of the lens. This is why I
concentrated on *differences* achieved by switching anti-shake *on*.
Thanks,
Ilya
Ilya Zakharevich
Bill Tuthill
> > Well, *I* did not see other on-equal-terms comparison. Do you have URL?
>
> Yes, here are a few, admittedly Minolta and not Sony:
> http://www.luminous-landscape.com/reviews/cameras/minolta-7d.shtml
> http://www.aliasimages.com/KM7D_AS_Test.htm
> http://www.epinions.com/content_174057361028
I have seen these.... The first one has no comparison. (Unless you
consider
I have no rigorous comparisons to offer, only the judgment of my own
eyes. But it seemed to me that...
something worth discussing...) The other two are hardly better.
So, with all the drawbacks, the fM article is the first "beyond
hearsay" on-equal-terms measurement of the advantage of stabilization
technologies. And while the only REAL metric is how "it works in YOUR
hands", very few people have enough time/equipment for this; thus even
semi-objective "pseudo-scientific" measurements are very useful.
Well, hearsay also has its advantages - it is in a different category.
Yours,
Ilya