I feel misled: Panasonic 20mm f1.7 II AF is SO SLOW!

Started 7 months ago | Discussions thread
Anders W
Forum ProPosts: 18,696Gear list
Like?
Re: I feel misled: Panasonic 20mm f1.7 II AF is SO SLOW!
In reply to micksh6, 6 months ago

micksh6 wrote:

Anders W wrote:

micksh6 wrote:

Anders W wrote:

micksh6 wrote:

Why do you feel mislead? Of course it's slow and it has been shown here long time ago. You just needed to look at the right tests:
http://www.dpreview.com/forums/post/52045407
http://www.dpreview.com/forums/post/52049352

A "controlled" test can also show that 20mm focuses much slower than other lenses and hunts more often too. It's not needed to refocus from close-up to infinity in order to see that. These tests show refocusing between about 5 and 7-8 feet.

Those tests are not controlled. They are biased in ways and for reasons spelled out in the thread to which you link. That's why they are not the right tests.

I didn't start this thread stating that 20mm is slow, so you might as well tell OP that he is biased and the lens is actually fast.

Don't worry. I already told him what the facts are a couple of times, latest here:

http://www.dpreview.com/forums/post/53583615

Whether you like my tests or not they showed exactly what OP is seing now. Your tests, on the other hand, showed that 20mm is as fast as other lenses.

My tests show exactly that yes, but only under the specific conditions I spelled out in the post where I originally reported my test results as well as in posts where I refer to them, latest in the post I link to above. As you you should be well aware by now, I have no trouble recognizing that the AF of the 20 is slower than that of other lenses under several other conditions. As you can see, this too is clearly pointed out in the post I link to.

The problem with your tests is that they don't show much at all due to unrobust test conditions that make the outcome very sensitive to minor variations along with a failure to control these variations sufficiently well.

BTW, a year ago I replied to your post with test results which you linked today: "the claim that 20mm focuses as fast as other fast primes is just plain wrong and it can mislead people". http://www.dpreview.com/forums/post/51100467

Perhaps, this is why OP feels misled now.

So you erected a straw-man. The unconditional claim made in the passage of yours that you quote is not and was not my claim. What I am, and have been, claiming is summarized in the post I link to above.

I think a curved focus target attributes to slow AF with this lens.

Any lens will have greater difficulty focusing on a curved than a non-curved surface since for the former there is no single point where the entire surface contained within the focus box can be in perfect focus. But on what grounds would the 20 have greater difficulty focusing on a curved surface than other lenses of similar FL?

This is what tests show.

No. For reasons already spelled out above, your tests do not show that and I don't know of any other ones that do either.

Explanation of test results can be a different topic.

One way of putting test results into question is to ask questions about them.

Some time ago I suggested that it might be because of noisy focus motor. Noise is caused by vibration. Vibration also leads to blurred images, and it's much harder for camera to find peak contrast on blurred images. Peak contrast of curved surface is already is harder to determine than of flat surface. And if some contrast samples are impaired by blur this can make AF fail, or it can lead to slow AF.

Have you ever tried focusing the 20 in 14x zoom AF using a remote and watching the screen as you focus? Did you compare with another lens? If so, what did that tell you?

Yes, I did. It told me that I'm not a superhuman. I can't see difference in blurrines on sequential live view frames because the frames are rapidly changing with 60fps frequency. Can you estimate blurriness of individual frames in fast-paced movie sequence? No? Why did you ask this question, then? (you don't have to answer, otherwise this will never stop)

I asked the question because, like you, I see no obvious sign of jitter in spite of the high magnification. So the amplitude of any movement caused by the focus motor must be very low and the impact therefore much smaller than that of handshake (for reasons further detailed below).

One way to see that would be to capture video with high speed camera. Since live view is changing with 60fps rate and there can be a visible transition between frames one would need a camera faster than 120fps to make sure that a single frame is captured without transition to next frame.

Because only few frames in focusing sequence would be more or less sharp these moments are hard to capture. I don't have such camera.

General logic says that vibration reduces quality of image (signal) that CDAF algorithm analyzes, and you can feel the vibration. Then it's easy to see why the algorithm needs more samples to produce good results on lower quality input.
Whether this quality impact is visible to human eye is another matter.

Further, have you examined the impact of hand-shake in the same situation (14x zoom AF)

There is no need to do that.

and how that impact compares in magnitude to any vibration from the focus motor (with the 20 or any other lens)? If so, what did that tell you?

Exposure time for AF is unlikely to be lower than 1/125s.

I am pretty sure that it is slower than that most of the time unless you run the EVF/LCD in high refresh. Normal screen refresh for the E-M5 is 60 Hz (1/60 s rolling shutter) and it probably drops to 30 Hz (1/30 s) when the light is sufficiently low. But the exposure time for each frame is of little importance for the AF signal for reasons spelled out below.

With 40mm equiv. lens one would need to have serious tremor in order to consistently produce fuzzy images because of camera shake.

As I have told you in earlier discussions about this, the point of importance with regard to the AF signal is how things compare between different frames. What the AF system is trying to determine is whether contrast is increasing or not from one frame to another. This depends not only on whether focus is actually improving or not but also on the extent to which you manage to keep the content within the AF frame constant. Of course, your handshake has a far greater impact on variations in that content than anything the focus motor might do.

AF motor of 20mm lens, on the other hand, can generate whole range of frequencies starting from 200Hz (period close to probable low-light AF exposure time) and that would directly affect optical elements.

The lens element shake has an order of magnitude more impact on image quality than camera shake. For example, 8 degrees lens tilt would rotate focus plane by 90 degrees.

This is just a theory, of course. One possible explanation of long-term observation that 20mm F1.7 lens hunts more often than other lenses in the similar conditions.

PS. I thought these questions were rhetorical but you kept reminding about them in later posts, so here are the answers.

Appreciate that you answered them. Now you also know why I asked them.

Because of that 20mm lens may require more AF iterations, and it's more likely to start hunting than other lenses in the same conditions.

This isn't happening in good light, but in low light exposure time during autofocus becomes longer and if it's around 1/125-1/250s the quality of AF samples is very likely to be affected by low-frequency vibration (around 200Hz or so).

You know very well about shutter shock. You can think of this as focus motor shock. You can feel vibration when 20mm is focusing, and camera is taking images when doing it.

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +21 more
Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow