Focus peaking behavior difference

Started by shayok, April 18, 2013, 09:59:19 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

shayok

Hi,

I feel the focus peak feature works much better on my T3i/600D with ML version 2011Dec22.600D101.alex than on my 5DMk2 with ML version v2.2.5D2.212. With the T3i, as I turn the focus ring and the subject comes into focus, it just goes wild with blinkies and focus is very accurate. On the 5D2 it sort of blinks feebly over a wide-ish range, making it hard to focus well.  Anyone else have a similar experience?

I'm using the same focal length and aperture (which is what drives DoF) on both, pointing to the same subject, and setting Focus Peak parameters to "D1xy, 1.0, local" on both. I briefly considered the possibility that with the t3i's 1.6x factor I'm simply seeing the subject more magnified which is helping, but even zooming out the lens 1.6x I feel the distinct difference in experience persists.

Thanks for any help. The only idea I have is to go back to the older version of ML on the 5D2 and see if it helps, but it doesn't sound fun.

-shayok

a1ex

Screenshots please (parameters + test scenes).

shayok

It seems 2011Dec22.600D101.alex doesn't have a screenshot feature. I could do a video of the LCD as I turn the focus ring if that helps.

nanomad

EOS 1100D | EOS 650 (No, I didn't forget the D) | Ye Olde Canon EF Lenses ('87): 50 f/1.8 - 28 f/2.8 - 70-210 f/4 | EF-S 18-55 f/3.5-5.6 | Metz 36 AF-5

shayok

I'm having the most difficulty focusing with the 5D2 on fine details not very close to the camera (at least a couple of meters away), but not at infinity either. The books in the bookcase in this example.

5DMk2 with ML version v2.2.5D2.212:

At 22secs  I say "even at infinity we have a little bit of blinking on those books". Actually that is not so much the problem as that the amt of blinking is not that different from when they're in focus.

T3i/600D with ML version 2011Dec22.600D101.alex

1%

i like the display filter peaking and haven't used blinking dots in a while. there is some lag but it seems more coherent and much less of that dots go all over the place in low light.

shayok

Quotelike the display filter peaking and haven't used blinking dots

Sorry I don't quite follow; in focus peaking there are only a few options: d1 vs d2; threshold; color. Where do I choose "display filter" vs "blinking dots"?


a1ex

That's a bleeding edge algorithm.

Can you do an apples-to-apples comparison (same camera, same lens, same subject, same exposure, both algorithms)? Also include the nightly algorithm, which was tweaked since the Jurassic era.

shayok

How do I choose different algorithms in the same camera? Do you mean d1 vs d2?

a1ex

Um.. I meant both versions (you said the older one works better). Actually, try all 3 of them: 2.2, 2.3 and nightly.

shayok

Let me first make a correction: the 5D2 is using v2.3.5D2.212 not 2.2. I don't know that 2.2 is in the picture at all (unless that's the same as 2011Dec22.600D101.alex). Where can I find the firmware corresponding to 2011Dec22.600D101.alex that I can run on 5D2?


a1ex


shayok

If I understand you correctly there was never a 2.2 for 5D2. So we can't do a meaningful comparison on the 5D2. So the only comparison that can be done is on the T3i, which doesn't serve my immediate selfish interest of getting good focus peak on the 5D2. (It might help developers understand the difference in behavior and fix if necessary the 2.3 behavior, which could eventually help me.)

nanomad

Given the fact that the algorithm is the same regardless of the camera, you should do a comparison on the 600D (and maybe on the 5D2 comparing 2.3 and the nightly builds)
EOS 1100D | EOS 650 (No, I didn't forget the D) | Ye Olde Canon EF Lenses ('87): 50 f/1.8 - 28 f/2.8 - 70-210 f/4 | EF-S 18-55 f/3.5-5.6 | Metz 36 AF-5

shayok

I'm a little scared of the nightly.

nanomad

Alrite, then at least do a comparison of 2.3 vs 2.2 .... although the algorithm in 2.3 is quite old
EOS 1100D | EOS 650 (No, I didn't forget the D) | Ye Olde Canon EF Lenses ('87): 50 f/1.8 - 28 f/2.8 - 70-210 f/4 | EF-S 18-55 f/3.5-5.6 | Metz 36 AF-5

shayok

2.2 vs 2.3 on 600D.



On the 2.3, I would almost claim (between the 52s and 54s mark) that we get two different peaks on the bookcase.

shayok

I was wondering if the developers had any thoughts on this, or if I should just invest in a clumsy monitor. It would be such a pity - I was so happy focusing with my lowly 600D, now with the 5D2 I have gone backwards!

nanomad

It would be best if you could include the nightly algorithm too since that's the current "best"
EOS 1100D | EOS 650 (No, I didn't forget the D) | Ye Olde Canon EF Lenses ('87): 50 f/1.8 - 28 f/2.8 - 70-210 f/4 | EF-S 18-55 f/3.5-5.6 | Metz 36 AF-5

shayok

I  tried http://nanomad.magiclantern.fm/nightly/archive/2013-04-20-ae7cc6b5f69e magiclantern-v2.3.NEXT.2013Apr20.zip

I tried it both on the 600D and 5D2. I feel the behavior is different enough that there can't really be an apples-to-apples comparison with 2.2 or 2.3. Maybe in time I'll get used to it, who knows?

I was really happy to see that on the 600D there is now a Magic Zoom fullscreen feature which with 1:1 magnification might actually do away with the need for Focus Peak for me. Then I was really bummed out to see that fullscreen is not available on the same version on the 5D2. Is that a bug? Also Magic Zoom (maybe with Large zoom window especially) is often too flickery to be usable on the 5D2.

Francis

Not a bug, just a feature not available on that particular model. A bug is when something does not work as it is intended to do. If it isn't there, it isn't there.