1. Reminder: Please user our affiliate links to get to your favorite stores for holiday shopping!

E-M1 AF Settings Test w/ 43 Lens - Settings Don't Improve AF Speed

Discussion in 'Olympus Cameras' started by tradesmith45, Oct 22, 2014.

  1. tradesmith45

    tradesmith45 Mu-43 All-Pro

    Dec 13, 2012
    Oregon
    Hi all,

    You're C&C is certainly welcome on the following.

    After getting a new E-M1 ( which now has Firmware v2.0), MMF-3 adapter & used ZD-300mm for wildlife work, I tried to figure out the "best" AF settings to use for action. Read many threads (likely from some of you) & did lots of shooting. My favorite testing was shooting RC model planes - great fun & great practice. All this practical testing(?) left me thinking that settings for Frame Rate, Release Priority, Release Time-Lag, C-AF Lock made no discernible difference in AF speed. But this testing involved so many variables, I couldn't be sure these results were meaningful.

    After some thought & trials, a relatively simple test procedure has confirmed my original conclusion - these settings yield no AF speed improvements that can be measured in some practical way and are very unlikely to provide any advantage in real shooting. So we can all relax.

    While all of the tests I report here were done w/ the 43 ZD300mm f2.8, tests w/ the m43 M.Z75-300mm @ 300mm showed that these settings make even less difference for such a slow focusing lens. The AF for the M.Z zoom is only about half the speed of the 43 ZD300.

    Before any of you get irate w/ me, there is an important caveat to these results. The standard deviation for these speed tests is 1/5th of a second. Not bad by human standards but still this is a coefficient of variation of about 25%. It is possible changing these settings makes an improvement that is too small to detect w/ the simple tests I've done. It may also be possible that the new Oly Pro lenses w/ faster AF speed will yield small but detectable improvements when settings are changed. But as you will see the old ZD300 is no AF slouch. Regardless, as the many debates about whether using the Vivid mode improves AF speed showed, there is such a thing as an AF placebo affect. These tests are sensitive enough to rule out all but the smallest practical benefits of different settings.

    And this set of tests address only raw AF speed for initial focus acquisition. It is possible some of these settings could improve the ability of the AF to stay focused on a moving target after initial focus acquisition or as interfering objects move into the frame - I'll try to test that later.

    The test procedure is simple. The target is the old Air Force resolution chart at 40' + an iPad running a stop watch app. I'd trigger the shutter on the 10's or 5's & both count the # of frames before acquiring focus & note the stop watch time captured @ focus. Exposures were ISO 500, SS=1/1250" & around f4.

    To test the affect of Release Time-Lag settings, I simply shot single frames of the target 20 times each using both Normal & Short Time-Lag w/ the lens already focused on the target in S-AF mode w/ Release Priority Off. The speed results & variation were identical for both settings: mean=0.20 sec. & Std. Dev=0.063 sec. This total time includes both how long it takes me to press the remote button & how long it takes the camera to fire as the clock hits say 1 min. & 10 seconds. Not bad in real world terms.

    Next were the tests of C-AF initial focus acquisition speed. For these the E-M1 was set to Sequential @ 6.5 FPS, 9 AF targets. The lens is first pointed at a target about 80' away & half-press the shutter button to get that target in focus. The the lens is then pointed at the chart/iPad @ 40' & the shutter button is pressed when the clock hits a 10. In 1 set of shots, I also tried initially focusing on a spot 50' away before shooting the chart @ 40' to see if closer pre-focus provided an improvement in final focus acquisition speed. The number of frames & time until target is in focus (@50% mag viewing) was recorded. For C-AF tests, at least 10 tests of each setting were completed.

    Here's a typical sequence of such shots where focus was acquired in the 2nd shot @ 0.70 sec & the AF settings are shown in the shot w/ Release Priority On:
    [​IMG][/url]_A210316 - 2014-10-21 at 12-14-34 by tradesmith45, on Flickr[/IMG]
    [​IMG][/url]_A210317 - 2014-10-21 at 12-14-34 by tradesmith45, on Flickr[/IMG]
    [​IMG][/url]_A210318 - 2014-10-21 at 12-14-34 by tradesmith45, on Flickr[/IMG]

    The end results w/ various combinations of AF settings were:
    Mean time to AF acquisition for all tests=0.86 sec. & 2.6 shots
    The range for all the tests was 0.89 to 0.82 sec. which is far smaller than the random variations of testing. Std. Deviation = 0.20 sec

    Here's a complete list of things I tested that provided no affect on AF speed:
    Release Time-lag - Normal & Short
    Display Frame Rate - Normal & Fast
    C-AF Lock - Off, High, Low
    Closer Pre-focus @ 50' vs 80'

    I found only 1 setting that made a reliable difference in C-AF mode. Having Release Priority Off reduced the number of captures w/o focus but had no impact on time to acquiring focus @ 0.82 sec. So turning C-AF Release Priority to On only fills you card faster w/ out of focus shots.

    EDIT: Please note that the AF speeds found for the ZD300mm were measured with the Focus Limits switch set for 6m to infinity.

    Hope all this is helpful to you.
     
    • Like Like x 5
  2. barry13

    barry13 Super Moderator; Photon Wrangler

    Mar 7, 2014
    Southern California
    Barry
    Hi, thanks for the info!

    Did you happen to test if higher sharpness and contrast settings have an affect on AF speed?
    https://www.mu-43.com/showthread.php?t=64939
    (I know you already posted a link to another report)

    Thanks!
    Barry
     
  3. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    Even if this rumour of faster AF with increased live view sharpness/contrast etc is true (which I personally doubt - I've seen no reliable evidence that it is), it wouldn't make any difference with 4/3 lenses since they use PDAF. The claim of improved AF is attributed to higher contrast on the live view improving CDAF - so with m43 lenses only.
     
    • Like Like x 1
  4. AussiePhil

    AussiePhil Mu-43 Top Veteran

    774
    Jun 1, 2014
    Canberra, ACT, Aust
    Phil
    One thing that seems to be missing is the direction of focus tests.
    What I mean is...
    How fast to focus going from 30->50 then from 50->30. Ie what happens if prefocus sent the lens from 30->60 yet target is at 50.

    I raise this as the behaviour of the lesser (more common) 4/3 lenses seems to be to hunt more if the new target is in the opposite direction to the original prefocus direction.
    I've noticed this most on the 70-300 on both 4/3 bodies and the EM-1. On the EM-1 the hunting seems to be less or maybe i've learnt to work around it over the years.

    Cheers
    Phil
     
    • Like Like x 1
  5. tradesmith45

    tradesmith45 Mu-43 All-Pro

    Dec 13, 2012
    Oregon
    No I haven't test this. I've been focused (pun intended) on AF speed @ 300mmm.

    But take a look at this "E-M5 Focusing Explained" at http://www.dpreview.com/forums/post/42140214

    The Oly AF system briefly boost contrast to the max for AF so there's nothing left that can be increased.
    Thanks!
     
    • Like Like x 1
  6. tradesmith45

    tradesmith45 Mu-43 All-Pro

    Dec 13, 2012
    Oregon
    Good question. In earlier tests, I included a long ruler in the frame. I could see that the ZD sometimes would overshoot the target & focus in front of it before getting focus lock. As mentioned, the tests where I pre-focused 10' behind the target produced no faster focus lock than when I had pre-focused 40' behind. I had assumed this would produce faster focus lock, another epic fail of my assumptions-:) So these 2 results together suggest pre-focus in front rather than behind probably doesn't matter unless the lens overshoots. Haven't figured out how often that happens.

    E-gad, more testing & we've started the rainy season here.
     
  7. Clint

    Clint Mu-43 All-Pro

    Apr 22, 2013
    San Diego area, CA
    Clint
    If you tested at 10' behind and 40' behind - "how do you come up with the conclusion "pre-focus in front rather than behind doesn't matter"? Just curious.
     
  8. DonParrot

    DonParrot Mu-43 Regular

    71
    Oct 7, 2011
    I really don't see any connection between shooting a chart board and the C-AF capabilities for action. That's a completely different affair, isn't it?
    And a test saying, that any FT lens focuses faster on the E-M1 than the m.Zuiko 75-300 completely contradicts all the experiences I made in the past 12 months with my FT and my µFT lenses.
    What counts is what happens out there when you are shooting and not what happens in artificial conditions in a test.

    And by the way: How can this...
    "The Oly AF system briefly boost contrast to the max for AF so there's nothing left that can be increased."
    ...be true if countless Oly users confirm that the opposite is the case?
    Just because it was described in a scientific way? say no more...
     
  9. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    Just because people say increasing contrast on live view improves AF doesn't prove it's true. After all, lots of people believe in homeopathy, but whenever it's subjected to scientific tests it's proved to be bogus. In fact, what tests I've read that try to look at the dependency of AF performance on live view settings usually can't find any correlation.

    Placebo affects are real - especially when we change something with the expectation that it will improve things!
     
  10. DonParrot

    DonParrot Mu-43 Regular

    71
    Oct 7, 2011
    Well, when I experienced this for the the first time - years ago with my E-Pl3 - I didn't expect anything of this to happen. It was just a grey foggy day and I switched to vivid and increased contrast and sharpness for nicer pictures - to find out that the AF worked significantly snappier. I really don't mind if people cope with a poorer AF as they just can't understand what's going on and opt for same old same old instead of using something even if they don't understand how it works. Although I think this is a kind of strange approach as arguably less than a percent of the motorists understand the electronics in their cars but drive them anyway. Nonetheless, anybody has to make their own decisions. But arguing against proven facts is just a kind of Don Quichote-ish, isn't it? And the fact that the mirrorless Olys' AF works snappier with vivid and increased contrast and sharpness has been proven by countless Oly users.
    Have you ever asked yourself why so many people claim that the E-M5's C-AF is useless while I regularly demonstrated for more than a year that the opposite is the case. Could be the adjustments, couldn't it?
    So, my recommendation: don't rely on tests but test yourself - and rely on your own experiences, afterwards.

    And by the way, even Brian Mosley from the Olympus Safari Group and many other well respected members of the Olympus family - who were extremely sceptical in the beginning - confirmed my findings after having tested the impact of the aforementioned adjustments themselves .
    Placebo effects - say no more.
     
  11. tradesmith45

    tradesmith45 Mu-43 All-Pro

    Dec 13, 2012
    Oregon
    Thanks for the question. Sorry if I was unclear. I am really only guessing based on the results I have - definitely no firm conclusion. But as w/ so much about the AF performance of this rig, my guesses could well be wrong.

    I'm making this particular guess based on what I've seen of the AF performance of the ZD getting to focus lock. Since there was no difference in AF times for the 80' & the 50' pre-focus cases AND the lens sometimes overshoots, I'm guessing that large pre-focus differences of all kinds have no impact on AF speed. That's a big disappointment for me. I did the 80' vs 50' comparison w/ the hope that AF would be faster if you can at least get close to the correct pre-focus distance. If that worked, we'd just need to make a small technique change to get better AF. Another assumption down the drain :frown:

    And this brings up something I'm still puzzling about - the variability in how long it takes to get to focus. W/ what seems like the exact same conditions, it may take 1, 2 or 3 frames to get to focus lock & I have no idea why. Could be the used lens is a bit worn & internal friction is variable. But as mentioned, I've also seen that the lens sometimes overshoots the target before getting into focus - why does this happen only sometimes?

    This variability also makes it very easy to reach incorrect conclusions from a few experiences.

    But there's lots I do not know about how the PDAF system in the E-M1 works. If you read the DPR thread "E-M5 Focusing Explained" & compare it to what at least I know about the PDAF system, we're really in the dark (pun intended-:).
     
  12. tradesmith45

    tradesmith45 Mu-43 All-Pro

    Dec 13, 2012
    Oregon
    As you will recall from my OP, I went to this more controlled test AFTER my field experience shooting birds in flight (some w/ feathers & some w/ motors) seemed to be demonstrating that these particular AF setting were not making any difference in AF speed. So yes, I do think this test is relevant to field conditions regarding raw AF speed.

    But this test also demonstrates something really important about judging whether some of these AF possibilities work or not. Without some control of variables, its just not humanly possible to reliably detect small differences in AF speeds that are only a few milliseconds. Even in these controlled circumstances, the coeff. of variation was over 25%. So even 10s of millisecond differences are just random. I find these results very humbling.

    Only reported AF speed difference between the ZD300 & M.Z75-300. I can't make any claims about other 43 lenses because I don't own any. I've seen too many online brawls about 43 vs m43 AF speed already & don't want to start another one.

    Please remember this test is really focused on how long it takes to get initial focus lock. That's important for BIF shooters where its common for something to come literally out of the blue. For most sports action, its often possible to follow the action as it moves around & the focus demands for that are different.

    This test does not reveal what may be important speed issues for subsequent shots after initial acquisition when needed focus changes may be very small. The DPR thread "E-M5 Focusing Explained" contained a revelation for me about Oly's design choice to use a CDAF system with a bias in favor of the last point of focus. This means that small focus changes after initial focus lock will be faster. As yet, I have no idea if the ZD300 is faster or slower than the M.Z75-300 for these small refocus changes.

    See above regarding variability & the AF placebo affect.
     
  13. tradesmith45

    tradesmith45 Mu-43 All-Pro

    Dec 13, 2012
    Oregon
    Hey Don, I for one would really, really love it if I could find settings that increased AF speed of my E-M5. Early on I casually tried Vivid & found no improvement. After reading E-M5 Focusing Explained (he also tested it) I understood why. You can see this looking through the EVF when you half press the shutter button. The screen brightens & increases in contrast momentarily. According to the author of Focusing Explained, this monetary increase in contrast is actually larger than users can achieve with camera settings changes. Don't know if the E-PL3 works this way.

    Regarding Vivid, here's a link to a simple but limited test of Vivid that I made https://www.mu-43.com/showthread.php?t=64939&page=7&p=699037#post699037

    But 43 Rumors says there seems to be a firmware upgrade coming for the E-M5. If there is some kind of consensus among Oly supported pros about setting that improve AF speed, we should expect to see this update to incorporate these changes since AF performance is an area where Oly cameras lag their competitors.
     
  14. DonParrot

    DonParrot Mu-43 Regular

    71
    Oct 7, 2011
    Well, when I found this out with the E-PL3 in 2010 or so, nobody knew that this was possible. So, a placebo effect ist just impossible as I wasn't searching for an AF improvement. This happened just by accident. And Brian Mosley from the Olympus Safari Group and all the other well established members of the Olympus fraternity were very sceptical when tresting my findings and confirmed them nevertheless, afterwards. So, I'd rather assume that there has something gone wrong with your BIF tests than that the settings don't work. So many experienced pgotographers arguably won't be wrong. it's clearly more probable that the small-ish group that can't cope with what the majority confirms is barking up the wrong tree, isn't it?
     
  15. tradesmith45

    tradesmith45 Mu-43 All-Pro

    Dec 13, 2012
    Oregon
    When we get some good weather (hopefully before July -:), I'll test it & I really hope your right. I'd love to get a free improvement in the AF speed of my beloved E-M5.
     
  16. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    OK, so can you point me to at least one test that has some rigour to it? At the very least it needs to:

    - Be well documented and repeatable - by others as well as the author. Anecdotal assertions don't count.
    - Shows a statistically-significant change in AF as the result of live view settings

    The closest I've seem to a systematic test actually shows no correlation - https://www.mu-43.com/showthread.php?t=64939&page=2&p=699037#post699037 (post 64 - by tradesmith45)

    Not only does the lack of a half-decent test raise the sceptic in me, but if you think about it, there's no good reason why it should be true (which doesn't mean of course that it's not, but I'm looking for an Occam's razor explanation here!). What I mean is this...

    - The data read of the sensor is a long way from being a displayable image. Constructing a live view image involves a complex processing chain including de-mozaicing, tone curve normalisation, colour balance, sharpening, noise reduction, ... This has to be done over and over n times per second. It uses a lot of processing.

    - OTOH, AF only needs basic edge analysis on a portion of the image. It also needs to be done quickly and repeatedly as the lens's elements move. Why go to all the trouble of doing the whole chain just for that? Moreover, a user could compromise AF performance by screwing up the live view rendering! I seem to remember somewhere that doing a fast Fourier transform on the image data and looking for a maximum in the high frequency elements was a typical approach for AF. If that is the case, doing the FFT on a portion of the image before the whole of the chain would be an obvious optimisation.
     
  17. DonParrot

    DonParrot Mu-43 Regular

    71
    Oct 7, 2011
    Hi tradesmith, if you are talking about C-AF, first of all you must downgrade the firmware to 1.2.. Then try vivid and - if you are of the opinion that the AF isn't snappy enough, afterwards - try increasing sharpness and contrast. But the basis for a working C-AF is firmware 1.2.
     
  18. DonParrot

    DonParrot Mu-43 Regular

    71
    Oct 7, 2011
    Well, action shoots aren't repeatable. Try it yourself and find out if it works or not. I have been shooting with these settings with both the E-PL3 (S-AF) and the E-M5 (C-AF with firmware 1.2) for years and know that they work. Anybody can find out for himself. So who needs a well-documented and repeatable scientific test if you just can take your camera, go out and find out what's going on. Just do it! it's not dangerous and nobody will get hurt. I really don't understand these discussions.
     
    • Like Like x 1
  19. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    Don - I appreciate your passion and conviction here, but neither of these things constitute proof. I appreciate that tests involving moving subjects are difficult, but in principle, they're still doable. The pharma industry spends billions on designing tests to remove placebo effects - and for good reason. Of course, no-one's life will be endangered by AF performance, but it's an interesting intellectual discussion anyhow :smile:
     
  20. Fri13

    Fri13 Mu-43 Veteran

    353
    Jan 30, 2014
    Die you notice any difference with Release Time-lag being normal vs short?

    As for me the Normal gives annoying lag when shooting macro and camera does autofocus first and then moment later take the shot, resulting lots of out of focus shots. But with Fast mode it works like expected, much more sharper photos as subject doesn't have aa time to move away.

    The Olympus 60mm macro gives strange results on me. As even the IBIS feels not to be working on release like it does with EVF only prior release. But macro photography is like having a 2000mm focal length and trying to catch a fighter jet at 800m distance. You just have very difficult situations with it.

    It is just sad that the Fast mode consumes more power, worth of 50 shots out of 350 when going from full to empty. So it needs to be considered when it is worth of it to be used.