Tuesday, December 22, 2009

Flicker test

The HD2000 is purely a NTSC/USA region camcorder since it records in 30fps and 60fps as opposed to those meant for the PAL/European market that records in 25fps or 50fps. But in this age of HD, the question of PAL vs NTSC doesn't really arise during playback on HDTV and I found that I can even playback my videos on my old PAL TV without any problem once I make the correct selection in the playback option.

However, there is some concern that there will be flickering when filming under fluorescent light at certain shutter speed since the mains frequency in PAL land is 50hz against 60hz in NTSC area. Actually I do not notice any flickering in my videos, maybe because most lighting now uses the energy saving bulb or the shutter speeds used by the HD2000 during filming do not result in any flicker or only slightly.

Anyway, at the request of one of my readers, Sergei, I made some short clips using the various shutter speeds that could be set manually. The shutter speeds that can be set manually are the following:

4s, 2s, 1s, 1/2s, 1/4s, 1/8s, 1/15s, 1/30s, 1/60s, 1/100s, 1/125s, 1/250s, 1/500s and 1/1000s.

However, according to the user manual, the lowest speed that would be used in the normal video mode is 1/30s and 1/15s in the high sensitivity (or lamp) mode. The highest speed possible is an astonishing 1/10,000s and I suppose this is seldom used in practice, if at all.

So what is the result of my test? You can see the result in the video below. This is a combined video of 7 clips of about 6-7 seconds each using increasing shutter speeds of 1/30, 1/60, 1/100, 1/125, 1/250, 1/500 and 1/1000 in this order. These were all shot using resolution of 1920x1080, 30fps (30p) and uploaded to Youtube as is without any rendering.





It shows that there is flicker, especially at the higher shutter speeds. There is slight flickering at 1/30 (very close to the light) and also at 1/60. At 1/100, there seems to be no flicker but the flicker starts again at 1/125 and gets more pronounced as the shutter speed increases. The results looks similar using the higher resolutions of 1920x1080, 60 fields/s(60i) and 1920x1080, 60fps (60p) though the flicker is now faster and the best shutter speed to use appears to be 1/100s in all the resolutions.

In practice, I don't see much flicker using the auto mode since I suppose the shutter speed used is 1/125 or lower and we don't normally shoot directly at the lights. Further more, fluorescent lights are now replaced more and more by energy saving lights so there are less chance of flickers while filming. If it is necessary to avoid or at least to reduce flicker, then set a shutter speed of 1/100s manually.

Below is a composite video of clips shot during my visit yesterday to a shopping mall showing the local Christmas spirit. All taken on auto indoor and totally hand-held with no sign of any flickering (probably no fluorescent lights were present). There is some sign of shakiness in the second last clip (dancers on stage) that employed fairly high zooms; otherwise not too bad for hand-held shots. I'll like to wish everybody a Merry Christmas and a Happy New Year with this video.

(Sorry, having problem to upload this second video in Youtube, will post again once OK. I am releasing this post without the video since Sergei seems to be desperate to have the flicker test result.)

Edited on 23 December, morning. This video is now uploaded at Vimeo here. The last song is complete in the original but the last few seconds were truncated during upload so hearts were lost! Merry Christmas, anyway!

Edited on 23 December, evening. At last, the video is uploaded in Youtube. Originally shot in 1920x1080, 30p but rendered to 720x480 for Youtube. This version is about 3 seconds longer than the one in Vimeo and is thus complete and more satisfying. Not sure if the missing seconds is due to Vimeo or to my line during upload. Nevertheless, Merry Christmas. Enjoy!





Ronald Kwok
http://ronaldkwok.atomicblog.hop.clickbank.net/

15 comments:

  1. Thanks Ronald
    Brilliant test! Best on the WEB!

    IMHO the test basically confirmed all my speculations in my replies to thread on avforums (I am blushing a bit actually ).

    Note: This test only applies to old-style fluorescent bulbs (usually a straight tube) and does not apply to compact modern screw-in low energy fluorescent bulbs which do not have any flicker (or very little at frequency 40-50 KHz - which for our purpose and shutter speed means none).

    For 30P as Ronald says we have:

    1)Shutter speed 1/30s = 33.33 ms - as expected we sometimes have just 3 flashes and sometimes 3 and a bit of 4th flash. Very little flicker (I expected about 5% difference between brightest and darkest frames - see my posts above for details). Still a bit annoying, but probably usable.
    I can bet that 1/50s would not have any flicker (the same as 1/100s) - the same as normal PAL camcorder. Sanyo - can you hear me? Please include 1/50s shutter for us in 50Hz contries! Or better still make a 25/50 fps version!

    2)1/60=16.66 ms is quit bad. We get 1.66 flashes on average which probably means about almost 2 flashes for brightest frames and as low as 1 flash (and 2 dark bits on both sides of it) for darkest frames.

    3)1/100s=10 ms is exactly one period of flash frequency. As we all know, an integral of a any periodical function taken over a period is constant whenever we start. There is no flicker (the same applies to 1/50 sec - exactly 2 periods).

    4)1/125s=8ms is getting worse again. It is now shorter then period so depending when shutter is open we can get all of the dark gap or may be none at all. It depends on how long dark gap between flashes is. If flash is very short (say less or equal 2 msec) then there is a possibility of missing it altogether now (as gap is 8 msec we can only have shutter open during the gap!).

    5)and 1/250=4ms, 1/500=2ms and 1/1000=1ms are definltely gettting worse - for the same reason as 1/125 - a frame can actually be shooting darkness.

    6)Actually at 1/250 - 1/1000s we can definitely see the 'Partial exposure' effect decribed in the article CMOS Rolling Shutter Rogs posted earlier.
    You can see them as horizontal brighter/darker areas on the picture.
    --------------------------------------------------------------------
    Thanks again Ronald - you are a star!

    Sergei Glasgow UK

    ReplyDelete
  2. As for practical implications for shooting indoor sports (which is what I will be doing 95% of the time) I am not quit sure what to do now.
    I really like ability to shoot 60P or 30P, but I at the moment I am only guaranteed a single shutter speed avalilable to me (1/100s) to not have any flicker.
    On the other hand there are other factors involved which may make flicker less or even result in none at all in an average hall:

    1)Sometimes bulbs are paired with a shift in flashes between the 2 bulbs arranged electronically - this is a safety feature as wee against stroboscopic effect (not the case in Ronald's video I beleive as results would not match my guesses so nicely if it was).

    2)In big(ish) halls lights would normally be running from all 3 phases of the AC mains, I beleive - if anyone is an electrician here - could you please confirm this?
    This would result in flash frequency trippled to 300Hz=3.33 ms so that shutters of 1/30=33.33 ms and 1/60 = 16.66 msec will be all over a few whole periods of this frequency. The only flicker which will be still noticeable (may be not) would be the fact that different bulbs will be at different points of the ceiling. This means that shadows and bright spots will be moving around a bit as relative moment of shutter opening is moving around (say relative to phase A sin wave).

    3)Newer fluorescent bulbs may be using High frequency (40KHz) AC generated internally and have no flicker at all (this wiki articale implies so):
    Fluorescent lamp - Wikipedia, the free encyclopedia

    The 3 points above are probably relevant to the second test video Ronald was going to post today (there was a problem - I am really looking forward to seeing that!) taken in a mall.
    This should be similar to an averags sport hall I'd guess (well some school halls may be much worse though ).

    Thanks again Ronald - you are a star!

    Sergei Glasgow UK

    ReplyDelete
  3. Hi Sergei,
    Thanks for all the superlatives and glad the video is useful for you. You can now watch the second video in Vimeo as stated in the edited post above. Merry Christmas!

    ReplyDelete
  4. Hi,
    Thank u for this graet post and comments.
    i am dealind with this issue for 7 months. my camera entered to lab to try fix this problem. they didnt find any problem with the cam :-). now i see the reason.
    Thank u guys.

    ReplyDelete
  5. Hi Dabari Family, glad that the test is useful for you so the effort is worth it. And thanks to Sergei for his explanation too. Hope you're still enjoying your HD2000.

    ReplyDelete
  6. An update:
    I have now bought a HD2000 and did some basketball shooting in 2 school halls.
    What I discovered was that in one hall I did get some flicker even using 1/100s shutter (!). I firmly believed before that 1/100s shutter should produce no flicker at all (we talking 60Hz mains power, of course). The lights in this school weren't very good, but there was no visible flicker noticeable by the naked eye.
    My guess is that one of the bulbs has some asymmetry in amount of light produced between positive and negative half-waves. This probably means that had HD2000 had a 1/50 shutter, flicker would go away completely.
    Also flicker exists in one end of the sport hall, but not the other.
    The good news was that flicker was not bad at all - I checked my video at home on big HD TV and basically it is there but you forget about it after watching the game for 30 secs. For a professional shot flicker is probably unacceptable though.
    The school hall's lighting uses big bulbs and has a yellowish tint to it.
    The other school I had no flicker at 1/100s and it uses most common tube-shaped flourescent bulbs you'd find everywhere.
    Basically I'd say that Sanyo must release a 50/25 fps version of HD2000. I would not buy this camcorder for UK if I knew that even 1/100s shutter produces flicker sometimes. Even if footage is still usable for myself it may be unacceptable for others.
    So if you need to shoot a lot of indoor sport semi-professionally, I have to say you will have to have a 25/50 fps camera. There is not a camera which does 1080p50 (as far as I know) though. I like 60P and 1/100s shutter a lot for basketball - this gives really usable slow play footage (e.g. I shoot 60P and press Pause then hold left arrow for 2 secs).
    Hope this helps someone - Hey Sanyo - are you listening?????

    ReplyDelete
  7. Oops, a typo in my previous post - I meant "we talking 50Hz mains power" (not 60Hz!!!)

    ReplyDelete
  8. Reread my post and thought that I'd add a bit more info about the flicker I got in one of the halls:
    As I said flicker was only visible in the far end of the court, but also it was only noticeable on the wall closer to the actual bulbs. Half way down the wall it was practically invisible and looking at the players/court floor it is impossible to detect, I'd say. For an idea of actual amount of flicker - watch first few seconds of Ronald's youtube test shot (at 1/30 sec shutter) - this is similar to what I got.
    Hope this helps

    ReplyDelete
  9. Hi Sergei, thanks for your follow-up comments, very thoughtful of you. Think it will be useful information for current and potential owners of HD2000. Just like to add that in my normal shooting, I hardly notice the flickering, if at all, even in auto mode. The user manual does mention about possible fliker thus: "If recording a video clip under fluorescent lights with the ISO set to 400 or higher, a noticeable flicker may appear in the image."

    ReplyDelete
  10. Hi Ronald
    In which software do you edit your movies?
    Does it have a multiple video tracks options?
    One more thing, in which resolution do you normally shoot?
    I've noticed that on 60p (the highest) it has trouble to downgrade the quality to 30 frames and it doesn't look good (missing frames) on external devices.
    I have to say that I've followed your blog for a long time until I finally purchased myself one (the price dropped in 300$ in just 3 months so it was worth waiting).
    thanks
    kobi

    ReplyDelete
  11. Hi Kobi,
    Presently I don't use any external software to edit my videos, I just use the built-in camcorder editing functions since I just do the basic cut and join. I ususally shoot at Full SHQ (1920x1080, 30fps) to avoid plyaback problems on my laptop/pc since they are not very high-end. Also not many editing softwares and media players can handle the highest resolution of Full HR but some of the latest versions may be able to do so. Hope you are enjoying your HD2000 and do share any interesting findings that you have on the HD2000. Cheers.

    ReplyDelete
  12. Hi Koby.

    My HD200 Just arrived today, I'm really excited. I am a huge video enthusiast I've shot with many different Pro/no-pro cameras, like; Canons, Panasonics, JVS ets... I have not gotten a chance to shoot or edit with it yet, but I have a few ideas and past experiences.
    No matter what I do (unless requested or need more pixels) I shoot @ 720p. Why? Because its smaller and easier to edit, and it typically looks good on a 1080p screen (Many TV shows are shot 720p (sometimes even @ 480p) and unconverted to 1080 or left at 720 and still look brilliant). Right now the "Film Look" is typically achieved by shooting @ 24p (fps) so I typically shoot @30p. 60p is really really nice, especially if you're looking for slow-mo, it also can be converted down to 24p.
    I primarily use Adobe Premiere Pro CS4 and experiment will Final Cut Express, but in both footage plays flawlessly (@30p) and I can main tain a pretty steady work flow (Not on extremely expensive computers).
    Again 1080p is not a must, 720 will enhance your computer performance. But unless you go Quad-core/Core i7 (which I don't) you won't be able to play full 1080 @60p flawlessly.

    (My editing Machine)
    This is what I use and it works great, although, I for see a need to upgrade maybe around 2011.
    Laptop (Primary editing machine)
    Core 2 duo 2.5ghz
    3bg ram
    512mb nVidia graphics card <<- Not really needed for editing but will be used in Adobe CS5
    2x internal HDDs. Primary - WD 320gb @ 7200rpm <<- 7200rpm is a must for HD video editing, its a standard in desktops, but typically 5200rpm come in laptops, this will enhance performance dramatically.

    This isn't perfect, but I am able to crank out some videos with out to much hassel.

    Hope this somewhat helps

    ReplyDelete
  13. *PS
    60p is always better for faster moving subjects, 30p will look like its missing frames if you or the subject is moving fast. And make sure your editor knows you are shooting @ 30p other wise it might choke

    ReplyDelete
  14. Nathan

    Thank you for the advice to use 720 probably at 30p. I could not play back 1080 at 60p at all.

    I wonder if 720p at 30p will produce better quality footage than that produced by my HD2.

    I am in Japan where the mains is at 60Hz and at 50Hz depending upon where you live. I live in the 60Hz area but I did not notice any flicker when using auto.

    ReplyDelete
  15. Hi Nathan and timtak,
    thanks for sharing. I am really surprised that in Japan you can have the mains at 50Hz or 60Hz depending on the area; I always thought that it is always only one standard in any one country. I suppose you'll notice the flicker only if you are in the 50Hz area and under certain conditions. BTW, I have found some media players that can play the Full HR videos at 60p. Watch my next new post on this!

    ReplyDelete