Personal View site logo
ETC Mode Discussion
  • 105 Replies sorted by
  • Only a quick test (ie only tried once as darkness fell) I set up my black-painted cylinder with fluoro-pink vertical stripe atop a dismantled air-cooler fan, synced to 78 rpm using strobe disk and a domestic power dimmer (don't try this at home)...

    image

    Both tests 720P 1/250 320 ISO. Lens 1 Som Berthiot Tele-Cinor 1:1.25 f75 Lens 2 Schneider-Kreuznach Xenon 1:1.5/25

    Don't pay much attention to results. I want to do some modifications; ie a narrower stripe.

    setup.JPG
    600 x 450 - 78K
    PinkStripeETC_OFF.png
    871 x 660 - 387K
    PinkStripeETC_ON.png
    891 x 555 - 357K
  • @Brianluce Agreed. I am looking forward to your test in the mode you'll be shooting in! :-)

  • @Roberto No, I already tested it a lot. In the real world. It's obvious is skews less in ETC.

  • Roberto, hopefully this is obvious to you, but you'll need to capture images of the barrel at the same phase of its rotation.

  • @balazer Yes, my testing will conform to scientific rigour. Capturing images at the same phase is easy, because the images match. Both camera and barrel apparatus are locked-off and level . In post, images must show little overlap.

  • If anyone has access to a professional vinyl-record turntable, you will find this testing easier to calibrate with the checker-board disk rim & variable speed knob. (I can't seem get one here). I really would like to see lots of comparable samples. Any wild deviations can be averaged-out and a pattern will emerge. My own barrel is turntable-size. Its 24 cm diameter provides a near-flat surface at the point of skew while still enabling a convenient 1.3 meter distance to sensor plane. In order be able to do dozens of clips at each setting, I'll need to speed up the process, hopefully by adding a bar with angle-of-skew calibration marks at the bottom of the barrel. That will cut down all this image-grabbing in post.

    My results should all be posted in a spreadsheet (no, I won't be cherry-picking some and conveniently forgetting others) and just a few, typical videos or pics posted. If others do the same, we've got some seriously powerful fast-testing happening here!

  • "Cherry picking"... You will be under the utmost scrutiny, guaranteed .....haha

  • @Rambo

    Don't scrutinise me personally. But if you can, please help constructively.

    Those who can, do. Those who cannot, teach

    Will you be contributing via experiment or armchair? ;-)

  • @balazer

    I find a very good way of double-checking the speed of objects crossing the screen is to count the frames in the time line. This way, I can reject all my samples which don't comply.

    @mpgxsvcd

    With your truck shots, All of trucks 1 to 3 have their bumpers crossing right to left in 8 frames. Truck 4 (white) at 3:004 with "Fine Printing Paper" marked on its trailer takes a full 15 frames. Red truck 6 at 05:007 through to truck 9 (blue & white front) at 8:005 are all 8 frames. After that everybody slows down.

    As for the 24fps stuff, I cannot view it correctly because of the 24-30 fps pull-down; too many re-doubled frames. But your original 24 fps mts can be tested.

    It looks like you're pretty right about truck speeds being consistent enough for skewv testing (that 8 frame-crossing at 30 fps) but you'll also have to time them and select like I wil.

  • Great discussion, as it goes I've always thought that the rolling shutter skew was a lot less in ETC mode. Also the Foundrys recommended values for their plug in to fix rolling shutter seem to suggest less correction is required in ETC mode, but things are different with 720p. http://www.thefoundry.co.uk/products/rollingshutter/training/

  • Their measurements seem to be correct. I read the topic more carefully and what I tested was 1080 while Roberto's focused on 720. Hence the difference in test results.

  • There is a good reason why this tests should be done at a higher shutter speed. That is because shutter speed has no affect on it. For instance if a faster shutter speed produced less of an affect then you would barely see the affect at 1/4000.

    The pictures I took at 1/4000th has the same angle as the ones at 1/250th and other ones that I took at 1/125th and 1/60th.

    The two things that really need to be tested here are ETC on and off for 720p and 1080p with a shutter speed of 1/4000th. I really think that the spinning can idea will work the best with a single camera. That is the most repeatable thing I can think of.

  • I just read the firmware 1.1 notes, "You can set to [EX. TELE CONV.] when recording still pictures or motion pictures." For a minute there I hoped it might mean while recording. [Afraid not]..

    -But wouldn't it be fantastic to be able to suddenly pull wide? I film dance a lot. When a camera pulls in tight I know instinctively there'll be trouble as soon as the dancers move.

    But imagine a one-button ETC-OFF and wham! Instant wide shot.

    Is this already on a wish-list?

  • Hi!

    I got some strange banding / noise in EVC mode. It's always visible in the dark areas of the video. Anyone know if this is normal or if my camera is broken? Could it be wrong ISO setting or any other setting i'm not aware of? I have tried different objectives but same result every time... (Panasonic Lumix 20mm 1.7 for example)

    The artifacts is clearly visible on a original file:

    http://janssonmusic.com/download/GH2_-_ETC_bad_quility.zip

    Settings: Quantum - Sedna C Setting - Q20 version.

    Frame rate: 24p

    Shutter speed: 1/50

    F: 8

    ISO: 160

    Objective: Canon 50mm 3.5 Macro with adapter

  • @jansson

    Yes that is normal with high ISOs. You might want to check the original with streamparser and make sure that it doesn't have a dip in bit rate during that segment. I see that a lot when I shoot the moon. Having a highly detailed and bright portion of the scene with a mostly dark border will show that affect.

    Try lowering the ISO, make sure Intelligent Dynamic range is off, and try different color modes. Smooth and Nostalgic boost the dark areas just like Intelligent dynamic range does. That is why most people think that they offer more dynamic range.

    Cinema is the only color mode that doesn't boost the signal. It underexposes by about 1/3 to 1/2 a stop. It clips the highlights though.

    B&W Smooth will boost the signal the least. It has contrast issues though.

  • Thanks for your reply mpgxsvcd! I used ISO 160 when I did the recordings. I have read about some ISO bug, but I havn't found any information regarding it. Dynamic range is off. I have just tried Smooth with -2 on all settings.

    When i'm not using the ETC mode, my picture quality is fantastic. So, I just wonder why it gets so bad in ETC mode... I think it's a nice feature, but unusable for me at the moment...

  • you could try it again at iso 200 to get rid of that noise bug. Interested to see if it helps!

  • So did this Rolling Shutter & Jello reduced in ETC bit ever get truly settled? Did @Roberto ever publish any later findings? When I last read this thread back in May of 2012, it seemed like the rolling barrel test was finally going to get this decided once and for all, at least for the actual skew, possibly not the perceived skew.

    I DID find that @Balazer said (in September of 2012): "Rolling shutter is a function solely of which mode you have selected and the ETC setting. 720p has less rolling shutter than 1080p, and ETC has less than with ETC off." Has any of that actually been established by experiment?

  • It seems to me that @Roberto 's pics say it all with there being a slightly smaller angle of skew using ETC.

  • The easiest most consistent way to test this would be to render a short animation of a vertical white stripe against a black background bouncing back and forth between either side of the screen of a monitor. It would have a controlled rate, and could be looped.

  • @Walker In Roberto's pictures the slant was nearly identical, with any tiny variations (less than a few degrees) easily accounted for by different speeds of the trucks driving by. His modification and analysis of @zigizigi's images don't convince me in the other direction either; I have no idea what speeds those sliders were going at all.

  • @neededandwanted I was referring to the spinning barrel, not the trucks

  • @Walker Yeah, that's the one I was hoping to see him follow up on. But I don't see enough difference in ANYONE's test to justify the constant assertion (in this board and elsewhere) that shooting in ETC reduces Rolling Shutter.

    Possibly a good test might be a spinning horizontal barrel with graph paper on it. It the barrel can spin at a constant rate, we can just measure the compression or expansion of the graph paper. Sort of measuring the jello instead of the skew. (Except that measuring an ANGLE in a vertical line/horizontal motion test seems to be easier to scale accurately -- the angle is the same no matter how much you zoom in or out).

    And maybe instead of a barrel, using the @jdenkevitz idea of utilizing a computer screen showing a constant-velocity pattern, possibly?

  • Maxbe these numbers help you: http://www.thefoundry.co.uk/products/rollingshutter/user-guides/

    If they dont't know it, I don't know who does. ;)