E-shutter and banding: what am I missing?

Discussion in 'Open Discussion' started by Klorenzo, May 18, 2017.

  1. Klorenzo

    Klorenzo Mu-43 All-Pro

    Mar 10, 2014
    Lorenzo
    As far as I know to avoid banding with the e-shutter you must shoot at multiple of the powerline frequency (50hz here). Because of the rolling shutter and all of that. Fine.

    These are my results (shooting a white wall with the E-M1):

    50 ok!
    60 banding
    80 banding
    100 ok!
    125 banding
    160 banding
    200 banding <--- ???
    250 banding <--- ???
    500 banding <--- ??? worst of all

    I suppose this is limited by the sensor readout speed but I do not know how (and it's midnight and I'm too tired to figure it out).

    Any ideas?
     
  2. eteless

    eteless Mu-43 All-Pro

    Jun 20, 2014
    Lights flicker at double the mains frequency as per cycle there is two peaks, one positive one negative.

    1/50 the exposure is long enough that it actually gets two peaks (a full cycle), no banding.
    1/60 the timing is out, so banding.
    1/80 out of timing
    1/100 you get one peak per cycle, however because each portion of exposure is the same part of the cycle there's no or very minimal banding.
    1/125 out of timing
    1/160 out of timing
    1/200 you have one exposure ending at the positive peak, then one starting at the peak and ending at the zero crossing, then the next area is zero to the negative peak, then the peak to zero crossing - so some banding because the light lags the curve (radiance doesn't follow exactly to voltage curve).
    1/250 out of timing
    1/500 out of timing

    Remember that the exposure time is per line (or multiple lines), the total exposure time still takes roughly 1/13th of a second or so (depending on camera, some models are faster). At 1/50th exposure time it will do around 4 total bands, at 1/200th this increases to 16 bands.
    Don't think of electronic shutter as flowing across the sensor like a mechanical, it's more like it steps across the sensor.
     
    Last edited: May 18, 2017
    • Informative Informative x 6
  3. eteless

    eteless Mu-43 All-Pro

    Jun 20, 2014
    Here's a representation of what is seen by the sensor when using 1/500th shutter speed drawn in napkinmath style to try and better visualize it, I apologize as I don't have access to anything fancier than pen and paper currently. The sine wave is voltage, the top lines represent the end of the last segment capture and the start of the next, the bottom lines show that 1/100th divides equally into the 1/50th mains frequency (and 1/50 gets the full two peaks). While 1/200th divides into this 'cleanly' the light emitted doesn't follow exactly the same curve (the rising edge has different radiance to falling edge)

    IMG_20170519_093921478.

    If you have any questions feel free to ask, I'll try and explain it as best I can.

    Edit: The above was also assuming a perfect model. You could just as easily assume with 1/200th that one band is the top part of the sine wave and the next is the zero crossing which is completely uneven (the highest radiance and the lowest). With 1/100th and 1/50th because they always include at least one half wave the impact of starting mid cycle is minimized.
     
    Last edited: May 18, 2017
    • Informative Informative x 3
  4. Klorenzo

    Klorenzo Mu-43 All-Pro

    Mar 10, 2014
    Lorenzo
    Thanks for your answer but there is still something I'm missing.
    I get it for 1/50s. Each line sees a full cycle: one peak and one valley. Each line is exposed independently so each sees a different cycle but all provides the same total amount of light. It does not matter if the exposure starts at the peak, at the bottom or at any other place.
    If I use 1/80s each line sees a different amount of peaks and valley so we have bands, depending on where we place the "exposure window" over the wave. Fine.

    But why 100? Maybe I found a hint here (with US frequencies):

    zvw18.

    "The light flicker frequency is twice the AC power line frequency, i.e. 100 Hz or 120 Hz, respectively."

    (the bottom of the power wave doesn't mean dark: zero means dark).

    If I got the chart right the whole red band (from full left to full right) covers 1/60s and sees two full peaks and two full valleys. So it works. And 1/120 works too. And 1/30s, 1/15s, etc. should be fine too. So 1/120s (or 1/100s in 50Hz countries) is the maximum usable speed. Is it correct?

    Yesterday evening I was also confused because I was thinking of 1/200s as a multiple of 1/100s, well, it is, but in the wrong direction.
     
    • Like Like x 1
    • Informative Informative x 1
  5. eteless

    eteless Mu-43 All-Pro

    Jun 20, 2014
    I think you've got it, as you've stated 1/100th would likely be the highest usable speed for 50hz power. :2thumbs:


    For anyone else reading along who might benefit I'll further expand on my previous posts with slightly better diagrams (I feel the slightly must be stressed, I used a lens cap and a clothes peg to draw them...).

    Mains voltage is generally AC (alternating current) and thus;
    It starts at zero and increases to the positive peak,
    Peaks and starts reducing,
    crosses over the zero voltage point,
    is then a NEGATIVE voltage potential,
    it then increases to the cross over point and the cycle ends (each cycle starts and ends at the same potential).

    Lighting will either not care* if the voltage is AC (incandescent) or rectify the AC into DC and use it**.
    Since the lighting doesn't care and it's easier to draw we can now look at rectified 50hz AC which would produce only positive peaks at 100hz, thus you get something which looks kind of like the following:

    rectifiedac.
    The image capture starts about 1/3rd of the way through the first peak.

    For 1/200th you can see that the first band of image sensor gets a high period, followed by the red area which is low - this produces the bright/dark banding.
    For 1/100th you can see that while the duration isn't long enough for a full cycle it still sees the same amount of light as each half cycle produces the same amount of light.
    For 1/50th you can see that it's going to get two full peaks.
    Slower shutter speeds would also likely produce little banding even if they're not multiples, this is because any variation is averaged out.

    If anyone has any remaining questions (or criticisms!) I'm happy to hear them and answer if/when I can.



    *=Fluorescent lights are kind of an exception as they take advantage of the changing voltage by using an inductor to induce a much larger voltage spike when the voltage reverses, this is the same behavior as a back EMF spike which occurs when you stop driving a motor suddenly. I'm ignoring them because it's not at all simple to explain it and it doesn't really change the flicker behavior. Also electronic ballasts because I would be here all day.

    **=I'm also ignoring smoothing as while it's not hard to explain it only lessens the impact of flickering (the troughs are generally not as low) and thus if a light is well smoothed you don't get flicker.
    I'm also ignoring switch mode power supplies for lights such as LEDs, a badly made driver will induce flicker no matter what you do and a well made driver will be operating at such a high frequency with enough smoothing that the camera can't detect it.

    Basically I'm ignoring all the fun stuff that most people won't understand without some background in electronics.
     
    • Winner Winner x 2
    • Like Like x 1
    • Informative Informative x 1
  6. dwig

    dwig Mu-43 Top Veteran

    672
    Jun 26, 2010
    Key West FL
    This is not a universal rule. Some lights (e.g. "Compact Fluorescents", etc) that use mains power but convert it to something different will sometimes be uniform or will flicker at a different frequency.

    Fluorescents and "Neon" lamps need high voltages, sometimes way above mains voltages, and therefore use some component to convert the voltage. Old style fluorescent ballasts and neon transformers worked by induction and don't alter the frequency so any flicker is at the same frequency as the mains power. Many modern fluorescent ballasts (all "compact fluorescents") and neon transformers are electronic. They convert the mains AC to DC, up the voltage, and then convert back to AC, usually at a much higher frequency totally independent of the mains frequency.

    LEDs operate on DC using a circuit to convert AC to DC and lower the voltage as appropriate. They will flicker at double the AC frequency unless the conversion circuitry alters it in some additional way (e.g. filtering it to smooth out the fluctuations, create its own on/off cycles to alter the "brightness" of the LED, ...).

    It should also be noted that fluorescent light do not emit that same color light as their brightness waxes and wanes with the fluctuating voltage. Banding can be more a matter of color bands than brightness bands. The glowing phosphors have "inerta" and don't stop glowing instantly when the voltage reaches zero.
     
    • Informative Informative x 1
  7. Klorenzo

    Klorenzo Mu-43 All-Pro

    Mar 10, 2014
    Lorenzo
    Moving back to a more practical level, when and how should I worry about banding? In other words, are there any general...well...expectations about where this problem may be worst? Any trick to spot dangerous lights?

    My current plan is to take a test shot at 1/200s every time I decide to use the e-shutter. Are there better alternatives? I'm sure I'm going so forget about it sometimes, so maybe a mnemonic association like (completely random): "outdoor -> safe, indoor -> bad, neon -> good, led -> good, concert -> bad", etc. may be helpful.

    And if I get banding, does it make sense to scan a few different shutter speeds to see if I'm lucky?

    Personally I used the e-shutter only two times: at a theater show with no problems, and at home with the banding problem that started this thread.
     
  8. eteless

    eteless Mu-43 All-Pro

    Jun 20, 2014
    Really the only time I worry about it is with LED stage lighting, it seems almost universal that smaller venues are cheapskates and the lighting on offer is often not smoothed to remove ripple. I've seen LED stage lighting which caused banding while using mechanical shutter (electronic would likely have been hilarious).

    Banding is so easy to avoid by just using mechanical that I don't think it's worth thinking about for the most part, just don't use electronic unless for video and forget about it.
     
    • Informative Informative x 1
  9. AussiePhil

    AussiePhil Mu-43 All-Pro

    Jun 1, 2014
    Canberra, ACT, Aust
    Phil
    As a nearly 100% e-shutter user now the very very simple thing to do is shot a short burst if you are unsure and review... any banding issues will be clear, and just switch to mechanical then. It's now my simple check
    Just to clarify something for modern LED lighting of all sorts.... nearly all LEDS are driven by custom driver IC's that operate in the kHz ranges so in a pure LED lighting setup banding will appear at completely different speeds and may never appear. This is especially try for commercial stage lighting that involves dimming.
    It's only non dimming half wave rectified led circuits running direct from AC that will see low Hz flicker.

    Here's a nice gotcha for 50hz countries though, mechanical shutter at 10fps and faster shutter speeds can capture the brightness difference between frames.
     
    • Like Like x 1
    • Informative Informative x 1
  10. PakkyT

    PakkyT Mu-43 All-Pro

    Jun 20, 2015
    New England
    I imagine this wouldn't be any different for 60Hz countries as I can not think of any reason why 50Hz would be any more susceptible to this over 60Hz.