2
\$\begingroup\$

What is the cause of those black bars on the picture? Their orientation changes by moving the camera.

The display is of a Samsung Galaxy S3 (i9300), an AMOLED display, set to minimum brightness, which involves PWM.

The camera is a Nikon 1 S2, picture taken at exposure of 1/16000 s. The PWM flicker should be in the range of 60Hz, so why the "fast" bars effect? Does the camera use a rolling shutter? EDIT: Yes it does, see my comment below

the image

EDIT2: The phone display flickers with 300-400 Hz and low duty cycle (15% or less). See photos:

exposure 1/40
exposure 1/100

\$\endgroup\$
2
  • 1
    \$\begingroup\$ It's possible this is related to the refresh of the LCD screen \$\endgroup\$ Commented Apr 6, 2015 at 17:28
  • \$\begingroup\$ The lines could be indicative of the timing of the PWM across the display. It would be impossible to perfectly synchronize the pulses on every pixel to have the entire display flickering uniformly. Without knowing how the driver is configured it would be difficult to guess at anything further, but you probably have a very fast clock driving the pulses to achieve the necessary variation. \$\endgroup\$ Commented Apr 9, 2015 at 1:50

1 Answer 1

3
\$\begingroup\$

The vertical refresh rate of the screen is around 60 Hz, but the horizontal refresh rate is much higher. Each line of the display is refreshed from left to right during the vertical refresh.

With a vertical resolution of 1280 lines, it means that the horizontal refresh rate is about 76800 Hz. As the exposure time is about 5 times that, it makes sense that you see five bands on the display.

Normally an LED screen doesn't have a blank between frames, but the PWM reduces the brightness by blanking the screen between the frames.

To capture the display without banding you would rather use an exposure time that is so long that you capture several vertical cycles.

\$\endgroup\$
9
  • \$\begingroup\$ I see 5 bands even at 1/640 (just a bit narrower as on the posted picture). Also as noted, it changes depending on camera orientation. I did some fast "panorama" shots and they confirm the camera has a rolling shutter, both for video (tested 1080p30 and 1080p60) and photos. \$\endgroup\$ Commented Apr 6, 2015 at 20:52
  • \$\begingroup\$ @Guffa are you sure AMOLED refreshes as CRT? Any references? Also, the explanation does not hold. If so, the photo should show one fifth (a rectangle) of the display and the rest black. \$\endgroup\$ Commented Apr 6, 2015 at 23:49
  • \$\begingroup\$ LCD and OLED displays are not CRT, and are not managed the same way. The screen does not have to be completely redrawn, but can be changed pixel for pixel. Thus a vertical refresh theory, which is incorrect based upon the orientation of the bars and the shutter speed, just doesn't fit. \$\endgroup\$ Commented Apr 9, 2015 at 1:40
  • \$\begingroup\$ @DavidBalažic: Yes, it's the rolling shutter in combination with the refresh rate that causes different patterns. The pattern will be multiple bands unless you align the shutter movement with how the screen refresh moves. \$\endgroup\$
    – Guffa
    Commented Apr 9, 2015 at 8:00
  • 1
    \$\begingroup\$ @Lathejockey81: In theory the pixels in an LCD screen could be changed individually, but that's not done. Why? Because it would mean that you would send as much addressing information to the screen as color information, doubling the data that need to be sent to update the screen. A mobile screen is updated just like any other LCD screen, that's why tripple buffering and VSYNC was introduced in Android 4 to make animations smoother. \$\endgroup\$
    – Guffa
    Commented Apr 9, 2015 at 8:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.