Tim42 wrote: ↑05 May 2022, 18:28
I have never noticed the effects John from DF describes. The only thing I noticed is that with unstable frame rates just below 60FPS, I don't perceive the picture stuttering quite as annoyingly at 120Hz as I do at 60Hz (but that's a different topic).
Correct.
Higher Hz can occasionally increase stutter of badly-framepaced software.
Some software have very erratic frame pacing.
- 60Hz fixed Hz VSYNC ON is like a one-dimensional snap-to-grid for frames with spacing intervals at 16.7ms (1/60sec)
- 120Hz fixed Hz VSYNC ON is like a one-dimensionasnap-to-grid for frames with spacing intervals at 8.3ms (1/120sec)
A lot of GPUs and software finishes frames very erratically (some frames take only 4ms to render, other frames take 10ms to render. On top of it, sometimes render-beginnings are also as erratic as render-endings, out of sync with refresh cycles).
60fps software with an 8ms rendering phase error (jitter of render starts/completions) will still look perfect at 60Hz because XBox and PlayStations are VSYNC ON, which rounds-off frames to the nearest refresh cycle (1/60sec = 16.7ms). So 10ms jitter/phase error in frame delivery internally in the GPU/software/drivers -- is cancelled out by the low snap-to-grid resolution of the 60Hz metronome (refresh cycles spaced exactly 1/60sec apart).
But play the exact GPU 60fps at 120Hz instead, and now you suddenly have stutter. This is because the refresh cycle intervals are finer granularity.
So depending on how rendering jitters, it might miss refresh cycles and generate a random pulldown of 1:3:1:1:3:2:3:1:3:1:2 instead of a perfect 2:2:2:2:2:2:2 pulldown for 60fps at 120Hz. GPU game rendering isn't perfectly framepaced. This shows up as stutter/judder at 120Hz that doesn't happen at 60Hz.
Occasionally, the processing in a GPU is cyclic (harmonic beat frequencies between multiple imprecise software clocks, imprecise driver clocks, imprecise CPU clocks, imprecise GPU clocks) and starts generating regular pulldown like 1:3:1:3:1:3:1:3 or even odd stuff like 1:1:3:3:1:1:3:3:1:1:3:3 or 1:3:2:1:3:2:1:3:2:1:3:2 pulldown that makes a false double-image effect easier to see. Other times it's randomized like 1:3:1:2:1:1:3:3:1:2:2:3 from background processing tasks of a GPU and CPU...
Unfortunately, sometimes this is a problem for video player software too! Not just games. Players are often software apps, and different apps are badly programmed. Some badly programmed video player software that framepaces 60fps video perfectly at 60Hz. But stutters at 120Hz.
Now, if framepacing is programmed properly and perfectly, this should never happen. But software aren't perfect. Graphics drivers aren't perfect. Video player software apps aren't perfect.
Sometimes pulldown judder also generates false double images, depending on how your human brain works. That's why 3:2 pulldown (24fps at 60Hz) sometimes generate a false double image effect for some humans, that doesn't occur as visibly during 24p mode. Every human brain sometimes perceive motion slightly differently (especially if GtG pixel response is super-fast).
3:2 pulldown is roughly like a harmonic between two different frequencies (e.g. 200Hz and 300Hz vibrating through the same guitar string). Although audio effects don't map exactly to visual effects, it is an apt metaphor because regular pulldowns are essentially harmonics.
It's important to ensure perfect TestUFO framepacing (green "VALID") so that you can rule-out possible pulldown-problems as an external non-display-related cause. We already know OLED makes 3:2 pulldown easier to see, because of fast OLED pixel response, and 24p becomes more visible improvement on an OLED than a slow LCD (where display motion blur can partially hide 3:2 pulldown judder).
So it's just a convergence of OLED fastness versus existing erratic pulldown now easier to see that's creating one of the many possible causes of 60Hz-vs-120Hz content differences.
Naturally, that's why 24Hz produces better 24p than 120Hz. Player software has a generous 1/24sec error margin before stutter occurs, instead of a 1/120sec error margin that's harder for software, because of the one-dimensional "snap-to-grid" behavior of lower-granularity refresh cycles helping smooth out 24p from flawed player software. An accidental hardware filter for software imperfections!
Note: GPU refresh rate perspective, not TV refresh rate perspective. Some TVs automatically frame-double/triple/quadruple/pentuple (e.g. use 48Hz or 120Hz to do 24p), but FPGAs and ASICs in TVs can do it perfectly. As long as the video source/console/GPU is outputting 24p 24Hz "on the wire", it will benefit from the stutter-smoothing of snap-to-1/24sec.
Note: Same snap-to-grid benefit applies to 60fps 60Hz, for TVs that use internal 60Hz-to-120Hz conversion, whether interpolation-based or non-interpolation-based to allow a 120Hz-only fabbed fixed-Hz panel to successfully simulate an uninterpolated 60Hz panel via repeating a 60Hz refresh cycle two times, for 60Hz "on-the-wire" sources.)
Fortunately with modern GPUs, TestUFO is currently capable of framepacing 60Hz perfectly at 120Hz, with some tricks (once properly setup). Then pursuit camera of 60Hz and 120Hz will at least rule-out TV-specific causes, allowing it to be traced to psychovisual causes instead, or to pulldown causes instead.