Remix.run Logo
▲ amluto an hour ago

I was expecting to read this and learn that the display was reasonably engineered and that there is some small subset of people who have extreme PWM sensitivity who have problems. But no, Apple modulates at 240Hz, according to plenty of articles.

We’re no longer in the dark ages of flicker. There’s actual industry guidance: IEEE 1789. And 240Hz, at any respectable depth of modulation, does not meet the low-risk limit, let alone the no-observable-effect limit.

I am wondering whether this is actually just an amazing firmware goof: the reports seem to suggest that operating at full brightness mitigates the problem, which makes little sense for an OLED display. An OLED showing 50% (post-gamma) gray ought to drive the pixels the same way as an OLED at 50% brightness showing white. And this display supports HDR, so even 100% brightness white isn’t really 100%.

I’m wondering if a little engineering mistake snuck in: some (most?) LCD displays dim by PWM-dimming the backlight, and plenty of backlights have current-to-light transfer functions that inherently low-pass filter the light. If someone naively ported the same logic to an OLED (which doesn’t have a backlight), they might have implemented brightness controls by applying a 240Hz envelope to the pixel excitation waveform. This makes no sense at all as an actual implementation, but I can totally see an AI doing it and writing tests to validate it :)

Even if there is a genuine engineering factor that requires low frequency PWM, I bet Apple could mostly mitigate the problem by deliberately mismatching the phases. For decades, pre-LED stadium lights had rather extreme flicker at 120Hz, but it was tolerable because the overall system was engineered to roughly evenly split the lights between three electrical phases, giving an effective modulation depth near zero.

I haven’t seen one of these displays myself. I find 120Hz flicker to be extremely obvious and very very unpleasant, and if these displays really do modulate in a nasty way at 240Hz, I expect I’ll be able to see it, albeit rather less obviously than at 120Hz. Also, pretty much any modern phone can see this in slo-mo video mode :)

▲Exuma 28 minutes ago | parent | next [-]

Wait... so operating at 100% brightness mitigates this? I bought this mac and have been reading about this. Stare at computer 15+ hours a day coding, but I also never use anything less than 100% brightness.

▲amluto 5 minutes ago | parent | next [-]

If you have a modern-ish phone to go with that Mac, try taking a slo-mo video of your screen (preferably with content with decent sized patches of something other than black or dark gray) at different brightness settings. Maybe even make a big gradient and display that at various brightness settings.

▲vitaflo 21 minutes ago | parent | prev [-]

PWM is usually only used for dimming. So at 100% brightness there is no PWM being used. I'm not sure how this apple monitor works, but that's the basic gist for most screens that use PWM for dimming.

▲formerly_proven an hour ago | parent | prev [-]

The "99% brightness has flicker" thing seems like a gaffe, but the general approach is common for low brightness levels.

▲amluto 23 minutes ago | parent [-]

I did some digging, and apparently there are serious designs using this sort of hybrid dimming:

https://www.imec-int.com/en/articles/displays-health-patches...

Apparently brightness consistency between pixels is poor at low drive current, so designers often use a separate PWM envelope to compensate.

And there are even papers studying the choice of frequency, e.g.:

https://pubmed.ncbi.nlm.nih.gov/41411543/

Unsurprisingly, 240Hz is not actually recommended (at least that’s what I get from the abstract, and I’m not about to pay to read the rest of it).

And, in an IMO pathetic example of how the US patent world works, there are apparently patents on obvious architectures like using a fast PWM clock and then discretizing the VRR frame interval to multiples of the PWM time. I thought of this myself about 15 seconds after contemplating that VRR was a complication and then I searched for it and immediately found two patents. WTF!