Shutter speed when using LED lights

I don't think that any of these panel lights could harm a sensor.

A flat field makes it easier to see what we are looking for, and can make sure that the whole frame is filled with a blanket of light. There are other ways, of course, like shining the lamp on a wall and photographing that, but that makes it harder to use fast shutter speeds without also getting a lot of noise or "under-exposing" and it might get easy to confuse sensor banding noise with light-source banding. One alternative is to put a smooth diffusing material in front of the lamp to get rid of any visibility of individual LEDs.

A bright (but not clipped) flat field, the fastest shutter speed, and the slowest rolling shutter will be the easiest context to see banding in. If you see none there, then you won't see any with a slower shutter speed, or a faster rolling shutter. If you do see it at the fastest shutter speed, then you can try various slower speeds and see where the practical limit is for that light panel. Of course, you don't want to try this will full panel intensity, as that could easily mean no strobing, even in a panel that does strobe at lower intensities. Keep in mind that if you choose one relatively low shutter speed and check for banding, and you don't see any, it may be just a lucky choice that is an exact integer multiple of the light cycle, and 1/3 stop slower or faster could reveal banding.

The 1MHz cycles that Bob speaks of would not cause visible banding, even at 1/16000. At 40KHz, like Jim was talking about, a panel without phosphors and partial intensity could show shallow banding at 1/16000, but he says the phosphors smooth things out, so that might not be an issue. Most panels are typically used for much slower shutter speeds, though, where you wouldn't expect a visible problem.

I *think* you and Jim are talking about different things.

I think you are talking about the frequency of operation of the switching power supply ("SMPS"). And assuming fixed-frequency SMPS operation, rather than, say Pulse-Frequency-Modulation ("PFM").
I think Jim is talking about Pulse-Width-Modulating ("PWMing") the output of the SMPS.

Some SMPS LED controller chips have PWM control inputs, and effectively save the state of the SMPS while the PWM control is "off".

One of the arguments for PWM dimming, rather than adjusting the SMPS output current, has been that low frequency PWM gives better colour constancy, because the LED temperature, and so band gap, and so colour, is more stable when the LEDs are on.
But PWM frequencies below 10s of kHz are undesirable because of flicker problems, and at 10s of kHz the LED temperature is likely to be quite stable, and dependent on average LED power dissipation.

I am.

I run PWM window at between 25 and 50 KHz, filler at between 200 and 2000 KHz.

What is "filler" ?
You design LED lights?

"Filler", switching frequency to control duty cycle.
I don't design LED lights, made some for myself, and they are also a part of our Spectron project github.com/Alexey-Danilchenko/Spectron/tree/master/hardware/Light_Source/LEDs

This the kind of thread I was hoping would start appearing here. Great technical discussion and I am learning some new stuff. Thanks to all for the insight and for shining a light on this topic!

Now, if we could just get venues to use these high-frequency PWMs that Bob, Jim, and Iliah have mentioned in public venues instead of those ridiculous, stupid 100 and 120 Hz systems that so many of them are fond of. Does the venue lighting industry not get the feedback about banding or do they just not care? Maybe they read that global shutters are right around the corner, and are already counting on them! (Just kidding). Even with global shutters, though, you may not get lighting modulation down the frame, but the lighting may black out during a frame (without "flicker fixer", which can break the cadence of a burst) or vary the exposure too fast for camera metering or users to respond, so for 100 and 120Hz lighting, even a global shutter works better or more consistently with longer exposures.

A lot more substance here than some of the other long-meandering threads that purport to be about photography but are mostly about semantics and psychology.

I've found that most stage lighting is higher in frequency than full wave rectified mains frequency, but, sadly, not nearly as high as lights made for photography.

Yes, most of the sad samples that people show in forums when they get luminance banding are from churches, rental halls, etc, and not so much from proper "stages".

Cost-cutting.
On a side note, recently I was shooting in a venue where they were running PWM at 25 KHz, and no inductor. EMC was off the charts, my 915 MHz remote wasn't working.

I think there must have been an inductor in there somewhere, or the power dissipation would be a bit mental.

I suspect that FCC class B compliance was not quite acheived...

This can happen with quite fancy lighting. Here's a DPReview thread about a Canon R5 picking up interference on its (unbalanced) audio input from a Nanlite Forza 500 LED light: www.dpreview.com/forums/thread/4671099

It seems that the Forza 500 generates a lot of RF crud, and the R5 has poor immunity on its audio input.

Here's an example from an - apparently - "proper" stage: www.dpreview.com/forums/thread/4699677

IIUC (maybe you have a better idea), the light is being PWM'ed at about 2.2kHz. And the PWM frequency is not very stable.