Avoiding camera flicker with PWM

[tom66]

I'm working on an art installation with 50 x 5W white LEDs which are all PWM controlled. Currently we are running at 200Hz PWM but we are flexible from around 100Hz to 5kHz. (Though at higher powers our MOSFET switches may dissipate too much power in the switching region, so the upper bound might be reduced.) This installation will probably be filmed by TV broadcast cameras.

Each LED is out of PWM phase with the other LEDs (by about 3.6 degrees each), so in theory, ripple current will be low.

It will be filmed on 576i or 1080i professional broadcast cameras, running at 50Hz, but we cannot line lock the cameras to our display. We also will probably not have any control over exposure settings etc.

One suggestion was for us to run the PWM 180deg out of phase at 225Hz but I suspect this would simply lead to beat frequencies.

What is a "safe" frequency for a video camera? Anyone have any experience with this?

[mikeselectricstuff]

I'm working on an art installation with 50 x 5W white LEDs which are all PWM controlled. Currently we are running at 200Hz PWM but we are flexible from around 100Hz to 5kHz. (Though at higher powers our MOSFET switches may dissipate too much power in the switching region, so the upper bound might be reduced.) This installation will probably be filmed by TV broadcast cameras.

Each LED is out of PWM phase with the other LEDs (by about 3.6 degrees each), so in theory, ripple current will be low.

It will be filmed on 576i or 1080i professional broadcast cameras, running at 50Hz, but we cannot line lock the cameras to our display. We also will probably not have any control over exposure settings etc.

One suggestion was for us to run the PWM 180deg out of phase at 225Hz but I suspect this would simply lead to beat frequencies.

What is a "safe" frequency for a video camera? Anyone have any experience with this?

Depends a lot on the camera, shutter time etc. Only way to be sure is test. It will probably help to have any LED modulation be an exact multiple of the 50Hz, so assuming the camera's shutter-open time is 50Hz (which you may be able to ensure by controlling the aperture) , it will see the same amount of light in each frame.
I usually find 250Hz to be a good compromise between flicker and switching losses & acoustic noise.

[NiHaoMike]

MOSFETs have very low losses into the few 10s of kHz. Additionally, if you're PWMing LEDs off a fixed supply with a current limiting resistor, linear would be just as efficient.

[tom66]

MOSFETs have very low losses into the few 10s of kHz. Additionally, if you're PWMing LEDs off a fixed supply with a current limiting resistor, linear would be just as efficient.

We're only driving them with weak GPIO drives - about 25mA per channel (specficially a PCA9685 for each 16 channels) so there is a potential concern that we could not turn the gates on and off fast enough. We are using PWM because each LED needs individual intensity control.

[mikeselectricstuff]

MOSFETs have very low losses into the few 10s of kHz. Additionally, if you're PWMing LEDs off a fixed supply with a current limiting resistor, linear would be just as efficient.

Except that when PWMing LEDs or any other power load, you generally want to have a slightly soft turn-on to reduce noise. I often put a 10K resistor in series with the gate to soften up the switching.
Something else that may help is to stagger the PWM phases between channels - PCA9635 allows this as you can specify start and stop times, not sure about the 9685.
 

[tom66]

Yep, our PWM channels are staggered.

[richard.cs]

It will be filmed on 576i or 1080i professional broadcast cameras, running at 50Hz, but we cannot line lock the cameras to our display. We also will probably not have any control over exposure settings etc.

Really? I would have thought the ability to line lock would be essential to any proffessional camera, it seems odd that you can't. Do you have any chance to lock it the other way and sync the lights to the camera rather than the camera to the lights?

[tom66]

It will be filmed on 576i or 1080i professional broadcast cameras, running at 50Hz, but we cannot line lock the cameras to our display. We also will probably not have any control over exposure settings etc.

Really? I would have thought the ability to line lock would be essential to any proffessional camera, it seems odd that you can't. Do you have any chance to lock it the other way and sync the lights to the camera rather than the camera to the lights?

It's for a news broadcast. I am guessing they will not have this capability.  I'd rather design it to not require it than to find out they can't do it...

[mikeselectricstuff]

I would have thought that any camera designed for this sort of use will  have some internal cleverness to avoid flicker.
You don't generally see flicker on news broadcasts where there is fluorescent lighting, so maybe it's actually integrating multiple frames - it probably needs to do this anyway to make the output look smooth & avoid the jerkiness that you see with faster shutter times.

[GK]

I don't have any experience with this kind of thing, but is the raw PWM being delivered directly to the LED lights? I could imagine that being a potential RFI nightmare. Couldn't you just use a variable buck regulator to deliver a smoothed DC to the lights? The switching frequency could be 10's of kHz and with additional stages of (post feedback) LC filtering I'm sure the ripple could be easily reduced to negligible levels. Make it a synchronous buck with voltage mode control (as in a class D amplifier) and the DC output will be fully and stably variable right down to zero volts and zero load current.
 

[tom66]

All been designed in already. With about 2 weeks to go it's not going to be changing. Also, each channel is fully & independently variable from 0 to 100%... so regulation of each would be a little too complicated.

As Mike has talked about on the Amp Hour, considerations such as RFI/EMI rarely come into consideration for short installations (in this case about 50 days) where you may not strictly meet standards. As long as it doesn't interfere with anything used in the nearby area, and it doesn't burst into flames, we should be fine.

Each LED is phase shifted (PCA9685 supports this) so our ripple current is fairly low - should not exceed that of one LED's current, about 350mA. Also, the channels will be soft-switched in a way, because of the weak GPIO drive.  We can increase this delay by putting a resistor between the source of the MOSFETs and our circuit & PCA9685 ground, if it turns out to be a problem. The LED current is kept in a separate loop to our main logic power and tied a single star point.

(It's a University project anyway - I've somehow got roped into doing the custom electronics & firmware.)