Syncronizing SMPS and EMC Noise

Normally not. The amplitude doubles if you sync two converters and they make about the same racket. The trick would be to sync them with the proper phase offset so they cancel each other's noise at least to some extent.

The usual strategy is much more sinister: Don't sync anything but wiggle the current into the timing resistor pin. That is the equivalent of sweeping something under the rug, hoping nobody saw that :-).

I have only sync'd stuff where it was functionally advantageous, for example to avoid internal noise or to fold the noise back to zero Hertz in the baseband.

Well, imagine a 100kHz converter and a 100.1kHz converter with equal noise in the same system, un-synchronized

Every 10ms it will coinside, so the phase noise will add up and that will be detected by the peak measurement, right?

Cheers

Klaus

We've FM swept switchers to keep the quasi-peak EMI down. A triangle wave from a simple Schmitt oscillator, squirted into the RC timing pin, works great.

Looks ugly, but the DC looked fine.

I do that all the time. Sometimes with the same voltage rail (multi-phase). ;-)

If no one saw it, it din't happen. ;-)

"I didn't do it. Nobody saw me. You can't prove a thing."

We do it all the time to avoid what we're trying to receive. I do it sometimes to minimize EMI in general.

Sometimes, unsync'ing PWM transition, just introducing a wee deadtime reduces CM noise emission significantly

Cheers

Klaus

Not sure I understand. Are you talking about the dead time between the upper switch and the lower (synchronous diode) in a synchronous switcher? This is a different thing but yes, shoot-through can be a real EMI (as well as efficiency) problem. In larger regulators, this is commonly a knob that can be turned. Not so much for regulators with integrated switches.

I am talking about 3 phase inverters, where coupling capacitance in the motor causes CM noise. If the PWMs are misaligned just slightly, the CM noise is reduced (quasi peak)

Regards

Klaus

Gotcha. Wasn't considering inverters.

The quasi-peak detector will catch that, it has a 1msec attack time and hundreds of msec decay. But in many civilian cases nothing under 150kHz is measured so you might get away with it.

Better would be to try to sync them up and have them counter-act each other.

Must be the reason for reactors being used with inverters, at least one of the reasons :)

Any way I would think a high frequency reference common to all

Jamie