How important is synchronizing?

[gearheadgene]

Say there are 2 or more switching supplies in a system. How important is it to synchronize them? Wouldn't it be better to have them running at different frequencies, not harmonically related, so that any transients produced during switching are spread out? It'll make the input look noisier, but the amplitude is otherwise smaller (sort of akin to spread spectrum).

I'm fairly certain that it's a good idea to synchronize devices running at very close frequencies to one another, however.

thanks

gene

[gearheadgene]

I don't know which crickets are louder - the ones out my door or in this thread

[fernando_g]

The problem is second order beat frequencies.

which mathematically can be expressed as follows:

fso = fa +/- fb +/- fc

Where fa is you music (which is spread spectrum), and fb and fc are the power supplies' switching frequencies.

If you are lucky, the beat frequencies will be out of audible range. But by Murphy's law, there could be instances that you will not be as lucky and then they will be within the audible range.

[Mooly]

Someones been doing some thinking beat frequencies in S.M.P.S.U.'s.
You mention spectrum spread. The chopper drive in an individual PSU can have this "spectrum spreading" applied to it as well to make the switching current appear more noise like.
I believe there were even some dedicated control I/C's for this very purpose, just don't ask me to remember the number
If I do turn anything up I will post it here.

[TechGuy]

Say there are 2 or more switching supplies in a system. How important is it to synchronize them? Wouldn't it be better to have them running at different frequencies, not harmonically related, so that any transients produced during switching are spread out? It'll make the input look noisier, but the amplitude is otherwise smaller (sort of akin to spread spectrum).

If they are sharing the same load, it may causes instablities between the feedback circuits, which could create excessive noise.

[gearheadgene]

That's pretty the answer that I was expecting - beat frequencies. It looks like I will have, at minimum, 2 switchers running that generate +/-15 (one single sepic) and 3.3 (another sepic, likely). I'm working on an 8-ch class-D amp and worry that the simultaneous audio switching will be a problem, add to it the switching of the smps. The notion of using 8 different switching frequencies for the class-D may be interesting. Or maybe using the same frequency, just phase shifted in 8 equal parts (or 10 or more, depending on the number of smps running). That would alleviate the beat frequency problem, and also keep simultaneous switching to a minimum.

[nonoise]

Phase shifting is the way to go, you could use a small CPLD to generate all the required clocks, fed from a master clock which also defines the class d amp switching Fc.
The resultant input ripple current on the SMPS will be much lower.
We used similar trickery at work and it made a big difference.

[star882]

Three flip-flops can be used to stagger the clocks for 2 phase operation. Use one flip-flop to drive the others from each output. Replace the first flip-flop with a sequence counter with n outputs and add flip-flops as needed for n phases.

[switchmodepower]

Originally posted by TechGuy


If they are sharing the same load, it may causes instablities between the feedback circuits, which could create excessive noise.

Can you elaborate? How does it cause instabilities?