5 switching regulators on the same PCB

[Frozenguy]

I am designing a PCB that will distribute power to various loads. I have thus far designed it to have five switching regulators. It has come to my attention I should worry about synchronization of the switching pulses. To that I suggested I design each power supply with a different frequency. I was told it seems like an ok approach off hand, but to look into related harmonics. Such as if the fifth or sixth harmonics are related between the various regulators. All regulators are identical part number, and the rest of each power circuit will be identical too except for the switching frequencies and current limit resistors.

How can I start to look into the harmonics of a particular buck regulator or design incorporating one. I am in the design phase.

Thanks for any help and insight.

 

[dknguyen]

Different frequencies won't get you around beat harmonics. Just different ones. Think about it...the reason beats happen in the first place is because even two regulators of the "same" frequency aren't actually the same frequency. They're different. So using choosing to use different frequencies doesn't get around the issue. Not only that, they drift over time and temperature so you even if you hand-selected them to be the exact same frequency they won't stay that way relative to each other.

It's a concern I guess...but frankly it can be really tough to find switching regulator ICs that both meet your requirements and have frequency synchronization. Lots of circuits make do without it. Frequency synchronization is not something that is easily done unless it's already built into the IC (at least I cannot think of a way without it already being accommodated by the designer of the regulator IC). I assume you're asking this question because your chosen regulator does not already have the functionality included otherwise it's kind of a no-brainer.

I would personally start by picking a regulator that already allows for it, and if not, just make accommodations for the appropriate filtering and go without the synch.

Are these regulators in series or parallel? And what kind of input/output voltages and currents are we talking about? What regulator is it?

What are the loads? Sensitive analog? Really high speed digital?

 

[Frozenguy]

They each have their own distinct loads but share common input and ground. Each regulator will be outputting about 6-8 amp at 24 V. Input is about 50 V. I was told there may not be a way around it as of now, but to look into it. May be something we deal with if it shows up later.

 

[dknguyen]

What type of loads?

And what is the specific regulator you have in mind?

 

[Frozenguy]

Vishay SIC461
Mainly resistive load but about 10% is inductive. I can look into those numbers. Might be a bit more inductive than I'm stating off hand.

 

[dknguyen]

Vishay SIC461
Mainly resistive load but about 10% is inductive. I can look into those numbers. Might be a bit more inductive than I'm stating off hand.

That's not what I mean. I mean are they analog? Digital? sensitive analog? high speed digital? Are they sensitive to noise or ripple at all? Because I don't think you need to give a crap about beat frequencies if they are insensitive to noise or ripple.

 

[Frozenguy]

Much of it is rather insensitive to noise but a few things like the ethernet switches and safety PLC may be a bit sensitive to noisy power input.

 

[dknguyen]

Much of it is rather insensitive to noise but a few things like the ethernet switches and safety PLC may be a bit sensitive to noisy power input.

Don't those types of devices have their own point-of-load regulators inside them? I would just use additional filtering in that case. If they're far away from the actual PCB then the wire inductance will also help filter the noise.

I also missed you saying that your regulator was the Vishay SIC461. No built-in frequency synch...that's going to be a PITA to make it synch. It might even be impossible. I don't see a way to access the clock.

 

[AnalogKid]

The part you have chosen has a constant on time, so it is more like pulse position modulation than pulse width modulation. As such, it does not lend itself to external frequency synchronization. Also, the noise signature varies as a function of load current, a particularly nasty filtering problem. Vicor DC/DC converters have a similar "feature".

ak

 

[ronsimpson]

Thanks for any help and insight.

I have used 5 PWM on the same PCB.
Here is an idea: There are two places where the 5V and the 3.3V regulator can talk to each other. Ground and +24V.
1) I show heavy traces in the grounds of each PWM. Then this ground connects to the ground plane in one place. Ground currents from cap to PWM do not flow in the ground plane.
2) Use a CLC filter to brake up the 24V into little pieces. Current in the PWM needs to come from the capacitor at the PWM and not from the 24V.