Help needed with multiple DC DC converters and noise spikes.

[homeward]

Hi All,

I have a wall mounted AC-DC converter feeding 3 DC-DC converters in parallel to give 3.3V, 5V and +/-15V power rails. The converters are switching at 294kHz, 495kHz and 347kHz respectively. Looking at the FFT of this input bus I'm seeing a large spike at 27kHz and its multiples and I just can't seem to get rid of it, or even reduce/increase it in anyway.

I assumed this was a beat frequency but none of those switching frequencies combine to give 27kHz, although the 294kHz and 347kHz give a beat frequency of 53kHz, which is (pretty much) the 2nd harmonic of 27kHz. This seems relevant?

Bypassing the inductor does nothing, nor does changing the value from 47uH to 10uH or indeed 220uH. Bypassing the 4.7r does nothing. I've added ferrites inplace of the fuses and added 10uF on the input pins of each converter but nothing seems to help.

Typing this out I realise the problem is maybe radiated rather conducted if changing this stuff does nothing, I'll post this anyway in the hopes someone can add something useful please. Schematic and FFT plot below.
 
Thanks.

[homeward]

Ok well probably not radiated, put the regulators on wires a few inches long rather than the PCB and that doesn't change anything either.   

[Bafflegab]

What happens if you only connect one of the converters at a time? Other permutations?

[moffy]

It could be noise through the ground connection of your scope. Any way that you could substitute a battery for the plug pack?

[Doctorandus_P]

PCB layout is extremely important for SMPS circuits. Do some research into "hot loop" for more info.

Also: KiCad does have normally closed (bridged) solder jumpers, both for schematic symbols and for PCB footprints.

[homeward]

What happens if you only connect one of the converters at a time? Other permutations?

Yes, but it turns out this was a red herring.

Tracked the issue down to the PSU! For the benefit of anyone reading in the future with the same problem: When I measured just the PSU it was quiet but that was with no load, when a dummy load was added then sure enough these spikes appeared. Changing the resistance on the load changes the frequencies of the spikes, which is why removing a regulator (as above) confused me. It's obvious in hindsight, as always.

Thanks for the help.

Also: KiCad does have normally closed (bridged) solder jumpers, both for schematic symbols and for PCB footprints.

Thanks, there was a reason I did it this way though I can't quite remember why, I think I couldn't find any 0805 nc footprints? And then making my own threw up DRC errors.

[Houseman]

Tracked the issue down to the PSU!

So, glad You found out. What did you do for solving on the PSU side?
Have you solved the issue?
BTW why did you put two resistors in seriers with the two caps of the input PI filter?
Regards

[homeward]

Tracked the issue down to the PSU!

So, glad You found out. What did you do for solving on the PSU side?
Have you solved the issue?
BTW why did you put two resistors in seriers with the two caps of the input PI filter?
Regards

I'm just kind of leaving it and hoping for the best really. The pi filter does a good job of preventing the spikes from reaching the regulators and the PSU is FCC/CE certified so is presumably ok from a regulatory perspective.

The resistors are there to damp the filter, it's very resonant otherwise. I'd advise anyone not to copy me though as I don't really know what I'm doing.