[time-nuts] potential low-RFI power supply

[ed breya]

Trying to look at the junk on the output voltage won't tell the story 
you need. Just connect a scope (1 meg) input to the DC output ground, 
and you'll begin to see the common-mode current situation. The actual LF 
and HF ripple on the DC out can be dealt with by sufficient filtering or 
post-regulation, and hardly matter compared to the common-mode currents, 
which will pervade everything connected.

First, put only a scope probe tip or direct input connection on the DC 
out ground or plus - shouldn't matter too much either way if the output 
floats. Do not connect the scope's ground to the PS yet - let its ground 
return to earth via its mains cord. This will likely indicate a very 
large signal, including the noises of the power supply and the scope and 
other local ground loop sources, against the high measuring impedance. 
This is the gross (in size and disgusting-ness) available CM driving 
voltage in your measurement setup, including line leakage, especially if 
the PS is fed by a two-wire cord.

With this setup, you should be able to see a pretty good representation 
of the actual signal on the main switching device in the PS. This is 
usually by far the biggest, fastest signal inside, and the cause of most 
EMC grief. Set up the scope triggering for LF Reject, and speed up the 
sweep to observe where the SMPS runs, say 20 kHz to maybe 300 kHz for 
newer type products, then play with the trigger level and hold-off. 
There should be plenty of crap available to trigger on - the trick is to 
get the right part. A dual timebase may help, but shouldn't be necessary.

You'll be able to tell if the PS runs in burst mode at no load, 
evidenced by chirping at low frequency, on the scope and maybe audibly. 
Now add some load, say 10 percent, and it will likely leave burst mode, 
and go to regular PWM at the proper frequency, which you now can 
measure. This frequency is the fundamental, for figuring out 
interference issues.

With a small load, depending on the topology, the PS may be running in 
discontinuous mode, evident on the switch waveform when the device shuts 
off, and a large ringing voltage appears, higher than the fundamental 
frequency. This represents energy in the tank circuit - transformer L 
and device and parasitic C all over the place. The resonant frequency 
and decay can change dramatically with conditions, and is not dependent 
on the fundamental, but is triggered by it, so imagine all the mixing 
products involved. This, plus line harmonics mixed in, tends to make 
what appears to be broadband noise, but a lot of it is actually finely 
spaced spurs that may move around with different loads and other 
conditions - what a mess..

As you increase the load, the PS may transition to continuous mode, and 
the big rings will disappear, since current through the energy storage 
inductance never goes to zero. Always present, are the very fast, 
smaller rings on all the sharp edges, due to leakage inductances and 
other device parasitics. These make up the very high frequency harmonic 
content, and easily escape into the environment.

As you can imagine, the EMC/spectral content can change a lot under 
different conditions. It's good to know as much as you can about what 
it's doing inside. Then it may be possible to avoid certain problems by 
operating in conditions where it's least offensive to your situation.

It's pretty slick to be able to see somewhat inside a black box, if you 
know what to look for. There's no guarantee this will work in any 
particular situation, but I'd say there's a very good chance of getting 
lots of details - or at least to see a lot of ugly waveforms that won't 
trigger properly. If you hardly see anything, then that means the device 
and you ambient environment are extremely clean - or more likely that 
your setup or scope have a problem.

The next step is to see some of your measurement limits. Now connect the 
scope ground (via BNC outer) or probe's ground clip to the signal point, 
which ideally results in no apparent signal, even with the sensitivity 
cranked way up. If this happens, then you're in great shape, 
ground-loop-wise. However, I'd guess or predict that you'll see plenty, 
because there is no true ground. The minimum signal is with the input 
shorted, and no connection to anything. When you hook it up, everything 
changes. Remember also, that a lot of noise will be around that's not 
from the DUT, so you need to determine what's what and where.

There are more measurements that can be made, of course, especially when 
you can think inside as well as outside the box. I'll have a bit more to 
say next time.

Ed