What is this noise?

[helix1]

I have a simple unregulated power supply, with a Schaffner FN9222R-1-06 Mains Filter going into a Myrra Transformer 45065, 2x115V primary, 2x18V secondary, 30VA. The secondary coils have a full rectifier and smoothing caps of 4700uF. It is set up as a split supply with +/- 18V output. With a 0.5 load applied between +18 and 0, or 0 and -18, in addition to the usual ripple I get a high frequency signal as shown in the attached scope shots. Can anyone tell me what the source of this noise is? It's about 1V pp which seems quite large. Is there any easy way to get rid of it?  This output will be the source for a more regulated supply, so would it be worth putting a low pass filter on the output to reduce it now, or catch it again further down the line?

[c4757p]

What is the source of the 0.5 (amp, I assume? don't omit units) load? And are you sure that load is stable?

[helix1]

Yes it's 0.5A. I'm using a dummy load which is based on the one Dave discussed in his blog.

[c4757p]

Try a resistor instead. Dave's circuit is horrible, he made no attempt to compensate the control loop and so it oscillates like a mofo. I bet the load is oscillating.

[helix1]

Ok will do! Can you recommend a good circuit to create an adjustable dummy load that is stable and won't introduce noise like this?

[dom0]

4700 µF at 18 V for a 15 VA winding is a lot for the transformer. Maybe ~5 % ripple or so. Especially if you intend to run it at full current (~.5 A).

[rdl]

Ok will do! Can you recommend a good circuit to create an adjustable dummy load that is stable and won't introduce noise like this?

Since you have an oscilloscope, you can probably just fix what you have. Here are my (not detailed enough) notes from fixing mine. My circuit was built on perfboard and wasn't the best layout, so you will surely need different component values, but maybe this will help. Trimmer resistors and a cap substitution box make things easier.

[helix1]

Thanks @rdl, that looks helpful. Is it just a case of trial and error, or are there any rules of thumb behind your choices of components for stability?

[rdl]

There are a couple of other threads here which discuss Dave's circuit. That's where I got the starting point, after that it was mainly just experimentation.

[c4757p]

It really is pretty trial-and-error - there's of course tons of mathematical practice around stabilizing systems (just google "control theory"), but it's pretty hard to make things mathematically guaranteed stable when they're test equipment like a load and you don't know what kinds of things they'll be connected to.

[helix1]

I will certainly give this a go, thank you!

It brings up another beginner's question though... is it possible to reduce a power supply's vulnerability to oscillations from its load?

[dom0]

Low output impedance, good phase margin, possibly additional impedance compensation if you have an idea about the load (e.g. capacitive Zobel networks on the outputs of audio power amplifiers - because the load -speakers- are inductive).

Some people prefer their systems "meta-stable": High output impedances (e.g. high side PNP), uncompensated control loops, shunted by tons and tons of capacitance to ground that attenuates all oscillations and regulation of the "regulator" to levels that are "below interest". A typical example of this "design pattern" would be so-called "rail splitter" circuits where a TL072 or some other unsuited op-amp directly drives a virtual ground with 10+ µF connected across it.