What does preamp PSU oscillation look like for linear regulators?

[shredhead]

Sup Y'all, I am wondering what PSU oscillation looks like on an oscilloscope. Specifically I'm trying to learn what it looks like for a linear regulator (+/-15V) under load. Does anyone have any pics of their scope capturing PSU oscillation?

I have an analog scope. Will the oscillations always be steady enough to trigger?

If you have pics of other power supplies oscillating like unregulated power amp's supplies, please post them as well.

[jitter]

I have some nice captures made with an Agilent (now Keysight) DSO-X 3054A at work of an oscillating LM337.
The oscillations were caused by too low ESR of the capacitors on the output of the LM337. LDOs (like LM317/337s) don't like that.

These oscillatons were strong and pretty steady and any scope should trigger them without any problem. I'll post one capture soon.

Edit: unregulated PSU's don't oscillate as they don't have any feedback. Did you mean ripple voltage?

[rayma]

Originally Posted by shredhead

I am wondering what PSU oscillation looks like on an oscilloscope.
Specifically I'm trying to learn what it looks like for a linear regulator (+/-15V) under load.

If you use a large enough cap, there's no issue.

[shredhead]

Originally Posted by jitter

I'll post one capture soon.

unregulated PSU's don't oscillate as they don't have any feedback. Did you mean ripple voltage?

Awesome, thanks jitter. That is a serious scope you have there.

Thanks for pointing that out about the feedback causing oscillations, I didn't think of that. Nevermind on the pics of unregulated then.

[anatech]

Hi shredhead,
It looks like a higher frequency signal riding on the DC supply voltage, which may or may not be the correct level depending on the waveform. If the level of oscillation isn't high enough to trigger with, the trace will look fatter than normal. Noise can also make a trace fatter, but whichever the cause, this needs to be fixed.

If you use a large enough cap, there's no issue.

No, not true. Larger caps can have higher inductance and will always create a reduced conduction angle which in turn causes radiated noise and bigger current spikes. Sometimes you are better off with a 120 Hz ripple that can be cleaned much more easily than the mess caused by large capacitance (worse with fast diodes).

Increasing supply capacitance has been in vogue for a few years now. Add that to fast diodes (now soft recovery diodes to fix that problem), and you set yourself up for problems. Notice that the "soft recovery diodes" came about after the dirty little secret came out that is reduced conduction angle and sharp current spikes. The real fix? Use the properly sized capacitor and normal diodes that will handle the current (+ peak current rating!).

[shredhead]

Originally Posted by anatech

The real fix? Use the properly sized capacitor and normal diodes that will handle the current (+ peak current rating!).

Do you have any suggestions of what to look at to properly size PSU ripple caps, regulator output caps and local PCB decoupling caps for a project?

Is there a worst case scenario of capacitance for demanding source material multiplied by how many op amps you are powering kind of formula or something like that?

[Elvee]

Originally Posted by shredhead

Does anyone have any pics of their scope capturing PSU oscillation?

I could find/make some, but that wouldn't help, quite the opposite in fact: parasitic oscillations can appear in so many shapes and guises that showing a few "typical" examples could in fact mislead you into thinking your circuit is OK, when in fact it isn't.

I have an analog scope. Will the oscillations always be steady enough to trigger?

No, but that isn't important: you are expecting to see pure DC. If you chose AC coupling and maximum sensitivity, the trace should stay perfectly flat, without any thickening or any other artifact.
Sometimes, you will see a clear phenomenon, like sawtooth oscillation or similar, but very often it will be a blurred or thickened trace., because the oscillation is not coherent, or too weak, or too high in frequency to be displayable, but it doesn't actually matter: you just look for signs of presence.
With the max sensitivity, the oscilloscope might give some thickening of the trace due to internal noise.
To make sure the noise you see is actually caused by the PSU you examine, just remove the mains power leaving the rest unchanged: if the artifact disappears, you can be sure the PSU is the culprit.
Note that an analog scope is ideal for this kind of job, because you don't have to worry about the sample rate/vs horizontal settings, sync, etc: just connect the probe and see if there are any visible changes, that's all

[yu3ma]

@shredhead

Take an look on this very nice document:
http://cds.linear.com/docs/en/applic...ote/an104f.pdf

Document explain how to test your PSU for transient response which is crucial to get picture about control circuit (regulator + FB).
In many cases, wrongly implemented FB and it's compensation lead to poor phase margin which consequently lead to oscillations on regulator's output.

Read whole document with attention and you will get an picture what you need to do.

[jackinnj]

In addition to what Chris points out, you can hear oscillation with an inexpensive transistor radio, or the error amplifier will get hot and possibly fail.

With Walt Jung's regulators on Jan's original "Old Colony" boards and the ADA797 as error amplifier you could hear the signal in the a.m. band.

You can use Fred Dieckman's regulator tester to examine PSRR and Zout. If you can find Zout, you can find phase/gain relationship:

[noSmoking]

hi,
Another good read and build to test snubbers and PS is here
http://www.diyaudio.com/forums/power...-test-jig.html
NS