Power supply noise and ripple

[knudsenuk]

Hi All

I wondered what views people have on measuring PSU noise and ripple. It seems to me that the only way to get accurate measurements (via a scope) is to not use the ground wire and either resort to a spring or the body of the probe (as ground). This sort of implies that you need to factor in this method of measurement into your PCB design. I spend ages trying to get a "real" reading, often made so much harder because the chance of being able to employ either of the methods I just mentioned rely on a ground contact being conveniently placed near the test point. And as we all know, the second you look at the screen, the probe moves or something 

It also made me question the validity of measurements made in such a way. I mean what does the actual circuit you are powering see?

Does anyone have any suggestions to make this process less painful? I am focusing on linear PSUs around the few mV neighborhood. Making improvements seems to be limited by my measurement techniques.

[dmills]

50 Ohm scope input, length of small coax with a BNC on one end and a small as possible 450 Ohm resistor soldered to the other...
Boom, 10:1 probe with bandwidth for days.

Scrape a bit of the solder resist off the ground plane next to the point you wish to measure, solder cable outer to ground plane, solder end of 450 ohm resistor to point you want to inspect.

For linear supplies where I am looking in the 1/f to a kHz region (Above that the action is usually on the load board) I often use an audio analyser or mic preamp to let me easily see it as noise density or in the frequency domain, usually far more informative.

[knudsenuk]

Interesting. Thanks I appreciate the reply. I will try this, although I never use a ground plane

[penfold]

[...]
It also made me question the validity of measurements made in such a way. I mean what does the actual circuit you are powering see?
[...]

Exactly that point, the ripple+noise spec of, say, a COTS PSU doesn't really make any guarantees about what you'll see at the load. From an 'evaluating the circuit' perspective, consistency and repeatability of measurement (I think) should be the top concern; there are so many ways of making the noise better or worse, can easily end up chasing your tail if the measurement is varying also.

I normally add a footprint for an SMA connector, route some lines to the board edge for an end-launch SMA or a PCB coax test point (Tektronix make them, at least for their scope probe sizes). For probes, over the years I've made up various configurations of attenuators inside old matching-pad and attenuator shells: DC blockers, ones with low-pass characteristics to remove some of the fuzz when looking for switching-frequency and first few harmonics ripple or transient artifacts. I'd often use them with a preamp... mine isn't exactly the best solution, it's a Mini-circuits ZFL-500LN+ (or similar, I forget what's in the box), very important to use a limiting diode with and TBH I've kinda painted myself into a corner making things to adapt what I had available when a high input impedance preamp and a scope probe would be better for a lot of cases.

A useful thing with ripple measurement can also be a signal isolating transformer so you can synchronize the scope trigger with the switching waveform without adding additional scope grounds to the circuit, used with some waveform averaging can show up quite a lot about resonances and EMC culprits and scrubs out anything that isn't linked with the switching waveform.