From: email@example.com (Tom Bruhns)
Subject: Re: linear power supply noise questions
Date: 12 Nov 2002 17:18:40 -0800
NNTP-Posting-Date: 13 Nov 2002 01:18:40 GMT
"Walter Harley" wrote in message news:...
> I have three questions:
> First, I'm using an LM337T as the negative regulator. It's bypassed at
> input with a 0.1uF ceramic, and at output with a small tantalum (don't
> remember value offhand). Am I right that the next thing I should do to
> reduce the supply noise is bypass the ADJ pin, presumably with a ceramic or
> polypropylene cap? Or would I be better off putting a poly or ceramic cap
> across the output (in parallel with the tantalum), or should I do both?
The best way, if you want _really_ low noise, is to use a circuit like
shown at http://www.wenzel.com/documents/finesse.html. Once you
understand the technique, you can see how to do it with an op amp, and
do it for negative as well as positive voltages. I just did it for a
-5V supply with very good results. It's possible to get power supply
output noise almost as low as the input-referred noise of the op amp
you use, if you are careful.
> Second, is this problem typical, and is that why these circuits usually show
> a cap to ground instead of a supply rail? It seems like it would be better
> engineering to use a cheaper, more widely available part in place of the
> nonpolar cap, but not if it means increased noise.
Oh, yes, power supply rails are never to be considered clean, unless
you've gone to pretty extreme measures to clean them. Op amps
generally have "power supply rejection ratios" listed in the data
sheets...they decouple to some extent from their power supplies.
Modern op amps have pretty low input offset drifts, and you can get
ones that have pretty low input offset voltages as well, so it's
generally not a terrible problem. What I'd recommend for your circuit
is to divide it into two gain stages, and just DC couple the first
one; let the second one take care of the DC offsets. Divide the gain
between them, and you'll end up with a much better bandwidth. An op
amp with 3MHz gain bandwidth product, run at G=100, will have about a
30kHz -3dB corner frequency...not too bad at first blush, but
typically it will have relatively bad distortion at the higher audio
frequencies, and will have noticable rolloff at the upper end of the
audio band. But a couple of G=10 stages cascaded will have 300kHz
bandwidth each, and much lower distortion.
> Third, my Hitachi V212 oscilloscope's max calibrated sensitivity is 5mV per
> division; with a x10 probe, that's 50mV per division, which means I can
> barely see the 0.5mV supply noise (instead, I discovered it with an Audio
> Precision analyzer). Do people generally use scopes with higher
> sensitivity, or outboard preamplifiers, or is a scope just not the right
> tool, or should I get my eyes checked?
I use spectral analysis equipment...a "Dynamic Signal Analyzer" or
"Vector Signal Analyzer." I suppose the Audio Precision analyzer you
used has similar capabilities. But those instruments aren't
necessarily very available to the casual hobbyist. On the other hand,
you likely have a computer with a sound card, and if you add an
external preamp and perhaps a way to get good common mode rejection,
you can pull audio-range signals into your computer and look at them
there. You can find inexpensive software...maybe even freeware...that
lets you do an FFT and display a spectrum, or perhaps show 1/3rd
octave bands. That can tell you a LOT about where to concentrate your
efforts. Plus, it can be very sensitive...a key advantage is that you
can "see" a signal buried in noise that you'd completely miss on your