From: Kevin McMurtrie
Subject: Re: linear power supply noise questions
User-Agent: MT-NewsWatcher/3.2 (PPC Mac OS X)
Date: Wed, 13 Nov 2002 03:19:04 GMT
NNTP-Posting-Date: Tue, 12 Nov 2002 19:19:04 PST
In article ,
"Walter Harley" wrote:
>"Kevin McMurtrie" wrote in message
>> What you saw was only half the problem. The power carrying lines are
>> modulated by the amplifier's own load, regardless of how good your
>> regulator is. It happens in the wires. It results in the pull-up and
>> pull-down phases having different levels of amplification, or
>> oscillating badly.
>Thanks, Kevin. I don't quite entirely understand this, though; maybe you
>can help me. I think what you're saying is that the supply line looks like
>a possibly-ideal voltage source, plus some finite source impedance. The
>power connection of the opamp, and any signals I refer to the power rail,
>connect to the far side of that impedance. So, any current that flows out
>the opamp's output causes a voltage drop on the supply rail, and therefore a
>corresponding induced voltage at the opamp input, in my case where I am
>referencing a feedback divider to the supply rail rather than to ground.
>But, it seems to me that induced voltage should show up only as a tiny
>(compared to the much higher impedance of my feedback divider) error term; I
>don't see why it would cause oscillation, since it's no different in nature
>(that is, it's not appreciably phase-shifted) from my feedback signal and
>it's much smaller. I do see why it could introduce distortion, because the
>error would exist on only one half the output cycle, but if my feedback
>network is on the order of 10k and my supply source impedance is on the
>order of 10 milliohms, that distortion should be on the approximate order of
>1ppm, right? Just trying to see if I understand.
The important thing is to have _one_ point being a signal ground. Pick
one spot and connect all the signal wires to it. It's OK for ripple to
appear on the signal ground but it must appear on all signal grounds at
once. What you don't want to do is place loads on your signal ground.
Line noise gets into circuits in ways that aren't obvious. It can even
happen outside the amplifier, say on the input or output signal grounds.