One of my Young Engineers (actually, relatively, they are all Young Engineers) designed a neat little logic fanout box, full of ecl and gaasfets and stuff. It runs on a +12 wall-wart and has a bunch of switchers inside to make the myriad + and - voltages that things like this use.
Two of the switchers use a chip new to us, National LM3102, a nice buck synchronous switcher, to step the +12 down to +8 and +2.5.
Problem: opamps clear on the other side of the board are showing huge amounts of spike noise, correlated with the switcher frequencies. The opamp layouts are very tight and no switcher stuff is anywhere close.
YE came to me for an opinion, and we played with it some, and it made no sense. After thinking it over, I suggested that the fets in the switchers were RF oscillating during their transitions, and the RF was being rectified by the LM7301 bipolar opamps.
That idea was weird, but not weird enough. This is what's actually happening at the switcher output node, just before the output inductor:
ftp://66.117.156.8/SwitcherRise.JPG
Note the roughly 1.5 ns rise time, and the 150 MHz ringing. That's equivalent to an RF burst plenty bad enough to get into the opamp front ends.
OK, here's today's puzzler:
- What's happening?
- How to fix it?
John