Injection-Locked Oscillators

Jan 28, 2015 23 Replies
J

We discovered the problem when we were calibrating the 8-channel VME version of the ARB. We'd cal it one day, and the DC offset would be off the next day. They asked me to look at it and naturally I poked fingers on/near the opamps, and the offset changed. Next step was the spectrum analyzer. All 8 channels are oscillating together at around

1.3 GHz, but not injection locked like on the 4-channel box. Touching channel 5 kills the oscillation on ch 0, etc. Like a xylophone.

So we checked the 4-chan box, and it's doing about the same thing, but locked.

These are power-pad MSOP8 opamps, a real pain to replace. On a dozen

8-channel VME boards.
John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Vote
Twitter Facebook LinkedIn
J

On a sunny day (Thu, 12 Feb 2015 17:37:02 -0800) it happened John Larkin wrote in :

Well, given that, I would experiment with a grounded copper strip over the packages?

Vote
Twitter Facebook LinkedIn
J

That might be interesting, but not nice for production. Replacing the opamps is easier.

We could have scaled up the Rf/Ri resistors, to reduce the opamp GBW. That would probably have killed the oscillations, but the bias current of the THS3201's is high, so the DC offset drift would have gone up out of spec.

John Larkin Highland Technology, Inc picosecond timing laser drivers and controllers jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Vote
Twitter Facebook LinkedIn
B

Or be more careful about decoupling the op amps that you use. Most op amps have got very little power supply rejection at high frequencies, so any dec oupling network that doesn't include a pair of ferrite chips for each op am p and a couple of capacitors per rail (one of them with enough ESR to kill any resonance with the ferrite bead) isn't going to hack it.

This ought to be telling my grandmother to suck eggs, but experience sugges ts that people don't think as hard about decoupling networks as they should . Quite a few of them think that 100nF of ceramic chip on each rail is all that's ever required, which is not thinking at all.

Bill Sloman, Sydney
Vote
Twitter Facebook LinkedIn

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required