Forgiveness - John Dunn, Consultant, Ambertec, P.E., P.C.

There is a really great application note AN-202 from Analog Devices that can be found on line at http://www.analog.com/static/imported-files/application_notes/135208865AN-202.pdf . Following the rail voltage bypassing example as shown in Figure 8 of that note, I set up a low-noise amplifier's op-amp rail voltage bypass capacitors as follows:

 

I wasn't actually using Analog Devices' parts. Instead, the op-amps were made by a still very highly regarded competitor. This was because low noise amplifier was a circuit I had taken from an application note of that competitor.

When I turned my amplifier on, there was pretty good low noise performance up to approximately 10 kHz, but there was this broad spectrum noise peak centered around 100 kHz.

I called the application folks over at the mystery company and they said: Oh, yes. Their op-amp's data sheet didn't show it, but there was a current noise peak in the 100 kHz area. Their suggested remedy was to use another op-amp from their product line which would make everything work just fine. I got one of those suggested op-amps, put it in the circuit and the whole thing just burst into vigorous and very noisy 100 kHz oscillation.

Their applications engineering guy and the local sales rep came to call and I demonstrated the whole problem. Their reply was that the rail voltage bypassing arrangement I had chosen was the cause of it all and until I did away with that, they could offer nothing else to help.

What I came to realize at this point was that the first op-amps I'd chosen had that 100 kHz noise peak because they were on the verge of 100 kHz oscillation and that the second ones were merely worse than the first.

I went next to using Analog Devices' own type AD797 low noise op-amp and lo and behold, everything worked great.

This all happened quite a few years ago, but to this very day, I still have this visceral distrust of that still very highly regarded and very prominent competitor's products.

I guess I don't have that much of a forgiving nature.

Comments

[Mark Nelson]

Hmmm. If you had done the bypassing with each rail's cap to ground (the way I was taught to always do it), you might never have experienced the problem, and you'd still have a high opinion of the company. The bypass cap connection from rail-to-rail doesn't seem like a good idea in general (though clearly specific conditions might make it necessary).

[John Dunn]

Hi, Mark.

Actually, I had tried it both ways and the op-amps did the same things. It was my choice to use the app-note's bypass plan to which the op-amp vendor decided to take exception, but the problem persisted either way.

Do check out that app-note though. It's quite edifying!

[George Storm]

Given the usual location of the "compensating" capacitor, I had expected the capacitor direct to ground would be between negative supply and ground (??).
Assuming "the other company" to have "merged" a couple of times since I first met them, I think I have used their products - and although they were sometimes more sensitive to layout there were other reasons to use them.
N.B. I love Paul Brokaw's humour.

[John Dunn]

Gotta take a look at that app-note again. It wasn't about a compensation capacitor of any sort, but about not unintendedly injecting rail voltage ripple into the current mirror circuits that in most op-amps are tied to the negative rail.

It's been too many years, so details elude me at this point, but that was the gist of the author's thesis.