Another op-amp stability question (LT1677)

?Maybe

o.

Where's the supply bypass? First guess is that 100 nF to ground will fix it . Keep the leads less than 1/4 inch long.

Not everything in the amp is as slow as 7 MHz--it looks like the two input stages (required for RRI operation) are fighting each other at the crossove r point.

Cheers

Phil Hobbs

Did you use a 1x probe? That could make it oscillate.

Datasheet indicates input stage NPN/PNP handover occurs at "approx 1.5V" above negative rail so that figures with your observed 1.2 threshold.

piglet

As I said, it doesn't matter what the load is. 330 ohms to ground, a 10x probe, a 1x probe, an active FET probe. Ditto supply bypassing.

-- john, KE5FX

Big glass-encapsulated axial ceramic in photo. I think it's 330 nF but it might be 100. Makes no difference.

-- john, KE5FX

I don't know either. I presume you've tried slotting in other opamps, were they ok?

NT

(If I don't make sense here, it's because I just had an endoscopy and I'm still metabolizing IV drugs)

Looking at the data sheet figs on sheet 8, the middle figure is input bias current vs cm voltage; if this is current out of DUT, isn't that a negative input impedance?

RRI opamps usually have those tricky PNP-NPN diff pair swapping circuits. I've always thought that the chip designers have to be very careful about how the offset voltage and bias current transition has to be handled as regards closed-loop behavior.

I have just built and tested it. The setup is in

I used a HP3325B as signal source, 500 Hz +- . The built-in offset voltage is nice here. There is a 49R9 termination resistor from Pin3 to gnd.

No problems whatsoever with Vcc from 3 to 15V. But with Vcc < 2.4V hell breaks loose. You can get any distortion or oscillation you can imagine by just playing with the voltages.

But the spec says 3V min, so that is ok.

regards, Gerhard

Thanks, that seems to be a good clue. I was mistaken when I said the source Z "didn't matter" -- I must not have tried anything very low. Can you take the 49.9 ohm termination out, and add a 10K series resistance from the function generator instead? Set the 3325B to

When I use a termination under a couple K, the instability goes away.

-- john, KE5FX

Hey, that's a better excuse than I usually come up with!

Not sure how to interpret it. If they measure Ib the way Analog Devices' MT-038 note suggests, they're plotting the average of the currents at both inputs. I imagine the effects of negative common-mode impedance would be reduced by the relatively high CMRR.

I'll bet the effect is related to that, in that the input bias current varies slightly when different transistors take over. With a high source resistance (as in my reply to Gerhard) the slight change in Ib might give rise to positive feedback, this time in differential mode. Also might explain why it's so temperature sensitive.

The OPA197 you mentioned the other day looks like a very good replacement for the LT1677. I'll have to give one those a try.

-- john, KE5FX

By the way, don't use any so-called "high resolution" modes or other averaging buttons on your DSO when you look for this effect. They clean up the oscillation nicely.

-- john, KE5FX

Hi, John,

you have found a real bug with the LT1677 !

cheers,

Gerhard, DK4XP

Thanks very much for checking, Gerhard. I was hoping I just got a bad batch of LT1677s or some midnight-shift specials out of Shenzhen.

Maybe this is common/expected behavior when driving RRI opamps from high-Z sources, but manufacturers need to document it if so.

-- john, KE5FX

Nah, there are high-Z things in the datasheet's applications circuits.

That sort of mistake is pretty rare for LTC in my experience--they seem to have learned from the LT1028 noise peak debacle. (Early LT1028 datasheets had noise plots that cut off at 20 kHz or so, neatly disguising the big peak at ~300 kHz that disfigures what would otherwise be almost the universal low noise, low-Z op amp.

(I much prefer the ADA4898 because it doesn't have the big peak and is much less squirrelly generally. Not as fast, though, nor quite as quiet outside of the peak.)

Cheers

Phil Hobbs

It's a known issue with some opamps, but I've stayed quiet cos I can't reme mber where a datasheet (IIRC) talks about it. IIRC the advise was to keep s ource impedance under 10k, and the cause is stray capacitive coupling from output to input plus a faster input stage than output stage. But as I say, I can't remember where this was from so ICBW.

NT

Doesn't seem like a typical phase-margin gotcha, or one that involves external coupling or feedback of any kind. There's no sensitivity to stray C at all. There must be a thermal component to the phenomenon, because it's only noticeable at very low frequencies and is quite sensitive to temperature. The oscillation itself is around 700 - 800 kHz but I've never seen it appear on signals above a few kHz. No electrical time constants are anywhere near that.

Given that people don't usually use pricy low-noise opamps with high- resistance sources, I guess it's not too surprising that the issue isn't well-known. That's no excuse for manufacturers not knowing about it and warning accordingly, though.

-- john, KE5FX

Yes, John. But I think its safe to say opamps going through input-pair transitions at certain voltages are suspect to be used at those voltages. Your experience illustrates that point.

But what's the use of a rail-to-rail opamp if, well, you can't use it rail to rail? I mean, really!

Sounds like LTC scrooed up that one.

what I described isn't. It's what happens when the input stage moves fast but the output stage doesn't.

You've tested for that?

that too?

If my memory serves me correctly, they have.

NT