Opamps as voltage sources (C-loading)

Sorry, the beginning of that needs some serious editing. GH

What about these... they come in 'p-dips'*

George H.

• to be said in the same voice as Pippin in the movie version of LTR, 'they come in pints!'

This sort of explains why many RRIO opamps are c-load stable.

The load cap doesn't add a pole, it just shifts the break frequency down.

Some RRIOs ring with modest cap loading but get very stable with more, like several uF.

ESR doesn't do any harm here.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Use this method. Most general:

(Left side is output to connector; the "BAV99" just cutoff betrays the ESD diodes for said connector.)

As mentioned, in large values the ESR dominates, at least when it does (probably avoid bare ceramics, unless stated otherwise by the amp datasheet?). That plays the role of the R57/R65 shown above. Since it's in series with the capacitance, there's no DC error and no DC feedback path necessary to correct its voltage drop, so the RC around the outside also goes away. Downside, you can't compensate the amp independently in this case, so the ESR and C must be within the range that the amp is comfortable with.

The effect of the above circuit is to add a pole and zero to the response (when the load is capacitive). When they coincide, the response is flat until cutoff, otherwise there is a peak or dip before cutoff. You can't, in general, compensate for _any_ load this way, only the worst case, and keep the wonkiness at high enough frequencies that are hopefully out of your passband.

Note that you can still use low-ESR capacitors, as long as the dominant capacitance, at the critical (crossover) frequency, is lossy (ESR = 1 / (2*pi*Fc*C)). It's not enough to throw in a few dozen 0.1 ceramics and a

10uF electrolytic of unknown ESR for any random amp, the crossover frequency needs to be right, too.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website:

Why does opamp R_out not dominate?

NT

Right I've done that. As you say below it's a little tuned to the load C.

George h.

Well your picture doesn't really help me. But it's the compensation cap that causes the phase shift and makes the output impedance look sorta inductive at higher frequencies...

Anyway I've got some opa2192's that are RRIO I'll see what they look like. figure's 32 and 33 here,

I can certainly add 50 ohms in series to the output.

George H.

right...

and take the DC feedback after the R and the AC feedback before the R.

m

LM8261 is cheaper; it's an official c-load opamp.

OPA197 is even better to have around. It's happy with a few (or many) uF of ceramic or film caps from output to ground. It will oscillate as a follower with 47 nF.

Pdip? Like this?

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Sorry, too cryptic.

In a classic opamp, there is an internal compensation cap that adds a

-6 dB/octave frequency rolloff, namely 90 degrees of phase lag. That's stable when you close a feedback loop. An external load cap adds another 90 degrees of lag, which makes 180, which turns negative feedback positive at some frequency, so it oscillates.

In a c-load amp, the external load c is *in parallel* with the opamp's compensation cap(s) so the frequency rolloff just gets slower as you add load caps, but remains single-pole, -6 dB/octave, still 90 degrees of phase shift, so still stable. Net output impedance goes down at all frequencies as load cap is added.

There are some great parts in the OPA1xx family.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

So looking back at your previous pic, for the RRIO opamp the load C is in parallel, because the one side (of C-comp) is common, and the other side is a base-emitter junction away from the power rail... (which is the same as ground for HF.)

The OPA192 worked fine at all C-loads... I messed around with the circuit and got it to ring.. worse was with ~0.1uF and then a ten ohm series R on the output killed that and made it sweet. 'scope shot

Hey as always, thanks for the help.

George H.

Right. Looking into the output pin, the opamp looks like a Miller cap. The added Cload is in parallel with that. Sort of.

Try a bigger cap. That will often stabilize an rrio opamp nicely.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Oh I did, 1 uF and 10 uF film caps... they still ring when I step the input. Hmm, just tried a 4.7 uF tant. It looks fine (a little over shoot) Here's the 10 uF film (no output R)

and 4.7 uF tant.

I think we have to point to cap ESR as the difference... film has ~0.03 ohms tant ~0.7 ohms

George H.

Tants have just the right amount of ESR for many apps, including switchers.

Not all RRIO amps have the topology I referenced. You've got to do what you did, try some load cap and a pulsed load to see if it oscillates or rings.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I was pulsing the input. I'm not sure how to pulse the load. Well, simply with a function generator. How about a 10 k ohm load R and I toggle the bottom from gnd to +1 V. (it's a 5V supply)

I'll see if that looks the same.

GH

That was in context of C only (and whatever ESR it has), no explicit R_out.

Do you mean the amp's output (source) impedance? The problem there is, it's usually inductive (a modest impedance, in the 10s or 100s of ohms, but typically rising with frequency), and that, plus some excess phase shift inside the amp, causes the pole to split, resonate and continue rightwards into the right half plane (i.e. pushing it into oscillation).

Tim

Input is just as good. I'd go from, say, +2 to +3 and not start at ground. The amp will slew into a big cap, and current limit, which can be more interesting. So play with the square wave amplitude on top of some DC offset.

You can AC couple a square wave, from a 50 ohm fungen, into the output too.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

not sure I'm following you there.

yes

yes - and R_out dominates massively over C_ESR in that situation.

NT

Well, silly me, I was coupling ground through the func. gen. oscillations everywhere ...wow.! (~1-200mV p-p) (with the FG at ~0V the oscillations were damped, but TC >1 ms. ~ 800kHz freq..weird DSO display :^) I added the series 10 ohms and everything got clean... huh? I'm not sure what I was testing with my circuit F-up... But I'm keeping the 10 ohms. :^)

So this 5V supply is going out into the world, and someone (most likely a physics student) is going to connect it wrong. For an opamp grounding is fine. But what about an over voltage* on the output? Maybe I want more series R? (or the lm317 in to92 clip and replace when someone fries it)

*Not AC voltages, but a 60 V DC lab amp... maybe 2 or 3 (4-6 total) lnd150's in parallel? 5V at 3-5 mA. or some beefier depletion fet and a resistor? (say 10 to 100 ohms)

Yeah I was trying to do a DC thing...

George H.