6
\$\begingroup\$

Suppose I want to make a simple non-inverting amplifier with a bandwidth sufficiently low.

Is it harder to make a stable non-inverting amplifier with a fast op-amp? When I read some documentation on the internet it seems that very fast op-amps like the OPA837 are more difficult to stabilize than a slow one. I do not really understand why. Why is the fastest not better than a slow one for doing the same job?

\$\endgroup\$
2
  • 1
    \$\begingroup\$ Do you mean you want a 1 MHz bandwidth unity gain amplifier and the options are (a) using a (say) 3 MHz BW op-amp or (b) a 100 MHz op-amp? Is that what you mean i.e. you want to use the faster BW op-amp in a unity-gain circuit that has low BW i.e. 1 MHz? Is that what you mean Jess? Are you asking why using the 3 MHz BW op-amp is easier to implement compared to the 100 MHz op-amp? \$\endgroup\$
    – Andy aka
    Mar 29, 2023 at 21:14
  • \$\begingroup\$ This is it ! :) \$\endgroup\$
    – Jess
    Mar 30, 2023 at 9:21

2 Answers 2

6
\$\begingroup\$

Op-amps in flat bandpass-to-DC amplifiers get unstable because of parasitics, including parasitic output to input coupling, capacitive loading, and excessively high source impedances combining with parasitics.

By default they should be stable and have gain margin per what datasheet indicates. Some op-amps are not unity gain stable, but the datasheet is explicit about it. In any case, instability with docile loads hints at a design too close to margins, and can be improved, before having to kill the gain loop too much by overcompensating the feedback loop.

Why is the fastest not better than a slow one for doing the same job?

Because, unless the load is variable, all that loop gain at high frequencies and fast slew rate capability contributes to quiescent current consumption and not much more, compared to a much slower op-amp in the same process. Even though modern op-amps are often faster and lower current than legacy types, although they often can't take as high a supply voltage. Loop gain excess to requirements doesn't improve performance, because you just stated that you don't care about high frequency out-of-band performance, and I had perhaps erroneously assumed that the op-amps are driving docile inputs with constant impedance and no periodic charge injection.

Now, an op-amp that's fast usually settles much faster - even orders of magnitude so - than the typical "1MHz, 1V/us" ballpark general-purpose types. That means that it will do much better driving variable loads. Any multiplexed A/D input is a variable load, unless there are internal buffers between the input pins and the MUX. Such "open mux" inputs are nasty and hard to drive while retaining accuracy, even when the overall bandwidth is small. They are even harder to drive when significant cabling is involved, since the cabling is usually mismatched to the source and load, and the mux transients treat the cable as a transmission line and happily try to do a job of a TDR on the cable, signal source, and interconnects.

I have had applications with DC-500Hz bandwidth where the op-amps driving the ADC had to settle in <<1μs, or else there'd be crosstalk between A/D channels, and additional nonlinearity. The op amps had to be 5+ orders of magnitude faster than 500Hz itself would suggest, and had to slew faster than you'd think too - about 25V/μs IIRC.

it seems that very fast op-amps like the OPA837 are more difficult to stabilize [...]

... if the design makes them unstable to begin with. That particular op-amp has splendid performance given its minuscule current consumption relative to bandwidth, and sure - if you're trying to use it on a breadboard or protoboard, it'll be harder to keep it stable. But on a well-designed PCB, it takes no more care than with any other op-amp. I have plenty of circuits where you could substitute OPA837 for a much slower op-amp in use, and the circuit would remain stable and often the overall specifications of the system it's in would not be degraded.

Just for kicks, I've had an opened up late 1970s vintage power supply on my bench, and replaced a random 741 with a 30MHz part on a DIP adapter. No oscillation. The feedback loop on the 741 was already suitably compensated well enough in spite of through-hole construction. It does oscillate with a 150MHz part though - the foil capacitor used for feedback, and its trace loop, have too much inductance it'd seem. Bypassing it at the socket with a snubber made of 100pF in series with 5kΩ made it quiet again. So, I wouldn't necessarily call it hard to stabilize in suitably designed circuit that doesn't destabilize it to begin with.

\$\endgroup\$
1
  • \$\begingroup\$ Thank you for this nice answer :) \$\endgroup\$
    – Jess
    Mar 30, 2023 at 9:28
5
\$\begingroup\$

The statement is incomplete.

It's difficult to frequency compensate an op-amp while not sacrificing much of the large bandwidth/loop gain it has.

Anyone can throw away loop gain by placing a resistor between both (+) and (-) inputs, but then you'd be wasting money on fast amplifier that now has been downgraded to the same capability of an slow one.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.