push-pull crossover + op amp as buffer

[fady232]

Hello all, can someone tell me if i understood this circuit properly?:

This circuit is basically identical to the last post i did, but with a difference, there is an op amp acting as voltage buffer, what he does here is this:

at pin "+" i connected the sine wave and the pin - is connected to the output.
Now what amplifier gonna do is a comparison beetwen the pin + voltage and pin -, and will try to match its output in order to make pin - voltage as close as possible to pin +, in this way with the op amp we compensate the drop of bjt internal diodes etc, cus we use our input as reference.

Secondly we got a low impedance as output.

Is this correct? for what other reason this circuit is used?

 

[D.A.(Tony)Stewart]

To make a circuit more useful than what you have shown, there must be some gain. By reducing load R you gain in current and power but limited by hFE of each transistor. This reduces the distortion of modulating diode voltage mismatch with Vbe and the less than unity gain errors from bulk emitter losses but with the Op Amp, the input error is amplified by the gain bandwidth of the circuit to reduce distortion with a lower Rout. Did you understand my answer in your previous question? Pls comment there.

 

[KlausST]

Is this correct? for what other reason this circuit is used?

What does the simulation say? Is it like expected?

Which circuit? The feedback only or the whole circuit?
But indeed I woud expect YOU to tell us what the circuit is for.

****
If you don´t need additional gain ... the circuits needs not.
For sure you may improve this ciruit .... you just have to tell us in wihc regard. Noise, distortion, part count, current consumption....

Klaus

 

[danadakk]

DC sweep close to crossover (w/o diodes one can observe the xcrossover) :

Small sig analysis

Zout seems constrained to ~ .1 mOhm at peak. Seems low to me, again how good
is the model.....not you have to subtract .1 ohms from curve due to method I used
to get at it.

Not sure about OP37 model completeness.....

 

[FvM]

OP37 is specified for stable operation only above G=+5, it should be never used in unity gain configuration.

 

[danadakk]

Actually I could not find a hard stop for G = 1, just a reference, with large Cload,
thats its unstable < G = 5. They do call out -

The OP37 provides stable operation with load capacitances of up to 1000 pF and ±10 V swings; larger capacitances should be decoupled with a 50 ohm resistor inside the feedback loop. Closed loop gain must be at least five

But doing a sim, with a 1V step, riding on step final value is a high freq osc, no Cload.

And datasheet G and Phase, does not extend to G = 1, maybe the author of datasheet
felt thats all that is needed. Although the slope of phase margin does, if extrapolated,
show in graph G = 1 not advised. Typicals, room T.

Kind of sloppy datasheet.

 

[D.A.(Tony)Stewart]

Actually I could not find a hard stop for G = 1, just a reference, with large Cload,
thats its unstable < G = 5. They do call out -



But doing a sim, with a 1V step, riding on step final value is a high freq osc, no Cload.

And datasheet G and Phase, does not extend to G = 1, maybe the author of datasheet
felt thats all that is needed. Although the slope of phase margin does, if extrapolated,
show in graph G = 1 not advised. Typicals, room T.

Kind of sloppy datasheet.

The simulated Bode Plots are small signal linear responses, do not test all the nonlinear aspects of voltage and current with this topology. Although it excels in many figures of merit for high GBW, low Noise, low Vio, good load response, it does have an unusual complementary push-pull stage that avoids static crossover distortion by cancelling NPN to PNP Vbe drops and vice versa. This is unlike the classic multi-stage complementary emitter followers with a Class AB bias. So there are some peculiar characteristics in the crossover circuit at full BW with Av ≤5 that are better left avoided than trying to compensate for different loads with mismatched junction capacitance with excessive feedback error gain.

I would expect step responses to be a better tell of the nonlinear signs of instability from ringing. This may require various step output voltages and step output loads for Av<=5.

Bravo to Frank for exposing this limitation on the datasheet.

 

[danadakk]

For sure their datasheet for large signal response shows non linear slew rate limiting.
Their phase margin at G = 5 implies the numerical value of 71 degrees aligns
with below measured ss response.

Sooo what is the phase margin at G < 5.....sloppy datasheet, given the G=5 phase
margin pretty high.

 

[FvM]

Actually I could not find a hard stop for G = 1, just a reference, with large Cload,
thats its unstable < G = 5.

Maybe you have a different datasheet. I read on page 1

The OP27 series is compensated for unity gain. The OP37 series is decompensated for increased bandwidth and slew rate and is stable down to a gain of 5.

It's possible to use a decompensated amplifier for lower gains, but you need to artificially reduce loop gain. The discussed circuit doesn't, thus it has a risk to fall into oscillations.

 

[D.A.(Tony)Stewart]

examples of |Av| <=5 on OP37
Resonance shows up with the non inverting input at unity gain at low input levels.

While inverting unity gain has 90% overshoot and is wide BW without slew rate limiting at low levels.
View attachment 199245

With Av=-5 small step overshoot is almost 50%.

With Av = -10 Overshoot only reduces to 45%.

The model in LTspice does not replicate the plot in the datasheet for Av = + 5 with small signal

But does match the slew rate + % overshoot response for large signal Av = +5 reasonably well.

Conclusion: LT spice shows how overshoot gets worse with Av = -1 and is unstable with Av = +1 and agrees with large step response with Av = +5 but not at all with the good but nonlinear step response in the datasheet.

 

[danadakk]

Maybe you have a different datasheet. I read on page 1

It's possible to use a decompensated amplifier for lower gains, but you need to artificially reduce loop gain. The discussed circuit doesn't, thus it has a risk to fall into oscillations.

Yes we are working off different datasheets.

What caught my eye was phase margin graph at G = 5 of 71 degrees, implied, but not directly stated for the
graph, Cload of 1000 pF as discussed in datasheet but no test circuit shown, so what was Cload on graph,
what parasitics used in fdbk loop.....ugh

The primary clue I saw was taking the graph and extrapolated the phase margin and G curves and
one could then see G = 1 no margin left. But that assumed no other unknown poles/zeros in uncharted
portion of graph. And graph was for typicals, and if Cload @ 1000 pF was involved then what was it at 0....,
but no test circuit called out or label on graph 1000 pF present.....

Sloppy datasheet.....unknown test conditions

Closest call out to your datasheet :

The OP27 series is compensated for unity gain. The OP37 series is decompensated for increased bandwidth and slew rate and is stable down to a gain of 5.

was -

I also did both ss and large sig transient sims and got similar results as Stewart for the ss
case. Large signal case looked like :

Clearly some RF instability, and this was for no Cload. Maye the PNP side limited
Ft had something to do with circuit response as well.....

 

[FvM]

Look sharp. TPC18 clarifies that 71° phase margin is obtained with Av=5. Your datasheet mentions the Av>=5 restriction already in the title. Stable operation with Av=1 isn't stated anywhere.

 

[danadakk]

Look sharp. TPC18 clarifies that 71° phase margin is obtained with Av=5

Yes, agree, never questioned that. What I question is TCP18 with 1000 pF load as mentioned
in earlier specs, or not ? And fact the margin is at 71 degrees, so what G is allowable
at 30 - 40 degrees, still a decent phase margin, or with no Cload if that graph was done with
Cload.

Your datasheet mentions the Av>=5 restriction already in the title. Stable operation with Av=1 isn't stated anywhere.

But it does not state, CLEARLY, the test conditions or schematic for test, especially for Cload and what
specs apply at the 1000 pF called out earlier in datasheet.

The title is not an adequate condition for design, but I agree its a caution to be looked at.

The datasheet is sloppy.....

--- Updated Apr 27, 2025 ---

Unity G follower (all below) , no Cload :

Cload = 1000 pF

Transient with 1 V in, Cload = 1000 pF

Transient, no Cload