Compensating a bootstrap amplifier

[tmadness]

I am playing around with a opamp bootstrap amplifier with a voltage follower bjt output stage in LTSpice. But I cant get the circuit to work together. individual components work fine, but when combined LTSpice fails to converge. 
I've attached my circuit. You may notice my frustrated RC compensation everywhere, a lot of it looks to be of dubious efficacy. 
Any help is appreciated, Thank you.

[Kleinstein]

For the bootstrapped supply there should be low loading to the bootrapped amplifier. So the emitterfollower at the ouput is good, but the boostrapped supply part give quite some load (though maybe still just acceptable).
The boostrapped supply should be slowed down to be considerably slower (e.g. faktor 10) than the BW of the OP. The OP should than get away without much extra compensation.
The filtering could be before or after the transistors driving the OP supply.

It gets more tricky if the amplifier has gain, because one than needs some way to drive an output signal with larger amplitude. So this usually involves another amplifier stage.

Attached is an example for an buffer-amplifier with supply bootstrapping. Ignore C79 that is marked to not populate.

[E-Design]

Generally with convergence problems, simplify the circuit if you can to see if / when it starts working. You can replace that whole circuit to the left with batteries (for example)

Does it work in a DC bias condition ?

Also, when I encounter similar bootstrapped simulation problems, its _usually_ the opamp model to blame. There are internal grounds or other stuff that you can't see that might be possibly causing trouble.

In fact, I have made my own "floating" models in the past to avoid such issues. Most vendor models don't actually float, nor do they obey KCL.


Try a different opamp, if you get one to work or it gives different errors, its a clue.

[tmadness]

Alright progress! but not really,

I swapped out the OPA277 for a LT1001 and it works... only on fixed power supplies.

The waveform of the bootstrap look good, however the circuit rails to the positive supply when I use the bootstrap as the power source.

[David Hess]

Not all operational amplifier macro-models model the supply pins properly to support bootstrapped operation.

[magic]

I'm pretty sure I have simulated similar circuits with CCS and zeners to produce the bootstrapped supply and it worked right away for a variety of built-in LT opamp models.

[E-Design]

I'm pretty sure I have simulated similar circuits with CCS and zeners to produce the bootstrapped supply and it worked right away for a variety of built-in LT opamp models.

Some do and some dont... I used to dig into the macro models trying to understand what the issue is.. eventually, I just made my own.

[magic]

LT tends to get things right in analog. And I doubt it was LTspices macromodels that you inspected because they are obfuscated.

I just tested my old sim with LT 1001, 1012, 1013, 1057, 1128, 1469 - all work

Though whether this would be stable in practice is another matter which I haven't tested because I realized that the overall concept of the larger circuit was a nonstarter.

[Kleinstein]

For testing the models one does not need to simulate a BS amplifier. It would be better to just simulate the case with the ground point moving around with another voltage source.  A correct model wold not notice the ground moving. For the test it is OK to use the spice internal ideal voltage sources.

[Renate]

For compensating a bootstrap amplifier I'd recommend $0.10/hour.
And definitely don't offer it any stock options.

[E-Design]

LT tends to get things right in analog. And I doubt it was LTspices macromodels that you inspected because they are obfuscated.

I just tested my old sim with LT 1001, 1012, 1013, 1057, 1128, 1469 - all work

Though whether this would be stable in practice is another matter which I haven't tested because I realized that the overall concept of the larger circuit was a nonstarter.

Well, maybe they have fixed them all. That would be nice. I assure you at least years ago, many LT models did not float correctly. I don't recall the parts.. I stopped using LTSpice for anything serious years ago.... LT also published models for other spice vendors & applications - those are not obfuscated and tend to have some of these issues.

I have complained to them about it so hey, maybe they cleaned them all up. The traditional macro models for opamps do not lend themselves to floating correctly which is unfortunate because so many vendors have embraced that style. Not many complaints are generated either because not many people bootstrap.

I like Kleinsteins idea of just testing by moving ground around.. but I dont think the simulators let you do that --ground is node 0 from which everything is referenced. You would need to drive another node as "ground" and then in the macromodel, replace all node 0's with your new "ground" thus replacing it with one that moves WRT to node 0.

Sometimes just looking at the macromodel that has a reference to node 0 is a sign of a "less than perfect" model - as real 5 pin opamps dont have a ground pin....

[tmadness]

I isolated the variables, kinda.
I added 2 behavioral voltage sources, one 15 volts above the output of the opamp and one 15 below. The "Behavioral" Bootstrap as it were.
The circuit works upto ~26 V on the output an then it rails.

[magic]

Well, maybe they have fixed them all. That would be nice. I assure you at least years ago, many LT models did not float correctly. I don't recall the parts.. I stopped using LTSpice for anything serious years ago.... LT also published models for other spice vendors & applications - those are not obfuscated and tend to have some of these issues.

I have complained to them about it so hey, maybe they cleaned them all up. The traditional macro models for opamps do not lend themselves to floating correctly which is unfortunate because so many vendors have embraced that style. Not many complaints are generated either because not many people bootstrap.

Maybe your complaints helped, then

The circuit works upto ~26 V on the output an then it rails.

The model undergoes phase reversal when input common mode leaves the supply rails. Not sure what the real IC would do but it ain't gonna work right.

[tmadness]

ThE sAgA cOnTiNuEs ...
I swapped out the  lt1001 for a AD8624 an opamp that has "... been carefully designed to
prevent output phase reversal ..." their words not mine.
And the circuit still rails, this time sooner.
Also I got the materials to build a prototype. No guarantees by when...

[magic]

The usual phase reversal protection saves you if the feedback network keeps at least IN- within the input common mode range.
Your circuit pulls both inputs out of the ICMR. It will never work, whether phase reversal occurs or not.

You need to bootstrap the opamp to the potential of the inputs and boost its output swing with additional discrete circuitry. Or come up with a completely different topology.

[E-Design]

ThE sAgA cOnTiNuEs ...
I swapped out the  lt1001 for a AD8624 an opamp that has "... been carefully designed to
prevent output phase reversal ..." their words not mine.
And the circuit still rails, this time sooner.
Also I got the materials to build a prototype. No guarantees by when...

You're assuming that the phase reversal claims are present in the models and behave like real world parts. Why would you assume that?
Opamp models can be extremely disappointing for anything other than their basic operation.. IMO, you may be just wasting time trying to get a model operate correctly with your circuit.

If you really want to know if your bootstrap can work in simulation, you need to develop your own model that has only the features you care about (say for floating, output swing, CMIR limits etc..).
Then later, when you care about other things that are modeled in the vendor model, then you can switch it and do non floating tests.

That or just start building a bench prototype. You could probably already have those tasks completed in the time it has taken trying to get your sim to work.

On the other hand, if you have a good ADI contact you can send them your schematic and they will explain why it cant work with the model and perhaps suggest something that will.

[Kleinstein]

A bootstrapped amplifier with gain has additional complications, as the supply for the ouput is limited and the inputs of the OP don't follow it's output directly.

In real world there is also parasitic capacitance that can upset the circuit, especially if high impedance. Here it makes a difference if the capacitance is to ground or the bootstrapped supply.

The shown circuit misses a load resistor to the negative supply - the emitterfollower can not  go below zero as shown.

[magic]

That or just start building a bench prototype. You could probably already have those tasks completed in the time it has taken trying to get your sim to work.

Nope, but there would be a pile of ICs with blown input stages
(Okay, maybe the high resistance of feedback networks would save them).

I will reiterate in plain English what I already hinted at and then Kleinstein also did: in this circuit, the inputs are at half the potential of the output. With enough output swing, it is inevitable that they will fall below the bootstrapped negative rail. No opamp, real or simulated, is going to work like that. If the model fails to work, that actually makes it an accurate model

[E-Design]

Well yes of course the circuit doesnt work -- but the OP seems uninterested in that. He seems focused on getting a simulation to run.  I am merely suggesting that the OP could have learned that his circuit doesnt work by building it on the bench must faster than getting his simulation to run well enough to show him that it doesnt work. A lot can be learned from a bench prototype - working or not!