Active Feedback on Parallel Chips

[Stocker]

I was looking through LB-44 from National, and this occurred to me. But I have a question where the master amp's feedback takes its signal- the load bus or sense bus? what's the difference? Drawing B to follow

[Stocker]

B

[runebivrin]

I must admit that I fail to see the point of adding unity gain buffers in the feed back path, but you're perfectly free to do that if you want to...

Either way, feedback has to be as in the app note. Remember, negative feedback serves to control the voltage at a particular node, and most of the time you'll want that node to be the output.

Rune

[Stocker]

Are you saying only that you are unaware of this concept?

"Lowering Buffer Opamp Distortion"
Lowering Buffer Opamp Distortion

The feedback in LB-44 is a wire from output to -in on each of the buffer amps. This is a schematic *based* on the brief's schematics. I can only wish I knew of a National brief or note or something that discussed active feedback in detail.

[runebivrin]

Well, yeah, now that you mention it I remember. OK. But I still don't quite understand what the issue is with the feedback point. As far as I understand, the point of the circuit in LB-44 is have one op-amp worry about the error on the output of the circuit as a whole (A1 in their schematic), whereas the others just track the direct output of A1 with unity gain. Any differences in actual output are absorbed by the 47ohm resistors. If the load resistance drops the voltage, A1 will attempt to raise the voltage, and then A2 through An will follow that, and thus help equally much in pulling the load.

It's unfortunate that this toplogy can't be used with LM38xx chips, as they're not unity gain stable.

Rune

[Stocker]

The high-power National op-amps are not unity-gain stable in non-inverting configuration, but this circuit is not limited to unity-gain on the drivers as far as I can tell, it's just drawn that way to pose my question. You could actually get some silly power levels this way. I leave it to the power-mad to try with inverting and gain with this configuration. Figure 3 in LB-44 shows how to bridge this puppy too.

This is however related somewhat to my earlier thread about using small-signal opamps to put out several amps in parallel by using dozens to hundreds of devices. Now that you mention it, it seems that the series resistors could be cut way down or out altogether which would be nice. I was thinking of using my pile of TL074s to be feedback for my pile of LM324s.

My question remains however: I am inclined to take the output directly from A1's output not after its resistor.

I suppose the new question is: how much can those resistors go down?
will this method even eliminate a DC offset?

[runebivrin]

As I understand it, unity gain is required, since the task of the additional op-amps is to track the output of the master op-amp as precisely as possible to avoid unneccessary dissipation in the IC:s when idling. The master op-amp can have any gain, but that doesn't matter to the additional drivers.

The difference between taking the feedback signal directly from A1:s output or after the resistor will ultimately affect output impedance, which may or may not be an issue. However, it makes sense to assume that if you use parallel op-amps it's because you want to be able to drive more current. Having resistors in series with the output outside of the feedback loop will kind of thwart that attempt, I'd say, unless the resistors are very small.

Rune

[Stocker]

Thank you, yes that makes sense about resistors in the loop. My instinct seems to be correct then.

As drawn, the circuit is as you understand and describe it. Could one not use an inverting configuration on the secondary load-sharing amplifiers? I freely admit to not having considered this application at all for non-unity-stable amplifiers. Gain of course would have to come in at A1's configuration and the rest follow the leader

[runebivrin]

I suppose it would be possible to use inverting amps, by preceding the secondary amps with an additional inverter, but then you'd be back with having to match resistors for overall unity gain.

I'm sure Circlotron could think of a current dumping circuit with discrete transistors for secondary amps

Rune