BridgeClone with current-dumping circlotron afterburner :-)

[Circlotron]

The saga of nutty amplifiers continues...

GainClones are good because they wrap a whole lot of circuitry into one tidy, easy to use package that has reasonable performance. Bridge amps are good because you can get twice the voltage swing and twice the slew rate than otherwise. Current dumping is good because you can easily use elephant-size output devices but without any concern about biasing, device matching, crossover distortion etc etc. Circlotrons are good because you can use two *identical* NPN or N-channel or tube devices, not an N-channel and an approximation of it in P-channel form. What's more, N-channels can be had in =seriously= big sizes and are cheaper than any wussy P-channel.

So then, what simply must be done is to roll all these ideas into one; the result being presented here. Actually I haven't built one yet because the idea only came to me in a flash this afternoon. The way it works may not be immediately obvious but I have had people way nore experienced than me look at it and it appears to be sound.

OK, I wont ramble on. The rails for the power opamps could be as much as +/- 40v because the opamps will not be drawing much current. The bulk of the load current is supplied from the two floating supplies. Their voltage should be at least as much as the opamp rails i.e 40v each at whatever current rating the load demands.

This amp should be able to drive almost any impedance load - Just keep on putting speakers in parallel. You need more output current? Just use bigger floating supplies and bigger mosfets. It's as simple as that! The sky is the limit. I'll try and post a SPICE cct in SIMetrix format soon. All comments welcome.

[phase_accurate]

Hi Graham

I think your circuit would work. But I assume that it would work better with BJT "afterburners" than using MOSFETs due to the lower opamp output-voltage needed to turn on the "afterburners".
Maybe differential feedback-takeoff would help as well, but I am not sure and have to think a little more on that one.

I would not call this one a current dumper. The original current dumping circuit uses frequency dependant parts to combine the two currents involved.
It resembles a principle called "Edwin amplifier" which was quite fashionable for DIY during the 70s because it didn't need any setting of bias-current and still achieve reasonable THD figures (though not high-end). I wonder why it isn't used more frequently nowadays because there are enough applications where its performance would suffice.

Regards

Charles

[Circlotron]

Hi Charles. Thanks for your observations. I'll see if I can find anything on this Edwin Amplifier. I use the term "current dumping" rather loosely, actually.

I ran a simulation with the opamps approximation LM3886's. The distortion comes out a little high but the waveform of the left side for example swings +23.012v and -22.968v. If I measured it differentially then it would probably be better because the RHS is a mirror image of the left of course. Output power was 132 watts on the imaginary workbench... It would go as high as 145 watts into 8 ohms just before clipping. The mosfets on the sim diagram are not really suitable, just what the demo SIMetrix simulator had.

[Circlotron]

Output waveform either side of load.

[Circlotron]

Put an ideal transformer across the output with one end of the secondary grounded so I could look at the voltage the speaker actually sees. Distortion is almost exactly the same. D'oh!

[phase_accurate]

Hi Graham

Below you see two basic variants of an Edwin output stage.

In the left variant the bias current is carried by the drivers and flowing through R1 and R2. This is a little like the circuit you want to use.
In the right version the bias current is flowing through the output stages and R3 and R4.
My guess is that the efficiency of the left variant is better at high levels than the right one. THD performance will most likely be better for the right one because the circuit properties change less when the circuit is going from A to B operation than for the left one.
I think there would be quite some possibilities for improvement with these circuits which went somehow forgotten lately.

Regards

Charles

[janneman]

Hi Charles,

I think your right-hand circuit wouldn't really qualify for an EDWIN. The significance of the EDWIN, IIRC, was that the drivers would actully provide the output signal through the lowish R1, R2 until such time that the (sluggish) output devices were turned on. In the right-hand circuit, the drivers can't do that.

Jan Didden

[phase_accurate]

Hej Jan

AFAIR the main motivation behind the Edwin was simple bias setting. Both circuits would quality for that purpose. But you may be right that only the left one is actually an Edwin. I would have to dig out a very old Elektor issue in order to check.

Regards

Charles

[janneman]

Charles,

Yes, I see your point with the right-hand circuit on the easy bias. Actually, as you said earlier, the (huge) drawback of both circuits is the step in gain when the main outputs kick in. Feedback can only do so much to linearize the thing. But I am pretty sure the left is the EDWIN; I had just started audio DIY when it appeared, and I was mightily impressed that anyone could come up with such a clever scheme.

Jan Didden

[phase_accurate]

Maybe the right-hand circuit might be improved by putting a series RC circuit across the diodes at the output, resembling a little closer the differential resistance of the diodes (O.K. they are current dependant but I guess you know what I mean) at higher frequencies. This might eventually lead to decreased crossover distortion.

Regards

Charles