another bizarre audio circuit

Bizarre??? Just a standard buffered input CE with negative feedback DC bias to stabilize the operating point against Vbe and reverse leakage collector current change with temperature- a textbook circuit...

Neat, like a parafeed amplifier. The choke will have to be pretty big for audio frequencies...

I can think of a modification that would give more gain for the negative feedback to "work" with: change the input transistor to a JFET, get rid of the capacitor and extra resistor in the bias stabilization circuit, and just stick a resistor in the emitter lead of the output transistor. Then, move the load resistor of the first stage down and connect an NPN transistor with a current limiting resistor to the output between that load resistor and the power supply. If I've figured things out correctly you then have a gyrator load for the first stage. The bias should still be self-stabilizing too, because of the DC drop across the output load inductor.

Yup!

Nope, sorry, cancel that, it won't work. The output stage has gain and the gyrator won't work properly.

You could just use a power resistor and tweak the drain voltage operating point. It's just not as elegant.

John

Which textbook?

John

The circuit I'm working on now as my "exercise in simplicity" from the thread a while back uses another transistor for the load on the output stage, simulating an inductor. This also has the advantage in Class A of being able to set the idle current to half of the desired maximum output current, instead of the usual need to set the idle at the full output current.

If I remember correctly I think if one uses a resistive load the quiescent idle current will have to be _more_ than the maximum output current to get the desired output current into the load when the transistor shuts down, because of the drop across the resistor.

I think I have an unhealthy obsession with simulated inductors.

I don't know about textbooks, but it was a pretty standard circuit back in

And voila, here's a schematic of a Dyna PAT-4 with essentually the same circuit:

The challenge that I had in 1965 was to build it using PNP germanium switching transistors which weren't all that linear, were speced for beta =

So, I built two amps with the circuit illustrated above, cascaded them, and used 4 darlington pairs for the transistors, rather neatly solving the beta problem. If memory serves the first stage was RIAA and the second was flat with a gain of 10.

The thing ran rings around any commercial circuit that we compared it to, whether SS or tubed both on the bench and in listening tests. In my travels through the lab's parts bin, I also found a stash of appropriate-valued mil spec metal film resistors, etc. As a commerical product, it might have had to sell for $100s.

I don't know about textbooks, but it was a pretty standard circuit back in

And voila, here's a schematic of a Dyna PAT-4 with essentually the same circuit:

The challenge that I had in 1965 was to build it using PNP germanium switching transistors which weren't all that linear, were speced for beta =

So, I built two amps with the circuit illustrated above, cascaded them, and used 4 darlington pairs for the transistors, rather neatly solving the beta problem. If memory serves the first stage was RIAA and the second was flat with a gain of 10.

The thing ran rings around any commercial circuit that we compared it to, whether SS or tubed both on the bench and in listening tests. In my travels through the lab's parts bin, I also found a stash of appropriate-valued mil spec metal film resistors, etc. As a commerical product, it might have had to sell for $100s.

I used to design language labs and recording studios/mixers, and we used that circuit a lot for tape heads and mic amps. It was called "the GE circuit", don't know why.

I think I used 2N3391s, silicon NPNs, which were the black [1] GE cylindrical epoxy transistors with the flange on the bottom and the concave top, where they poured the epoxy in. 1968, approximately.

People here were discussing a low-parts-count headphone amp, and it occurred to me that this configuration is a lot of bang for a few parts, as a power amp. I thought the bipolar-mosfet thing was cute, but then I'm easily amused.

John

John Larkin a écrit :

As the other Fred said, it's a pretty standard circuit, from the 60's, with its structure inherited from the toobs era.

It's nothing else than a CFB opamp which made good use of the low available gains and FTs of that era.

Yes, you can swing almost all the way to the rails with a constant-current load, as opposed to a resistor, so it's more efficient... like an inductor. The inductor allows close to 2x Vcc p-p voltage swing, if that matters.

But the constant-current load bumps the active parts count by 50% !!!

I think you could do a nice headphone amp with one mosfet. Or go nuts and use a GaN fet.

John

Just about any textbook that goes into bias point sensitivity analysis of transistor circuits- you remember the S- functions, mainly ICQ stability. The big three were HFE, VBE, and ICBO. Then the rest of your circuit is just ac-bypass and the shunt-series feedback for signals. I've seen it dozens of times.

I bet you haven't seen the bipolar+mosfet version, with inductive pullup, used as a power amp.

John

Here is a couple of very low noise audio circuits for the same purpose:

I like BJT and JFET in cascode.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

The one on the left, with the base and gate connected, is sure cute.

John

..

I believe the collector inductor and/or transformer are used to significantly increase the efficiency of the output stage, where that consideration outweighs the additional cost.

I'm having a little trouble with the operation of that one. The way it's set up, assuming the output sits at 0 volts won't the input BJT be cut off?

Jfets usually run at negative gate voltage. Assume the transistor base is at +0.7. The jfet source will be at some more positive voltage,

Actually, you can cascode a transistor into the source of a fet that has a grounded gate. In that case, the source/collector voltage might be a volt or two. You would have to look at the fet transfer curve, and know the design operating current, to see exactly what that voltage might be.

The problem with jfets is the huge part-to-part variation in Idss and transfer curves. A 10:1 datasheet spread in Idss isn't unusual.

For a typical vanilla jfet in this circuit, Vs might be +1 to +4 volts maybe.

John