Discrete Opamp Open Design

[coluke]

Following suggestions by Bonsai and JC, and inspiration by SW: 0.0008% THD @ 100 mW into 30 ohm, 1 kHz; loop gain ~ 53 dB up to 30 kHz, Zout < 20 mohm. 10 devices (ignoring fet ccs) - now I have only to find those damned J113s...

Ciao,

L.

[Wavebourn]

Originally Posted by RNMarsh

I am interested in the practical application to a real audio amp circuit, of course. Can you show it in a working circuit so others can make it if you dont want to -- or at least talk in more detail about the proposed appl to a circuit which can have new ideas generated from it and so it can be applied to other toplogies.... [sorry got my manager hat on.] -RNM

What I talk about, is basic. There are many ways to implement it. For example, take the majority of power amp schematics shown in Solid State forum, they all contain so called "Vbe multiplier", that is one transistor between bases of output stage (buffer). This "Vbe multiplier" is one example: it is in direct contact with an output transistor, and when it is heated by an output transistor it's Vbe drop decreases. It decreases voltage between it's collector and emiter, hat is applied as bias between bases of transistors of the output buffer, decreasing bias voltage that decreases idle current. Decreased idle current decreases temperature of output devices, henc temperature of the sensor that is Vbe multiplier in this particular case.

Similarly, as I described already few times, in that schematic that John posted, temperature of output devices can be sensed by diodes, or similar Vbe multipliers, and the result used as a voltage reference for the current source in the tail of the input diff-pair. What to draw here? It is all obvious. Also, please remember that I am lurking here from work computers that run Linux and can't install on them schematic drawing software.

This "thermal stability" discussion has nothing to do with the topology that Scott Wurcer presented. It is the parallel thread about "Practical Example" offered by John Curl few pages before, the reply to his question, how to improve the stability of the schematic. It is not intended to give you complete designed, breadboarded, and ready to manufacture opamp.

[scott wurcer]

Originally Posted by stinius

Hi Scott
I agree with everything you are saying here.
I think your circuit is the most interesting here so far.
I'll put it in to my simulator and have a closer look at it.
BTW: is this the model you are using for the BF862?


Cheers
Stein

I don't use the package parasitics here otherwise essentially the same.

[scott wurcer]

If for the moment we allow complimentary JFET’s there is complimentary symmetry on the circuit I posted. Here for those that are interested. Load the output with 90k and the open-loop BW is 20k while the OPEN-loop distortion is -70dB. I thought these things were good folk medicine for the sound. The triple-darlington is virtually no load to this and can be biased to drive a heavy load too. At 20k the base current re-capture lowers the open-loop distortion by 30dB.

[Joachim Gerhard]

This is simple enough so i can build it.

[Joachim Gerhard]

Should J18 gate not show to the inside ?

[Joachim Gerhard]

NOO, i get confused, J17,18 N-channel and J1,J2 P-channel ?

[kevinkr]

Guys let's focus on all of the many very cool discrete op-amp designs being surfaced here. Thanks..

[PaulBysouth]

Here is yet another circuit for a discrete opamp. It has the front end of the AD797 opamp, but with a single ended Class A Mosfet output stage. This circuit can be varied considerably. The input devices can be Jfets, Mosfets or BJTs. The output could have a short circuit protection added if required. Offset correction could be added to vary the current through R5, R6.

It has been a year or so since I've modeled this circuit in detail, but it seems to work very well. THD is very low, and H2 dominant. Note that I haven't built one to actually test (and don't plan to). I've suggested suitable sot223 surface mount devices, but almost anything can be used. No unobtainable P channel Jfets are required! The compensation probably needs some work!

Paul Bysouth, August 2012

[Chris Hornbeck]

Is the I2 connection to the output follower correct? Don't think I've ever seen that before.

Thanks,
Chris