Discrete Opamp Open Design

[RNMarsh]

Originally Posted by john curl

Well, you are going to stand alone with this, Richard. The VERY IDEA of not using a servo is artificial. The effort, in many examples, to eliminate either the servo or a feedback capacitor, is just too much to make the best possible audio quality.
Let me give an example:
About 15 years ago, I visited a company called Threshold, then located down in Southern California.
It was apparently the successor of Nelson Pass's Threshold.
They built a very formidable solid state power amp, that they INSISTED be DC coupled without a servo. What they had to do with parts selection and temperature controlled adjustment was FORMIDABLE! They put MORE effort in device selection for low drift than virtually anything else in the design. Then, they went out of business. Did not surprise me at all.
Now, what are we trying to do here? What do we have that can be truly 'servo unnecessary', over a variety of applications. A low gain line stage is about the only example.
For example, IF you want the lowest intrinsic offset, then get a monolithic matched pair of bipolar devices. They should always be better than monolithic fet pairs, that will always have more offset. Then carefully build the circuit from there.
IF you insist on using a jfet pair, then get perhaps 100 2SK389's. Build a special offset tester and measure each and every pair for amount and polarity of offset, and put each fet in a separate box. Then, take two IDENTICAL 2SK389's with the same offset and wire them in parallel, facing each other, thereby canceling the inherent offset to less than 1mV. Add a relatively large copper heatsink and shield the 2SK389's from temperature changes and air currents. NOW, you don't need a servo. Is it worth it?

John, I know ALL of this... nothing new. I am challenging everyone to think outside the existing boxes. It can be done without crippling the amplifier... but not with the same old linear brute force methods. Note: When I conceived the dc servo idea... it was, at that time, outside the box, too.

[Wavebourn]

Originally Posted by RNMarsh

Althought thermals could cause current/bias changes to be propagated from input to output (or visa-versa)... depending on topology, they could be used to create negative bias feedback that stabilized the currents with temp changes. This technique could then stabilize the output dc offset as well. -RNM

It is exactly what I proposed for JC's unstable FET drawing: to add diode sensors of temperature of output MOSFETs and control reference voltage for current source in the tail of an input differential pair. If you want "all FET" design you can use JFETs instead of diodes. And additional opamp in this "thermal servo" loop.

[stinius]

EUVL asked for a diff in diff out thing, I designed this one today using p and n jfets. (use your favourite jfets, BJT's and output buffer and you are up and running, pretty low THD+N)

I'll design another one using only BF862 in the ips.

Cheers
Stein

[Mike Gergen]

RNM and WB
There is an amp that does this; The GB150.

You'll have to have Greg explain the rest of it.

[flg]

John, in my collection of circuits from where ever, I have a dwg that looks like your 2nd topology suggestion (below). Was that used in the Vendetta?

[Wavebourn]

Originally Posted by RNMarsh

Note: When I conceived the dc servo idea... it was, at that time, outside the box, too.

Do you mean, it was used before everywhere outside of the box containing the certain topology that you used?

[flg]

Wavebourn, a term for inovative thinking, outside the norm.

[Wavebourn]

Originally Posted by flg

Wavebourn, a term for inovative thinking, outside the norm.

What is "The Norm"?

It is very personal definition. Is "OpAmp" topology "The Norm" for an audio amp? Not for me. But yes, for somebody else.

[bcarso]

Originally Posted by john curl

Well, you are going to stand alone with this, Richard. The VERY IDEA of not using a servo is artificial. The effort, in many examples, to eliminate either the servo or a feedback capacitor, is just too much to make the best possible audio quality.

[snip]

What do we have that can be truly 'servo unnecessary', over a variety of applications. A low gain line stage is about the only example.

Just looking now at the Linear Audio article to which Richard refers, the voltage gain of the headphone amp presented is about 4.3. A 2N5457 and a 2N5460 with local degeneration form the input. There is a trim adjustment potentiometer.

So, yeah, you might get away with that without a servo. But try a phono stage, or some other circuit of comparable gain

[dirkwright]

Originally Posted by RNMarsh

John, I know ALL of this... nothing new. I am challenging everyone to think outside the existing boxes. It can be done without crippling the amplifier... but not with the same old linear brute force methods. Note: When I conceived the dc servo idea... it was, at that time, outside the box, too.

It's not a level playing field if you're not going to participate. It's also no fun if you don't participate. I suppose you already have a solution, so what's the point of this exercise to you then? I could just pick up a copy of the magazine with your article in it, make a few changes, claim it as my own, and post the schematic here. I don't see the point of this at all. It's just a bunch of babble leading no where. Mr. Curl has already pointed out that IC's are going to be superior to just about anything an amateur can come up with, and unless said amateur is willing to spend hundreds, or even a few thousand, dollars on this investigation, then it won't come close to what anyone can buy in an IC opamp.