Discrete Opamp Open Design

[scott wurcer]

Scott - you have the floor/ball..... take it and run. Are you going to do a discrete version of the 627 with us? Or? -RNM

I have something simple to start with as my discrete 627 is complicated, but I need to do it from home tonight.

 

[dirkwright]

Since virtually all of the likely designs to pour forth are going to want an output buffer, I dredged up and dusted off the attached.

It can be scaled back to lower rails without too much trouble. The bandwidth per se from a 1k source R is about 35MHz. There is the advantage of reduced thermal distortion when used with about a 515 ohm load, since the output devices are equidissipational*, that is, their dissipations roughly track. But when, as we anticipate, this is inside of an overall global loop with plenty of loop gain, the rest of the works will easily correct for dissipation-induced shifts, as the shifts should be fairly slow, and thus the thermal concerns should be minor.

The bootstrapping of the input devices raises the input Z, but also requires some lumped C to reduce peaking. Since such C is constant, the distortion at high frequencies at the input is reduced.

Short-circuit protection is left as the proverbial exercise to the reader.

Brad

*the approximate condition for this is that the peak output current into the load at just before onset of clipping is twice the quiescent current. Things are slightly more complicated by the emitter ballasting and the fact that the input devices draw their operating current from the output devices due to the bootstrap arrangement.

What I thought was meant by "opamp" was that it had both + and - inputs that were balanced to ground and basically required a differential input stage. There's absolutely nothing wrong with this circuit, but I assumed we were limited to a certain kind of topology. Did I make a wrong assumption?

 

[Wavebourn]

Short-circuit protection is left as the proverbial exercise to the reader.

It is limited by beta of output transistors and currents supplied by sources.

 

[Wavebourn]

What I thought was meant by "opamp" was that it had both + and - inputs that were balanced to ground and basically required a differential input stage. There's absolutely nothing wrong with this circuit, but I assumed we were limited to a certain kind of topology. Did I make a wrong assumption?

This circuit is an output buffer that Brad proposed.

I have another version, can't find the picture...

 

[dirkwright]

This circuit is an output buffer that Brad proposed.

I have another version, can't find the picture...

Oh, my mistake. Sorry Brad!

 

[dirkwright]

If we are not counting current sources, what about current mirrors? Do we not count them also?

 

[john curl]

You should quit 'counting' and start designing. These arbitrary rules are self limiting.

 

[bcarso]

What I thought was meant by "opamp" was that it had both + and - inputs that were balanced to ground and basically required a differential input stage. There's absolutely nothing wrong with this circuit, but I assumed we were limited to a certain kind of topology. Did I make a wrong assumption?

This is presented as a PART of a final opamp design. Note the triangles in the various Groner-presented topologies? Those are where this would go --- they are ~unity-gain buffers.

EDIT: sorry read and replied before I saw Wave's reply.

 

[bear]

In terms of buffers I like Chris Paul's buffer design... not sure where to find the schematic though... a similar on was suggested in a diy thread a year or two ago and I piped in with the requisite moving of the input devices to bootstrap them to the outputs... maybe I can find that if I search enough...

_-_-bear

 

[bcarso]

It is limited by beta of output transistors and currents supplied by sources.

It might survive a momentary short, like a brush with a clumsy test lead ground, but it's not likely bulletproof for magic smoke releasing probers

 

[john curl]

Later, you can 'prune' the circuit of any extras that are not necessary.

 

[bcarso]

In terms of buffers I like Chris Paul's buffer design... not sure where to find the schematic though... a similar on was suggested in a diy thread a year or two ago and I piped in with the requisite moving of the input devices to bootstrap them to the outputs... maybe I can find that if I search enough...

_-_-bear

I think the only things I've seen of Chris's are hollow state, so that would be interesting.

 

[dirkwright]

You should quit 'counting' and start designing. These arbitrary rules are self limiting.

Yeah, um, I'm not that talented. Thanks for the encouragement anyway.

 

[bear]

I think this may illustrate somewhat:

http://www.diyaudio.com/forums/solid-state/189123-anyone-built-power-diamond-buffer-amp.html

and here:

http://www.diyaudio.com/forums/soli...iscrete-medium-power-opamp-3.html#post1549752

May or may not have application here.
I shoot from the hip, usually loaded with blanks or duds...

_-_-bear

I am led to believe that this is Chris Paul's work, and was perhaps published in a mid 80s (?) audio amateur if not elsewhere. I do not have the original source. But have used this exact circuit altered with jfet inpoot and mosfet outpoot... (I did use a servo, iirc to keep the DC offset nil, but that is the easy way out...)

 

[bcarso]

I think this may illustrate somewhat:

http://www.diyaudio.com/forums/solid-state/189123-anyone-built-power-diamond-buffer-amp.html

and here:

http://www.diyaudio.com/forums/soli...iscrete-medium-power-opamp-3.html#post1549752

May or may not have application here.
I shoot from the hip, usually loaded with blanks or duds...

_-_-bear

I am led to believe that this is Chris Paul's work, and was perhaps published in a mid 80s (?) audio amateur if not elsewhere. I do not have the original source. But have used this exact circuit altered with jfet inpoot and mosfet outpoot... (I did use a servo, iirc to keep the DC offset nil, but that is the easy way out...)

I can ask Chris about it.

 

[Wavebourn]

It might survive a momentary short, like a brush with a clumsy test lead ground, but it's not likely bulletproof for magic smoke releasing probers

All depends on appetites.

Yo can reasonably limit this current, without worry to loose too much on speed of rechargings.

I still can't find schematic of my buffer. It relies on the current in the tail that increases with signal voltage level, so protection works as protection from lower than threshold load resistance that is more relevant for protecting of output transistors from excess power dissipation than protecting from current on any level of signal, from zero to the rail. It gradually shifts from voltage follower to voltage to current converter. I discarded it in late 70'Th as an output stage for IC because selection/adjustment were needed, but for custom design using discrete components it is fine.

 

[stinius]

Agreed.
Nice to have, not must have (less than 8 active devices).

The same applies to extendability to a fully balanced configuration, a la OPA1632.




Patrick

How many transistors is allowed?
IMHO it's not possible to design a low THD circuit with only 8 transistors, for the fun of it I designed one today, JFET ips, BJT "VAS", MOSFET "buffer" THD is in the "Guru" range.

BTW: I have dersigned a discrete differential balanced "opamp", but used more than 8 transistors, it's an all BJT design, so I guess that rules me out of the "competition" and the THD is also far too low for some.

Apart from that I guess it's not easy for you to start a GB thread with a "stinius" design.
"The worlds best discrete opamp, designed by "SW" approved by "JC" applauded by a lot of "engineers" and GEB's" Is in fact a much better sales argument.

Maybe someone could make a spec? It's not possible to design anything without a detailed spec.

Cheers
Stein

 

[benb]

I presume performance would be optimized by matching the beta of the LTP and complementary transistors in these circuits? Most of these signal-level transistors are pretty cheap, and one can buy lots of them for matching. Might even want to connect complementary pairs' cases together physically to give better tracking with temperature - a consideration when making a layout.

 

[dirkwright]

I presume performance would be optimized by matching the beta of the LTP and complementary transistors in these circuits? Most of these signal-level transistors are pretty cheap, and one can buy lots of them for matching. Might even want to connect complementary pairs' cases together physically to give better tracking with temperature - a consideration when making a layout.

Yeah, that's usually the case.

 

[dirkwright]

I think this may illustrate somewhat:

http://www.diyaudio.com/forums/solid-state/189123-anyone-built-power-diamond-buffer-amp.html

and here:

http://www.diyaudio.com/forums/soli...iscrete-medium-power-opamp-3.html#post1549752

May or may not have application here.
I shoot from the hip, usually loaded with blanks or duds...

_-_-bear

I am led to believe that this is Chris Paul's work, and was perhaps published in a mid 80s (?) audio amateur if not elsewhere. I do not have the original source. But have used this exact circuit altered with jfet inpoot and mosfet outpoot... (I did use a servo, iirc to keep the DC offset nil, but that is the easy way out...)

I have some power amp designs that use a modified diamond buffer output. Bob Cordell mentions them in his book.