diyAudio (
-   Solid State (
-   -   High loop Gain Composite Op Amp Circuits (

jcx 15th November 2004 05:16 PM

High loop Gain Composite Op Amp Circuits
2 Attachment(s)
Walt Jung has long been an advocate of 2 op amp (or op amp + buffer) composite circuits and points out their advantages – check his web site; “Op amp Audio” series

Gerald Graeme has also advocated 2 op amp composite amplifier, with high loop gain from 2 pole responses – “Amplifier Applications of Op Amps” 1999 devotes a chapter to composite op amp circuits and their compensation

This sim circuit topology is close to circuits I have built and I have verified that they can achieve measured 1KHz harmonic distortion (and 1KHz IMD from 1:1 10KHz + 11KHz) below my 160dB instrumentation limit

Composite amplifiers can reduce active device contributions to audio frequency distortion to unmeasurable levels (1 V input, 20 KHz sim results: )

The output op amp is a modified integrator that provides large audio frequency gain, those in the know will object that the loop phase response, which exceeds 180 degrees around 100 KHz invites oscillation or poor recovery from clipping

(the Loop Gain Probe is available from the LtSpice Yahoo group files area - just cut it out and delete .prams lines if you don't want to use it)

The clipping response is greatly improved by the symmetrical clamping diode bridge that holds the input op amp to +/- 1 V output when the output op amp saturates, this means that the input op amp is held in its linear operating region, its contribution to loop gain and phase shift is minimized, and its output only has to slew ~ 1V to when recovering from saturation ((1.6 V @ 20 KHz input, V(n005) is Lt1022 output)

Simple back-to-back diodes would also provide for the improved clipping response, but the bridge reduces diode capacitance and nonlinear conduction current that add directly to the feedback error voltage and cause nonlinear distortion

(distortion and phase margin # may be off slightly, I plotted those with all 1N418 diodes before it occurred to me to reduce clamp voltage with Schottkys, this means diode feedthru may be slightly higher due to higher capacitance form lower reverse bias on D1,2)

jcx 29th June 2006 03:26 PM

2 Attachment(s)
I thought I might collect some of these multiloop/gain boosted composite op amp ideas, this thread is already here...


Originally posted by jcx
Floating supply op amps may be modeled with some of the “multiple pole-zero” models that TI and Analog devices have presented in their app notes

But they don’t use them uniformly or often, many Boyle style macromodels are still being released

So you have to look inside the model to tell – Spice node 0 should not appear inside a op amp model

Once you start digging in you realize how poor many models are and you might just roll you own with 1-2 poles and a output resistance to generically play with bootstrapping

Here I point out some of the problems I had with the author of the AD8610 op amp model:

The OPA227 model seems to get gain right – I‘ve not tested whether output current appears plausibly in the ps terminals – yet another layer of modeling accuracy

I have been playing for several years with power supply bootstrapped op amp circuits

One previously known composite op amp with power supply bootstrapping showing loop gain enhancement from the bootstrapping of the internal voltage gain stage is from Mohapatra and Sandman’ 1980 GB patent

The sim shows Mohapatra&Sandman’s topology with the OPA227 model, I call this a feedforward bootstrap topology

The loop gain is substantially boosted by the bootstrap connection when the load impedance Rload is high

Because the OPA227 is in series with the bootstrap amp there is no improvement in its output impedance so you can see the gain boost disappears with heavier loading

The “loop error” I’ve plotted (in green) is essentially the inverse of the loop gain – at 10 KHz with the high load impedance you can see 100 dB loop gain => 1 GHz “GBW”

jcx 29th June 2006 03:33 PM

2 Attachment(s)
over at headfi I'm encouraging the use of a multiloop topology with the TPA6120 (= THS6012)

posting here with the asc for those wanting to play with LtSpice

mikeks 29th June 2006 04:01 PM

Good stuff jcx...Cheers!

Christer 29th June 2006 04:22 PM

Rod Elliott used to have an article at his site with a headphone amp consisting of a VFB + CFB combo. It was never a build project, but something he did to try out some samples of CFB ADSL drivers he had got. Unfortunately, that article seems not to be around anymore, and he now uses only the CFB amps as they are. Anyway, as far as I remember, his idea with the combo was to get the precision of VFB combined with the speed of CFB, and he said then it was the best headphone amp he had heard. I don't remember the details, whether there was any local feedback around each op amp, for instance.

Samuel Jayaraj 1st July 2006 12:36 PM

Elektor carried an article using VFB+CFB to make a very fast Composite Opamp. It was a concise article plus schematic by T.Giesberts.

mfc 2nd July 2006 12:00 AM

Re: High loop Gain Composite Op Amp Circuits

Originally posted by jcx
(the Loop Gain Probe is available from the LtSpice Yahoo group files area - just cut it out and delete .prams lines if you don't want to use it)
Hi jcx,

I'd like to use the Loop Gain Probe. I downloaded it from
the LtSpice yahoo group. Using your circuit, I set .param prb=-1
and ran it. No errors in the transient analysis with .asc and .asy
files in their proper place.

I then did an AC analysis run and I can get a plot of Vout in
magnitude and phase. it looks flat out to ~300Khz.
How do I tell LtSpice to show a plot of loop gain in magnitude
and phase?



jcx 2nd July 2006 03:59 AM

the loop gain probe uses both V and I sources to test return difference which gives the correct result anywhere in the feedback loop - the usual simplified test with just a V source works best when the impedances are mismatched by the largest amount possible so that information/energy transfer is well approximated as flowing in one direction only

to use the loop gain probe you have to run 2 analysis, as in the commented out .step in the example above, then the correct value for the loop transmission (Bode's T commonly (and sometimes approximately) "return difference", “return ratio”, “loop gain”, ”excess gain”, “feedback factor”) is given by the complicated expression at the top of the sheet, you have to copy that into the plot as the trace definition you want to display (clik on the expression, copy its text, then go to plot>add trace and paste the expression at bottom of the box)

I don't know where all of the Spice syntax for that expression is documented - the @1, @2 suffixes refer to the trace "steps" - the multiple passes of sim caused by the .step command

Loop Gain probe theory:

middlebrook paper:

mfc 2nd July 2006 05:06 AM


Originally posted by jcx given by the complicated expression at the top of the sheet, you have to copy that into the plot as the trace definition you want to display (clik on the expression, copy its text, then go to plot>add trace and paste the expression at bottom of the box)

I cut and pasted the expression:

(note: :p is actually the text chars colon followed by p)

and LtSpice complains about unknown current requested: I(V:p:i)

so I found a .plt file that had this expression:


and I cut and pasted this expression and was able to get
it to work. It duplicated your results.

Looks like the names used in the loop probe changed or something.

Thanks for the help,


jcx 30th December 2010 09:29 PM

Black's Error Feedforward is a Composite amp too - is sub ppm good enough?
3 Attachment(s)
I grabbed Bob's example sim from the TPC/TMC debate in the Cordell Book thread

my poking at it seems to indicate that the distortion is limited by the slow output Q and simply can't be made much better by feedback - especially the THD 20KHz that Bob likes as a simple quality metric (at least without going deep into conditional stability territory with 3-rd order or higher loop gain schemes)

if you can't fix up a amp with more feedback then the other tool is Black's Error Feedforward - rare in audio because of the difficulties power output combiner but famous none the less due to Quad’s "Current Dumping" amplifier

a error feedforward amp is defined by having 2 independent power amp paths to control the output - and the consequent difficulty of combining the outputs without their mutual loading causing unwanted additional (destabilizing) feedback

The VanderKooy Lipshitz Error Feedforward paper shows several schemes - "Current Dumping" actually wraps negative feedback around the "dumper" amp stage - but 2 other schemes are shown - I chose the Parallel Feedforward from their fig 4b - which doesn't intentionally add loop feedback to the main amp and the coupling through the output impedances of the amps is minimal when used with a frequency selective power combiner

so the following is a Composite Amp using a large, slow discrete amp and a fast error feedforward op amp path
the instrumentation op amp measures the error at the diff pair input of the main amp with a gain of 10x, and then the CFA DSL driver op amp integrator provides the required error correction V drive for the Resistive leg of the L,R power output combiner network

Bob's THD 20KHz number is improved by ~20x in this sim

you will need Bob's zip file and to put my asc in the same folder with Bob's Cordell Models.txt file or you can download Bob's transistor models from his
book's site

All times are GMT. The time now is 03:13 PM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 18.75%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio