LM1875 nested Gainclone

[carlmart]

Originally posted by linuxguru After playing around in LTSpice for while, I came up with another nested gainclone concept similar to Walt Jung's composite op-amp/buffer.

Have you tried in LTSpice using WJ's multiloop concept putting a chipamp at the output?

I mean something following the original Jung's values:

http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_4.pdf

My idea had been to replace the output chip for an LM3886, but I am not sure if they are unity-gain stable.

[AndrewT]

Originally posted by linuxguru

In the sims, the LM1875 can be used below the recommended gain of 10 with suitable compensation. However, I'm not sure that it extends all the way to a gain of 2. For unity gain or similar operation, the TI/Burr-Brown OPA549 family may be more useful, since they're unity-gain stable.

[carlmart]

I did read that comment on the OPA549 being more suitable for that unity gain job, and in fact the 3886 datasheet advises a minimum of x10 closed loop gain.

So it looks as if the Jung concept might not be applied if using that chip.

[AndrewT]

another member says it's possible to stabilise the 3886 down to 3times. But, he did not specify the phase margin.

[jcx]

I worry that stability considerations for "nesting" slow, high minimum stable gain, power chip amps in effective multiloop amps are non-trivial

Jung's multiloop examples all use much faster output amps such that the input op amp's dominant pole compensation is fine for the whole loop since the fast output amp doesn't contribute very much phase shift at the unity loop gain intercept

with power chip amps you can't afford any extra phase shift at their (GBWmin )/ (Min stable gain)

that means that the input op amp's contribution to the outer loop gain must be flat from way below that frequency to >4x the power chip amp's max GBW/min stable gain

on the lower side this can mean you have to roll off the input op amp's gain to 0 dB near 20KHz - meaning your nesting isn't able to improve the power amp's behavior at or beyond that frequency

some have relied on the power amp never being worse than the typical spec on GBW - not a safe practice for diy builders without oscilloscopes and some knowledge of how to test for stability

I have built several multiloop amps with fast output devices – the most “powerful chip amp” I’ve used for the output is the LT1210 – but with its typical op amp Vsupply limits it wouldn’t go far with 8 Ohm loads

I would look for faster power amps – anyone know the GBW min spec for TDA729x DMOS chip amps?

[mwmkravchenko]

Anyone think about the LM12 as an output device?

Mark

[carlmart]

Originally posted by mwmkravchenko
Anyone think about the LM12 as an output device?

Things might get a bit expensive if you do. The LM12 is far from cheap.

[linuxguru]

I had put this on the back-burner for a few months, and then took another look at it over the last week. I added a lifted ground and tweaked the compensation schema for better stability, and the THD20 is even lower and mostly even-harmonic. It now looks good enough to build a prototype.

A few notes on the schematic below:

1) Vi and Ii are injection sources using the Tien-Vishwanathan method for the open-loop gain simulation plots only. In the real circuit, Vi is a short and Ii is open.

2) R6 is required for the simulation to converge quickly - it's the equivalent of a base-stopper in discrete topologies.

3) The outer-loop gain and compensation is set by R8, R9 and C7.

4) The inner-loop gain and compensation is set by R4, R5 and C6. C6 is not very critical - any value above 22pF seems to work fine, with increasing stability (as well as THD) as C6 is increased. The 33pF value shown is a good compromise; higher values like 47 pF or 68 pF can be attempted on the first prototype for improved stability.

5) THD20 simulates at ~ -130 dB at 20W into 8 ohms, with a unity-gain phase margin of ~85 degrees.

6) No Boucherot/Zobel network is required.

[linuxguru]

Here is the THD20 FFT plot at 20W into 8 ohms - THD20 is ~ -130 dB, mostly even harmonic. At lower swings, the odd harmonics almost vanish.

[blueskynis]

Little suggestion,

I don't like high impedance feedback circuits. Lower a bit the impedance of feedback circuits to decrease noise (main reason), DC offset and possibly THD (because input currents are highly nonlinear). Also, R1 could be lower to decrease noise. Besides, nice work, keep going

Cheers,

Nesa