Problem with parallel LM3886 (PA 100)

[mnturner]

hey guys,
i've just finished a paralleled LM3886 (PA-100 style) based power amp and i'm having a weird problem.

The two LM3886 amplifiers are on separate boards and have their inputs connected by a wire between the two boards.
One board has the zobel network and damped series inductor, the output of the other board again is connected to this one by a wire.
The resistors are all 1% metal film type, hand matched to about 0.5% in the signal paths.
The two sets of four 1Ohm paralleled resistors at the outputs of each LM3886 were used because I couldn't get the 0.1Ohm, 1%, 3W resistors suggested in the application note.

At this stage, both boards work perfectly when not connected to each other, but whenever i connect them, they oscillate rail to rail at high frequency and as a result heat up very quickly.

I have connected the positive and ground terminals of my scope probe to each output (positive end of probe to output of one board, ground lead of probe to output of other board) with the boards unconnected, in an attempt to get an idea of how different the ouput signals were, and there was no visible difference *except* when the volume pot (25k, log) was between about 2 and 4 0'clock, when there was a high frequency signal of about 150-200mV difference between the two outputs. When the volume was turned up more that this, this signal disappeared and the difference once again became negligibly close to 0mV.
I tried to test this because I was under the impression that the biggest problem with the parallel setup is that differences in the outputs cause the two amps to drive each other, although I don't know if that would actually cause the oscillation I'm seeing at the output when the boards are connected.

The power supply is an unregulated 24-0-24 t/f with 9,400uF at each rail, which is putting out about +/-38V.

I have tried with 4Ohm and 8Ohm loads as well as unloaded and the same problem occurrs.

I'm pretty confused. I'm open to the possibility that the fact that the two amps are on two different boards is causing problems due to interference etc.

Does this sound like a grounding issue? I was under the impression that ground loops caused hum, not huge oscillation.

Anyway, if anyone has any ideas or suggestions I'd love to hear them.

[infinia]

1) A picture of your test setup would be helpful>
2) use 2 probes instead of one w/ground/earth at an output of an amp. Even when your PS is floating there is a HF ground loop! Or Measure gain difference with an RMS tone and a DVM. After the gains are matched as close as you can also look at square waves for signs of instability.
3) Your problem is most probably you need a seperate RC zobel at the output of each amp closer 0.1 UF and 4.7 ohm.

[mnturner]

hi infinia,
sorry, the zobel is actually 0.1uF and 4.7Ohm, that was an error in the schematic.
i'll get pics up if i can.
what did you mean by there being a HF ground loop when my power supply is floating?
i tried testing the difference between the outputs that way because i only have a single trace scope with the one probe.
I have measured the gain difference with an RMS tone and a crappy DMM, as far as i could tell the gains were very well matched.
I'll give the extra rc zobel a shot and see if it helps at all, but i thought zobels were intended to counterract the changing impedance of a speaker load; this problem occurrs even when the amp is unloaded.
Thanks for the reply,
mitch

[leadbelly]

The 1st thing I would look at is better termination at the input. Terminate both inputs separately: don't rely on a pot, it could easily be the source of the instability itself.

[tangmonster]

first thing I would do is lower R1 and R13 to about 1k.

if that doesn't help then:

Are you using a 47k pot?

try and change your input to look something like this schematic:

PA100 DIY 2x LM3886 in parallel gainclone audio amplifier

i am guessing you used that as a design guide anyway?

You can put the pot before the input.

I don't see the use for two zobel networks on the output? and c7 should be 100nF and not 100pF.

[infinia]

Hi Mitch
the RC zobels are to keep the amp happy near the unity gain freq ie stable phase margin. They like to be located near the IC as well as PS decoupling 220uF + 0.1uF.

cheap DMM sb be ok using low freq tones 400Hz and near full scale. 1.990 V
Let's calculate the gain mismatch if you want. Don't forget to look at square waves.

[infinia]

Originally Posted by tangmonster

first thing I would do is lower R1 and R13 to about 1k.



I don't see the use for two zobel networks on the output? and c7 should be 100nF and not 100pF.

Hi TangM
Agreed about input circuitry. I think the impedance looking out from non-inv input should be close to Rf at DC for offset and noise.

Normally 1 zobel should be good, but this is on 2 separate boards with unknown to me (wired ground return paths and output inductance).

[semi932]

This may not be sceintific but try tying both boards speaker grounds together. seperately.

[Sebastiaan]

Dear,

1: Measure the DC offset from each individual amplifier.
2: Route all grounds to a separate star ground, (give separate ground tracks to the Zobel, decouple caps, and line signal ground).
3: like others suggested, make the input impedances equal.

In my opinion it is never wise to build chip-amps in parallel without a DC servo. A few DC mV. difference can cause huge current loops between the two chips.

With kind regards,
Bas

[infinia]

Input mods