LM3886 guitar power amp design - adjustable output impedance

[Delta]

I know this is basic stuff, but I never underestimate my ability to make an arse of things.

I'm building a 25W guitar amp, the preamp / tone stack / soft clipper is from Rod Elliot's design with minor tweaks.  I'm using a LM3886 as the output stage, with +/- 24ish supplies from a 18-0-18 80VA toroidal transformer.

I'm also using some current feedback to raise the output impedance, in order to reduce damping and let the speaker flap around, giving a more "valvey" sound.
From the equations given on another of Rod's articles, with a 0.2R shunt and a 18k / 1.2k feedback network, this should give an output impedance of around 3.2 ohms.  I will make this switchable - when the switch contacts are travelling, it will result in the LM3886 operating at unity gain for a split second, this shouldn't be a problem, should it?

By my calculations, 25W into 8ohms requires a peak voltage of ~20v, the preamp/clipper can provide around 1.2Vp, and the feedback network of 18k / 1.2k gives a gain of 16, this should allow it to hit 25W but prevent the LM3886 from being driven into clipping.

A feedback capacitor of 4.7uF gives a low frequency pole (is this the same as the -3db point?) of 28Hz, I might even decrease the value of this cap to raise the freq, as the lowest note on a guitar is 82Hz anyway...

Is my 220pF cap across the FB resistor correct to avoid HF oscillations?
Is 100pF the correct power supple bypass for a power opamp?

Please give me your feedback (shit pun intended!).

[fcb]

I'd be more concerned about the risk of thumps from the gain switch.

Probably keep the LM3886 at a fixed higher gain and switch an attenuators on the front end - possibly with a buffer amp.

[Delta]

I'd be more concerned about the risk of thumps from the gain switch.

Probably keep the LM3886 at a fixed higher gain and switch an attenuators on the front end - possibly with a buffer amp.

The switch doesn't adjust the gain, it adjusts the output impedance....  Will it cause an audible thump?

[Audioguru]

The datasheet says the LM3886 is stable when it has a gain of 10 or more so yours will oscillate all the time with your capacitor parallel with the feedback resistor and will oscillate again when the switch causes a momentary gain of 1.

Will the increased output impedance of the amplifier cause the speaker to sound like a bongo drum?

[Delta]

The datasheet says the LM3886 is stable when it has a gain of 10 or more so yours will oscillate all the time with your capacitor parallel with the feedback resistor and will oscillate again when the switch causes a momentary gain of 1.

Will the increased output impedance of the amplifier cause the speaker to sound like a bongo drum?

Cheers for the info!

So the feedback cap will cause it to oscillate as the gain at HF will be unity (which is < 10!)?  I'll get rid of that cap then...

When the switch contacts make, will the oscillation cease immediately?  At what frequency is it likely to oscillate?  Audible frequencies?

I wouldn't have thought it would sound like a bongo drum, just a bit "looser", and maybe with a more even frequency distribution.  I'll let you know after I've built it!

[Audioguru]

The oscillation frequency will be far above audible frequencies but might overheat the IC and will burn out a tweeter.

[Delta]

The oscillation frequency will be far above audible frequencies but might overheat the IC and will burn out a tweeter.

Well given that it will only be at unity gain for a matter of milliseconds, I don't think I'll have to worry!

Does the rest of the circuit and my calculations seem OK?

[Kleinstein]

If the amplifier oscillates, this would be likely  in the 100 kHz - 1 MHz range. Not audible, but maybe a problem to some components like RC damping parts or filter caps.

It is rather simple to avoid the problem: just ad a resistor (e.g. 100 Ohms) to GND on the left side of the switch.

[fcb]

Why not use the switch to short out the 0.2R resistor?

[Delta]

Why not use the switch to short out the 0.2R resistor?

1) It would have to carry the fully speaker current.  (Not really an issue to be fair)

2) It would result in more power being delivered to the speaker load (no longer the resistor drop) with a slight increase in volume.  I want to be able to switch between low/high Z with the volume staying constant, in order to hear what difference it makes to the sound.  ie I want all other things to remain constant when switching.

[langwadt]

Why not use the switch to short out the 0.2R resistor?

1) It would have to carry the fully speaker current.  (Not really an issue to be fair)

2) It would result in more power being delivered to the speaker load (no longer the resistor drop) with a slight increase in volume.  I want to be able to switch between low/high Z with the volume staying constant, in order to hear what difference it makes to the sound.  ie I want all other things to remain constant when switching.

put a small, say 10R,  resistor before the switch. when the switch is OFF you get 10R in series with your 1.2K i.e. not significant
when the switch is ON you have 10R in parallel with 0.2R also not significant

you could also replace the switch with a low value pot

[Delta]

put a small, say 10R,  resistor before the switch. when the switch is OFF you get 10R in series with your 1.2K i.e. not significant
when the switch is ON you have 10R in parallel with 0.2R also not significant

That probably the best idea so far; 10R from the switch common down to GND?

But are we worrying about nothing here?  The amp oscillating for 50ms or something?  There are no tweeters to damage, and I assume the oscillation will stop as soon as the contacts are made and it's back running at a gain of 16...?

[langwadt]

put a small, say 10R,  resistor before the switch. when the switch is OFF you get 10R in series with your 1.2K i.e. not significant
when the switch is ON you have 10R in parallel with 0.2R also not significant

That probably the best idea so far; 10R from the switch common down to GND?

10R from the 0.2R to the switch, switch wired to be open or short to ground

[Delta]

put a small, say 10R,  resistor before the switch. when the switch is OFF you get 10R in series with your 1.2K i.e. not significant
when the switch is ON you have 10R in parallel with 0.2R also not significant

That probably the best idea so far; 10R from the switch common down to GND?

10R from the 0.2R to the switch, switch wired to be open or short to ground

A-ha!  Got ya!  That seems like the way forward.  It will prevent any problems should the switch contacts fail too... 

[Synthetase]

A feedback capacitor of 4.7uF gives a low frequency pole (is this the same as the -3db point?) of 28Hz, I might even decrease the value of this cap to raise the freq, as the lowest note on a guitar is 82Hz anyway...

Personally I wouldn't for two reasons. You may want to play down-tuned at some point, or play a bass through the amp, or use an octave pedal. Also since it's the -3dB point, it means the frequency response as you go down has dropped by 3dB at that point, this means that you're losing response even at frequencies higher than the nominal value. You're not losing much, but you don't really want to lose anything, so keep the -3dB point well away from anything even remotely likely to go through the amp. 28Hz sounds like a good compromise to me.

By the way, I like the idea of adjustable output impedance. I use both the P27 pre and power amp circuits and mine is currently set to ~8R. The ability to play around with it at the flick of a switch is pretty damn cool.