LM3886 GC with LM4562 tone control

[Mooly]

Originally Posted by blu_glo

Mooley also pointed out quite correctly that putting the feedback capacitor in lost your power amplifiers DC reference (can be on either of the two input pins but must be on one or the other). This is remedied by a >>22K say 100K resistor value (somewhat arbitrary but must be much greater than the volume resistor in value).

Pin 10 is the one that must be referenced to ground.

Pin 9 may or may not be according to the design requirements. If it is DC coupled (no 33uf) then the amp will have a large(r) DC offset than needed and will "amplify" its own internal offsets as they drift. AC coupling returns the gain to 1 at DC and minimises these issues. The input bias currents (the tiny DC current that flows out of or in to the input pins) should be equal for both inputs. Or more correctly the volt drop developed by those currents across the input and feedback resistors should be equal. Keeping the resistor values equal achieves that. So 20K feedback resistor and 20K input bias resistor.

For example, 0.1 ma bias current (just to illustrate, it would never be this high) would develop 0.0001 * 20K or 2 volts at pin 10. It would also develop the same voltage at pin 9 with the bias current flowing in the 20K feedback resistor. Same voltage on the inputs means no difference in voltage between the inputs. So the output is zero volts DC... all good. Now change either 20K to some other value, say 30K. We now have 0.0001 * 30K which gives 3 volts on the appropriate pin. Result... a DC offset error, one that is made worse if the feedback is also DC coupled because the chip can amplify that error.

So keep the DC currents equal by keeping Rin and Rf equal and AC coupling the feedback return.

[blu_glo]

Thanks for tidying that up, I forgot to finnish saying where that resistor should go; and yes I agree it should remain same value as the feedback resistor!

[Mooly]

All he has to do is go and build it now

[mxriddler]

Originally Posted by Mooly

The "output end" of the chain (the 10 ohm or 2.7 ohm in your new diagram) is at low impedance and unaffected by whats going on around it and any extra PCB length. It is the inputs of the chip that would be sensitive to extra length and the risk that brings of stray pickup. I wouldn't like to say yay or nay to the technique without verifying it and designing a suitable PCB to match. Stick to what you have drawn.

The times 4 buffer would bring a huge gain change as you say.

C1 and C2 (C3 ???? in the tone control) The inverting buffer (in the tone stage) has an input impedance determined by the 20K. So that determines how big C1 has to be and that in turn is determined by you and what you deem a suitable cut off point. Typically a designer would work to <1Hz for these. So that implies at least 10uf. Why not make the 20k's bigger ? Say 180K. That brings the added bonus of a higher input impedance too and allows a 1uf to be used. C2 needs to be bigger as it drives the low impedance tone network. I would use a 100uf electrolytic here. There is no right and wrong in all this. They are just choices. The 1k and 33uf in the power amp come in at around 5Hz (which is fine for the final power amp stage)

Just spotted an error... The 2 uf cap in the power amp. You need a 20K (or 19 to be correct) to ground from this cap. The pin is floating at DC otherwise and the amp can not establish its DC conditions.

Sorry, I meant C6 (not C3!). Does C1, 2, 6 need a resistor to ground too, like Cin? I guess I'm a little confused what C1,2,6 are doing. I thought they were there to block dc. A 5hz or less cut off point is fine for my usage.

thanks!

[mxriddler]

Originally Posted by blu_glo

also the "bypass line" should go to the other side of C1.

Why should the bypass line go on the other side of C1? Wouldn't I want as few capacitors on the signal path as possible? With Cin after the pot, that should block any dc offset from the input right?

[mxriddler]

Originally Posted by Mooly

All he has to do is go and build it now

I'm looking forward to it!

[Mooly]

Don't be afraid of caps in the signal line

It should look something like this... I've left the bypass as it was.

The 470K's define the DC conditions on the caps when they could otherwise be "floating". It stops load thumps and bangs when switching and connecting.

[mxriddler]

Originally Posted by Mooly

Don't be afraid of caps in the signal line

Thanks a lot. Working on making the changes now.

Not afraid, just cautious based on what I've read, but of course I should always be cautious of what I read... Caps seem to be a hot topic for some. I'm learning by reading/experimenting, not by any actual electrical training...

Since the values are so high, they'll need to be electrolytic, would there be any benefit to a poly cap bypass on the electrolytics in the signal path? I built the Millet Starving Student Hybrid headphone amp and used a poly bypass on the output capacitors and did notice a difference, maybe that's a different case though.

Nichicon KZ or FG a fine choice here?

BTW, here's my MSSH I built last year. Other than a cmoy, this was my first real amp project. I'm pretty happy with how it turned out. I use it on my desk at work most of the time. Funny how many people are somewhat afraid of glowing tubes for some reason...

[Mooly]

That looks nice Excellent job.

The big thing to concentrate on with the chip amp is the layout and grounding of it all.

Did I post a link to this ?
3 stage LIN topology - NFB tappings?

you have to get this right.

[blu_glo]

Originally Posted by mxriddler

Why should the bypass line go on the other side of C1? Wouldn't I want as few capacitors on the signal path as possible? With Cin after the pot, that should block any dc offset from the input right?

I did this to isolate your "direct" circuit from potential DC offset from an external source that could otherwise cause a speaker thump. As Mooly says do not worry we are not trying to convince you to fit TOO MANY capaitors, just enough to do the proper job!