Crosstalk headaches!!!

[dschwartz]

Hi there!
I have been struggling to debug my latest design..its a two channel high gain preamp (with a tl074).but one of the channels is always ON, for stereo purposes..
But if i use channel A clean and set channel B (which is always on) on high gain, i hear the distorted sound of channel B on the background of channel A..hence..crosstalk.
My first guess was the jfet switching leaking signal, but the buzz is there independently of the volume, so no..
Then i suspected V+ is being modulated by the clipping..added a 220uF cap between + and gnd..improved a little, but no gold....
Then, i suspected Vref is leaking signal between the tl074 inputs (all stages are inverting).. I'm buffering Vref with a tl072 for all stages and other circuitry, so i can't use a capacitor to filter Vref more (i asume because of the capacitive loading?)
So i added a 10k ohm resistor for each "+" input of each inverter stage to try to isolate each stage Vref..
No improvement.. even with both volumes at 0, if i crank the gain of any channel, i get faint signal on the output..
I have the volume control referenced to Vref, so I'm still thinking vref is being modulated, and its shared by the rest of the circuits.

Is it a good idea to have Vref buffered? Or should i use simple resistor /cap based Vrefs for each opamp package in the system?

[MikeA]

A schematic and physical layout would be helpful in answering your questions...  crosstalk is usually a higher level signal bleeding into a lower level part of the circuit, which is then amplified.  Without more info, I'm initially thinking to look at the locations of traces, wires, jacks, etc. for a proximity of high and low level signal paths. 

[anotherjim]

It might be worth trying a resistive Vref temporarily, but unless you have something deliberately running signal current via vref (like a diode clipper or volume pot return) I wouldn't expect interaction that way with what you started out with. Probably more chance of xtalk in the quad amp itself. If you suspect that and the quad is socketed, you might make a daughterboard to plug a pair of duals in -  probably easier to start again?

[duck_arse]

if your circuit has a dual and a quad, and the quads is the problem, why not reorganise and shift the always on to the dual, and the Vbias buffer to the quad? of course, only as a last resort fix.

[dschwartz]

I don't think it's internal xtalk on the quad opamp because if i turn both channels volume to zero, i still have bleed through the output (volumes are referenced to vref)
Maybe a tl072 is a terrible vref buffer.. I'll try a ne5532 to see if it improves..

[iainpunk]

if i were a betting man, id bet that its due to close traces, close wires or other close proximity signal paths.

do you have it on
A> 'vero board'/'strip board' (i strongly dislike this)
B> 'pad per hole'/'perf board' (the only proper type)
C> breadboard (only acceptable for prototyping or temporary circuits [although i used to gig with a breadboard pedal])

because A and C are notorious for crosstalk, since the traces are mostly long and parallel and provide big amounts of capacitance which in turn can cause crosstalk.

cheers, Iain

[dschwartz]

Well, it's on a double sided pcb with ground pours and smd components..
It could be wire crosstalk, but it would change if i move them around..
I'll have to discard Vref interference before searching more obscure reasons

[iainpunk]

post your schematic!!!!!!!!!!!!!!!!!

where did your PCB come from, designed it yourself? post pcb layout!
we need way more info

cheers, Iain