MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

Mixer Crosstalk
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Author Mixer Crosstalk
cackland
Recently amongst many projects, I built an 8 channel stereo mixer. Its the usual inverting summing amplifier with a unity gain feedback network. 100k ohm input impedance.



I've noticed there is crosstalk between the adjacent channels. For example:

I patch a square wave into one channel, with the level potentiometer all the way closed. If I open the channel above or below, a high passed version of the adjacent signal comes through.

It only happens on the adjacent vertical channels, not the the others.

How would one correct this? Is there a general rule of thumb that I've missed.

EDIT: After reading up, I've come to the conclusion its my signal routing. I've routed each channel in a vertical fashion which explains why its only the vertical channels and not the adjacent column channels receiving crosstalk.

Would you recommend having a decent ground plane between each channel trace to reduce crosstalk? Other recommendations?

Schematic.



Kampfzwerg
By no means am I an expert, but the schematic looks ok to me.

Could there be PCB traces with the inputsignal running in parallel?


'cuda
cygmu
I had crosstalk issues on a banana-format output module which were much improved by reducing the input impedance from 100k to 10k. In the thread where I asked for help it was pointed out to me that pro audio gear uses 10k as a matter of course.

I wonder if you might also get some improvement from grounding the switch connections on the inputs.
cackland
Kampfzwerg yes, I edited my initial post with that info. The input traces are running parallel and in close proximity.

That is a good question gygmu, and wonder.

I would say from my last hour of crosstalk research, I bet its my routing. I'm redoing the board layout with ground plane between each trace and decent distance between input traces.

smile
BugBrand
Agree on dropping the resistor values.
Routing-wise - keep the virtual ground traces short - ie. pin 2 of your mixer opamp where the summing resistors meet. So put the summing resistors near the opamp rather than near the pots.
cackland
BugBrand wrote:
Agree on dropping the resistor values.


From my understanding, a higher input impedance can have a higher noise influence, however less load on the input source.

A middle ground of 47k should work? As have a bunch of 0805.

If I went 10k, would I change the potentiometer resistance as well? Or stay at 100k?
cygmu
My naïve understanding of this is that the goal of reducing the resistor values is to lower the voltage induced by crosstalk current.

Long parallel traces that are close together behave as if there is a capacitor connecting them. (Well, there is, I suppose.) So a changing voltage on one will induce a current on the other. This current generates voltage when it passes through a resistance, and that voltage is what you hear as crosstalk.

It’s quick and easy to try lowering the resistors in the op amp sections just by putting something in parallel on your existing build. Might as well try. The pots could also be worth changing but that’s a bit more hassle of course.
cackland
Thanks cygmu.

I've rerouted a new pcb and added it to an existing order.

I'll try the parallel resistor suggestion, or just bake a new board with new values.

I will compare the two and see the difference.
Kampfzwerg
Hey cackland

Please Keep us posted about your Progress.

I will have to build a similar cirquit in the future for my
Oberheim 8voice clone.

Thanks
home_listening
I'd also move the 100nF cap to pre potentiometer
cackland
home_listening wrote:
I'd also move the 100nF cap to pre potentiometer


Ive never seen this in all the schematics I've come across. Reasons?
djs
General question- why even have the 100nf capacitors at all?
cackland
AC coupled mixer - It’s an audio mixer. Removes any dc in the signal and acts like a input high pass filter of around 15hz.
cackland
Update:

So in my quest to improve my design, reducing the crosstalk and the noise floor to a minimum, I adjusted my pcbs layout to allow for some grounded between the signal traces.

I received the updated board, assembled and tested.

Results:

No crosstalk between inputs when the adjacent channel knob was opened.

However, I've introduced some other anomalies which believe could be my layout again. Channel 1, 3 and 5 all have noticeable bleed. Channel 1 and 3 are noticeably on the left, and channel 5 is significant through both left and right. I can only assume is now my power traces that might be inducing the bleed, especially for the equal bleed on channel 5 which runs parallel to the +-12v traces. If so, that can be fixed with another revision, however the LEFT ONLY bleed for channel 1 and 3 doesn't make sense.

I've attached my eagle files for those who want to dive in deeper. Or see the attached layout.

Eagle files (Google drive)
https://drive.google.com/open?id=1m068AS37C4HCsIH-g9fRyHlfs_jt9VeC



Any advice or recommendation on my pcb layout, welcome smile

P.S... I like symmetry.
guest
in what way are those channels bleeding? is it L/R crossover on the channel, or is it channel 1 bleeding into 3, etc?

also, as to AC coupling caps before or after a potentiometer, i tend to put them before. its not super critical, but i think there are some advantages to putting them before. if you have a DC signal being applied, then there is no longer DC across the pot. this is said to extend the life of the pot and reduce long term noise, as the constant electric potential increases corrosion. not sure how big of an effect this is. it also gets rid of DC thump when rotating the pot. if you have DC on the pot, and you rotate the wiper, the voltage at the wiper will modulate with its position. since this is changing with time as you rotate, this gets passed through the AC coupling capacitor, and is seen at the output.
cackland
guest wrote:
in what way are those channels bleeding? is it L/R crossover on the channel, or is it channel 1 bleeding into 3, etc?


No, there is no bleed into other channels.

For example, each channel is normalized. When I patch a signal into the left input of channel 1 and the pot is completely closed, some signal is heard in the left output. This is the same for channel 3.

Channel 5 has equal bleed for both left and right.

The bleed is only high frequency content from the source. My test is directly from an oscillator using a square or saw output.
guest
i would be suprised if it was capacitive coupling to the powersupply, as you have decoupling caps that should swamp any of that out, and then the opamp itself has a high PSRR. for ch1,3 there is a fair bit of trace overlap as the signal enters the pot, and then leaves. there could be some coupling going on there. for ch 5 i cant see anything egrecious, but it could just be the pot itself radiating to components underneath it.

one other area of bleed is finite off resistance of the pots. the wiper lead, wiper contact, and the end of travel of the pot all have some small resistance asoociated with them. this varies from pot to pot. so when you turn it all the way off, there still might be a few ohms down there, which will allow some signal through.
cackland
Any improvements to the routing or additional circuitry that might help?
guest
well, you could try disconnecting the output capacitor of ch1,3 and reroute with wires to see if its the culprit. then if it is, you move that trace so its not right on top of the input trace. in general, you dont want traces running any appreciable distance parallel to one another. if you can put some ground between them, that helps. but, when they are on opposite sides of the board, and its a 2 layer board, you cant really do that.
cackland
You mean the decoupling caps from the pots?

So even though the traces are on opposing layers, they can still interfere with one another?
jorg
cackland wrote:
So even though the traces are on opposing layers, they can still interfere with one another?


Yup. I've encountered that recently - they weren't even parallel; they only crossed over each other on opposite sides of the board. In high-end boards, interior ground & power planes are used to reduce this problem. In two-layer boards, avoid parallel lines even on opposite sides, use ground fill, pay attention to large signals near small ones, keep your impedances reasonable.
Mungo
guest wrote:
for ch1,3 there is a fair bit of trace overlap as the signal enters the pot, and then leaves. there could be some coupling going on there. for ch 5 i cant see anything egrecious, but it could just be the pot itself radiating to components underneath it.

one other area of bleed is finite off resistance of the pots. the wiper lead, wiper contact, and the end of travel of the pot all have some small resistance asoociated with them. this varies from pot to pot. so when you turn it all the way off, there still might be a few ohms down there, which will allow some signal through.
Well spotted, these two actually combine with each other and could easily be 100dB or worse of crosstalk.
cackland
Ok thanks for the advice guys smile
devinw1
Good posts everyone! I've learned a lot about good board layout from the friendly folks here and the internet in general. The devil is indeed in the details!
guest
another thing to keep in mind, is that the signal traces are only carrying half of the signal, and the ground plane is usually carrying the other half (on a fully differential line its not the case). so you have to keep track of where those signals are flowing as well. i tend to put a lot of ground plane vias on the board, just so that current can find its way back without having to take some long journey around the board. at lower frequencies and lower power this typically isnt a problem, but it also isnt too much trouble to add a bunch of vias (assuming youre not drilling your own board!).
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Page 1 of 2
Powered by phpBB © phpBB Group