Author Topic: Audio isolation, am I doing something stupid  (Read 825 times)

0 Members and 1 Guest are viewing this topic.

Offline loki42Topic starter

  • Frequent Contributor
  • **
  • Posts: 290
  • Country: au
Audio isolation, am I doing something stupid
« on: November 30, 2022, 10:09:49 pm »
So I have a board with pwm driving rgb leds, matrix cap touch and a micro combined with an analog audio signal path of filters and jfet gain elements.  The analog path has a lot of gain ( from milli volt to 8 v p to p) and I'm picking up noise from the pwm leds and the cap touch charging. 

I have tried many board layouts (4 layer board) but the only way I've found to suppress it is to have transformer isolated power for the noisy pwm / touch elements and then communicate with them using an isolator (digital or opto). If there another option?  I feel like I'm making this more complicated than needed but I'm following all the neat how to lay out your analog board and care about return paths guidelines and they don't help.
 

Offline ConKbot

  • Super Contributor
  • ***
  • Posts: 1381
Re: Audio isolation, am I doing something stupid
« Reply #1 on: November 30, 2022, 10:44:56 pm »
Obligatory show layout and schematic. But the problem with gross audio frequency return currents is that they just go wherever they want unlike well behaved high frequency return currents that just stay under the trace.

If you end up with a ground gradient of even tens of microvolts, that can be gained up to mv of noise on your signal depending on your circuity.  I.e. let's say a 2nd stage amp with a gain of 100 is being fed from the first stage amp 25mm away.  The 1st stage amps output voltage is G*VIn, but only from where the feedback resistors attach to vout and GND.  stage twos Vin will be "measured" from it's input to the GND of the 2nd stage feedback divider.  If there is a 100uv gradient over that 25mm from the GND pin of the two feedback dividers, from that PWM, it couples into the signal, gained by 100 and shows up as 10mv of noise in stage two's output.
 
The following users thanked this post: thm_w

Offline ajb

  • Super Contributor
  • ***
  • Posts: 2582
  • Country: us
Re: Audio isolation, am I doing something stupid
« Reply #2 on: November 30, 2022, 11:05:04 pm »
Would need to share your schematic and PCB as ConKbot says.  Full isolation of power and signal is kind of an extreme solution, so chances are there are better options.  If it is a problem of noise appearing in the ground plane, then one strategy worth looking at is slotting the plane to control where the noisy return currents flow.  Note that this is not the same as *splitting* the plane, IE having two separate planes only joined at one point, rather it's strategically placing gaps in the plane to block return currents from flowing through specific areas where your sensitive circuitry is.  Same idea applies to power planes, if you have those and if the sensitive stuff has insufficient PSRR and/or you have signals referenced to the power plane (including voltage dividers).  But all this is just guesswork without being able to see the hardware you're working with.
 

Offline DonKu

  • Contributor
  • Posts: 40
  • Country: us
Re: Audio isolation, am I doing something stupid
« Reply #3 on: November 30, 2022, 11:34:11 pm »
Although my audio mixer amp uses isolation, you may get some mileage out of the Bill Whitlock link shown at Note 2.
 

Offline loki42Topic starter

  • Frequent Contributor
  • **
  • Posts: 290
  • Country: au
Re: Audio isolation, am I doing something stupid
« Reply #4 on: December 01, 2022, 12:04:37 am »
Obligatory show layout and schematic. But the problem with gross audio frequency return currents is that they just go wherever they want unlike well behaved high frequency return currents that just stay under the trace.

 Is a possible solution to make sure I only reference a virtual ground that I create off something like a shunt voltage reference? I'm using a bipolar power supply currently though. My negative rail is a little noisy in very high frequencies because it's generated by a charge pump but it's got a low noise LDO on it, and the system performance seems the same when I swap it for my bench supply.


Same idea applies to power planes, if you have those and if the sensitive stuff has insufficient PSRR and/or you have signals referenced to the power plane (including voltage dividers).

I've tried slots and splitting the ground and power planes in a range of ways. I've currently got power on the outer 2 layers and ground on the middle 2, routing signals on the outsides. I found continuous planes slots a little more effective than any of my attempts at star style grounds, but I'm not very familiar with how stars are actually meant to be routed. I've got a bunch of power (3.3, 5, -12, 12) so I am a bit confused about how to route the power correctly. I am biasing the JFETs with resisters relative to power and ground.

Have other people resorted to fill isolation in this case? I've seen a few audio products that do (with vactrols / optos for signals) so I assume it's not totally crazy but I'm wondering if the cost / complexity trade off is worth it. Is it the easiest solution given I've tried at least 10 board revisions... :)
 

Offline Geoff-AU

  • Regular Contributor
  • *
  • Posts: 144
  • Country: au
Re: Audio isolation, am I doing something stupid
« Reply #5 on: December 01, 2022, 12:25:57 am »
The search space of things that won't work can be quite large, and it stands to reason that if you are doing all the things you do know how to do, and it isn't working, then you're not doing something and you have no awareness of it.  Therefore it's impossible for you to describe to us what your circuit lacks, which is why providing the schematic and layout is vital.  It's a direct path to others seeing what you don't.  Otherwise we are playing guessing games just as much as you are, and it will take a long time to figure out what's happening.


 
The following users thanked this post: Someone, ajb, Wolfram

Offline loki42Topic starter

  • Frequent Contributor
  • **
  • Posts: 290
  • Country: au
Re: Audio isolation, am I doing something stupid
« Reply #6 on: December 02, 2022, 06:47:10 am »
I'll attempt to make a basic circuit that demos the idea, but I assume it should be common to any design the shares a digital and analog part like this.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16553
  • Country: us
  • DavidH
Re: Audio isolation, am I doing something stupid
« Reply #7 on: December 02, 2022, 08:01:18 am »
I have tried many board layouts (4 layer board) but the only way I've found to suppress it is to have transformer isolated power for the noisy pwm / touch elements and then communicate with them using an isolator (digital or opto). If there another option?  I feel like I'm making this more complicated than needed but I'm following all the neat how to lay out your analog board and care about return paths guidelines and they don't help.

A single point ground might be enough to solve the problem.  CLC decoupling might also be needed.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf