Yet another Ground/Opamp Decoupling thread

[dirtyhanfri]

Hi

I'm finishing the design of a PCB for a project involving more than a pair of inputs and outputs, not a big mixer, but 6 Line input and 6 Output circuits and some relays, I'm working in a 2 layer design, and have some doubts about grounding.

I know this have been covered many times, and I applied many tips read here, but I'd like to check if I'm misunderstanding something...

There will be 2 psus, one for +/-24 and one for +/-18 both with their 0V connected together to the chassis. That will be the "main chassis point" also I'll connect there 0V from the main board, I'm doubting if run each xlr to it or connect them to the chassis in their mounting screw, maybe I should connect them trough their mounting screw and also daisy-chain every xlr in a wire going to "main chassis point"?

In the main board, where the audio circuits will be, I have separated ground for every audio circuit block, I've drawn traces for GND and then outlined a zone surrounding the circuit which is copper filled, every circuit block has it's ground isolated from other circuits, they connect to the chassis via xlr's. Is this OK or should I connect every ground plane via a single wide trace? Or just connect the input circuit to it's associated output circuit and so?

For the relays, (fed by the +24V PSU) I've drawn  traces for each 0V, joining together to make a wider trace which ends in a little plane where I'll put a Faston connector or whatever to run a thick wire to that chassis point mentioned earlier. Right?

My main concern is about the opamp decoupling caps (all of them running from the +/-18V PSU) should I connect their 0V leg to the audio GND? I have it this way right now, but I could make a copper plane for them and join it to the 0V of the+/-18V PSU/Chassis. What will be better in noise terms?

Thanks in advance

 

[dmills]

Pin 1 (and the shell) of the XLRs should be connected directly to chassis right at the connector, pin one does NOT connect to your sensitive audio ground (except at one point, see later).

The relays should be laid out to minimise loop areas, especially the loop involving the catch diodes, and should not connect to the audio ground plane.

Personally I am not a fan of playing clever games with audio grounds in the first instance, a nice solid plane is usually better unless you are prepared to do a lot of hand tweaking and spin many prototypes.

Basically you have an internal plane used as a reference for all the single ended audio inside the box, which should be connected to the  chassis at ONE point to control common mode voltage, and you have the chassis which carries all the pin 1 and circulating earth loop current, as long as these things meet at only a single point it is all good.

The current loop around an opamp output stage flows via the decoupling caps, so I would urge that they be connected to audio ground (as in the reference plane) directly, as this will minimise loop area as well as providing the stiffest tie between analogue power and the reference which will minimise PSRR issues.

Don't go for too low an ESR on the bulk electrolytic caps, this is not somewhere that is helpful and it can cause problems.

Regards, Dan.

 

[dirtyhanfri]

Personally I am not a fan of playing clever games with audio grounds in the first instance, a nice solid plane is usually better unless you are prepared to do a lot of hand tweaking and spin many prototypes.

Well, I thought this approach can be more problematic, as read here it's prone to current loops, or I'm totally wrong? Actually big and solid plane attracts me, but went the other way thinking it was technically better...

Maybe it's just something strongly dependent on self experience?

P.S. Thanks for taking your time in reading and response.

 

[dmills]

Thing is, circulating current mostly does not matter as long as the voltage it develops between the nodes you care about is small enough, a low Z plane is better here then several cm of thin track to a star point.

Now a full up hierarchical ground scheme can work, but  that is black magic and unless you are trying to get the last 0.001% on the numbers it is usually far more trouble then it is worth. 

It is noticeable how spiders webs of star grounding are very much a tweaky hifi thing to do these days, you seldom see it in studio or broadcast kit, where it is all planes (But also where balanced lines rule for IO).

Regards, Dan.

 

[abbey road d enfer]

There will be 2 psus, one for +/-24 and one for +/-18 both with their 0V connected together to the chassis. 

These 0V should be connected together; that will be the main reference point. Now the connection between this reference point and the chassis is not critical; it is not essential for this connection to be stiff, any length of PCB or even wire will do.

That will be the "main chassis point" also I'll connect there 0V from the main board,

Your priorities are in the wrong order. It is is essential to define the 0V reference first, then connecting it to the chassis. You may even find that joining chassis at the ground near the inputs is preferrable.

I'm doubting if run each xlr to it or connect them to the chassis in their mounting screw, maybe I should connect them trough their mounting screw and also daisy-chain every xlr in a wire going to "main chassis point"?

The XLR's should be connected to the chassis with a stiff connection, and then only one point of the chassis connected to 0V.

In the main board, where the audio circuits will be, I have separated ground for every audio circuit block, I've drawn traces for GND and then outlined a zone surrounding the circuit which is copper filled, every circuit block has it's ground isolated from other circuits, they connect to the chassis via xlr's.

What? The XLR's pin1 should not be connected to 0V.
The audio grounds should be connected hierarchically. The concept is "ground follows signal". I know it's a little compicated when there are multiple inputs and outputs, but it is doable.

For the relays, (fed by the +24V PSU) I've drawn  traces for each 0V, joining together to make a wider trace which ends in a little plane where I'll put a Faston connector or whatever to run a thick wire to that chassis point mentioned earlier. Right?

Is 24V used for something else, or just the relays? It's a very bad idea to have any currents shared between audio and relays.
Good practice is to run relay+ and relay- as completely separate nets, and join one of them to 0v at one point only. Now if this +/-24V is also used for audio, there should not be any other connection.

My main concern is about the opamp decoupling caps (all of them running from the +/-18V PSU) should I connect their 0V leg to the audio GND?

Yes, but make sure that the junction of each pair of decoupling caps is very short and coincides with the return current from that stage, i.e., if a stage drives e.g. a fader or a string of potentiometers, the bottom wire of the fader/pots should coincide with the junction of the decoupling caps.

I have it this way right now, but I could make a copper plane for them

You should have several copper pours following the hierarchy; that is to say the copper pour should be a very large track

and join it to the 0V of the+/-18V PSU/Chassis. What will be better in noise terms?

Indeed it should join the 0V in hierarchical order; the output stages grounds connected to the 0V reference, then the input stages ground connecetd to the output stages ground at the input of these output stages.

 

[dirtyhanfri]

What? The XLR's pin1 should not be connected to 0V.

Well, I meant I can connect audio circuit's GND to chassis via the XLR, I'm not intending to connect Pin1 directly to 0V, I didn't explain that point correctly.

Is 24V used for something else, or just the relays? It's a very bad idea to have any currents shared between audio and relays.
Good practice is to run relay+ and relay- as completely separate nets, and join one of them to 0v at one point only. Now if this +/-24V is also used for audio, there should not be any other connection.

Just for relays, audio stuff will run from the +/-18 PSU

You should have several copper pours following the hierarchy; that is to say the copper pour should be a very large track

Yes, that's what I have, I was in doubt about joining different sections altoghether.

Indeed it should join the 0V in hierarchical order; the output stages grounds connected to the 0V reference, then the input stages ground connecetd to the output stages ground at the input of these output stages.

Thanks for that, it is the explanation I was missing.

Thanks a lot to both for your answers, I'm gonna run KiCad and join some GND planes right now.