Please criticise my PCB-design

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

living sounds

Well-known member
Joined
Jul 26, 2006
Messages
3,931
Location
Cologne, Germany
For my multichannel DAC I drew up this balanced line out with some added gain, using two halfs of a NE5532 set up as inverting inputs. It's mostly based on the GSSL output schematic, with the gain staging modified and an added input RF filter. I mostly wonder wheter I got the grounding right. DC blocking caps should be bipolar, but the PCB program didn't have those. The converter board is stereo, the analog grounds of the outputs on the converter PCB are already joined there. I could use only a single ground wire, but then shielded cables wouldn't work.

Thanks!
 

Attachments

  • Conv_analog_out_1_0.jpg
    Conv_analog_out_1_0.jpg
    443.8 KB · Views: 93
living sounds said:
For my multichannel DAC I drew up this balanced line out with some added gain, using two halfs of a NE5532 set up as inverting inputs. It's mostly based on the GSSL output schematic, with the gain staging modified and an added input RF filter. I mostly wonder wheter I got the grounding right.
For utmost isolation between channels, the power connector should be located in the middle, between the two inputs.
How the rest is interconnected may create unwanted effects, though. That may be quite critical in a multi-channel system.

DC blocking caps should be bipolar, but the PCB program didn't have those.
What is a PCB software that does not allow creating devices as simple as a non-pol cap?  :)

The converter board is stereo, the analog grounds of the outputs on the converter PCB are already joined there.
The way the PSU "ground" (0V) is connected matters much more.

I could use only a single ground wire, but then shielded cables wouldn't work. 
Why wouldn't they?

Now I think you're applying much thinking to a less-than-optimum design.
The two-opamp CCOS relies on extremely well matched resistors to perform.
I have long since ditched them in favour of the THAT1646.
 
abbey road d enfer said:
For utmost isolation between channels, the power connector should be located in the middle, between the two inputs.
How the rest is interconnected may create unwanted effects, though. That may be quite critical in a multi-channel system.

Thanks!

The GSSL has a similiar asymmetric power connection and even more asymmetric ground connection. I am currently using an old GSSL pcb for testing and didn't notice any problems with isolation.

What is a PCB software that does not allow creating devices as simple as a non-pol cap?  :)

ExpressPCB, I am still a beginner here. I'll look into this and see if I can edit the drawing.

Why wouldn't they?
Well, I would have to leave the cable ground floating on one side. Probably good practise anyway, right?

Now I think you're applying much thinking to a less-than-optimum design.
The two-opamp CCOS relies on extremely well matched resistors to perform.
I have long since ditched them in favour of the THAT1646.

But this doesn't provide the gain I need and is fairly noisy. And I don't know if it sounds as good. Despite all the attention to power supply, grounding, decoupling etc. every op amp still sounds different in every design I tried. And I've got that one type of NE5532 that sounds good to my ears and nulls well, so I want to use it for the output...
 
I agree with Abbey, I would like to see more symmetrical power distribution, perhaps with a meaty ground trace running down the middle of the PCB...

These may only make subtle differences if at all... but you asked.

JR
 
Not routing on a grid is making my trace OCD flare up.  ;D

Is there a specific reason you're trying to do this on a single layer board?  If you get these done by a real PCB house, there is zero cost savings.

Agreed with the others:  I would place the power connector in the middle, and then fanout the ground traces on the bottom, along with the power traces, and then try to reserve the upper layer for the interconnects.  Since the layout is partly symmetrical you can "mirror" the layout around the center connector.
 
I would definitely find a toolchain that can do a ground plane. That will make your layout way nicer and it will be much more forgiving to someone not fluent in pedantic grounding principles.
 
living sounds said:
The GSSL has a similiar asymmetric power connection and even more asymmetric ground connection. I am currently using an old GSSL pcb for testing and didn't notice any problems with isolation.
It is probably measurable but not perceptible.

But this doesn't provide the gain I need and is fairly noisy.
Gain should be added with a simple opamp; anyway, the 1646 must be driven by a very low impedance source.
But noisy ???? I can't agree with that.
 
I am not familiar with 2- and more layer designs. I see potential for problems (traces crossing, I don't know the rules). From what I've read ground planes can have their own problems.

And, obviously, this design makes it possible to self-etch a balanced board if needed.

Does the joining of grounds make sense the way I did it here?
 
living sounds said:
I am not familiar with 2- and more layer designs. I see potential for problems (traces crossing, I don't know the rules). From what I've read ground planes can have their own problems.
Crossing traces on different layers is pretty much the benefit.
And, obviously, this design makes it possible to self-etch a balanced board if needed.
You can self etch 2 sided boards but you have to manage the vias without plated thru holes. It isn't that hard to coordinate vias with component leads.
Does the joining of grounds make sense the way I did it here?
This may be a matter of style/taste but routing the ground around the outside edge of the board results in a longer path than a strong ground in the middle that radiates out... The longer ground path is not likely to cause an audible problem.

JR
 
JohnRoberts said:
This may be a matter of style/taste but routing the ground around the outside edge of the board results in a longer path than a strong ground in the middle that radiates out... The longer ground path is not likely to cause an audible problem.

Thanks John!

The idea there was to keep the "dirty" power decoupling ground as far away from the "clean" audio ground. This notwithstanding, I've had stability problems in a console where the "dirty" power decoupling ground was too far removed from the audio ground.

In this design I could easily cut the paths at the sides and join the audio ground through the center. The return paths for PSU decoupling would still not cross the audio ground. My current design allows for the audio ground to be far removed from the other traces.

Or would it be preferable to just create a lower layer ground plain and simply route all ground connections there?
 
living sounds said:
Thanks John!

The idea there was to keep the "dirty" power decoupling ground as far away from the "clean" audio ground. This notwithstanding, I've had stability problems in a console where the "dirty" power decoupling ground was too far removed from the audio ground.

In this design I could easily cut the paths at the sides and join the audio ground through the center. The return paths for PSU decoupling would still not cross the audio ground. My current design allows for the audio ground to be far removed from the other traces.

Or would it be preferable to just create a lower layer ground plain and simply route all ground connections there?
I wouldn't over think it...

I try to think of PCB traces as having high resistance and think about voltage drops from current flowing in the traces. This is me overthinking it...  ;D

JR
 
Here's version 3.0.

This one has a few 2nd layer traces (green). This also made it possible to shorten length of the trace from the first op amp output to the second op amp input.

I kept the grounding the way it is because it makes it possible to neatly contain the psu ground return paths. The ground connectors at the bottom (outputs) won't be connected anyway, output cable shield will go to the enclosure and the enclosure will be connected at the PSU board.

Does this make sense?
 

Attachments

  • Conv_analog_out_3_0.jpg
    Conv_analog_out_3_0.jpg
    615.1 KB · Views: 29

Latest posts

Back
Top