Audio: balanced signal lines on pcb ?

[Zaaappp]

[semi-n00b question here, so pls be nice ]

Consider a simple, audio-effects box used in a professional recording studio:

• An audio in, two-layer pcb's doing both analog and digital processing, then audio out.
• Both in and out are balanced, line-levels, commonly used in sound gear.
• The audio must be the highest possible quality.

The circuit will have several controllers and processors running clocks at MHz speeds. (Assume I do all the correct things in layout, e.g. ground plane between noisy clocks, supplies buffered, etc.)

Is there any reason to keep the signals balanced on the pcb traces themselves?

Meaning differential, two signal traces referenced to circuit gnd per signal (versus just one signal line referenced to gnd).


What I'm unclear about is if things like, say, the common-mode rejection effect would even be a factor when things like micro-strip cross talk and injected noise (from adjacent circuits) could be a bigger issue? … said another way, would I be shooting myself in the foot to use balanced signals on the board when unbalanced (but clean) signals would work just as well?

- My (overly-optimistic?) Ideal specs :laugh:

• SNR (Signal to Noise Ratio): -70 dB (Audio Level) to -75 dB(AL) … that's around a 24-bit master.
• Headroom: about 10 dB(AL) max
• Frequency response: 10Hz to 25kHz

Thanks for the assistance!
- Howard in Florida

{edaboard rocks, the others merely flubber}

______________________________________
References:
https://en.wikipedia.org/wiki/Microstrip
https://www.skottanselektronik.com

 

[BradtheRad]

Digital processing has an advantage, because the 1's and 0's are noiseless. As you state, you plan to convert formats properly.

differential, two signal traces referenced to circuit gnd per signal

Does this imply you would mirror every function performed on Line A, so that Line B also gets the same function performed at opposite polarity? This sort of duplication might be difficult.

 

[FvM]

Balanced audio signaling is a requirement in professional audio for the external interfaces. I think we should just take it as granted in this discussion.

Secondly high resolution audio AD and DA converters mostly utilize differential analog signaling. The feature has been introduced latest when the mixed signal chips proceeded to use a single analog supply voltage of 5V or below. So your question focusses to the point if the entire audio signal path should use differential signaling or convert to single ended and vice versa, e.g. to reduce the bill of material.

There are different crosstalk mechanisms to consider. For low and medium impedance audio circuits, common mode voltage differences ("ground loops") and inductive coupling are dominating. They can be effectively eliminated by differential signaling, I believe however that well considered circuit design and layout allows single ended signals inside an audio instrument without affecting the dynamic range.

It's probably instructive to analyze the design features of recent professional audio instruments. I stopped designing audio devices long ago, so my present knowledge is more on a general literature review level.

 

[Dan Mills]

This is actually a bit subtle.

Firstly, get the basics right, PIN 1 goes directly to chassis (Along with the first stage of the RFI filtering) and there is only a single connection between chassis and any single ended audio reference plane (The 'ground' plane).

Now on the real question, if you go differential all the way (and it can sometimes make sense) then every component in that chain has the potential to hurt CMRR, and unless you strip out the common mode voltage early you can run out of headroom due to common mode noise even with no audio present. Also, fully differential usually ends up costing you some noise performance.

Whatever you do you want to remove common mode voltages early, and you want to do it with a very large common mode impedance to ground, as this helps CMRR when sources have slightly unbalanced source impedances (THAT corp have some clever line receivers with a bootstapped common mode impedance that are excellent). Usually at this point you have a single ended signal referred to some reference pin.

Now lets talk about current loops, there a a common (and incorrect) view that an opamp is functionally three terminal device, but if you think about the current loop starting at the opamp output this is clearly not the case, and worse the loop switches route depending on which quadrant the opamp is driving in, so even with a single ended signal it pays to track the return path back to the junction of the opamp decoupling caps to minimise the loop area (This net ties down to ground plane at that point). It is not really differential, more of a hierarchical ground arrangement, but minimising that loop makes a big difference even if going fully differential is inappropriate (and it often is within the board).

Figure out which supply the opamp Vas integrator is referenced to, this is the rail that really needs attention.

Personally for pro stuff I would not bother with trying for a two layer design, 4 layers is better behaved in all ways, and importantly lets you get a solid ground plane in (I am not a fan of split analogue/digital ground planes, but that is a religious issue).

So in summary:
Get the basics right.
It depends.
Think about loops and loop areas.
Current loops do not magically end at the driving opamp unless you take measures.
4 Layers is not significantly expensive and is almost always better then two when EMC time rolls around.

Regards, Dan.

 

[schmitt trigger]

Concur with the 4 layer board option.

Several years ago, 4 layer was far more expensive than two layer.
Nowadays, you may be talking perhaps a 20 to 30% premium.

But the improvement, specially in the type of circuits you are planning to design, is very significant.

Also, since one does not require to run beefy ground and/or power lines on the signal layers, you have more area to place components and/or traces. The result is a more compact layout.