5
\$\begingroup\$

On a circuit board with a mixer chip (and some auxiliary components like coupling capacitors), I have to lay a number of tracks (35 µ copper layer on epoxy substrate; max. track length: 120 mm). The amp is not part of this board. Neither is a DAC.

Now I'd like to know how wide tracks transmitting HiFi analog audio data (up to 20 kHz; including classical music) must be. Standard width would be 0.5 mm.

Here are my schematics and preliminary PCB layout. Please note that neither the 12V–to–5V converter, nor the Schmitt trigger have anything to do with the audio. They are just there because there is some space left over for them:

Schematics

PCB layout with routings

ORIGINAL FILES: If anybody is interested in the original Kicad files, please go to this folder: Mixer PCB

\$\endgroup\$
18
  • 5
    \$\begingroup\$ That it is carrying music and audio has nothing to do with the size of the tracks. \$\endgroup\$
    – Rob
    Dec 8, 2021 at 21:40
  • 7
    \$\begingroup\$ What you want to avoid is injecting noise INTO your signals. You want those traces to be isolated. You want to KNOW where their ground return current is going. You want to keep them away from any AC signal traces (i.e. digital power or signals), and not run parallel with them. Don't create ground loops (or ANY loops really). Hard to read your PCB layout, but it SURE LOOKS LIKE you've blown off the idea of using a star ground topology. That may be a big mistake. i.e. resources.pcb.cadence.com/blog/… \$\endgroup\$
    – Kyle B
    Dec 8, 2021 at 21:53
  • 7
    \$\begingroup\$ The signal trace width is not that important or relevant. I would be much more concerned about your ground routing (why no ground plane?) and loop area (the traces don't seem to be routed with much care to reduce loop area). This looks autorouted (if it is, the autorouter didn't seem to do an awesome job) and would benefit from a new manual routing with care. \$\endgroup\$
    – nanofarad
    Dec 8, 2021 at 22:15
  • 4
    \$\begingroup\$ Some folks are calling for 'ground planes'. There's a school of thought that this is NOT what you want for AUDIO. You really want that star-ground topology. You want to be able to control the path of every current, which a ground plane won't let you do. FWIW, I fix alot of high-end audio gear, including professional mixing boards (i.e. like for bands doing concerts or soundstudios doing recording kind of mixers). There are NEVER solid planes on these boards in the audio sections. There ARE solid planes under anything digital. So that's empirical evidence I know, but a good datapoint. \$\endgroup\$
    – Kyle B
    Dec 8, 2021 at 22:36
  • 4
    \$\begingroup\$ I use Altium. The autorouter seems programmed to maximize EMI. Totally useless. @Neppomuk You can implement a star ground. I am CERTAIN you can do it, you have PLENTY space on the PCB. The trick to successful power & ground routing is to do it FIRST, before you route signals. \$\endgroup\$
    – Kyle B
    Dec 9, 2021 at 22:33

3 Answers 3

13
\$\begingroup\$

Line level audio has tiny currents, so trace resistance (and width) doesn't matter.

Now, the pot on LM317 is a problem, because if you set it wrong (or if the wiper develops a bad contact) your LM317 can be stuck at its maximum output voltage, which will fry your 9V chip. It would be preferable to use fixed resistors. If you want it adjustable, you can put a 7.8V zener between ADJ and GND to make sure the output voltage of LM317 never exceeds 9V. You can also use a fixed resistor for the lower leg of the voltage divider, and a pot wired as adjustable resistor for the upper leg, with proper values to make sure it can't be set to a voltage that would fry the chip. Note adding a feedback cap to LM317 will improve PSRR.

Unused inputs on CD4093 should be grounded, otherwise they will pick up noise.

I'd definitely use a ground plane on this board, and thicker traces for power.

Edit:

Since this is for a car, there will be vibration, which means it is a good idea to use non-microphonic coupling capacitors. So I'd recommend replacing all the 2.2µF film caps with non-polar electrolytics.

\$\endgroup\$
2
  • \$\begingroup\$ ...film caps are microphonic?! \$\endgroup\$ Dec 10, 2021 at 12:39
  • \$\begingroup\$ Oh yeah, if there's DC on them and they vibrate, they'll definitely put some noise in your signal. How much is another question, that's not specified in the datasheet. \$\endgroup\$
    – bobflux
    Dec 10, 2021 at 12:43
11
\$\begingroup\$

The short answer:

It doesn't really matter, the only thing that the traces will be contributing is inductance and it will be really really small. A 1" 1oz (35um) weight copper trace that is 0.5mm (~20mils) wide will have ~7nH/in of inductance, hardly anything to worry about since the tolerances of any inductors, resistors or capacitors will have many many times more Henries of error.

One thing you may want to worry about is having a good solid ground layer and using large traces and bypass caps on the Vcc lines\power rails of any components that are sourcing large ammounts of power.

\$\endgroup\$
4
\$\begingroup\$

up to 20 kHz

You're in a car, and you've got connectors going somewhere. You've got noise on those inputs easily into the 100s of MHz range, never mind kHz.

Novices are usually unaware of how important it is to eliminate noise. I would strongly recommend filters on all analogue inputs to this mixer. If you don't, your signals are going to be in a world of hurt (or Hurtz).

Even if (per your comment back) this is inside the same enclosure, there really is a substantial amount of electrical noise floating around in a car. The more you can stop this propagating through your analogue, the better your results will be.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.