-3

I am currently routing a pcb where there are some differential pair: Speaker, Mic. How do I decide what should be the space between both tracks and what should be the width for the traces ?

I add a screenshot of the guide for this application:

enter image description here

Source: Quetel UMTS/HSPA Module Series WCDMA UGxx Audio Design Note

user92481
  • 313
  • 2
  • 6
  • 12
  • ... should have added the screenshot earlier. Without it, it really sounded like you were talking about analog voltages. – Marcus Müller Jun 11 '16 at 14:20
  • This question begins with an incorrect assumption. That differential audio-frequency layout is sensitive to track geometry. Since it is not, the question itself is incorrect. @user92481 is confusing RF protocol with low-frequency (audio) dependencies. – Richard Crowley Jun 11 '16 at 14:23
  • @RichardCrowley I don't think so richard. The guide say those audio line have to meet the rule of differential signal. Although those lines are labelled with PCM in it, those lines are actually going between audio codec and Speaker and between audio codec and mic. – user92481 Jun 11 '16 at 14:48
  • @user92481 you still don't appear to comprehend the difference between baseband audio frequencies (as close as you can get to DC) with even digital audio (which is measured in KHz and has no special layout requirements). Thousands of audio-frequency PC boards are in use with no special layout geometry because audio just doesn't require it. It won't hurt to use RF-style layout but it is not necessary and simply an unwarranted pain. – Richard Crowley Jun 11 '16 at 15:56

3 Answers3

3

For audio frequency applications, length matching will mostly make no difference. Think about it: assuming an electromagnetical wave travelling along your transmission line travels at roughly \$\frac23 c_0\approx 2\cdot10^8\frac{\text m}{\text s}\$, and audio happens below 100 kHz, your wavelength is > 2 km – a couple millimeters more or less are really irrelevant.

The same goes for spacing: At these frequency, the energy from the audio signal is really not exchanged through electromagnetic fields. So just space the two lines far enough apart that the capacitance starts to matter (it won't, don't worry).

In fact, it's more important what kind of energy you don't get into your signal, at least for the MIC input: keep your lines close together to exploit the fact that all spurious emissions coupling into those lines will probably hit both lines of the differential pair the same; that way, they will end up as common-mode noise, and be rejected by the differential amplifier (or whatever is at the end of your differential pair). That's why these signal pairs should end up the same length: any additional length on either line might lead to differential noise.

Keep sensitive signals away from high current, fast switching signals. That goes for differential signals like it does for single-ended ones: Don't route your signal lines through the power supply and keep your multi-ampere amplifier output lines the hell away from your microphone input.

Marcus Müller
  • 88,280
  • 5
  • 131
  • 237
  • This isn't an analog audio signal, per the net name it's some kind of (digital) Pulse Code Modulation – pericynthion Jun 11 '16 at 14:16
  • 1
    .... "Speaker, MIC" really really does sound like an analog signal. – Marcus Müller Jun 11 '16 at 14:17
  • So these are digital clock data and clock lines. Mention that in your original question next time :) So, rule of thumb: Necessary to be safe Signal frequency is 10* clock frequency. I don't know the clock frequency on these lines, but let's act like it is 10 MHz . Well, then apply the typical digital line layout. I won't write a second answer explaining that. – Marcus Müller Jun 11 '16 at 14:19
  • But still, at 10 MHz, don't worry about length matching (look at the screenshot...). Use any coplanar microstrip line calculator out there to match the line impedance to what your ICs have in source/sink impedance. Done. – Marcus Müller Jun 11 '16 at 14:21
  • There is a mic and a speaker going to an audio codec..., then the audio codec will send speaker and mic signal to module via PCM interface – user92481 Jun 11 '16 at 14:37
2

Track and gap widths are of concern in controlled impedance, but unless your speaker and mic connections (I assume audio frequencies) are going a long way you should not have any particular requirement for controlled impedance.

Using a differential pair to avoid common mode noise is common in audio circuits.

The track widths should be wide enough to support the necessary current (which will be more important on the speaker interface, I suspect).

Note that single layer routing is preferred as going through the board (with vias, although vias to inner layers at source and destination is common) can cause differential to common mode conversion which partly defeats what you are trying to achieve.

Without more specific details I cannot give any more guidance.

Community
  • 1
Peter Smith
  • 21,923
  • 1
  • 29
  • 64
  • I have added an extract of the guide supplied by the manufacturer. This is all what they say, they don't give any metric.. – user92481 Jun 11 '16 at 13:53
1

The manufacturer's guide you quoted is about routing RF signals. Specifically so that it won't interfere with nearby audio signals.

You seem to be asking about audio, not RF. There is no rigorous requirement for audio frequencies as there is for RF. You are misreading the guide.

Richard Crowley
  • 14,382
  • 2
  • 20
  • 37
  • No, this is about audio design, but associated to RF environment (here it is a GSM module where there is some audio port working with a codec for audio) – user92481 Jun 11 '16 at 14:37
  • Unless you are designing a cell phone, the information is irrelevant to your application. Ordinary audio circuits do not require precision differential signal routing on PC boards. You are over-thinking this. – Richard Crowley Jun 12 '16 at 12:51
  • this is almost a cell phone. It use GSM module connected to this audio codec.. – user92481 Jun 13 '16 at 06:22