GSM buzz and gsmhandset.state

Andy Green andy at openmoko.com
Fri Oct 24 04:52:58 CEST 2008


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Somebody in the thread at some point said:
| On Thu, 23 Oct 2008 22:04:36 +0800, matt_hsu wrote:
|> Can I know which hardware version do you use? A5 or A6? I used your
|> state files to try my A5. It seems that it could reduce the buzzing
|> noise.
|> But it could not get rid of it all.
|
| /proc/cpuinfo says it's A6 (0360), while frameworkd (Debian) says A5.
| It's the one with capacitor on SD clock lines. Should be A6 if I remember
| the text on the PCB correctly.

It should be an A6, right.

| Seems that I was lucky with getting no buzz at all, most of the time it
| is still audible (not hurting ears anymore though).
|
| Investigating further, I recorded Mic1 and Mic2 while calling directly on
| the FreeRunner (switched to a modified capturehandset.state with
| increased recording levels while calling):
|
| http://wiki.openmoko.org/images/3/34/Lmic1_Rmic2-interference-call.wav

Hey great work doing and documenting all this!

| (no headset connected, both mic capture volumes 3, only mic1 routed
| through ALC_MIX, mic2 directly to right PGA).
|
| Snippet of the waveform:
| http://wiki.openmoko.org/wiki/Image:Lmic1_Rmic2-interference-call.png
|
| A voicebox recording of the buzzing before the "fix":
| http://wiki.openmoko.org/images/7/71/FR_GSM_buzz.wav
|
| After the "fix", the voicebox did not record endlessly anymore and
| recognized the remaining buzz as silence.
|
| Looking at the first recording, it is visible that Mic1 signal is
| inverted by the path it takes through the wolfson chip. Sadly it does not
| seem to be possible to mix inverted Mic1 and Mic2 in hardware without
| going through ADC - doing so on the PC reduced noise somewhat... I wonder
| if that could be done in software on the FreeRunner, recording both mic
| signals, deducing from Mic1 when contamination occurs, extracting the
| according waveform from Mic2 signal, and then subtracting the extracted
| (average) buzzing waveform from Mic2 whenever Mic1 levels get over a
| certain threshold - and then playing back the resulting signal over
| either PCM or voice codec to the GSM modem. I think that plan might be
| thwarted by full duplex restrictions by the chip, as I did not succeed
| yet to play back recorded audio while still recording without heavy
| artifacts (clicking, a lot of noise).

That is a cool idea, but the match is bad comparing mic1 and mic2
waveform from your great image on the wiki page.  Mic2 seems to have
more capacitance that smooths the response to the impulses.

| Turning off sidetone mic on MONO MIXER (control 77) and playing back
| music over Hi-Fi DAC/PCM to mono interface / GSM resulted in buzz-free
| (though a little distorted, did not tweak much) music experience on the
| phone called. This basically confirms that signals get contaminated
| through MIC path.

Yes it seems to be this way: nice test.

| I did not succeed yet in playing back recorded MIC signals over the voice
| codec/voice DAC without the signals leaving the wolfson chip (loopback) -
| is that actually possible?

If I understand you mean to capture the DAC output at the ADC input?
This does seem to be possible via this recmix thing, it can bring
monomix output back into the right ADC input.

- -Andy
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iEYEARECAAYFAkkBOAAACgkQOjLpvpq7dMqxqQCfQIfjT2xxQiHAAK1drokSA+fE
7+IAn29qNaCvRF82UZ5iR1dRZ4cCHZVp
=+4lA
-----END PGP SIGNATURE-----




More information about the community mailing list