issue of vs1063's encoding gain

[jwc1979]

Hello,

I met an strange issue when tuning the audio of vs1063. I have set vs1063 to auto gain mode and the volume is suitable, sound is clear. but sometimes, the volume become loudly sunddenly with noise when I just repower the board. And it can not recover when repower again. it seems like some state of chip is not stable. how do i debug this issue?

kind regards

Wayne.

[jwc1979]

after some time, I repower the board, it recover again.

[pasi]

Which encoding format are you using? Are you using the vs1063a patches package?

Do you get the noise in the analog outputs (monitoring), and/or to the encoded data?

Double-check the voltages you use, and that you use high enough (but not too high) clock multiplier for the encoding parameters you use.

[jwc1979]

Hello pasi

I use the PCM format currently. And I have used patchs package for vs1063a.

I got the noise in the analog outputs. And the input clock is 12Mhz, clock multiplier is 5(sc_mult mask is 0xe000). The encoding stream is 8kHz, 16bit mono PCM data.

Are the encoding parameters suitable to the clock?

kind regards

Wayne

[pasi]

PCM doesn't take almost any CPU - especially at such a low samplerate, so almost any clock is fine. With 5.0x clock though, make sure your CVDD is stable 1.8V .

It wasn't quite clear from your original post if the noise can also appear in the middle of encoding after being started correctly, or only comes up when you start encoding?

Check the "regular suspects", for example that xTEST isn't floating, and measure the voltages.

[jwc1979]

Hello pasi

Thank you for your advice. We will check it.

We met a new scenario about this issue. After vs1063 power up, and the voice is good, keep vs1063 powering up, restart the other side, the voice become loudly on both side(vs1063 is set on codec mode). At this moment, the microphone becomes sensitively. I must speak gently and keep away from the microphon, or the other side can not hear what I say.

kind regards.

Wayne.

[pasi]

Byte-order issues are possible when you use 16-bit PCM. If you miss or insert a byte, the byte order of the data will get swapped, and you'll effectively get 256x gain to the signal.

Maybe you could test with IMA ADPCM or g.722 to see if you still get the same issues.

[jwc1979]

Hello Pasi

I am not sure what is your means.
What result would be if testing with ADPCM or G.722? The isuue rate would be small than PCM, or it would not occour the issue?

kind regards.
Wayne.

[pasi]

The suggestion was to rule out byte-order / transmission issue.

You won't have byte-order issues with IMA ADPCM, because it can synchronize to its block header if there are too many or too few bytes received.

g.722 is byte-based, so it also cannot have problems with the byte-order.

[jwc1979]

Got it.

thanks for your reply.

kind regards.

Wayne.