TLV320AIC3254: Codec Programming Issue

Chandler KI

Part Number: TLV320AIC3254

Hi,

We are having an issue with the AIC3254 codec on approximately half of the boards of a batch recently manufactured. When the codec is programmed by our host MCU, it seems like some noise is generated inside the codec itself, outputting high voltage noise into our amplifier, producing loud sounds. We isolated the codec inputs to verify this. Any audio input either does not come through to the output, or the noise just overpowers the input.

In the attached picture is our miniDSP process flow of this codec, which has worked in the past for these boards. Also, only about half of our boards are experiencing this problem with the codec.

We discovered that removing the Biquad Filters and the parametric EQ ("ProtectLeft" ,"ProtectRight" , and "PEQ_1" in the process flow) and using this modified .cfg file will solve this strange issue, and allow the non-working codecs to operate normally.

Any ideas as to why removing these components from the process flow allows the codec to operate normally? We still require the biquad filters and parametric EQ, so we are hoping to solve this issue with removing those components.

Thanks

over 3 years ago

0 Jeff McPherson over 3 years ago

TI__Mastermind 47905 points

Hi Chandler,

Our codec expert is out of office and will be returning Monday. But in the meantime, can you create a process flow with the same ins and outs, play the same signals, but only have the filters?

Best regards,

Jeff

0 Chandler KI over 3 years ago in reply to Jeff McPherson

Prodigy 190 points

Hi Jeff,

Thanks for the quick response. I created a process flow with the same ins and outs, with only filters and the parametric EQ components.

If only the parametric EQ is used (with or without the front-end biquad filters), I see/hear the same issues as above, so it seems isolated to the parametric EQ.

If only the biquad filters are used, no issue occurs and the signals correctly goes from the input to the output.

Also, this new process flow with only the parametric EQ causes the same issues on any of our boards with codecs not working with the original process flow, but the process flow works OK on any boards with codecs that were previously working fine.

0 Jeff McPherson over 3 years ago in reply to Chandler KI

TI__Mastermind 47905 points

Hi Chandler,

That's good info to know. Thank you for checking that. I'm redirecting this to our codec expert.

Best regards,

Jeff

0 pdjuandi over 3 years ago in reply to Chandler KI

TI__Mastermind 29765 points

Hi,

So you have narrowed it down to the EQ, can you send the EQ settings?

Is it possible to place/swap the not working unit unto the working board?

Regards.

0 Chandler KI over 3 years ago in reply to pdjuandi

Prodigy 190 points

Hi,

The settings below are used in the parametric EQ for our process flow. It is just the default component settings (flat gain and 10 bands).

The parametric EQ gain settings are changed during runtime by the host MCU, applying different gain values for those 10 bands.

Also, it is not currently possible to swap a non-working aic3254 to a working board. But, we have tried replacing our host MCU on a couple boards and some of them seem to be working after this MCU swap. The actual codecs were not replaced. So it seems like removing the Parametric EQ from the process flow or replacing the host MCU solves some of these issues so far.

0 pdjuandi over 3 years ago in reply to Chandler KI

TI__Mastermind 29765 points

So MCU is tuning them on runtime and it's not working as expected, but when EQ is removed audio quality is good and working as expected - correct?

This looks like host is not configuring correctly or board issue.

0 Chandler KI over 3 years ago in reply to pdjuandi

Prodigy 190 points

Yes exactly. We are still investigating any host MCU problems, but I still do not understand why removing the parametric EQ component allows the codec to work correctly.

0 pdjuandi over 3 years ago in reply to Chandler KI

TI__Mastermind 29765 points

EQ is just band filter so it's based on the configuration and since the configuration works on other boards it's likely from the setup of the bad boards.

0 Chandler KI over 3 years ago in reply to pdjuandi

Prodigy 190 points

Are any pages of registers I should view and compare from a working codec to non-working codec?

We looked at all the registers in Page 0 and Page 1 to be sure the right settings were being set by the host MCU. The only difference in register values happens in Page 0, Register 2. On every good board tested, that register has a value of 0x60. In every non-working board, that register of the codec has a value of 0x50. The application guide says this is Reserved Register.