TLV320AIC3104: Unexpected Noise on customer board and EVM

Hi, Alberto,

The noise issue you are experiencing is probably related to the routing of the device. In general we recommend to connect directly the DAC output to the HP driver through the mux when noise performance is not important, but lower power consumption is desired. For best performance operation, routing through the mixer path is recommended. Could you please ask customer to route the output through the output mixer and see if the noise gets improved?.  For further reference about this, please refer to section 2.5 of  SLAA749 

Best Regards,

  -Diego Meléndez López
   Audio Applications Engineer

Diego,

thanks for your reply.

Customer tested routing to output mixer.

Their feedback:

Basically your suggestion helped to reduce the noise amplitude about 30%.

But the current 200mV p-p value is still quite large.

I would highlight that what we see on the CODEC coutput does not seems to be a real noise – look on the picture below.

From the signal waveform can bee seen that this signal is directly generated by DAC output – you can see even the DAC steps in the signal.

However when this signal is being played by the earmuff speaker it is perceived as a white noise.

It seems to me that this must be caused by some bad setting that DAC generates such pseudo-noise signal.

I hope that this behaviour is not considered as typical for this CODEC otherwhise itthe values of SNR stated in datasheet could be hardly achieved.

Can you kindly provide your feedback?

Thanks

Regards

Alberto

Hi, Alberto,

Thanks for the feedback. Could you please share the clock settings customer is using?. I took a  quick look to the register setting and it seems that the order followed is not what we usually recommend. In general for these devices, it is recommended to follow the following order to avoid possible pop,/noise issues:

1. Provide valid clocks and power supplies
2. Configure the internal routing
3. Power-up the DAC/ADC
4. Unmute DAC/ADC
5. Power-up and unmute drivers.

I will try customer's code and see for a different issue, in the meantime, could you please check if changing the sequence of the code could help with the issue?

Thanks and best Regards,

  -Diego Meléndez López
   Audio Applications Engineer

Hi Diego,

thank you for advice, if I understand it correctly it is needed to have clocks on codec, before setup of registers is made, correct?

Now order of customer setup is:

1. Power is provided.
2. HW reset is done
3. Setup via I2C is made:
Register Value
0 0x00
1 0x80
7 0x0A
43 0x04
41 0x02
37 0xE0
38 0x10
47 0x80
61 0x80
65 0x0D
51 0x0D
19 0x3F
22 0x3F
15 0x00
16 0x00
4. Clock are provided and data transfer start

Interface is set to I2S Standard with 16kHz

Measured values are:
Master clock is 4,5MHz
Audio serial data bus bit clock 554kHz
Audio serial data bus word clock 17,29kHz

Please let me know your feedback.
Best regards,
Alberto

Hi, Alberto,

Thanks for the feedback. Correct, it is recommended to provide the clocks before configuring the device. I reviewed the clocks used by customer and the clock combinations are not valid. From the code I understand the device is acting as a slave, so the clocks used for the I²S bus should be adjusted for a proper operation. Can you please check with customer why they are using those clock values and see if it is possible to provide valid audio clocks?. Please refer to section 10.3.3.1 Audio Clock Generation from the datasheet to find out how to configure properly the clocks of this device.

Best Regards,

  -Diego Meléndez López
   Audio Applications Engineer

Diego,

customer was able to set the frequency of the SAI to exactly 16 kHz (pls see below). Unfortunately the noise is still the same. It did not help. Noise levels are up to 200mV Peak-peak.

Do you have any suggestions what they can do to make it better?

Thanks and regards,

Alberto