Linux/TLV320AIC3104: audio microphone has been heard extremely strong noise, sounds like a running engine, much more intensive than the signal

hello Cloud Yun,

please attach your register configuration and Schematic and I will take a look.

I The TLV320AIC3104 GUI does have some preset examples that you could look at. Do you have an EVM?

If you are using the AGC, I suggest disabling it to see if the noise goes away. Additionally, you can disable the input blocks to see if the noise goes away, followed by disabling the PGA.
This is certainly not normal function, so I surmise that there is a configuration error.

best regards,
-Steve Wilson

Cloud Yun,

I'm unable to see the image you attached. Can you try to post it again?
best regards,
-Steve Wilson

Hi Steve,

Can you see it now?

Cloud yun,

I still cannot see the image. Can you try a different format?

best regards,
-Steve Wilson

Hi Steve,

We seem to have found part of the problem, as our schematic design, we use mono output.  The software is expecting stereo output from the TLV320AIC3104, can you help to check whether we can still give out a stereo output?

7635.schematic.pdf

Cloud Yun,

I am not sure I understand. are you talking about the I2S Out?

-Steve Wilson

Hi Steve,

Yes, my concern is the I2S digital output channel, could we output stereo based on the current schematic?

Cloud Yun,

The I2S output for your device should be Stereo, but the Left channel will have all zeros. it is not possible on the AIC3104 to route MIC2R to the Left ADC.

It is possible to set the dac data path to play a mono mix of the LEft and right Data channels, but it doesn't sound like that will help you.

One thing that might help.... It looks like you are not using the Line1LM, Line1LP, line1RM, and line1RP inputs. is that correct?

because if you used the TLV320AIC3105, your input routing capabilities are greater. aside from the inputs the AIC3104 and AIC3105 are very similar and the pinout is the same for everything but the inputs.

best regards,
-Steve wilson

Hi Steve,

Hi Steve,

Is it possible to make use of the PIN14 of TLV320AIC3104 by making it as MIC2L signal to drive left ADC instead of taking it as MICDET?

Does the functionality difference of  MIC2L or MICDETonly come from AC coupled or DC coupled? 

Hi Steve,

Thanks for your clarification.
Could you share a schematic example of DC couple input to AIC3104? I will check whether we will have plan to redo the schematic design.

Hi Steve,

Thank you for your support, we have figured out the previous problem of MIC input in software.

Now we are focusing on the speaker, can we loop the DOUT to DIN directly from the system feasibility perspective? I have implemented that, but found that the noise is much stronger than the signal, could you share some practical register configuration based on our schematic,  two channels with the single output of HPLOUT and HPROUT.

Best regards,

Hi Steve,

Cloud Yun,

can you explain what you mean by "I shorted the DOUT to DIN in a Xilinx FPGA"?

When I am shorting the Dout of the AIC3104 to the DIN of the AIC3104, I am just shorting the pins together. Can you get a scope capture of the DIN? with WCLK and BCLK?

-Steve

Hi Steve,

We have both DOUT and DIN routed to a Xilinx FPGA pin, in order to make a stable and reliable connection, I just shorted the PIN without any register pipeline, the overall delay is only 4.534ns, I am assuming it doesn't matter too much for a 48KHz signal, equivalent to shorting directly, unless you have a different opinion in terms of noise introduction.

What do you expect for a "good" waveform or a "bad" waveform respectively? Do you have a "good" waveform to refer to?

Cloud yun,

The expected I2S data/Clock waveforms are in the datasheet.  I have attached  scope shots of a 1khz test tone on the HPROUT of the AIC3104.  

The first scope shot is from TP26.  This is an unfiltered output.  The Second scope shot is from TP33.  There is a passive LPF on this output to reduce out of band noise. 

Best regards, 

-Steve Wilson

Hi Steve,

It  seems that the testing point is HPRCOM rather than HPROUT.

Does your testing result come from your golden reference board? How should we generate 1khz test tone to get a similar testing result as you did?

Besides, we see in our case, if we adjust the ADC volume a bit lower, the noise of speaker has been sort of faded, do you think the input volume is still another factor causes noise if we short DOUT to DIN?

Thanks,

Cloud 

Cloud Yun,

This testing result came from our EVM, I'm not sure I would call it a "golden reference board" but it is designed with all of the layout considerations described in the datasheet.
my test tone was generated with program on my PC and the EVM was used to bridge USB to the codec. Any I2S output source would be acceptable. do you have access to an AP?

using the input PGA will definitely add some noise. try muting the PGA and see if the Noise goes away.

best regards,
-Steve Wilson