This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TLV320AIC3204: TLV320AIC3204 has high frequency noise on MICBIAS will can cause a audible noise in system

Part Number: TLV320AIC3204

Hi, Team

Customer reports a audible noise issue in their walkie talkie system when they use AIC3204 replace AIC29.

I've done some tests on a TLV320AIC3254EVM to verify it, because I don't have a AIC3204evm-k right now.

Here is the test  summary I have done.

Can you please kindly help to check the root cause?

TLV320AIC3204 MICBIAS noise issue.pptx

  • Hi, Harson,

    The tests can be done on a AIC3254 EVM as it shares the same micbias structure, so on that you are fine.

    I have read the report and was able to replicate the behavior on the EVM, and actually seems that it is expected. I made more tests, by using the codec with external power supplies and using a different sequence, and found that this small signal is seen when the lower micbias values are used. This behavior might be related to the internal structure of the on-chip micbias generator, which is actually an amplifier whose output is derived from a band-gap reference. As you pointed out, the solution would be to use a higher micbias value or connect it directly to the power supply. I am looking with my team to see if this is actually a native behavior. The noise level seems really small and is out of the audible spectrum (around 5MHz), can you provide more details about how it is being recorded as audible?. 

    Best Regards,

      -Diego Meléndez López
       Audio Applications Engineer

  • 答复 RE EXTERNAL 3204芯片MICBIAS咨询.msgHi, Diego

    Thanks for the quick confirmation.

    The noise I measured is about 10mV.   The frequency is equal to ADC MOD CLK.

    In customer's system, the FS is 8KHz, AOSR = 128, then the ADC MOD CLK is much lower than we see in EVM.

    The MIC output is also about a couple of mV, so this noise will also been amplified by MICPGA.

    There is no external analog LPF in from of input of ADC.

    I've attached the I2S data read from ADC output, the number is how many LSBs.

    If there is noise, the output will have higher peak to peak values.

    In customer's walkie talkie, if the user just press Push To Talk key, and don't speaking, the receiver will hear a much higher noise.(they compared to old solution of AIC29).

    AIC29 doesn't have this problem in customer's experience.

  • Hi, Harson,

    Thanks for the additional feedback. Regarding the noise frequency, if AOSR is 128 and the sampling rate is 8KHz, then the noise frequency would be 1.024MHz, which is outside the audible range, but this could suggest that there is an aliasing problem. You mention that there is no LPF in the input, is it possible to test an external anti-aliasing filter to remove the high-frequency noise at the input of the ADC?. There is an anti-aliasing filter integrated in the codec, but it is possible that the cutoff frequency is too high to reduce the frequency of the noise (1.014MHz).

    I am still waiting for the confirmation to see if this behavior of micbias is expected. A possible solution to remove micbias noise would be to create an RCR circuit in the location of micbias resistor, as suggested in the wiki article I mentioned before.

    Best Regards,

      -Diego Meléndez López
       Audio Applications Engineer

  • Hi, Diego

    Yes. The alias of the high frequency may cause the problem.

    Customer is going to add a 1st order RC filter after MICBIAS pin, they will verify this today.

    From my test in our EVM, there is 47uF in EVM which is working with external bias, I tried it.

    It will helps.

    From the information of datasheet and technical reference, I compared the 2-nd order analog filter of AIC29 & AIC3204.

    AIC29 could have -20dB for the noise. But AIC3204 is 0dB.

    fin

    AIC29 attenuation

    AIC3204 attenuation

    10000

    0

    0

    100000

    0

    0

    200000

    0

    0

    225000

    0

    0

    300000

    0

    0

    316000

    0

    0

    450000

    -6.12

    0

    675000

    -13.17

    0

    900000

    -18.17

    0

    1000000

    -20

    0

    1197000

    -23.09

    0

    4200000

    -44.9

    -21.81

    6000000

    -51.12

    -28

    10000000

    -60

    -36.87

    Customer still waiting to close this by figure out the root cause.

    Right now my thought is the in-stability of the MICBIAS amplifier. I've also ask them to try remove the 1uF directly connected in MICBIAS pin.

     

     

     

  • Hi, Harson,

    Thanks for the additional information. The 47µF cap you mention is used as a decoupling cap for the external bias supply, by connecting the micbias output of the codec to the EXT_MICBIAS test point, you will effectively filter out the noise as an RCR circuit will be implemented. I assume this is the connection for the test you did, can you confirm?.

    Your theory seems correct, I haven't heard back from my team but will ping them again. You mention that they have a 1µF connected to the micbias output, it is important to remove it as it is not recommended to use it when the micbias is generated from the internal amplifier, the cap is only recommended when micbias is connected to VCC directly.

    Best Regards,

      -Diego Meléndez López
       Audio Applications Engineer