• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 146 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

TLV320AIC3262: Background noise observed in handset mode

akshi malhotra
Prodigy 245 points
Part Number: TLV320AIC3262
Other Parts Discussed in Thread: AM5728

Tool/software:

Hello,

We have been testing audio quality in different scenarios after the changes suggested by TI. There has been significant improvement in audio quality. 

But we are facing background noise in a particular scenario explained below:

During the call between A and B, if B is in hands-free mode and A is in handset mode then A is hearing a background noise. 

We have tried to decrease the microphone gain in Hands-free mode, but this leads to decrease in audio level in Hands-free mode.

The following adjustments were also tested and fine-tuned but did not resolve the issue:

   1. AGC Hysteresis Setting and AGC Noise Threshold (B0_P0_R87 and B0_P0_R95) Current Value: 0xAE
   2. AGC Maximum Gain Setting (B0_P0_R88 and B0_P0_R96) Current Value: 0x44

We have attached the audio scripts for Hands-free and Handset mode. The sample recordings for this issue are also attached :

* File 1: Phone_B_Microphone_Recording.pcm — Audio recorded using the microphone of Phone B.

* File 2: Phone_A_Audio_Received_From_Phone_B.pcm — Audio received on Phone A, originally recorded by Phone B.

Could you please recommend configuration that might resolve this issue?  Phone_mic_capture.zip

  • 0
    TI__Mastermind 35220 points

    Hi Akshi,

    Could you give some more detail of the signal chain in these modes? What is the difference between headset and hands-free mode? When you say background noise, do you mean white noise as in poor SNR? or noise from behind the person speaking into the microphone, as in the microphone is too sensitive?

    Thanks,
    Jeff McPherson

  • 0 in reply to Jeff McPherson
    Prodigy 245 points

    Hello Jeff,

    Handset Mode: This mode usually features a microphone built into the handset of a VoIP phone (similar to a landline), placing the microphone closer to the speaker's mouth.

    Hands-Free Mode: This mode usually employs a speakerphone with a separate microphone from the handset, placing the microphone farther away from the user. As a result, it tends to capture more ambient sound.

    Detail of the signal chain :

    Analog Mic → Audio-codec (TLV32AIC3262) → SOC(AM5728)→→ → → Network → → → → Audio-codec (TLV32AIC3262) → SOC(AM5728) → Handset Speaker / Hands-free Speaker

    Background noise refers to white noise (ambient noise). In the previous query, the zip file containing scripts for hands-free mode and handset microphone, as well as the voice sample, was attached.

  • 0 in reply to akshi malhotra
    TI__Mastermind 35220 points

    Hi Akshi,

    I listened to the audio. I'm not familiar with the language so its hard to tell exactly what is happening, but it sounds like the AGC is suppressing most of the noise correctly, and then boosting the signal when the speaker talks, where there is background noise. Am I understanding correctly? Depending on how far away the user is from the microphone, by definition the SNR will be worse, meaning the noise is going to be very difficult to eliminate without more powerful tools such as Active Noise Cancellation or Beamforming algorithms. With more basic tools like the ones in the codec, you can increase performance by keeping the microphone as close to the intended source as possible, and consider using a directional microphone.

    Some suggestions for tuning the codec, rather than reducing microphone gain, you could reduce the max PGA gain applied by the AGC preventing the PGA from boosting the noise too far. You can boost the signal with digital volume instead, meaning you don't add any additional noise from the PGA. You could also implement the first order IIR filter in the signal processing chain to help reduce the signal band around the voice. Typically voice signals have no meaningful content past 8kHz or below 200Hz.

    Best regards,
    Jeff McPherson