AEC example Works on GSM Calls but Not on VoIP Calls (WhatsApp etc.)

[batuhantsdmr]

Checklist

• Checked the issue tracker for similar issues to ensure this is not a duplicate.
• Provided a clear description of your suggestion.
• Included any relevant context or examples.

Issue or Suggestion Description

Hello,

I am using the ESP-SR AEC algorithm(https://github.com/espressif/esp-adf/blob/release/v2.x/examples/advanced_examples/algorithm/main/algorithm_examples.c) on an ESP32-LyraT-Mini V1.2 board.

When I make a regular GSM cellular call, the AEC noticeably reduces the echo and the far-end user hears much less acoustic feedback.

However, when I make an internet-based voice call using applications such as WhatsApp
the AEC appears to have almost no effect. The far-end user hears nearly the same echo level as when AEC is disabled.

The setup is:

A smartphone is connected to the ESP32-LyraT-Mini V1.2 via Bluetooth HFP.
GSM calls are initiated on the smartphone.
WhatsApp calls are also initiated on the smartphone.
The ESP32 acts only as the Bluetooth hands-free audio device.
The same microphone, speaker, Bluetooth connection, and AEC configuration are used in both cases.

In other words, the only difference is whether the smartphone call is a GSM call or a WhatsApp/VoIP call.

With this setup:

GSM call over Bluetooth HFP → AEC works well.
WhatsApp call over Bluetooth HFP → AEC has little or no effect.

What is the reason for this?

How can I fix this?

[sun-xiangyu]

I guess the audio codecs between GSM call and WhatsApp call is different, please check it

GSM call → Phone and headset negotiate a Narrowband (8 kHz) CVSD codec over HFP.

WhatsApp call → Phone prefers Wideband (16 kHz) mSBC for better speech quality.

[batuhantsdmr]

--audio state connected ← "connected" = CVSD/8kHz
AFE: Input PCM Config: ... sample rate:8000

Also, if there's a problem like 16kHz <--> 8kHz, wouldn't there be distortion in the sound? There's no problem with the sound itself, but there's an echo in the WhatsApp calling.