XVF3510 echo cancellation not working for VoIP call

[Marius]

I purchased an evaluation board for XVF3510 chip (XK-VF3510-L71 dev kit board).
The purpose is to integrate the XVF3510 chip into our product, to cancel the echo for a VoIP conversation (Acoustic Echo Cancellation).
But, despite the fact that we are using XVF3510, the echo is still present and disturbing.

Using the XVF3510 development kit I followed the "getting started" pdf guide on XMOS website and configured hardware/sound settings on my PC and MacBook.
With XVF3510 board connected to USB, speakers near the board, music was playing from youtube in the speakers (XVF3510 also receiving the audio via USB).
I recorded myself with Audacity speaking over the music. Indeed, when I play the recorded audio from Audacity (only left channel) - my voice was clear, with (almost) no music.

I updated the firmware inside the XVF3510 with the i2s version of binary on XMOS website (XM-013711-SM-G-XVF3510 I2S Firmware binary.xe).
Using an USB-i2s board (this model) I was able to make the same test as above - the audio recorded with Audacity was mostly my voice (the music was at a quiet/soft level).

Next, I made a test with a VoIP call, using a PC softphone (3CX).
From the PC I am calling a mobile phone user.
The PC with 3CX softphone has the XVF3510 dev board connected to USB - so 3CX is using the output from XFV3510 as microphone input.
The incoming voice call from 3CX softphone is played on local speakers (also, this audio stream goes to XFV3510 board, for audio reference).
Locally, the VoIP call is normal - I can speak in the microphone of the XFV3510 and the mobile phone user can hear me, I can hear in the speakers the voice of the mobile phone user.

The problem is that the mobile phone user hears his own voice on the mobile phone.
His own voice is played like an echo on this phone, after about one second after he speaks.
What I understand is that the XFV3510 receives the voice (of the mobile phone user) from the local speakers. And XFV3510 sends back, over VoIP call, the voice of the mobile phone user.
But this means the AEC (Acoustic Echo Cancellation) is not working on XFV3510 ?
How can we get rid of this annoying echo ?
Is it something that we need to configure on XFV3510 (set some parameters) ?
How come the XFV3510 is doing a good job when I make the test like in the "getting started" pdf guide (recording my voice over the playing music), but when I make a VoIP call, the echo cancellation function is not working ?