TIDA-01470: Changing coefficients for beamforming using AER bf_tool

Rafael

It appears that this got posted on the wrong E2E forum. Somehow your quest got posted on the low power amplifier and comparator forum. I can try to move this for you but you may want to repost if you need immediate assistance.

Chuck

Hello Chuck,

This is very weird as to me it shows the thread on the TI Designs forum.

Anyway, I am trying to repost the question to the Audio Forum.

Thank you,

Rafael

Hi Rafael,

I see you have posted your question to another forum therefore I am going to close this thread to reduce duplicate forum threads. If you still need support from this forum please reply back within 30 days and it will reopen the thread.

Thank you,

Tim Claycomb

Hi Tim,

I am still not getting any kind of response for this issue, even on the Audio forum. Maybe I should be posting to a different forum?

It is very confusing since this reference design from TI deals with many different components and technologies. What do you suggest?

Thank you,

Rafael

Hi Rafael,

I apologize for you not getting a response in a timely manner. I'll attempt to locate the author of the document or someone else who will be better suited to help you.

Thank you,

Tim Claycomb

Dear Tsvetolin,

please do not close this thread.

As you may observe, the other one was only created due to a recommendation on this thread, but it has not gotten any replies so far. Moreover, Tim commented on the previous post that he is checking if he can find someone that may help me.

If you need to close one of them, please close the other.

Thank you,

Rafael

Rafael,

As I'm seeing you are using coefficients generated by bfgui.exe tool as it is described at:
training.ti.com/.../Voice-Processing-Tools-Software-SLIDES.pdf
but the coefficients can vary depending on Beamformer parameters as it is described in TIDEP-0088 section 4.1.2 Calculating Filter Coefficients at:
www.ti.com/.../tiducr7a.pdf
If bf_tool generated coefficients are different then these in demo example it shoold not be a problem because the input conditions (beamformer parameters) can be different.

BR
Tsvetolin Shulev

Hello again Tsvetolin,
You are correct, I am trying to follow the steps on TI's documentation. But I am using an LMB setup instead of the CMB used in the documents you quoted.
I understand that different beamforming parameters will generate different coefficients, but I tried using the same conditions as described in www.ti.com/.../tiducy1c.pdf since I am using the C5517 EVM with the EVM from TIDA-01470 reference design. I believe in such conditions I should obtain the same coefficients.
Maybe I am not able to generate the same values because I do not understand the angle parameter I need to use in bf_tool. Could you clarify that for me? Especially relating to the board from TIDA-01470, this angle is relative to what?
Any help here is appreciated since so far I was not able to get promising results from this setup. The processed audio still has a lot of noise that we would like to eliminate from the microphones inputs.

Best regards,
Rafael

Rafael,

The beamforming filter coefficients depend on the geometry of the microphone array and the angle of the direction of the source with respect to the microphone array.
You can find some clarifications and diagrams at:
www.labbookpages.co.uk/.../delaySum.html

BR
Tsvetolin Shulev

Dear Tsvetolin,
I am aware of this parameters and their importance to the coefficients.
As I said in my original post:
===========================================
I tried generating the coefficients using:

Sampling rate: 16 kHz
# Mics: 4
Mic distance: 2.125 cm
BF angle: 90
Geometry: 0

I got the following coefficients:

-22 98 -241 471 -817 1331 -2120 3470 -6434 20542 20967 -6847 3861 -2478 1649 -1089 692 -409 213 -88 20 0 0
0 -22 98 -241 471 -817 1331 -2120 3470 -6434 20542 20967 -6847 3861 -2478 1649 -1089 692 -409 213 -88 20 0
0 20 -88 213 -409 692 -1089 1649 -2478 3861 -6847 20967 20542 -6434 3470 -2120 1331 -817 471 -241 98 -22 0
0 0 20 -88 213 -409 692 -1089 1649 -2478 3861 -6847 20967 20542 -6434 3470 -2120 1331 -817 471 -241 98 -22

While the ones in the demo file are:

#if NUM_OF_MICS==4
/* SYS_BF_ANGLE_P90 */
Fract sysBfFltP90[][SYS_BF_FILTER_LENGTH] = {
{ -50, 150, -319, 576, -948, 1485, -2296, 3662, -6639, 20754, 20754, -6639, 3662, -2296, 1485, -948, 576, -319, 150, -50, 5, 0, 0, 0},
{ 0, -50, 150, -319, 576, -948, 1485, -2296, 3662, -6639, 20754, 20754, -6639, 3662, -2296, 1485, -948, 576, -319, 150, -50, 5, 0, 0},
{ 0, 5, -50, 150, -319, 576, -948, 1485, -2296, 3662, -6639, 20754, 20754, -6639, 3662, -2296, 1485, -948, 576, -319, 150, -50, 0, 0},
{ 0, 0 , 5, -50, 150, -319, 576, -948, 1485, -2296, 3662, -6639, 20754, 20754, -6639, 3662, -2296, 1485, -948, 576, -319, 150, -50, 0}
};
===========================================
So, I am trying to generate coefficients using the EXACT same specifications, which means same geometry (linear array), same distance between 2 mics, same sample rate, same number of mics and same angle (positive 90º), and even then I get different values.
So my question remains: Why are the coefficients different?
The problem is that if I am inputing wrong parameters, I do not understand what is wrong and therefore will not be able to use the tool correctly. As far as I am concerned I am setting the parameters correctly, so if I am not I need a clear explanation of what should be different.

Regards,
Rafael

Rafael,

You are welcome. All the best with the product development!

Lali