IWR1443BOOST: CFAR_THRESH / mmWave Demo confusion

Enric Gardein

Part Number: IWR1443BOOST

Hello everyone,

I need to know which value is exactly used for the CFAR threshold, based on register CFAR_THRESH.

I looked through e2e forum, documentation and code, but found no solution - it was more confusing.

In the User Guide for the Demo Visualizer it is stated, that the value given in dB needs to be represented in Q9 format.

I understand the given formula (Q9 conversion + 20log(10) to log(2), discussed here for example e2e.ti.com/.../2322209 ).

But I dont understand, why it should be Q9.

The HWA Manual (HWA 2, page 14 and 15) says, that the CFAR_THRESH is a 18-bit register, which is interpreted as 14.4 (lin) or 7.11 (log) value.

Neither 14.4 nor 7.11 is Q9 format, right? Could you please clearify this?

Additionally, I am confused about line 1500 in data_path.c of mmWave demo:

hwaCommonConfig.cfarConfig.cfarThresholdScale = obj->cliCfg->cfarCfg.thresholdScale;

On the left hand side, we have uint32 (which is enough to store the 18 bit value for CFAR_THRESH) and on the right hand side we have uint16.

I guess this is because we do not need larger threshold scales for mmwave demo, so we dont need to worry about the first 2 bits of CFAR threshold (so they are 0). Is this correct?

One last thing - just for clarification: The HWA can run CFAR in log mode. This means, that all values (2D FFT) are log2(abs(x)) values, right?

Thank you very much, I really appreciate all the support you give here at e2e,

Enric

over 7 years ago

0 Cesar over 7 years ago

TI__Guru*** 148954 points

Hi,

Please see CFAR related threads

e2e.ti.com/.../632673

e2e.ti.com/.../631790

Thank you
cesar

0 Enric Gardein over 7 years ago in reply to Cesar

Intellectual 500 points

Hi Cesar,

thanks for your reply.

I already saw those threads and read through the documentation and code (as mentioned in my opening post).

But in the User Guide for the Demo Visualizer is stated, that the threshold is represented as Q9,

but in the documentation for the HWA is stated, that the threshold is represented by 7.11 (log).

These to statements do not match.

Please clearify this.

Thanks,

Enric

0 Enric Gardein over 7 years ago in reply to Enric Gardein

Intellectual 500 points

Hello,

could someone please look at this?

Thanks,

Enric

0 Michelle Liu over 7 years ago in reply to Enric Gardein

TI__Genius 11405 points

Enric,

Q11 format comes from log2 output. In the HWA user guide, the threshold scale is Q11 because it is assuming a linear input and followed by log2 before CFAR. So the CFAR threshold and input to CFAR are both Q11 format.

However, in mmw demo, log2 and summation steps are taken before CFAR (no log2 inside of CFAR block).
log2 output is Q11
Summation step: the input scale (srcScale) is 2, meaning 2 redundant bits are added to MSB and 6 bits padding to LSB, then it's in Q17 format.
the dstScale is set to 8, meaning summation step output to local memory will have 8bit dropped at LSB, giving Q9 format.

This is the value being used for CFAR so CFAR threshold needs to be Q9 format as well.

Hope it's clear enough to your question. Let me know if further support is needed.

Thanks and Regards,
Michelle

0 Enric Gardein over 7 years ago in reply to Michelle Liu

Intellectual 500 points

Hello Michelle,

thank you very much!
This was very clear and I understood.

Thanks,
Enric

Sensors

Sensors forum

IWR1443BOOST: CFAR_THRESH / mmWave Demo confusion