AWR1642: The function "convertSNRdBToVar" in SRR demo

Rata Zhang

Expert 2840 points

Part Number: AWR1642

Hi,

In the function convertSNRdBToVar the frequency estimate fVar is computed as below

fVar = (float)invSNRlin * (12.0f/((2.0f*PI_)*(2.0f*PI_))) * recipsp((n_samples*n_samples - 1));

which is according to CRLB in the following formula:

var(f)=12*snr^(-1)/ ((2pi)^2*N*(N^2-1)

But in the demo,the denominator is ((2pi)^2*(N^2-1) according to the above equation, why is the parameter N missed?

Thanks in advanced,

Regards,

Rata Zhang

over 7 years ago

0 Cesar over 7 years ago

TI__Guru*** 149044 points

Hi,

Please check if in the demo code the factor N is not cancelled out.

We will also check if this is not an error

thank you
Cesar

0 Anil Mani over 7 years ago

TI__Expert 4720 points

Hi Rata,

The SNR formula given in the CRLB formula is a 'per sample' SNR (call it SNR_per_sample). However, the SNR that is given as input (call it SNR_in) to the 'SNR to var' function is computed after processing n_samples. ,Thus SNR_in = SNR_per_sample*n_samples.

Regards

Anil

0 Rata Zhang over 7 years ago in reply to Anil Mani

Expert 2840 points

Hi Anil,

The SNR formula given in the CRLB formula is for time domain signals,but in the demo,the SNR is for frequency domain. That is to say, the SNR after performing FFT is N times that of original signal.

Is that right?

Regards,

Rata

0 Anil Mani over 7 years ago in reply to Rata Zhang

TI__Expert 4720 points

Hi Rata,

Your understanding is correct.

Regards

Anil

0 Rata Zhang over 7 years ago in reply to Anil Mani

Expert 2840 points

Hi Anil,
I don't understand why the SNR after performing FFT is N times that of original signal.Can you provide some information to be refered?
Thanks in advanced
Regards,
Rata

0 Cesar over 7 years ago in reply to Rata Zhang

TI__Guru*** 149044 points

Hi,

FFT has a scaling factor N

thank you
Cesar

Sensors

Sensors forum

AWR1642: The function "convertSNRdBToVar" in SRR demo