Hi All,

I was toying around with the rfft() of the SHARC 21369 with a digital sinewave data. After the rfft(), the complex number is then converted to magnitude by rfft_mag(). The signal is been conditioned by a kaiser window with beta of 18.

The question is, after the conversion the noise floor of the signal is around -150 dBFS, when i do the same process on the same data with matlab it yielded -200 dBFS.

So is this the limiation of the rfft() function? as the function only accept float type but not long double type.

Anyone experience this before?

I have attached the signal file. Sampling rate of 48 kHz if you are wondering about it.

I finaly figure out what is causing the high noise floor. It is due to the kaiser window that is generated by gen_kaiser(). I have dumped the value of the gen_kaiser() from the SHARC and have it do windowing in matlab, the end result is -150 dB noise floor.

It seem like increasing the beta of the kaiser window in gen_kaiser() does not have improvement in the noise floor.

Is this the problem with the 32 float number for gen_kaiser() ? Cause Matlab is using 64 bit double to generate kaiser.