Re: Signal Package pwelch

[Thomas D. Dean]

On 8/13/19 1:32 PM, Thomas D. Dean wrote:
> I have a application using the signal package pwelch function that seems 
> to produce an incorrect plot.
>
> [s, f] = pwelch(iq,[], overlap=0, NFFT=2048, Fs=sample_rate, 
> detrend='none', range='shift', plot_type = 'dB')
>
> iq is read from an rtl-sdr
>
> knowing the location of some signals, like the Hydrogen line, and, 
> comparing this to python, pwelch puts the "hydrogen bump" about .3MHz 
> too high.
>
> pwelch(iq,[], overlap=0, NFFT=2048, Fs=sample_rate, detrend='none', 
> range='shift', plot_type = 'dB');
>
> How do I upload an image to show this?

More information,

The "hydrogen bump" should be between 1420400000 and 1420500000 Hz.
But, pwelch places it between 1420700000 and 1420850000

octave:100> [s, f] = pwelch(iq,[], overlap=0, NFFT=2048, ...
        Fs=sample_rate,  detrend='none', range='shift', ...
        plot_type = 'dB');
octave:101> freq
freq =  1420400000
octave:102> x = freq + f;
octave:103> idx=find( ( x >= 1420700000 ) & ( x <= 1420850000 ) );
octave:104> plot( x(idx), 10*log10(s(idx)));

I can post the .cc and the .m code, if anyone is interested.

Tom Dean