Resampling a binary data

TEXBAT is a recorded dataset about spoofing scenarios for evaluating GPS signal authentication techniques.

On this project, there are recorded GPS signals stored as complex 16- bit samples at a rate of 25 Msps.

I want to change its sample rate because maximum sample rate of my SDR (RTL-SDR) is 3.2Msps.

How can I change its sample rate from 25Msps to sample rates below than 3.2Msps?

[Star Strider]

I have not looked at the file and I am not certain that this approach would apply to this problem sinc e I do not understand it, however when I do resampling in a signal processing context I prefer the Signal Processing Toolbox resample function.

[Bugrahan Ustundag]

Thanks @Star Strider for the suggestion, to be more clear,

The data file is a binary file comprising of I and Q of the sampled signal. They are complex sampled. For more info: I/Q files (.bin files)

I want to resample a .bin file and create a new-sampled .bin file.

Is

resample

function valid for complex sampled signals too?

[Walter Roberson]

I don't think I quite understand some of these outputs

data = repelem([1+2i, -3-4i], 1, 4).'

data = 
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
  -3.0000 - 4.0000i
  -3.0000 - 4.0000i
  -3.0000 - 4.0000i
  -3.0000 - 4.0000i

resample(data, 1, 2)

ans = 
   0.6532 + 1.3483i
   1.2706 + 2.4499i
  -1.9323 - 2.4326i
  -3.4466 - 4.6259i

ifft(fft(data), 4)

ans = 
   4.2426 - 1.8284i
  -0.0000 + 1.0000i
  -8.2426 - 2.1716i
  -4.0000 - 5.0000i

interp1(1:length(data), data, 1:2:length(data)).'

ans = 
   1.0000 + 2.0000i
   1.0000 + 2.0000i
  -3.0000 - 4.0000i
  -3.0000 - 4.0000i

data = repelem([1+2i], 1, 8).'

data = 
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i

resample(data, 1, 2)

ans = 
   0.7370 + 1.4740i
   1.0880 + 2.1760i
   0.9318 + 1.8636i
   1.0880 + 2.1760i

ifft(fft(data), 4)

ans = 
   2.0000 + 4.0000i
   2.0000 + 4.0000i
   2.0000 + 4.0000i
   2.0000 + 4.0000i

interp1(1:length(data), data, 1:2:length(data)).'

ans = 
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i
   1.0000 + 2.0000i

I guess the fft result suggests that the power is being redistributed over the output, so the ifft output should probably be divided by the decimination factor

Answer 1

[Star Strider]

The input vector has to be a real double array. So resample the real and imaginary parts separately.

Doing the experiment —

Fs = 1000;
t = linspace(0, 9999, 10000).'/Fs;                      % Assume Column Vectors
sz = exp(2*pi*1i*t*100);
% Resz = real(sz);
% Imsz = imag(sz);
[zrr,tr] = resample(real(sz),t,500);
[zir,tr] = resample(imag(sz),t,500);
szr = zrr + 1i*zir;

figure
plot(t, real(sz), '.-', t, imag(sz),'.-', 'DisplayName','Original (Fs = 1000)')
hold on
plot(tr, real(szr), 's-', tr, imag(szr), 's-', 'DisplayName','Resampled (Fs = 500)')
hold off
grid
legend('Location','best')
xlim([0 1]*5E-2)

It appears to work as expected.

.

[Bugrahan Ustundag]

should I apply a FIR filter before the sampling?

[Star Strider]

I am not certain what you are referring to.

If you are sampling an analog signal, it is appropriate to do analog filtering of the signal prior to digitising it using a Bessel lowpass filter with the cutoff frequency slightly less than the sampling frequency of the A/D converter. Bessel filters are characteristically IIR filters, and are preferred here because they do not exhibit any phase distortion in the passband.

I generally prefer IIR filters unless I need them to have several different passbands or stopbands in the filter. In that situation, FIR filters are generally easier to design and implement, and I usually use them only as discrete filters with a signal that has already been digitised

Answer 2

[Image Analyst]

If you can read it into a time, t, array and gps (y) array, you might try interp1

gpsNew = interp1(t, gpsOld, tNew);

The old signal will have about 8 times as many samples as your new signal. Of course if you want to average the values in a certain window so you'll take the average of all the extra values instead of just picking a sample, you'll have to do some other things, like preprocessing the signal with movmean to get a signal averaged over the window.

[Bugrahan Ustundag]

Thank you for your answer, but I do not believe that I can read it into a time.

What I have is a complex sampled signal in a binary file.

You can find the binary files in this link: https://rnl-data.ae.utexas.edu/datastore/texbat/

For example, I want to resample ds4.bin file from 25Msps to 2.5Msps

[Image Analyst]

OK, no problem. Looks like Star solved it for you (since you accepted his answer).