Please or Register to create posts and topics.

Time dependence of measured signals

In the measurement we have, a 1.71 GHz microwave signal is sent to RSA250X (high pass filters in front of RSA250X to reduce P1f), and we measure the power at 1f, 2f and 3f.


As we measure P1f, P2f and P3f repeatedly (2100 times), we find that the signals show clear time dependence (page 1 of the pdf file). On the contrary, such a time dependence disappears and everything works fine as we measure P1f, P2f and P3f separately (page 2 of the pdf file). The mission file and the data (pdf) are in the attachment. The MatLab code is shown in the following.


It seems that something is mysteriously accumulating somewhere in the spectrum analyzer and results in this time dependence artifact. Any idea how to get rid of this artifact?


f = f_input*(10^9)*Harmonic;

address = 'http://localhost:54664/';
blockName = 'Block_Spectran_V6B_0';
RTSASuite = AaroniaHTTP(address);

%Get the Blockgraph configuration attached to the HTTP Server block
[remoteConfig, statusCode] = RTSASuite.getRemoteConfig();

%Set Spectran V6 paramater
[response, statusCode] = RTSASuite.setValueParameter(blockName ,'main', 'float', 'centerfreq',f);

RTSASuite.setEnumParameter( blockName ,'calibration', 'preamp',5 , 'Disabled,Auto,None,Amp,Preamp,Both');

RTSASuite.setEnumParameter( blockName ,'main', 'decimation',1 , '1 / 512');


% Get Spectra data
r = RequestMessage;
resp = send(r,URI(append(address , 'sample')));
spectrum = resp.Body.Data.samples;

n_spectrum_center = 30; % delta f = 5.5 Hz => check [f-165Hz,f+165Hz]
n_spectrum_start = 0.5*length(spectrum) - n_spectrum_center;
n_spectrum_end = 0.5*length(spectrum) + n_spectrum_center;
spectrum_center = spectrum(n_spectrum_start:n_spectrum_end);

peak = max(spectrum_center);

Uploaded files:

This is a measurement on the edge of quantum noise, I assume that there are two frequencies overlapping, i.e. the frequency at which the measurement is carried out, and another frequency in the system e.g. calibration, temperature, PLL etc. An attempt would be several samples per to take a frequency hop and see what happens or to wait longer/shorter after the hop