Measuring the delay spread with a network analyzer.

[manos87]

Hello to all,
I would like some help from you in a project that I work now.
I built two monopole antennas and I measured reflection coefficient of each one and the S21 for the received power.
The two antennas operate at 900MHz. What I wanted to ask you is how I can find out the delay spread using the VNA?
Any ideas?
Regards,
Manos

 

[FvM]

First prerequisite is that a the antenna configuration actually involves multi-path transmission. Does it?

Secondly the "other" transmission pathes must be strong enough to show up when S21 is transformed to time domain.

 

[manos87]

First prerequisite is that a the antenna configuration actually involves multi-path transmission. Does it?

Secondly the "other" transmission pathes must be strong enough to show up when S21 is transformed to time domain.

How can I ensure that the antenna cofiguration actually involces multipath-transmission?
Also, I can set the power to 30dBm to be sure that the other paths are strong enough.
If I make sure that the two conditions above are valid then I can set the number of points to the highest point in my case is 1601 points.
I concetrate in the range of frequencies from 600MHz up until 1200MHz and I get the data from the VNA.
Here I want to ask you if I should set the sweep time to 1s.
Then I use the ifft function on matlab like below:
(Z is the complex number that I get from VNA data.

yifft = ifft(Z, N)*N;
figure(1);
amp =abs(yifft); % plot
time = 0:1/N1-1/N); % ^^ the same <- sweep time of 1 second. is that correct?
plot(time, amp);
grid(on);
title('Time- Domain signal');
xlabel('time(s)');
ylabel( 'Amplitude');




The amplitude is pretty big up to 35, where as the received power in 900MHz is -28dBM.:???:
Any thoughts?
Regards,
Manos

 

[FvM]

A real multipath setup could be created by reflecting surfaces (e.g. walls) in the antenna's enviroment.

Typical VNA output power is around 10 dBm, you'll hardly find an instrument that provides +30 dBm without external wideband amplifier.

I'm not sure about the relevance of a increased sweep time. I would expect steady state with regular sweep time as well. But you would want to set small bandwidth for increased SNR if it's selectable with your VNA, it sets a slower sweep as well. Depending on the instrument type, automatic sweep time selection may be manually overridden.

 

[manos87]

Yeah you are right FvM the instrument provides output power up to 10dBm. My mistake.
Also, the measurements take place at an indoor lab with 4 walls, a ceiling and a floor so I think it should be ok.
Above I have the impulse response from the newtowork analyzer.
What should I do to find the delays spread now?

 

[manos87]

Any help guys?

 

[FvM]

Applying common sense to the result:
All wall echos can be expected in a several 10 to maximum 100 ns range. Anything beyond this range is just noise.

 

[manos87]

I found a mistake that I made.
First of all, when I measured the received power I used a Frequency span of 600MHz and number of points(samples) equal to 1601(it is the maximum value on my VNA).
I think that is more sensible to use a frequency span of 100MHz and 1601 number of points.
Now, the problem is that as FvM said the time delay should e from 10-100ns but as you can see in my plot this is not the case.
My question is,how can I compute the time vector to plot the impulse response?
Should it be t = 0:1/1601:1?
ot t = 0:1/100MHz: 1601*(1/100MHz)?
Thanks in advance for your answers!

---------- Post added at 13:45 ---------- Previous post was at 12:03 ----------

FvM one last help?

 

[manos87]

Okay I am gonna write a part of matlab code that I use in order to get the impulse response from the ifft.

% Gathering data -irrelevant with our result
N = 1601; % Number of points
K = importdata('D0-5ST1SPAN250.S1P',' ');
freq1 = K.textdata(4:1604,1);
real = K.textdata(4:1604,2);
im = K.textdata(4:1604,3);

for i=1:N
Freq1(i) = str2double(freq1(i));
Real = str2double(real(i));
Im = str2double(im(i));
Z(i) = Real + Im*1i;
Phase(i) = angle(Z(i));
% ABsval = abs(Z);
end

%Frequency span
Span = (Freq1(N) - Freq1(1))*10^6;
step = 1/Span;


%%% Impulse response

yifft = ifft(Z, N);
figure(1);
amp =abs(yifft);
power = amp.^2;
dbpower = pow2db(power);
time = 0:step:1600*step;
plot(time, dbpower);
grid on
title('Time - Domain signal d = 0.5 sweep time = 1s');
xlabel('time(s)');
ylabel( 'Power(dBm)');

%%% THE FREQUENCY SPAN IS 250MHz.

The file that we read as an input is also attached along with the plots for the channel impulse response.
Do you see something wrong in my results?
Thanks in advanceView attachment SPAN250.pdfView attachment untitled.pdfView attachment D0-5ST1SPAN250.rarView attachment SPAN250.pdfView attachment untitled.pdfView attachment D0-5ST1SPAN250.rar