AnsweredAssumed Answered

Extracting Demodulated IQ samples from VSA to Matlab

Question asked by Bilal141 on Oct 11, 2017
Latest reply on Nov 7, 2017 by Bilal141

Like • Show 0 Likes0
Comment • 25

I am facing a problem with extracting demodulated IQ samples from VSA 89600 to Matlab.

I have 16QAM demodulated signal in VSA and i want to extract IQ samples to Matlab for post analysis.

I use 2 methods (actually 3). First is through saving trace in .mat file. When i try to analyze the saved IQ samples in Matlab they should be the same as transmitted samples but they are not, they are not correlated to the transmitted samples.

Second method... when i use binblockread to read the VSA buffer from the trace format IQ i loose half of the samples (i dont know why) and this is with 1 point/symbol. and when i use binblockread to read from trace format Real and Imaginary i get all the samples (exactly same as that of saving the trace to .mat file) but then again it have no correlation with the transmitted samples.

My question is what is the right way to extract the demodulated IQ samples from VSA to matlab and these samples should be the same as transmitted samples? Any help will be appreciated.

BR
BK

No one else has this question

Outcomes

25 Replies

hognala

Oct 11, 2017 12:00 PM

What format are you using? You can query it with:

:TRACe:FORMat?

The doc says:

Command Syntax:

:TRACe[1:N]:FORMat[:Y] <value>

<value>

::=

Query Syntax:

:TRACe[1:N]:FORMat[:Y]?

Returns

::=

Remarks:

<none>

Example:

:TRAC:FORM "LogMagnitude"

Version:

13.00 or later

.NET Equivalent:

For more operation and usage information, see:

Agilent.SA.Vsa.Interfaces.Trace.Format

Agilent.SA.Vsa.TraceFormatType

Actions

Bilal141 Oct 11, 2017 2:13 PM
Mark Correct
Correct Answer
As i said when i use IQ format i loose half of the sample. Then i use both the Real and Imaginary separately and get the data into two buffers and then add them as bufferI + 1j*bufferQ so i get all the data and the constellation looks the same as that in VSA but this does not correlate to the transmitted data. I need to have the IQ samples correlate to the transmitted data.

BR
BK
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Oct 11, 2017 2:54 PM
  Mark Correct
  Correct Answer
  Can you describe more about what you mean by "I lose half of the sample?" Maybe some screenshots would help us to visualize that. Also, you could use IO Monitor to capture the traffic and that might give you some insight.
  
  You might need to contact one of our Technical Contact Center's directly to get access to one of our experts (www.keysight.com/find/contactus)
  Like • Show 0 Likes0
  Actions
Bilal141 Oct 12, 2017 1:26 AM
Mark Correct
Correct Answer
So basicaly what i am doing is ... i have IQ samples (from 16QAM) which i download to ARB(AWG) then it goes to PSG then recieved on UXA and visuallized on VSA. Now i demodulate the data in VSA and want to extract IQ samples from VSA to Matlab. I use the following code

fprintf(vsa,'FORM:DATA REAL32');
fprintf(vsa, ':TRAC1:FORM "IQ"');
fprintf(vsa,'TRAC1:DATA?');
bufferdata=binblockread(vsa,'double');
iqdata_vsa1= bufferdata(1:2:end-1)+1j*bufferdata(2:2:end);

With this i get half of the samples (using 1 point/symbol)
Let say i transmit 1000 samples so i should extract 1000 symbols from VSA but i get only 500 samples.

Then i use another method shown below

clear bufferdataI
            fprintf(VSA,'FORM:DATA REAL64');
            fprintf(VSA, ':TRAC1:FORM "Real"');
            fprintf(VSA,'TRAC1:DATA?');
            bufferdataI=binblockread(VSA,'double');
                        clear bufferdataQ
            fprintf(VSA, ':TRAC1:FORM "Imaginary"');
            fprintf(VSA,'TRAC1:DATA?');
            bufferdataQ=binblockread(VSA,'double');
iqdata_vsa = bufferdataI + bufferdataQ*1j;

With this i receive all the samples but the problem with this is the received samples have no correlation with the transmitted samples. I need to receive exactly same samples as transmitted ones.

Also i try to save the trace which give me all the samples as this second method but then again no correlation with the transmitted samples.

I also try to check sample by sample like taking only few samples lets say 10 samples (samples = symbols with the case of 1 point/symbol) and then when i use the first method to extract to matlab i get only 5 samples.

What am i doing wrong here?

BR
BK
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Oct 12, 2017 7:19 AM
  Mark Correct
  Correct Answer
  Try this code instead (notice I changed double to single. This aligns with the 32-bit real32 format you are telling the instrument to use):
  
  fprintf(vsa,'FORM:DATA REAL32');
  fprintf(vsa, ':TRAC1:FORM "IQ"');
  fprintf(vsa,'TRAC1:DATA?');
  bufferdata=binblockread(vsa,'single');
  iqdata_vsa1= bufferdata(1:2:end-1)+1j*bufferdata(2:2:end);
  Like • Show 0 Likes0
  Actions
Bilal141 Oct 12, 2017 7:34 AM
Mark Correct
Correct Answer
I got all the samples but now the constellation is wrong meaning that this data which i got is not what i transmitted. I have attached the figure of constellation from iqdata_vsa1
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Oct 12, 2017 7:51 AM
  Mark Correct
  Correct Answer
  Ok, so what if you do this. Notice I am using :TRAC1:FORM "IQ"
  
  fprintf(vsa,'FORM:DATA REAL64');
  fprintf(vsa, ':TRAC1:FORM "IQ"');
  fprintf(vsa,'TRAC1:DATA?');
  bufferdata=binblockread(vsa,'double');
  iqdata_vsa1= bufferdata(1:2:end-1)+1j*bufferdata(2:2:end);
  Like • Show 0 Likes0
  Actions
Bilal141 Oct 12, 2017 7:47 AM
Mark Correct
Correct Answer
This 'single' is same as 'float' with 'real32' which i have tried, and this results in the same constellation as i have attached in previous comment, and no correlation with the transmitted samples.

And even if i mention Real32 and by quering the format using query(vsa,':FORM?') It always says Real64. and also from VSA help only Real64 and Ascii is mentioned(see attached picture)

SO does this mean i can only use Real64 and Ascii format?
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Oct 12, 2017 7:52 AM
  Mark Correct
  Correct Answer
  ya Real64
  Like • Show 0 Likes0
  Actions
Bilal141 Oct 12, 2017 7:58 AM
Mark Correct
Correct Answer
Using this

fprintf(vsa,'FORM:DATA REAL64');
fprintf(vsa, ':TRAC1:FORM "IQ"');
fprintf(vsa,'TRAC1:DATA?');
bufferdata=binblockread(vsa,'double');
iqdata_vsa1= bufferdata(1:2:end-1)+1j*bufferdata(2:2:end);

I loose half of the samples meaning that i transmitt 1056 samples i receive 528 samples and also the constellation of iqdata_vsa1 is different than that of VSA software.

SO this means that i am not having the right data in iqdata_vsa1.
Like • Show 0 Likes0
Actions
Bilal141 Oct 12, 2017 8:22 AM
Mark Correct
Correct Answer
From the VSA software the constellation looks fine, the EVM looks fine (ofc with the reference constellation) then i try to save the rx data from VSA to .mat file to compare the received samples with the transmitted, but i found no correlation b/w rx and tx samples also the constellation doesnt match with that of VSA and also the number of samples i extract to matlab are reduced by a factor of 2 (for a reason i dont know) ..... so this means i am not having the right data in matlab.
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Oct 12, 2017 8:54 AM
  Mark Correct
  Correct Answer
  What code are you using to plot it?
  Like • Show 0 Likes0
  Actions
  - hognala @ hognala on Oct 12, 2017 8:58 AM
    Mark Correct
    Correct Answer
    Have you tried this:
    
    fprintf(VSA, ':TRAC1:FORM "Constellation"');
    Like • Show 0 Likes0
    Actions
Bilal141 Oct 12, 2017 9:22 AM
Mark Correct
Correct Answer
I am using plot(real(iqdata_vsa1),imag(iqdata_vsa1),'o'); And yes i have tried this constellation format and this doesnt solve my issue.
I am wondering does it has something to do with the sampling rate in the VSA.
How to define sampling rate in VSA it is just from points/symbol? and if i have 1point/symbol it means my sampling rate is same as my symbol rate?
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Oct 12, 2017 9:53 AM
  Mark Correct
  Correct Answer
  Aha! TRACE:DATA? is only returning the Y data, so thats why you are only getting half the data. Instead, try this:
  
  fprintf(vsa, 'TRAC1:DATA:XY?');
  Like • Show 0 Likes0
  Actions
Bilal141 Oct 12, 2017 10:51 AM
Mark Correct
Correct Answer
I will try this tomorrow as i am out of Lab now but one thing more

Isn't this fprintf(vsa, 'TRAC1:DATA:XY?'); the same as the code shown below

fprintf(VSA,'FORM:DATA REAL64');
            fprintf(VSA, ':TRAC1:FORM "Real"');
            fprintf(VSA,'TRAC1:DATA?');
            bufferdataI=binblockread(VSA,'double');
fprintf(VSA, ':TRAC1:FORM "Imaginary"');
            fprintf(VSA,'TRAC1:DATA?');
            bufferdataQ=binblockread(VSA,'double');
iqdata_vsa = bufferdataI + bufferdataQ*1j;

I have tried this. With this i get the correct constellation as that of VSA, meaning both constellation of VSA and Matlab are same but with this, the data in iqdata_vsa does not correlate with that of the transmitted data.

Or this fprintf(vsa, 'TRAC1:DATA:XY?'); is different than what i have done using Real and Imaginary format?

BR
BK
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Oct 12, 2017 11:01 AM
  Mark Correct
  Correct Answer
  They are different. The TRAC1:DATA:XY? returns both the X and Y data explicitly. In your code above, both sections are doing TRAC:DATA? which gets converted to TRAC:DATA:Y? on the instrument-side (the Y? is the default query on that SCPI tree node, so if you don't supply an explicit override, it is used). So, even though you aren't adding the Y, it is implicitly getting added to the query. You can tell this from the documentation where it lists the SCPI command as:
  
  :TRACe[1:N][:DATA][:Y]?
  Anything in brackets is optional/default, so it can be omitted and the instrument will added on the back-end.
  Like • Show 0 Likes0
  Actions
Bilal141 Oct 13, 2017 3:07 AM
Mark Correct
Correct Answer
This fprintf(vsa, 'TRAC1:DATA:XY?'); works the same as this one given below

fprintf(VSA,'FORM:DATA REAL64');
            fprintf(VSA, ':TRAC1:FORM "Real"');
            fprintf(VSA,'TRAC1:DATA?');
            bufferdataI=binblockread(VSA,'double');
fprintf(VSA, ':TRAC1:FORM "Imaginary"');
            fprintf(VSA,'TRAC1:DATA?');
            bufferdataQ=binblockread(VSA,'double');
iqdata_vsa = bufferdataI + bufferdataQ*1j;

Real is I data which is X-axis and Imaginary is Q data which is on Y-axis so these both works the same

This gives me correct constellation as that of VSA but the data in iqdata_vsa does not correlate with the transmitted data. I dont know why is my received data different than my transmitted data even though i have specified the demodulation properties in VSA correctly.

BR
BK
Like • Show 0 Likes0
Actions
Bilal141 Oct 13, 2017 7:12 AM
Mark Correct
Correct Answer
OK i solved my issue It was basically related to how i download my input signal to ARB (AWG).

Thank you for your Help. I appreciate that

BR
BK
Like • Show 1 Like1
Actions
- hognala @ Bilal141 on Oct 13, 2017 7:37 AM
  Mark Correct
  Correct Answer
  Excellent!
  
  Thanks for the followup
  Like • Show 0 Likes0
  Actions
Bilal141 Nov 6, 2017 10:26 AM
Mark Correct
Correct Answer
Hi I hope you are doing well. I need to ask something more related to the same task.

As i am now able to read the downsampled, demodulated data from VSA to Matlab. Now i need Raw data (data before demodulator, which is upsampled data) from VSA to Matlab, which is in the Raw main time trace (without timing correction) and Time trace (with time correction).
When i read this data into Matlab and downsample it in Matlab with same filter setting as used at transmitter side, and then i calculate EVM. This EVM is not the same as EVM in VSA for downsampled data. And i need the EVM in Matlab and VSA the same.
Am i doing something wrong here? According to what i understand from reading.. Time trace gives the raw data and the number of samples depends on the frequency span selected which i use Fs/1.28

BR
BK
Like • Show 0 Likes0
Actions
- hognala @ Bilal141 on Nov 6, 2017 10:33 AM
  Mark Correct
  Correct Answer
  I'll see if I can find someone who can answer this piece.
  Like • Show 0 Likes0
  Actions
Bilal141 Nov 6, 2017 10:37 AM
Mark Correct
Correct Answer
Ok Thank you I will be waiting for your response here.

BR
BK
Like • Show 0 Likes0
Actions
hognala Nov 7, 2017 8:19 AM
Mark Correct
Correct Answer
Bilal141 - Here is the feedback that I received:
"Raw main time provides uncorrected data as received from the hardware. Main Time provides data which has time corrections applied and is resampled for the user selected span."

To get assistance with the MATLAB demodulation issue, we recommend that you contact your local Technical Contact Center (www.keysight.com/find/contactus) and a domain expert can be assigned to help.
Like • Show 0 Likes0
Actions
Bilal141 Nov 7, 2017 9:31 AM
Mark Correct
Correct Answer
Hi

Thank you so much. I will try to contact these expert see if i can get any help.

BR
BK
Like • Show 0 Likes0
Actions

Extracting Demodulated IQ samples from VSA to Matlab

Outcomes

Related Content

Recommended Content