Crosstalk Out to In

Placement, modules, components and accessories; the ones that exist and the the nice-to-be's
Post Reply
gmb
Posts: 4
Joined: Sat Mar 18, 2017 11:05 pm

Crosstalk Out to In

Post by gmb » Mon Apr 03, 2017 7:55 pm

I just did some measurements about the crosstalk between Out1, Out2 and the inputs. All Outputs and Inputs are terminated with 50 Ohms. This is basically the coupling from the outputs to the inputs, not so much the couling In1 to In2. I swept Out1 and after that Out2 and plotted the coupling to both inputs giving 4 curves.

I put the Red Pitaya into the metal casing (the original one) and made a second run of measuring. See both files - one before, the other one after. There's not too much change.

Any idea where the maximum at exactly 200 kHz comes from? Is there any inductors somewhere that are resonant at 200 kHz and cause magnetic coupling?
You do not have the required permissions to view the files attached to this post.

gmb
Posts: 4
Joined: Sat Mar 18, 2017 11:05 pm

Re: Crosstalk Out to In

Post by gmb » Wed Apr 05, 2017 10:51 pm

Did some more ... if I disconnect 50 Ohm termination at signal generator output, the crosstalk gets better by almost 10 dB. Red and blue is In1 and In2 as before, green and magenta is after disconnecting 50 Ohm termination only at output. All Inputs still terminated.

Does not seem to be capacitive coupling ... should get worse not better as the voltage is higher without load.
You do not have the required permissions to view the files attached to this post.

User avatar
redpitaya
Site Admin
Posts: 880
Joined: Wed Mar 26, 2014 7:04 pm

Re: Crosstalk Out to In

Post by redpitaya » Wed Apr 12, 2017 9:26 am

Hi,

Can you please describe in more details how you performed measurements
(sweeping of out1 and out2).
I.e have you used SCPI commands etc.

Regarding the second part:
If you have inductive coupling then in case of larger current going from outputs (load connected to the output) you will get more crosstalk.
and if the output current is small (no load) the crosstalk will be also smaller.

Also, which version of STEMlab you are using
STEM 125 -14 or STEM 125-10?

Best, RP team

gmb
Posts: 4
Joined: Sat Mar 18, 2017 11:05 pm

Re: Crosstalk Out to In

Post by gmb » Thu Apr 13, 2017 7:50 pm

I am using Octave to communicate with SCPI commands. I did use both out1 and out2 but not simultaneously. I swept out1 (out2 off) and measured in1 and in2 and after that I swept out2 (out1 off) and measured again both inputs. This way I get 4 curves. I am using STEM 125-14.

In the meantime I got new interesting results. I compared the crosstalk out2 to in1 and a second time out2 to in1 but with out1 also switched on at same amplitude and frequency but 180 degrees phase. Both outputs are left open. It does help quite a lot in the range 30 kHz to 300 kHz (around 10 dB better!).

I am still not sure how the coupling works ... I also tried to connect 47nF to the +3,3V to ground on the connector, but that did not have any visible effect on cross coupling. My idea was that a ripple on the supply might cause this.
You do not have the required permissions to view the files attached to this post.

User avatar
redpitaya
Site Admin
Posts: 880
Joined: Wed Mar 26, 2014 7:04 pm

Re: Crosstalk Out to In

Post by redpitaya » Thu Apr 20, 2017 11:58 am

Hi,
measuring crosstalk is quite tricky thing.
you must be sure to have well defined setup and on that setup you define the crosstalk levels.
So, I think that correct way of measuring crosstalk for example
between OUT1 and IN1 is:

1. OUT1, OUT2 , IN1, IN2 are all terminated with 50OHM. OUT1 and OUT2 are designed for 50Ohm.
2. OUT1, OUT2 amplitudes are set to 0V.
3. Acquire signals on IN1 and calculate FFT. When calculating FFT you need to use correct equations for calculating dB.
4. Save this FFT "trace" as "noise_fft"
5. OUT1 amplitude is set to 0.7V (don't use max amplitude in order to avoid non-linearities of DAC)
then you make freq. sweep for OUT1 frequency - from 100Hz to 50MHz.
6. For each f. step of OUT1 acquire signal on IN1, calculate FFT -> you get "cross_talk_FFT(f.step)"
7. for each f. step subtract "noise_fft" from "cross_talk_FFT(f_step)"
8. plot all "cross_talk_FFT(f_step)" one above another. (don't average them!!!)
8a. instead of 8. you can pick up max values for each f.step from "cross_talk_FFT(f_step)" matrix.
with that you will extract worst cases.

9. When you start STEMlab board start Oscilloscope application and then close it.
This will load fpga filter parameters.
then you can go with SCPI and remote control.

Now, when doing acquisition you should allays have same sampling rate in order to get
correct FFT calculations.
Preferably you should use max sampling rate.
But if you wish to look at lower freq. range you can use lower sampling rate
and repeat steps from 1-8 while max OUT1 freq is limited by Nyquist.


"I am still not sure how the coupling works ... I also tried to connect 47nF to the +3,3V to ground on the connector, but that did not have any visible effect on cross coupling. My idea was that a ripple on the supply might cause this."

- Yes, power supply ripples and ground plane can affect outputs performances.
but I think this will not affect crosstalk noticeably.
https://ln1985blog.wordpress.com/2016/0 ... rformance/
This is the reason for step 4. exclude environment noise form crosstalk measurements.

The code bellow is for calculating power noise using FFT.
When dealing with amplitude ratios your FFT equations will look different
In equation 1 (look at code) you should use magnitude of complex value and not complex conjugate multiplication
Amplitude spectrum in volts RMS = sqrt(2)* mag(FFT(A))/N
in our case
Amplitude spectrum in volts RMS = sqrt(2)* mag(xdft_hann)/N
and equation 2 should look like

A= 20*log10( Amplitude spectrum in volts RMS/OUT1 amplitude);

Code: Select all

%% data  - acquired signal in volts  
Fs = 125E6;
t = 0:1/Fs:1-1/Fs;
N = length(data);
freq = 0:Fs/N:Fs/2;
R_load=50;

%% Hann window;
w = hann(N)';  
K = 64;
tmp = hann(K)/K;
tmp = fft(tmp);
tmp = tmp.*conj(tmp);
dcspan = sum(tmp(1:K/2)>1e-6)-1;
win_gain = sqrt(sum(tmp));


xdft_hann = fft(data.*w)/win_gain;
xdft_hann = xdft_hann(1:N/2+1);
%% Equation 1
Pmag_hann=(xdft_hann.*conj(xdft_hann))./(N^2);       % Calculate power spectrum to Fs/2   

Pmag_hann = Pmag_hann./R_load;                                 % Calculate for 50 Ohm
%% Equation 2
Pdb_hann  = 10*log10(Pmag_hann);                               % Power in dB 1W as reference
PdBm_hann = 10*log10(Pmag_hann./1E-3);                 % Power in dBm  1mW as reference

Best, Zumy
PS 1: Double check the new equations for amplitude - I didn't test it
PS 2: A lot of people use same equation for FFT spectrum calculation and then getting wrong results.
mag(a+jb) is not equal to (a+jb)x(a-jb) (complex conjugate)

Post Reply
jadalnie klasyczne ekskluzywne meble wypoczynkowe do salonu ekskluzywne meble tapicerowane ekskluzywne meble do sypialni ekskluzywne meble włoskie

Who is online

Users browsing this forum: No registered users and 15 guests