De-embedding for SMA connectors

[shubhangibhosale]

Hi all!
Attached is a picture of meander lines that were fabricated for measurements. however there is a difference between the measurements and simulated results for s parameters.
SMA connectors(top mount) were used, and only two of the pins of the connectors were actually soldered. To get a match between the actual measurements and simulations, i need to de embed the effect of theseSMA. also will the unsoldered vias have any effect on the measurements?

Thanks & Regards,
Shubhangi.

 

[shubhangibhosale]

Also Rohde & Schwarz ZVR Series VNA (45 MHz – 20 GHz) was used for measurements and full 2 port TOSL calibration was carried out. Do i need to use a different strategy for calibrating for PCB?

 

[milind.a.kulkarni]

Open Via's will have definate effect bcz ....it acts as inductance....also your simulation need to considered lot more parameters in details when you are dealing with S Parameter.... like parity capacitance of board, resistance of track, load that you are putting for calibration etc.

Refer these two links may be helpful to you -
http://www.ece.rutgers.edu/~orfanidi/ewa/ch13.pdf
**broken link removed**


Good Luck

 

[flanello]

How big is your difference between the measurements and the simulation?
A phase shift because of the SMA connectors could Dee-embeded by with a "port-extension", probably the network analyzer has such a function built in.
An other way is to include the SMA connectors in the simulation.

 

[shubhangibhosale]

there exists an average of 15% difference between the simulated and measured data over a frequency range of 1 to 6 GHz.
Could you shed some more light on how to execute a "port extension"?
Also i would appreciate any guidance on making a seperate calibration kit for the PCB's.

thanks a lot!

 

[flanello]

Normally you calibrate to end of the cable, and thus a SMA connector introduces to type of errors:
At first it extends the electrical path from the end of the cable to the end of the SMA connector, thus the signal gets a phase shift. If you know the extension (the first estimate is the length of the SMA connector) you can calculate the phase shift and correct the result. This has to be done for each frequency point. Modern network analyser often have a built in function for this.
The second at is a loss from the SMA connector, if this should be corrected you need to determine the loss and then deemed it from the measurement.
A good description on that topic is found in **broken link removed**

 

[FvM]

You didn't clarify, what the reference planes (respectively port locations) in your simulation are.

I guess the PCB bottom side is a continuous ground plane, without solder mask? In this case, I doubt that unsoldered pins have a large impact, although I wonder, why you didn't solder it?

A considerable effect can be expected by the vertical mounted SMA connector, edge mounted connectors would be preferable. Correcting for the connector length, which has been discussed as port extension, would be easy anyway. You would want to consult your VNA manual or a profound network analyis tutorial about the procedure. It's basically a delay correction.

A common method to compensate for connector effects would be a reference PCB with a straight microstrip line used for the T calibration.

I guess however, that most simulation to measurement mismatch is caused by the sharp bends of the microstrip and the coupling of adjacent trace segments, unless you have covered these effects in your simulation, utilizing a full featured EM solver. Did you?

 

[shubhangibhosale]

Hey flanello
Thanks for the link, found it quite useful.

I am considering making new PCB's with end launch SMA connectors. As regards the ground plane, it is continuous and without the solder mask. Also all the bend effects and coupling were considered for simulation.

Thanks,
Shubhangi.

 

[FvM]

I guess, you can change the PCB to edge mount SMA by simply cutting the board.