Single-ended two-port VNA measurement issue__need help

[rf1008]

Recently,i was measuring balanced load's impedance,such as dipole antenna. I used Agilent E5061B VNA for the single-ended two-port measurement and got 4 S parameters,then the balanced impedance could be calculated by formula.
I found the result has much difference with simulation. One 75R 0603 resistor has been measured to verify the difference.S11=S22=0.429+j0,S12=S21=0.571+j0,they're the simulation data.But the measured results are:S11≈S22=0.423+j0.008,S12≈S21=0.511+j0.235. S11/S22 are quite similar ti simulation,S12/S21 have much difference in Real and Imag.
I finally found the doubt curve after VNA calibration,three pictures are attached here. N1021B is my single-ended two-port measurement fixture as differential probe. OPEN/SHORT/LOAD calibration is done at two coaxial ports and then use port extention function to compensate N1021B,the final calibration result is correct,pls refer to attachment which name is "calibrated". All actions above are correct,I think so!
But when i shorted the two tips of N1021B,the curves of S12 and S21 changed much.In picture which name is "Tips short",you can find it.When two tips are short,S11 and S22 should be at 50 OHM point,S11=S22=0+j0,S12 and S21 should be at OPEN point,S12=S21=1+j0.The S11/S22 on Smith chart have a little variation from 0+j0 and can be accepted,S12/S21 however have too much variation from 1+j0, especially at 900MHz S12/S21=0.874+j0.483,the Real part is less then 1 and the Imag part is too much larger then j0.
I think the full calibration and compensation are correct, but, why S12/S21 changed too much according to OPEN point when N1021B two tips shorted togather. This instrument is purchased from Agilent this year, i don't think it has any problems.

Thanks,may someone could help me.

 

[D.A.(Tony)Stewart]

It might be useful to review your test fixture and your calibration method with swept return loss and difference in resonant frequency. Naturally you need an anechoic environment to reduce reflection induced effects.

 

[rf1008]

It might be useful to review your test fixture and your calibration method with swept return loss and difference in resonant frequency. Naturally you need an anechoic environment to reduce reflection induced effects.

Thanks! I use the standard OPEN/SHORT/LOAD method for calibration and port extention to compensate the length of differential probe(N1021B). I don't understand your words of "with swept return loss and difference in resonant frequency" .

I want to measure impedance at 900MHz,and chamber is good for antenna measurement. But i just measure a pure resistor as DUT and find the s12/s21 variation at higher frequency according to the curve in attachment. thanks.

 

[kayaoo]

I guess port extension is not accurate enough and the tip length is not included.In your case, port extension when "short" may be more accurate. But the calibration error should not bring much trouble in your antenna test except the phase should be included.

 

[hxtsz15]

Kayaoo is right, your extension is wrong, if you could apply calibration directly on the tip of the probe, it would be much better.

Or you can measure a well known short line (delay), and modify the extension.

 

[D.A.(Tony)Stewart]

I suggest onboard coax connector with calibration cable adapters.

Is the antenna tuned to your f or offset? That's why I suggested broader sweep with results on return loss S12.

Then isolate which components affect resonance from passive elements and tracks. Perhaps your permittivity constant is less than you estimated or probe lead calibration is not accurate. You can also solder onboard miniature Murata coax connector with/without switch.

**broken link removed**

Coaxial Connectors (Chip Type Receptacle)
MM
Microwave Coaxial Connectors
(Chip Type Receptacle)
HSC Type, 4000pcs. /Reel (ø178mm)

also
HSC Type
JSC Type
SWG Type
SWF Type
SWD Type
GSC Type

 

[rf1008]

thanks to all.

I really do the calibration well and make the port extention carefully,you can see the final curves after calibration,S11/S22 at OPEN point and S12/S21 at Center point.
I use the N1021B as differential probe and very hard to calibrated at the tips interface for SHORT/LOAD,

 

[D.A.(Tony)Stewart]

Then you need to analyze your design for sensitivity to calculations and determine what caused the error. List all the possible variables and do a worst case tolerance analysis and simulate.

 

[rf1008]

Thanks to all!

I think i have fond the reason. I use attached coaxial pigtails as differential probe to replace N1021B and calibrated at two tips interfance so that NO port extention is needed, and the final calibrated curves are correct.