Calibration error in PNA

Hi, I am in process of measuring S11 of a network. However my accuracy requirement in measurement is 10^(-4) in linear scale. I am using Agilent's PNA in frequency range of 10-250 MHz. Due to use of 10m co-axial cable, I am getting ripples in the S11 at the level of 10^(-3) which I attribute to calibration error. I have put on averaging for 1000 traces to achieve this.

Is there a way I can reduce the calibration error by an order of magnitude? Any references into insight of calibration process of PNA will also be appreciated.
PS: I am calibrating in ECAL mode.

How do you calibrate the measuerement setup? I presume ECAL module connected behind the 10 m cable.

How do you see the 1e-3 S11 ripple? After calibration with the same ECAL load? If yes, it may be that the expected accuracy is beyond PNA specification. If the deviation is stable, you may get off by performing a reference measurement after calibration and performing a correction of measured S11 with an external tool.

FvM said:

How do you calibrate the measurement setup? I presume ECAL module connected behind the 10 m cable.

How do you see the 1e-3 S11 ripple? After calibration with the same ECAL load? If yes, it may be that the expected accuracy is beyond PNA specification. If the deviation is stable, you may get off by performing a reference measurement after calibration and performing a correction of measured S11 with an external tool.

Click to expand...

Yes, I calibrate the setup with the 10 m cable and ECAL module is placed after this. So for return loss, I have some expected functional form. From the measured data, I fit for this form and subtract it. In principle, I should be left with measurement noise. However, the residuals show ripples which corresponds to cable length. I have attached a sample of residuals that I get.



For sure, it is not random measurement noise; rather it is inaccurate calibration which has not removed the effect of cable at this level. The deviation is not stable enough to use a reference subtraction. However, I will give it a try and see if it removes significant portion of the error.

I am interested in knowing the actual calibration process of PNA using ECAL. That might give us hint of ways to reduce calibration error further. Out requirement is to bring it down by a factor of 10.

Thanks for the reply.

Global_Hawk said:

However my accuracy requirement in measurement is 10^(-4) in linear scale.

Click to expand...

-80dB accuracy on return loss?

It is easy to ask for this.
It is very, very, very difficult to reach this level of accuracy.

The impedance (load) accuracy of ECal modules is specified as worse -40dB = 0.01 in linear scale.

Global_Hawk said:

Due to use of 10m co-axial cable, I am getting ripples in the S11 at the level of 10^(-3) which I attribute to calibration error.

Click to expand...

If you require only good repeatability (not absolute accuracy), you still need cables where the phase and amplitude of reflection and transmission is very stable (unchanged) when you move the cable. This stable behaviour is a requirement for good calibration results also. These cables exist, but they are REALLY expensive and usually kept to the minimum length.

https://www.gore.com/en_xx/products/cables/microwave/test/gore_vna_microwave_test_assemblies.html

I think your biggest problem is that your cables not stable/repeatable, so that phase/amplitude changes when you move the cables. And your expectation what accuracy and repeatability you can expect from Ecal (simple to use) is wrong.