THE IMPORTANCE OF USING CALIBRATION STANDARD DEFINITIONS FOR A VECTOR NETWORK ANALYSER page 5

The characteristics of the BNC(m) connectors (see Table 2) were entered into the 3.5 mm cal kit  definition table of the VNA and a one port calibration performed. The test aids were then used to find the calibration error of the BNC(m) OPEN connectors. Figure 13 shows the resulting OFFSET LOSS errors for OPEN standards and it can be seen that the BNC(m) connector is usable up to 317 MHz  for OFFSET LOSS within the previously set limit of 0.033 dB.

The OFFSET LOSS measurements for the BNC connectors were originally performed using the N 50Ω kit definitions in the analyser but this caused OFFSET LOSS errors. When the USER definitions were correctly entered into the definition table  the OFFSET LOSS errors for an OPEN standards were reduced but the OFFSET LOSS errors for a SHORT standard were greatly increased. This was suspected to be due to the mathematical model in the VNA using a simplified method for calculating the phase corrections for a N SHORT which would normally have an OFFSET DELAY value of zero. The 3.5 mm cal kit was then selected for all measurements because the 3.5 mm SHORT has an electrical length which has to be taken into account and this resulted in the VNA performing the calibration differently and to give a better over all result even though the identical cal kit definition values were entered. Although it was not possible to ascertain exactly what was happening in the VNA processing, it was a demonstration that the mathematical model in the VNA can make certain assumptions in relation to the cal kit selected and using a kit with different characteristics can cause errors even when the USER DEFINITION is correctly entered.

The process was repeated for the BNC(f) OPEN connectors and then both the SMA OPEN connectors (all using the 3.5mm kit with USER definitions in the VNA) and the results are shown in Figure 13 and Table 3.

 

[Graph showing measured OFFSET LOSS of OPENS using USER definition]

The OFFSET LOSS measurements were repeated for the SHORT standards. It can be seen in Figure 14 that the OFFSET LOSS errors have slightly increased from those in Figure 9. This could be due to connector repeatability or because there was no selection of Inductance coefficients  The process of measuring the residual error was repeated for all standards and the usable upper limits entered into Table 3.

[Graph showing measured OFFSET LOSS of SHORTS using USER definition]

Table 3 Offsets with USER definitions.

Type Sex Upper OFFSET LOSS Limit of 0.033dB (MHz) Upper PHASE ERROR Limit of 0.33 deg (MHz) DELAY ERROR at 0.3 MHz (ps)
BNC SHORT Male 176 1,920 -2.194
OPEN Male 317 >2,000 -0.181
SHORT Female 1380 >2,000 -2.756
OPEN Female 1770 >2,000 0.219
SMA SHORT Male 318 >3,000 -1.83
OPEN Male 2,860 >3,000 -.045
SHORT Female 900 >3,000 -0.954
OPEN Female 2,750 >3,000 0.045

CONCLUSIONS

    The DELAY ERROR @ 0.3 MHz of SHORT standards with a user definition was always greater than that of the corresponding OPEN standard showing that the definition for the SHORT standards would have benefited from the inclusion of INDUCTANCE coefficients. These coefficients would have also improved the OFFSET LOSS and PHASE ERROR performance.

 The large reactance values of the connector standards tended to "pull" the analyser and produced large differences between the two measurement methods. This shows that the Agilent policy of reducing the calibration standards to the absolute minimum physical size is correct. There was also error produced by the inclusion of adaptors and it was not possible to remove this error with the methods used. The absolute minimum number of  high quality adaptors are required when characterising a calibration standard.

 The calculations and measurements made have shown that entering the standard characteristics into the VNA definition table improves the phase accuracy of the BNC 'connector standards' dramatically and in general improves the maximum offset loss of OPENS. The large errors occurring when the definitions are not used are not visible to the engineer after calibration and it is very easy to be fooled into thinking that we are making an accurate measurement when we are not. It should be remembered that the process of characterising a calibration kit from first principals is time consuming, tedious and requires a high skill level. Even when the correct USER definition is entered into VNA the mathematical model used by the analyser may not be suitable for the calibration standards in use. Unless you are working well below the upper limits shown in the tables above it is far preferable to purchase the correct kit for the VNA being used. It will be cheaper in the long run, you will also have traceability and confidence in your measurements.

W J Highton 6/2/2011

Technical Article Index    < 4  5    Page Number