


Uncertainty Analysis
of an RF Test
Platform (page 5) 

←
The E4413A Sensor Calibration Expanded Uncertainty analysis assumes a 1.2:1 VSWR into the PNAX output cable. This value represents the uncertainty of the power measured during calibration at the PNAX cable to RFIU interface calibration plane.
PNA Gain measurement uncertainty is computed using the “PNAX Measurement Uncertainty” calculator, downloaded from the Agilent Website. Calibration was performed using the N4692 ECal module, and the effects of the thru were included.
The insertion gain uncertainty from +30 to 40 dB is on the order of ±0.33 dB, worst case. For the measurement of a fixture with 30dB transmission coefficient the data shows an uncertainty value of +/ 0.27 dB.
Among the primary drivers of repeatability in a typical RF test system are the electromechanical switches in the RFIU. To calculate the Root Sum Square (RSS) percent (%) error for switch repeatability across N switches with x dB repeatability the following formula may be used.
Finally, the expanded uncertainty is 2 times the standard uncertainty error (%). The 2
sigma uncertainty corresponds to a 95% confidence level.. In addition, if we assume the test equipment is held to 23 ± 3 deg C, then the PNAX amplitude stability is ±0.03 over this temperature range. The system being analyzed includes a power amplifier that may be switched into or out of the measurement path.
To calculate the output power measurement uncertainty we need to combine the power meter uncertainty, the uplink fixture S21 uncertainty, and the mismatch loss between the fixture and the UUT. The data in Table 4. assume a fixture output VSWR and UUT VSWR of 1.2:1 for the frequency ranges listed below.
Input power can be measured using the PNA, Spectrum Analyzer (SA) or the internal power meter. The following discussion data shows the power measurement uncertainty of the PNA and the SA.
Calculations for switches assume the switches are specified for 0.03 dB repeatability over their lifetime. In addition, if we assume the test equipment is held to 23 ± 3 deg C, then the PNAX amplitude stability is ±0.03 over this temperature range. The formulae used for the calculations below are the same as in B. 4).
The calculation of the PNA measurement uncertainty is similar to the calculations shown for output power uncertainty, however; in this case the calibration output power of the PNA Port 1 cable to RFIU calibration plane is connected at the UUT downlink power interface plane. To calculate the downlink power uncertainty we need to combine the power meter uncertainty, the downlink fixture S21 uncertainty, and the mismatch loss between the fixture and the UUT. The data
assume a fixture VSWR and UUT VSWR of 1.2:1 for the frequency ranges listed below.
The calibration output power of the PNA Port 1 cable to RFIU calibration plane is connected at the UUT downlink power interface plane and measured with the SA. As before, to calculate the downlink power uncertainty we need to combine the power meter uncertainty, the SA measurement uncertainty, and the mismatch uncertainty between the fixture and the
UUT.
Gain is measured using the PNA. The uncertainty will be the combination of the PNA S21 measurement uncertainty at a transmission coefficient of 0dB (assuming fixture loss and UUT gain cancel), uplink fixture S21 uncertainty, the downlink fixture S21 uncertainty, the input fixture to UUT mismatch uncertainty, and the UUT to output fixture mismatch uncertainty.
VIII. CONCLUSION
Upon first analysis it may seem that generalpurpose RF automatic test
equipment would cost more than specialpurpose systems. Solely in terms of rack hardware and instrumentation this may be true as the generalpurpose system must by definition be a superset of all the special purpose systems. As anyone that has designed and implemented an automated RF test system knows, however, racking the instruments and cabling everything together is but a tiny fraction of the effort of bringing the system online. The real costs are associated with documenting the system, characterizing the system, calibrating the system, creating the system selftest, writing the test executive software and writing the test algorithms. These costs can equal and in many cases exceed the cost of the hardware. When these costs are taken into account, the Common RF Test Platform may well become the lower cost solution.




