The uncertainty in a receiver calibration is important since it forms a bound on possible uncertainty in many of the measurements listed in the introduction. Since there are many possible setups, we cannot provide a blanket result but can show how this uncertainty can be computed.
Source Side
• Power calibration accuracy (assuming it is used): set by the power meter/sensor and mismatch between the power sensor and the source reference plane.
Receiver Side
• Mismatch between source reference plane and receiver reference plane
• Linearity of the receiver
• Trace noise of the receiver
Receiver linearity is typically better than 0.05 dB as long as one is a few dB away from the compression limits. See the VectorStar VNA data sheet for more information:
• VectorStar MS464xB Series VNA Technical Data Sheet (TDS)—11410-00611
Trace noise is typically less than 0.01 dB as long as one is above –10 dBm port equivalent power. The mismatch contribution depends on where the reference planes are. If both return losses are on the order of 20 dB, there could be a contribution of 0.08 dB (best case scenario). The source side uncertainty can be, at best, 0.1 dB but is commonly several tenths.
Typically, these components are combined in an RMS sense and may be as low as 0.5 dB composite. As suggested above, the actual value will be a strong function of the setup being used. The power meter and VNA uncertainty application notes should be consulted for additional information.
The receiver calibration is automatically interpolated and extrapolated (always and independent of the ‘interpolation’ switch available on the CALIBRATION menu). Generally, the receiver response is a smoothly varying function of frequency so this (linear in magnitude) interpolation does usually not introduce substantial additional uncertainty. If the initial frequency list at the time of receiver calibration is, however, sparse relative to the electrical length of the setup, the errors will increase. Above a few GHz with a simple 1 m cable setup, a transmission receiver response may vary only by 1 dB over a few GHz, so a sparse density will suffice. With a mismatched fixture and multi-meter long cable runs, the step size may need to be in the 10s to 100s of MHz if substantial interpolation is needed and receiver calibration uncertainty needs to remain well under 1 dB.
Another interpolation exception occurs at low receiver frequencies (< 50 MHz) where the system switches to alternate IF frequencies for spur avoidance in certain sub-bands. Interpolating across these edges will result in distortion. In general, major bandswitches (2.5, 5, 10, 20, 38, and 40 GHz for base VNAs) can result in similar issues.