Hi everyone,
My question is: do VNA error terms typically behave in a nature that would allow for RSS as a reasonable method of combining terms as opposed to worst case addition. From the GUM procedures I think they usually specify a dependency or sensitivity coefficient? We're trying to determine if RSS is acceptable with the VNA.
Andrew
sorry for such a late reply but yes, in general we also take an RSS approach. Single errors such are source match and load match are essentially uncorrelated, as is noise, repeatabilty, etc. There are some correlated errors, but it is extremely complicated to compute their effects (correlation can reduce or increase the error relative RSS). We actually make a real-time uncertainty analysis option for the PNA (option 015) that implements the full correlation effects. It is really complicated and the documentation for it is more than 140 pages of dense Linear Algebra of Complex Variables, and so almost no one can repeat it in practice (we had 4 guys working for 2 years or so on it). So, almost everyone uses the RSS approach.