<azonenberg>
Blue curve is actual S-parameters of a probe
<azonenberg>
Red is Sonnet model of the actual probe geometry + resistors, with a somewhat guesstimated transmission line geometry for the tip socket
<azonenberg>
Cyan is S-parameter model of resistors string
<azonenberg>
if you switch from one model to the other around 2.25 GHz it's actually quite a close match
<azonenberg>
By no means perfect, but it's the closest i've got yet in simulation
<sorear>
how much do you know about the response of the VNA? feels weird to see traceable metrology without error bars
<azonenberg>
I have a full cal report showing response on a known line
<Nero_>
you call out the VNA response with a calkit
<Nero_>
cal*
<Nero_>
one of the problems with error in VNAs is that the uncertainty of the fixturing is going to be waaaay bigger than that of the VNA itself if you have any decent VNa
<azonenberg>
Nero_: thoughts on the sim?
<azonenberg>
i found that the impedance of the probe tip has a *huge* impact on performance past a few GHz
<azonenberg>
in particular, changing the stub length i modeled from 50 ohms to ~220, which is a closer match to the parallel cylinder geometry of the actual tips, gives way different -and closer to measured data - results
<azonenberg>
Unfortunately i'm not sure what i can do to improve that
bvernoux has joined #scopehal
<Nero_>
sorry, had to drive home
Nero_ is now known as NeroTHz
<azonenberg>
NeroTHz: So basically what i think is happening is that from about DC to 2.5 GHz performance is limited by reflections between the resistors in the attenuator string
<azonenberg>
which i think i can improve quite a bit
<azonenberg>
From there on out, it seems like the probe tip plays a greater role
NeroTHz has quit [Read error: Connection reset by peer]