On Tue, Aug 18, 2020 at 08:20 AM, <biastee@...> wrote:
The key benefit is the sex at a
reference plane can be changed AFTER calibration - e.g. by switching
APC7-SMA(f) adapter to APC7-SMA(m) adapter. This technique known as "swap
equal adapters" enables measurement of DUT with same sex connectors (e.g. a
DUT with two N females) which is known as a "non-insertable device" in VNA
parlance.
Ah, I see. I was thinking about mentioning swap-equal adapters, but didn't because they are mostly obsolete and expensive. They are not exclusive to APC7 though. The nicer 3.5mm calkits e.g. tend to come with a matched set of f-f, m-f and m-m adapters. Even used these adapters still go for a pretty penny though. I suppose that if you go with APC7 you might have better chances of getting a bargain...
I guess there is no motivation for Keysight to continue supporting "swap
equal adapters" because e-Cal and "adapter removal" are now standard on all
their offerings, unlike the 90s till early 2000s, when "adapter removal" was
only available on the higher end models like 8510x but NOT on the plain
vanilla 8753x.
Essentially, yes. I didn't realize the older 8753 didn't support adapter removal. I think 8753D/E do though. In any case, adapter removal calibration is cumbersome and the newer analyzers probably support unknown-thru, which is both more convenient and more accurate (see below).
As to "adapter removal", the big impediment is that it is not supported by
firmware nanoVna, right? Additionally, the user will need to know the
adapter's electrical length at each frequency and modify the cal-kit
definition.
Well, yes and no. If you know the electrical length (and potentially loss), you wouldn't need adapter removal calibration, you can just modify calkit parameters accordingly ("defined through"). This could just be implemented in firmware. Right now, as far as I am aware, there is no support for calkit parameters whatsoever, so if you use the internal calibration (rather than some PC software), I would be more worried about the open and short definitions.
For proper adapter removal, you don't need to know the exact electrical length (it is enough to know it within a quarter-wavelength), it is determined during calibration. Unfortunately, like some of the other more advanced options, it needs a full 2-port analyzer.
In general there are three differnent types of VNAs:
Four-receivers, full 2-port (=source can be switched between ports)
These can do unknown-thru (or reciprocal-thru, SOLR, UOSM, depending on where you buy your VNA). For this you measure S/O/L on each port as usual and then connect an "unknown" thru adapter between the ports. The analyzer can calculate the full S-Parameters of the adapter from this and remove any effect of the adapter. You may even be able to use the DUT itself as the unknown thru, no need to even reconnect the cables. This is very accurate and very convenient (it might even be a better option than flush thru for some insertable DUTs!).
Classical architecture (like the 8753)
As I mentioned, these cannot do the internal measurements needed for unknown-thru. It is still possible with an additional calibration step (two-tier calibration), but it won't be as accurate and you will probably need to do it "offline".
Otherwise you need to do adapter removal calibration. You do a normal 2-port SOLT calibration, but _twice_. Once with adapter connected to port 1 and once with the adapter connected to port 2. The VNA compares both calibrations and can remove the effects of the adapter. But you need calibration kits for both sides of the adapter. If you have a DUT with e.g. female SMA on both sides, you need both male and female SMA kits. You also need to make a total of 8 calibration measurements.
Transmission/reflection "1.5 port" with source fixed to port 1 (like 8714, NanoVNA)
- 1-port adapter removal: you do a one-port calibration at the test-port, then put on the adapter and then do another one-port calibration with the adapter. From the difference between these calibrations you can characterize the adapter and then later de-embed it (with some reasonable assumptions). I don't think this is commonly implemented on commercial analyzers, but it is just a matter of software. You need calkits for both sides (so e.g. male and female).
- defined thru: If you the length of your thru (specified by the manufacturer, measured using one of the other methods once or estimated from the mechanical dimensions), you're good and need only one calkit. If you don't care about phase accuracy you can also just set the length to 0. On a full 2-port this may degrade error correction significantly, on a T/R VNA it doesn't really matter.
- swap-equal adapters are indeed still a practical option if you have them (and don't need high accuracy), no software needed, only need one calkit
- ecal is, of course, accurate and convenient (if you have it and it is supported...)
