开云体育

A few years ago I made an open and a short female SMA standard for calibrating my HP 8753C VNA (for the load I used a very nice 85052 female 3.5mm load). I never used the open and short above about 30 MHz, so I didn't worry too much about their accuracy.

With my purchase of a NanoVNA I thought I could put them to use in lieu of the male standards the NanoVNA comes with. And so I thought I'd check their accuracy, using my 8753C and the Rohde & Schwarz T-check method.

Of course, I had not characterized my standards, and so I thought, as a first attempt at checking them, I would use the "stock" 3.5mm cal-kit definitions that are stored in the 8753C.

The key parameters of the "stock" 3.5mm definitions (to match HP's 85033C kit) are:

Open: C0 = 53, C1 = 150, C2 = 0, C3 = 0, and Offset Delay = 14.491 ps
Short: Offset Delay = 16.695 ps
Thru: Offset Delay = 0 ps

Much to my surprise, the results were very good, as you can see in the attached T-check plot.

But, I wondered, do the "stock" HP delays bear any resemblance to the actual delays of my SOLT standards?

Assuming HP's Offset Delays were not the same as my SOLT Offset Delays, why not change HP's "stock" Offset Delays (as defined above) to move the reference plane to the end of my standards (i.e. at the location of the actual open and short).

So, as an experiment, I decided to subtract 14.491 ps from HP's "stored" 3.5mm definitions. In other words,

o The Open Offset Delay changed from 14.491 ps to 0 ps.
o The Short Offset Delay changed from 16.695 ps to 2.204 ps
o The Thru Offset Delay changed from 0 ps to -14.491 ps

The results weren't quite as good as I expected (see attached T-check PNG).

Maybe, I thought, the Thru's delay change needed to be positive, not negative, so I changed its Offset Delay from -14.491ps to +14.491ps. Even worse results!

So I tried doubling the Thru's Offset Delay. That is, rather than making it -14.491 ps, make it -30 ps. The results now look pretty good.

But why does doubling the Thru delay give these results? And so my questions are:

1. Why did doubling the Open/Short Offset Delay delta of -14.491 ps to be the Thru's Offset Delay give the results it did?

2. Why are the Thru Offset Delays in HP's Cal Kits all spec'd to be 0 ps? (Clearly the length of the thru's are not zero, so what is HP referencing to determine that an Offset Delay is 0?)

Thanks for any insight provided!

- Jeff, k6jca

Mechanical construction of standard have a relationship to the plane of calibration.
Basically you can measure that as distance or time they use time as its easy to measure.

What does it mean is that you short has a longer transmission line to the shorted piece
and as a result looks longer, same for the open. For 0-30mhz its likely not an issue
as the distance is very small at 300mhz it gets more so and higher has greater impacts.

Most people make a bigger deal of the 50 ohm where the short and open are most critical.
For special cases I have short, open, and loads that are flange mount SMA as I want the
plane of calibration to be the back side of the connector. The purpose would be when I
measure strip line circuit on PCB and the board and the circuit is under measurement
not the already known connector..

FYI most calibrated opens have a shell and removable pIn. Least my sets do (all HP
for N, BNC, and SMA).

-----------------
I do not accept private email due to forum scraping groups.io

On Sun, 15 Dec 2019 at 15:19, Jeff Anderson <jca1955@...> wrote:

A few years ago I made an open and a short female SMA standard for
calibrating my HP 8753C VNA (for the load I used a very nice 85052 female
3.5mm load). I never used the open and short above about 30 MHz, so I
didn't worry too much about their accuracy.

Fair enough.

With my purchase of a NanoVNA I thought I could put them to use in lieu of
the male standards the NanoVNA comes with. And so I thought I'd check
their accuracy, using my 8753C and the Rohde & Schwarz T-check method.

I'm a bit sceptical of the T-check method. R&S don't sell the device to do
this any any more, but they used to. They only sell verification kits based
on airlines and attenuators. Ken Wong at Keysight looked at it, and was not
impressed.

There's not the proper error correction in a nanoVNA to perform a T-check
anyway.

Of course, I had not characterized my standards, and so I thought, as a
first attempt at checking them, I would use the "stock" 3.5mm cal-kit
definitions that are stored in the 8753C.

The key parameters of the "stock" 3.5mm definitions (to match HP's 85033C
kit) are:

Open: C0 = 53, C1 = 150, C2 = 0, C3 = 0, and Offset Delay = 14.491 ps
Short: Offset Delay = 16.695 ps
Thru: Offset Delay = 0 ps

Depending on how you construct the standards, you are likely to get delays
longer than those. They are very short.

If you make your own standards, the parameters of the males and female will
almost certainly be different. I would suggest you modify the 85032B N cal
kit, as that has different parameters for the opens and shorts, which you
will have.

2. Why are the Thru Offset Delays in HP's Cal Kits all spec'd to be 0
ps? (Clearly the length of the thru's are not zero, so what is HP
referencing to determine that an Offset Delay is 0?)

Because they are based on a *"flush thru".* In other words, one standard is
male, and the other standard is female, so they connect together with zero
delay. HP can't guess the delay of your thru. For a "typical" female-female
thru the delay will be 41 ps, which should be entered into the calibration
standard #4 as 41 ps.

Thanks for any insight provided!

- Jeff, k6jca

Dave, G8WRB

Hi Jeff

It is not a mystery why it work ?

You open offset delay is not 14.491ps but 17.15ps with the C0 and C1 included

Then the difference between the short and open is 17.15-16.695=0.45ps so the open is a bit longer than short by open air displacement of 0.45/.3=1.5mm You homemade female open has a fringe c so also a bit longer than the homemade short and 0.45ps correspond to 9fF as 1 ps=20fF. The fringe C for your open can be simulated by the program FEMM and will be about 25fF (I just simulated a female SMA adaptor) so in reality you home made open has a delay 1.25ps longer than the short. Now comes the fun part. If the electrical length of you homemade adaptor is from calibration plane to rear of adaptor was 16.695ps then you had as HP85033C male clone with the exception that the open was 1.25-0.45=0.8ps longer. Find the electrical length of the homemade adaptor as the length from calibration plane (2 mm recessed from the front) to the read of adaptor and divided by 0.3 and divide once more with 0.695 being the VF of Teflon. The you can figure out how much to compensate as X - 16.695ps being positive or negative

The reason for Thru set to 0 is the VNA probably has adaptor removal included

Kind regards

Kurt



-----Oprindelig meddelelse-----
Fra: [email protected] <[email protected]> P? vegne af Jeff Anderson
Sendt: 15. december 2019 16:19
Til: [email protected]
Emne: [nanovna-users] Homebrew Female SMA standards and T-Check



A few years ago I made an open and a short female SMA standard for calibrating my HP 8753C VNA (for the load I used a very nice 85052 female 3.5mm load). I never used the open and short above about 30 MHz, so I didn't worry too much about their accuracy.



With my purchase of a NanoVNA I thought I could put them to use in lieu of the male standards the NanoVNA comes with. And so I thought I'd check their accuracy, using my 8753C and the Rohde & Schwarz T-check method.



Of course, I had not characterized my standards, and so I thought, as a first attempt at checking them, I would use the "stock" 3.5mm cal-kit definitions that are stored in the 8753C.



The key parameters of the "stock" 3.5mm definitions (to match HP's 85033C kit) are:



Open: C0 = 53, C1 = 150, C2 = 0, C3 = 0, and Offset Delay = 14.491 ps

Short: Offset Delay = 16.695 ps

Thru: Offset Delay = 0 ps



Much to my surprise, the results were very good, as you can see in the attached T-check plot.



But, I wondered, do the "stock" HP delays bear any resemblance to the actual delays of my SOLT standards?



Assuming HP's Offset Delays were not the same as my SOLT Offset Delays, why not change HP's "stock" Offset Delays (as defined above) to move the reference plane to the end of my standards (i.e. at the location of the actual open and short).



So, as an experiment, I decided to subtract 14.491 ps from HP's "stored" 3.5mm definitions. In other words,



o The Open Offset Delay changed from 14.491 ps to 0 ps.

o The Short Offset Delay changed from 16.695 ps to 2.204 ps

o The Thru Offset Delay changed from 0 ps to -14.491 ps



The results weren't quite as good as I expected (see attached T-check PNG).



Maybe, I thought, the Thru's delay change needed to be positive, not negative, so I changed its Offset Delay from -14.491ps to +14.491ps. Even worse results!



So I tried doubling the Thru's Offset Delay. That is, rather than making it -14.491 ps, make it -30 ps. The results now look pretty good.



But why does doubling the Thru delay give these results? And so my questions are:



1. Why did doubling the Open/Short Offset Delay delta of -14.491 ps to be the Thru's Offset Delay give the results it did?



2. Why are the Thru Offset Delays in HP's Cal Kits all spec'd to be 0 ps? (Clearly the length of the thru's are not zero, so what is HP referencing to determine that an Offset Delay is 0?)



Thanks for any insight provided!



- Jeff, k6jca

Hi Jef
One more comment. The load being 50 ohm it is not sensitive to a small displacement of the calibration plane.
You will soon see a perfect T-Check
Kind regards
Kurt

-----Oprindelig meddelelse-----
Fra: [email protected] <[email protected]> P? vegne af Jeff Anderson
Sendt: 15. december 2019 16:19
Til: [email protected]
Emne: [nanovna-users] Homebrew Female SMA standards and T-Check

A few years ago I made an open and a short female SMA standard for calibrating my HP 8753C VNA (for the load I used a very nice 85052 female 3.5mm load). I never used the open and short above about 30 MHz, so I didn't worry too much about their accuracy.

With my purchase of a NanoVNA I thought I could put them to use in lieu of the male standards the NanoVNA comes with. And so I thought I'd check their accuracy, using my 8753C and the Rohde & Schwarz T-check method.

Of course, I had not characterized my standards, and so I thought, as a first attempt at checking them, I would use the "stock" 3.5mm cal-kit definitions that are stored in the 8753C.

The key parameters of the "stock" 3.5mm definitions (to match HP's 85033C kit) are:

Open: C0 = 53, C1 = 150, C2 = 0, C3 = 0, and Offset Delay = 14.491 ps
Short: Offset Delay = 16.695 ps
Thru: Offset Delay = 0 ps

Much to my surprise, the results were very good, as you can see in the attached T-check plot.

But, I wondered, do the "stock" HP delays bear any resemblance to the actual delays of my SOLT standards?

Assuming HP's Offset Delays were not the same as my SOLT Offset Delays, why not change HP's "stock" Offset Delays (as defined above) to move the reference plane to the end of my standards (i.e. at the location of the actual open and short).

So, as an experiment, I decided to subtract 14.491 ps from HP's "stored" 3.5mm definitions. In other words,

o The Open Offset Delay changed from 14.491 ps to 0 ps.
o The Short Offset Delay changed from 16.695 ps to 2.204 ps
o The Thru Offset Delay changed from 0 ps to -14.491 ps

The results weren't quite as good as I expected (see attached T-check PNG).

Maybe, I thought, the Thru's delay change needed to be positive, not negative, so I changed its Offset Delay from -14.491ps to +14.491ps. Even worse results!

So I tried doubling the Thru's Offset Delay. That is, rather than making it -14.491 ps, make it -30 ps. The results now look pretty good.

But why does doubling the Thru delay give these results? And so my questions are:

1. Why did doubling the Open/Short Offset Delay delta of -14.491 ps to be the Thru's Offset Delay give the results it did?

2. Why are the Thru Offset Delays in HP's Cal Kits all spec'd to be 0 ps? (Clearly the length of the thru's are not zero, so what is HP referencing to determine that an Offset Delay is 0?)

Thanks for any insight provided!

- Jeff, k6jca

One last, many of the HP loads are air, the most often home made open and
short are PTFE loaded, there is a VF difference.

The HP load is definately air (minimal dielectric) to the resistance.
The open and short are for my set air.

I have them to calibrate the HP4191A I have and work well for the nanovna.

Allison
--
-----------------
I do not accept private email due to forum scraping groups.io

Hello Jeff,

You have made an interesting T-Check measurement in post 8236, date 2015-12-15 16:19.

Is your MATLAB script "Cal_T_Check.m" also usable with Octave under Linux?

73, Rudi DL5FA

Allison, Dave, and Kurt, many thanks for your replies.

Dave -- thank you very much for the "flush thru" description. That explains a lot, and it was a key bit of info that I lacked.

Kurt -- I'll attempt some accurate characterization when I return home (I am "on the road" at the moment). I found some APC-7 terminated airline in the back of a drawer a few days ago, so that should help me characterize C0. Thanks very much for your advice.

Allison -- I am intrigued as to how you are using your 4191A. I have one, too, but I've been using it with the stock fixture. Am I correct in assuming your are adapting its APC-7 to an SMA cable, and then making your measurements at the end of that cable? And where are you defining your reference plane?

Again, thanks to all.

- Jeff, k6jca

Hi Rudi,

The MATLAB T-check script is part of a package of scripts written by Dick Benson, W1QG, and downloadable from the Matlab Central:



Please note that it requires MATLAB's "RF Toolbox", so I do not know if it is compatible with Octave. Give it a try, maybe it will work.

And note that the T-check script works with an S2P file.

Best of luck!

- Jeff, k6jca

Hi Allison
Sure they will work fien with thee NanoVNA as the difference in offset delay is very small as the NanoVNA is designed for ideal calibration kit except the open is compensated by 50fF. But you are not calibrating to the calibration plane of the Female adaptor of the NanoVNA or the Male adaptor at the end of the test cable if using a such. However the Display/Scale/Electrical delay in the NanoVNA can fix it if you enter the offset delay of the male versus female short.
Kind regards
Kurt
-----Oprindelig meddelelse-----
Fra: [email protected] <[email protected]> P? vegne af aparent1/kb1gmx
Sendt: 16. december 2019 02:29
Til: [email protected]
Emne: Re: [nanovna-users] Homebrew Female SMA standards and T-Check

One last, many of the HP loads are air, the most often home made open and short are PTFE loaded, there is a VF difference.

The HP load is definately air (minimal dielectric) to the resistance.
The open and short are for my set air.

I have them to calibrate the HP4191A I have and work well for the nanovna.

Allison
--
-----------------
I do not accept private email due to forum scraping groups.io