On Wed, 30 Oct 2019 at 19:16, <erik@...> wrote:
After calibration with the correct calibration error parameters, a 1
meter cable connected to the reference plane with short and open at its
other end, should give perfect circles spiraling inwards in the smith chart.
This however was NOT true for my nanoVNA
So I added a small error model for the bridge with only a constant offset
for the reference R, both real and imag)
The bridge will have some imperfections, which will be frequency dependent,
so I don¡¯t think it is sensible to try to attach an impedance to the bridge.
The coaxial cable will have loss which will rise with frequency. The fact
that the coax has loss means that it must have a complex impedance. A lossy
coax can¡¯t have an impedance which is purely real.
I see a couple of issues
1) Your coax is going to be rather lossy.
2) The impedance of typical cables have a specification of +/- 2 ohms.
Only a OpenC0 of 52fF is needed to correct the calibration set to this
level. The bridge correction is 0.29 - 0.6i ohm.
Any suggestions?
Yes - I would give up with your length of RG58. A few improvements
1) Use a 3.5 mm to APC7 adapter with a proper APC7 airline. I bought a 20
cm APC7 airline for about $50 off of ebay. Not cheap, but probably the
best solution. A 3.5 mm airline would be better still, but I am unaware of
any long ones being available. I have a few, but they are too short for a
900 MHz instrument.
2) Use 1 m of RG401 semi rigid. That has a very low loss, so will have an
almost a real impedance, but it could be in the range 48-52 ohms.
3) Buy an SMA T-piece. That¡¯s a 3 port network. Terminate one port is some
reasonable impedance - any old nominal 50 ohm load will do, but in theory
it can be any impedance. Then use the T-checker method:
This method assumes that the T-piece is lossless, which I think is a better
approximation than assuming a bit of RG58 is lossless. However, that
doesn¡¯t mean that x dB of loss messes things up the same in each method.
*Is your intention to add proper support for calibration kits in the
NanoVNA FIRMWARE?* I would really like to see that. I would like to use N &
7-16 calibration kits, but those are impossible to do well without taking
into account their characteristics, as for example the delay on a male N
short or open can¡¯t be made anywhere near zero.
*If that¡¯s your intention,* drop me a private email and I will send you an
SMA calibration kit
free of charge *if you want one.* (My company produces VNA calibration
kits). The S-parameters of each standard is individually measured using an
HP VNA and HP calibration kit.
A verification attenuator is also supplied,
which means that you can actually measure something with known properties.
The calibration kit used to measure the attenuator is a 26.5 GHz 3.5 mm HP
85052B - *not* the calibration standards supplied. That essentially means
detects 99% of all possible errors.
* A damaged calibration standard
* incorrect entry of calibration data
* Errors m in the VNAs calibration algorithm
* Most hardware faults in the VNA
will all be detected.
Small discrepancies between your measurements of the attenuator and our
measurements of attenuator are to be expected. It will however be hard to
know if that¡¯s due to minor problems in the calibration algorithms, or
hardware limitations of a $50 VNA.
Dave
--
Dr. David Kirkby,
Kirkby Microwave Ltd,
drkirkby@...
Telephone 01621-680100./ +44 1621 680100
Registered in England & Wales, company number 08914892.
Registered office:
Stokes Hall Lodge, Burnham Rd, Althorne, Chelmsford, Essex, CM3 6DT, United
Kingdom
