Calibration coefficients - can one change them?


Dr. David Kirkby from Kirkby Microwave Ltd
7/02/19  

I've got nanoVNA on order, but don't have it yet. It's not due to at least
4th July, so that's no big surprise.

Can anyone tell me if one can change the coefficients of the
open/short/load kit? I just have done a test with 4 sorts of opens. The
connector was a male N, so the calibration standard a female N. But
anyway, here are the results at 902.625 MHz, which is the nearest frequency
I have any data. (I had calibrated to 18 GHz for another job, so don't have
data at exactly 900 MHz, but this is as close as makes no practical
difference)


*Male N plug just open. No female standard*. Phase = +4.64722 degrees @
902.62500
Female open standard from HP 85032B open standard, used without any
extender, as it is NOT supposed to be. Phase = +4.64014 degrees @ 902.62500
MHz
Female open standard from HP 85032B calibration kit, with the extender, as
it SHOULD be used -3.65161 degrees @ 902.62500 MHz
Female Open standard from HP 85054B 18 GHz calibration kit. Phase =
-18.18457 @ 902.62500 MHz

IMPORTANT
The male N plug was a metrology grade male-male adapter from an HP 85054B
VNA calibration kit. Since the shape of the male pin is not well defined,
and commercial grade connectors are recessed by various amounts, I would
not expect this result to be reproducible with different connectors, but
fairly substantial variations

Anyway, the conclusion is that for the female N calibration standards, the
phase varies from +4.6 to -18.2 degrees at 902 MHz. So a phase variation of
22.8 degrees depending on what calibration standard one uses. The variation
would be smaller at lower frequencies, but depending on what the nanoVNA
assumes about the calibration standard, one could get quite different
results.

Ideally, one needs to be able to enter the offset delay of the calibration
standards as an *absolute minimum.* The fringe capacitance would be nice
too, but could just about get away without that, as one could do a rough
job of compensation via changing the offset delay. Better still would be a
third order polynomial, which would make entering coefficients from a
commercial calibration kit easy, without trying to work out any
compensation values.

If the firmware is dumb enough to assume the open standard is ideal (phase
= 0 degrees), and the short is ideal (phase = 180 degrees), then it would
cause significant errors.

Likewise, if the firmware made assumptions about the SMA kit supplied,
those assumptions would be wrong if a different SMA, 3.5 mm, N or APC cali
kit was used.

Even if you are not interested in phase measurements, be aware of the fact
the point of the vector correction is to correct for amplitude variations
too. Anyone that believes that they don't need to worry about vector
correction, as they are only interested in amplitude measurements, is
seriously mistaken.

Dave

--
Dr David Kirkby Ph.D C.Eng MIET
Kirkby Microwave Ltd
Registered office: Stokes Hall Lodge, Burnham Rd, Althorne, CHELMSFORD,
Essex, CM3 6DT, United Kingdom.
Registered in England and Wales as company number 08914892
https://www.kirkbymicrowave.co.uk/
Tel 01621-680100 / +44 1621-680100



Dave, as far as I can tell to this point, there is no ability to enter your own polynomial coefficient of the standards. The unit comes with SOL 3.5 mm and they are described as fairly simple construction. Using independent calibration of a 3.5 mm cal kit on another VNA, I determined the quality of their standards. And based on that response I believe for HF applications the unit is quite acceptable.

Again, my earlier question are the standards they provide described within the firmware, don't know. My initial guess is they are taken as ideal, but I may be wrong. Alan

________________________________
From: nanovna-users@groups.io <nanovna-users@groups.io> on behalf of Dr. David Kirkby from Kirkby Microwave Ltd <drkirkby@...>
Sent: Tuesday, July 2, 2019 12:06 PM
To: nanovna-users
Subject: [nanovna-users] Calibration coefficients - can one change them?

I've got nanoVNA on order, but don't have it yet. It's not due to at least
4th July, so that's no big surprise.

Can anyone tell me if one can change the coefficients of the
open/short/load kit? I just have done a test with 4 sorts of opens. The
connector was a male N, so the calibration standard a female N. But
anyway, here are the results at 902.625 MHz, which is the nearest frequency
I have any data. (I had calibrated to 18 GHz for another job, so don't have
data at exactly 900 MHz, but this is as close as makes no practical
difference)


*Male N plug just open. No female standard*. Phase = +4.64722 degrees @
902.62500
Female open standard from HP 85032B open standard, used without any
extender, as it is NOT supposed to be. Phase = +4.64014 degrees @ 902.62500
MHz
Female open standard from HP 85032B calibration kit, with the extender, as
it SHOULD be used -3.65161 degrees @ 902.62500 MHz
Female Open standard from HP 85054B 18 GHz calibration kit. Phase =
-18.18457 @ 902.62500 MHz

IMPORTANT
The male N plug was a metrology grade male-male adapter from an HP 85054B
VNA calibration kit. Since the shape of the male pin is not well defined,
and commercial grade connectors are recessed by various amounts, I would
not expect this result to be reproducible with different connectors, but
fairly substantial variations

Anyway, the conclusion is that for the female N calibration standards, the
phase varies from +4.6 to -18.2 degrees at 902 MHz. So a phase variation of
22.8 degrees depending on what calibration standard one uses. The variation
would be smaller at lower frequencies, but depending on what the nanoVNA
assumes about the calibration standard, one could get quite different
results.

Ideally, one needs to be able to enter the offset delay of the calibration
standards as an *absolute minimum.* The fringe capacitance would be nice
too, but could just about get away without that, as one could do a rough
job of compensation via changing the offset delay. Better still would be a
third order polynomial, which would make entering coefficients from a
commercial calibration kit easy, without trying to work out any
compensation values.

If the firmware is dumb enough to assume the open standard is ideal (phase
= 0 degrees), and the short is ideal (phase = 180 degrees), then it would
cause significant errors.

Likewise, if the firmware made assumptions about the SMA kit supplied,
those assumptions would be wrong if a different SMA, 3.5 mm, N or APC cali
kit was used.

Even if you are not interested in phase measurements, be aware of the fact
the point of the vector correction is to correct for amplitude variations
too. Anyone that believes that they don't need to worry about vector
correction, as they are only interested in amplitude measurements, is
seriously mistaken.

Dave

--
Dr David Kirkby Ph.D C.Eng MIET
Kirkby Microwave Ltd
Registered office: Stokes Hall Lodge, Burnham Rd, Althorne, CHELMSFORD,
Essex, CM3 6DT, United Kingdom.
Registered in England and Wales as company number 08914892
https://www.kirkbymicrowave.co.uk/
Tel 01621-680100 / +44 1621-680100



Yes, calibration data can only be modified with electrical delay. For better accuracy, I customized the shortest and most accurate calibrations kits possible. Naturally, the accuracy of professional calibration parts cannot be achieved. Unfortunately, some clone makers do not understand the role of the calibration kit, providing a poor quality load to act as a calibration kit, which can result in worse measurement results.

hugen
gen111.taobao.com


Dr. David Kirkby from Kirkby Microwave Ltd
7/03/19  

On Wed, 3 Jul 2019 at 04:04, <hugen@...> wrote:

Yes, calibration data can only be modified with electrical delay.

What exactly do you mean by that? Are you saying that the firmware has the
ability to allow one to adjust the delay of the calibration standards? If
so that’s good news. Just being able to enter a delay for the standards
will be a vast improvement over having no control whatsoever.

But someone else wrote ideal standards were assumed.

For better accuracy, I customized the shortest and most accurate
calibrations kits possible.
There’s no need to make the shortest calibration standards possible, *IF*
one can enter the delay of the standards.

*The internet is full of people claiming incorrectly the need to have the
shortest possible delays on calibration standards. * If you look at the
delays on professional calibration kits, you will find they are *longer* on
modern high-end kits than they were on obsolete kits!

The obsolete HP/Agilent 85032B type-N calibration kit had delays of

http://na.support.keysight.com/pna/caldefs/85032BE.htm

Female Short = 0.000 ps
Female Open = 0.093 ps

Those are the shortest possible delays one can make on female N.

Now if you compare those figures to the current $20,690 high-end 18 GHz
85054B kit, you will find the delays are much longer in the modern kit.

http://na.support.keysight.com/pna/caldefs/85054B.htm

Female short = 27.990 ps
Female open = 22.905 ps

For the lower cost ($2410) 9 GHz 85033F 9 GHz kit, the delays on the female
parts are:

http://na.support.keysight.com/pna/caldefs/85032F.htm

Female Short = 45.995 ps
Female Open = 41.170 ps

*So very approximately, for female N calibration standards:*

Obsolete 6 GHz kit - delays around 0 ps
Current $2410 9 GHz kit - delays around 43 ps
Current $20,690 18 GHz kit - delays around 24 ps

Clearly on female N parts, it’s possible to make delays close to 0, but
Keysight don’t use them.

The same general principle is true for the male N standards, but in that
case it is impossible to make a zero delay. If you check the links I gave
above, you will find the following.

*Very approximately, for male N calibration standards:*

Obsolete 6 GHz kit - delays around 17 ps
Current 9 GHz kit - delays around 43 ps
Current 18 GHz kit - delays around 60 ps

For obsolete HP 3.5 mm kits, the delays were from memory were around 17 ps,
but any of the modern kits (85033E, 85052D or 85052B) all have delays of
about 30 ps.

Naturally, the accuracy of professional

calibration parts cannot be achieved.


That statement is a bit ambiguous.

If you are saying that a high end VNA is best used with a professional
calibration kit, I would agree with you.

However, if the NanoVNA firmware always assumes idealised parts with a
delay of zero, then things would change *dramatically*. One could easily
make more suitable opens and shorts than would would have if one spent
$2410 on an 85032F. Spending even more, buying the $20,690 85054B would
give you even less suitable open and short standards.

The loads from the $20,690 kit would always be the best loads, but of
course it would be crazy to use loads that probably cost $1500 each on a
VNA costing less than $100.

Unfortunately, some clone makers do not understand the role of the
calibration kit, providing a poor quality load to act as a calibration kit,
which can result in worse measurement results.

Yes, clone makers often don’t appreciate or care what crap they turn out.
But a lot of self-proclaimed experts, writing web pages about how to make
calibration kits don’t understand what they are doing.

My own company, Kirkby Microwave, does care and we have enough knowledge to
understand most of the intricacies. I cringe at some of the stuff I see
written on the internet about VNA calibration kits.


hugen
gen111.taobao.com

Dr. David Kirkby
Kirkby Microwave Ltd.
https://www.kirkbymicrowave.co.uk/


<http://gen111.taobao.com>

--
Dr. David Kirkby,



Hello David,

I mentioned in a much earlier post, I have no idea what is in the firmware definition for the standards supplied with the nanoVNA. In addition, based on the physical description of the standards provided, i.e. the 50 ohm load a pair of 100 ohm chip resistors soldered in parallel, that they may not represent the state of the art! However, if there were a way to measure and characterize them as a entry to the unit, such as a classification table, that would be helpful,

I did measure their 50 ohm load after an independent 3.5 mm SOLT cal on a hp 8753 VNA and found their load standard at 900 MHz was at best 17 dB return loss. Another measurement of a second independent 50 ohm load from Weinschel on the 8753 at 900 MHz demonstrated a significantly better return loss. However, on the nanoVNA the Weinschel showed the same 17 dB return loss.

Is there an issue here? YES!

I might add, folks who use this instrument should consider obtaining SMA connector savers. It will not be long before the poor SMA connectors on the nanoVNA are tainted.

Alan



I should add, that after a calibration with their standards, the resulting short, open and load are PERFECTLY located on the chart. Really! Wow, I have to pray that a set of APC-7 or my sophisticated cal standards could do so well. That said, this is a neat box and applying a little care to the measurement and compensation for the short comings can rectify some of these issues.


Dr. David Kirkby from Kirkby Microwave Ltd
7/03/19  

On Wed, 3 Jul 2019 at 13:19, alan victor <avictor73@...> wrote:

I should add, that after a calibration with their standards, the resulting
short, open and load are PERFECTLY located on the chart. Really! Wow, I
have to pray that a set of APC-7 or my sophisticated cal standards could do
so well. That said, this is a neat box and applying a little care to the
measurement and compensation for the short comings can rectify some of
these issues.
I'm afraid to say you are stepping into the error many people make. They
first calibrate with a set of standards, then see those standards look
perfect on the Smith Chart. I have hayfever at the moment, but made my
first YouTube video yesterday to show the problem.

https://youtu.be/9HEchzY0Gvw

Please ignore the sniffles I have and the fact I only used a mobile phone.

https://www.kirkbymicrowave.co.uk/
Tel 01621-680100 / +44 1621-680100



I am quite familiar with the error. Please read my post carefully. I point this measurement observation out as this is a STRONG HINT that something is NOT correct!

In any case, I measured their "standards" independent and I found the 50 ohm load at 58.5 + j 11.5 ohm at 900 MHz.

The short and open are 412 pH and 190 fF respective.

I suspect no polynomial L or C coefficients are loaded to a table in the nanoVNA to reflect these deviations.



I guess to make the point clear, 3.5 mm SMA standards DO NOT APPEAR as IDEAL SHORT, OPEN on the chart. The fact that they DO on the nanoVNA SAYS IN ITSELF... There is an issue!


Dr. David Kirkby from Kirkby Microwave Ltd
7/03/19  

Okay, I get what you are saying now.

That problem should I guess be resolvable in the firmware, although it will
take a very careful look at the firmware to see how to resolve it. But the
problem should be resolvable.

Dave

On Wed, 3 Jul 2019 at 14:31, alan victor <avictor73@...> wrote:

I guess to make the point clear, 3.5 mm SMA standards DO NOT APPEAR as
IDEAL SHORT, OPEN on the chart. The fact that they DO on the nanoVNA SAYS
IN ITSELF... There is an issue!



--
Dr David Kirkby Ph.D C.Eng MIET
Kirkby Microwave Ltd
Registered office: Stokes Hall Lodge, Burnham Rd, Althorne, CHELMSFORD,
Essex, CM3 6DT, United Kingdom.
Registered in England and Wales as company number 08914892
https://www.kirkbymicrowave.co.uk/
Tel 01621-680100 / +44 1621-680100



Spot on Dave, I agree. Addressing this process would provide a nice value added capability.

Alan

________________________________
From: nanovna-users@groups.io <nanovna-users@groups.io> on behalf of Dr. David Kirkby from Kirkby Microwave Ltd <drkirkby@...>
Sent: Wednesday, July 3, 2019 2:57 PM
To: nanovna-users
Subject: Re: [nanovna-users] Calibration coefficients - can one change them?

Okay, I get what you are saying now.

That problem should I guess be resolvable in the firmware, although it will
take a very careful look at the firmware to see how to resolve it. But the
problem should be resolvable.

Dave

On Wed, 3 Jul 2019 at 14:31, alan victor <avictor73@...> wrote:

I guess to make the point clear, 3.5 mm SMA standards DO NOT APPEAR as
IDEAL SHORT, OPEN on the chart. The fact that they DO on the nanoVNA SAYS
IN ITSELF... There is an issue!



--
Dr David Kirkby Ph.D C.Eng MIET
Kirkby Microwave Ltd
Registered office: Stokes Hall Lodge, Burnham Rd, Althorne, CHELMSFORD,
Essex, CM3 6DT, United Kingdom.
Registered in England and Wales as company number 08914892
https://www.kirkbymicrowave.co.uk/
Tel 01621-680100 / +44 1621-680100



Hello, Dr. David.

You can set the electrical delay with "DISPLAY, SCALE , ELECTRICALDELAY".
Yes, commercial calibrations are not required to be as short as possible, they can provide accurate correction data. But at a limited cost, it is easier for manufacturing to be as short and accurate as possible than to provide more accurate correction data. At the same time for HAMs with no VNA experience, it is easier to provide calibrations that no longer require additional data to be entered, although this is not precise enough, but the measurement below 1GHz is sufficient.

BTW,please forgive me for my poor English expression.
Thank you!

hugen
gen111.taobao.com