开云体育

On Tue, 20 Aug 2019 at 22:21, Warren Allgyer <allgyer@...> wrote:

Dynamic range is the difference between the calibrated reference, in this
case “0” dB and the noise floor. Nothing more complicated than that.

WA8TOD

Warren,

I am afraid to say it *is* more complicated than what you state. There are
actually at least *two* dynamic ranges on a VNA.

The lower limit of useable measurements is usually set by the noise floor
of the VNA, although Internal spurs sometimes exceed the noise floor.
That’s relatively simple concept, and is what I believe you are
talking about. I

The upper limit of usable measurements can be set by one of two very
different characteristics of the VNA.

1) The *system dynamic range* is what you are referring to, with your “0”
dB reference. This is relevant to measuring passive devices.

2) The *receiver dynamic range*. Is set by the maximum signal the receiver
can tolerate before it goes into unacceptable levels of compression. For
professional VNAs, “unacceptable” is usually considered the 0.1 dB
compression point of the receiver, but for low cost VNAs, it would not to
define “unacceptable” differently . This is relevant if the source power is
amplified, by an amplifier added for that purpose, or by the DUT.

I have not done any serious measurements on my NanoVNA. However, if after
calibration the difference between the 0 dB reference and the noise floor
is 70 dB, then the *system dynamic range* is 70 dB. That is what is usable
with a passive device.

If after calibration the VNA can measure an amplifier with up to 20 dB of
gain, the the receiver dynamic range is 70+20=90 dB.

So it is not as simple as you think!


--
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

Excellent!
I had been looking for a concise description of those correction calcs.

This seller was pretty quick and packaging was done with great care. Mine was packed in anti-static bag layed over a foam deposited in a aluminiun tin can wrapped in bubble then in envloppe....(the tester where in a separate bag cable and the two cable layered under the foam.)

Yes that makes sense. I am sure you will have no problem either finding a
library for Matlab, or just implementing the calcs in Matlab.

The 10/12 point cal will need some sneaky workarounds - as the Nano is not
a 'full' 2 port VNA, and so you can't measure the OSL standards on port 2.
There is a workaround cal type called 1 path 2 port though, where you use
port 1 for both and reverse the DUT manually. This is implemented on the
8753 and is for use with the simple 85044A T/R test set.

The simpler place to start is with the S11 correction calcs. Here is a page
with info on the standard equations to use for a 1 port VNA:


Here are some equations I used in an older program a few years ago, which
might save you a bit of time - if only to give you a point of comparison to
see if it matches your result.
These should be correct, but they are from early on my VNA journey :)

// e00 + A1*M1*e11 - A1*Deltae = M1
// e00 + A2*M2*e11 - A2*Deltae = M2
// e00 + A3*M3*e11 - A3*Deltae = M3

deltae = (((A2*M2 - A1*M1)*(M1-M3)) + ((A3*M3 - A1*M1)*(M2-M1))) / (((A2*M2
- A1*M1)*(A3-A1)) - ((A3*M3 - A1*M1)*(A2-A1)))
e11 = (M2 - M1 + (A2-A1) * deltae) / (A2*M2 - A1*M1)
e00 = M1 - A1*M1*e11 + A1 * deltae

Where the Ax are the ideals and the Mx are the measurements.
Then you can apply the error corrections using this:
A = (M - e00) / (M * e11 - deltae)

I'm out of action this weekend too, look forward to catching up next week
sometime.

Roger

Given that it works at all, my guess is that it will work just fine for you.
There is some question as to whether or not some of the clones do as well at extremely high frequencies,
but most of us won't be using it at 900mhz and expecting the best possible accuracy.

Hard to make a blanket statement about the clones, as they come from a number of different shops,
and they may have evolved over time.

In what way does your clone differ, other than lack of shielding?
Could be that the ebay seller obtains stock from several different shops
as a function of availability and the prices quoted to them.

Jerry

No one has been able to define any performance differences. Only cosmetics. I too have. “Bad” one and it’s performance is fantastic and consistence with my much higher priced instruments. Ignore the cosmetics and enjoy!

WA8TOD

Having spent a career in digital design, analyzing this thing is exercising a bunch of stuff
that was explained in classes I slept through some 45 years ago.
So you RF guys should get out your red sharpies and correct this as you see fit.

Those magazine references Alan gave will be well worth looking up, have not yet done so.
Some may not be easy to find without access to a university library.
Available (legally!) on the web are the following designs that I consider precursors to the nanoVNA:
AQRP VIA, kit #25: Uses same si5351, SA612A's tlv320aic3204 and an ARM processor with LCD
Baier's VNWA: Uses same resistive bridge into three SA612A's
Also of note are the AIM430 by W5BIG: (same V/I front end as the AQRP VIA)
and of course, what seems to be the granddaddy of amateur VNA's, the N2PK:


The ASCII diagrams below will display correctly when formatted to use a mono-space font such as courier.
Curious that other groups.io forums allow me to format my text to use a specific font, but this one does not.
Perhaps to discourage egregious SHOUTING?

Here's my take on how the nanoVNA bridge works, starting with this very simple resistive bridge.

Vs o----+--------+
| |
50 50
| + - |
+---Vd---+
| |
Z 50
| |
GND GND

We apply a known AC voltage at Vs and measure the voltage at Vd.
You can follow the math by pretending Z is a pure resistance and
that Vs and Vd are DC voltages, but remember that Vs, Vd, and Z
are actually vector quantities having both magnitude and phase.

The two 50 ohm resistors on the right divide Vs by 2,
so the voltage at the top of Z is Vz = Vs/2 + Vd
The current through Z is the current through the resistor above it,
so Zi = (Vs - Vz)/50
The impedance is Z = Zv/Zi = (Vs/2+Vd) / ((Vs-(Vs/2+Vd))/50)

There will be an unknown impedance Zx between the bridge and the top of
the unknown device due to traces, connectors, and cabling.
We could find Zx by putting a short across the connector provided
for the device under test Z and measuring impedance as described above for finding Z.

With an open at that connector, the voltage at Vd should be equal to Vs.
We can use this measurement to determine any scaling and phase shift
between Vd and Vs.

The above two correction factors interact, the actual correction factors
to be used would require solving a couple of simultaneous equations.

A 50 ohm load at the connector should give a Vd reading of zero,
since the bridge is now be balanced.
Exactly how that gets used as a correction factor is not obvious to me.

The nanoVNA bridge has some extra resistors hanging off it, looks more like this:

Vs o---150---+------+--------+ Vt
| | |
56 50 50
| | + - |
GND +---Vd---+
| |
Vz +---75---+ Vz-Vd
| |
Z 50
| |
GND GND

The 75 ohm resistor in parallel with Vd is the combined effect of
resistors R9,18,19,20,21 and the 2*1.5k differential input impedance of U8.
82 || (2*390 + (2*50 || 2*1500)) = 74.99 ohms

From the network shown above, we can solve for Z
by setting up three simultaneous equations using Kirchoff's current law:
(Vt-(Vz-Vd))/50 + Vd/75 - (Vz-Vd)/50 = 0 # Currents into node Vz-Vd
(Vs-Vt)/150 + (Vz-Vt)/50 + (Vz-Vd - Vt)/50 - Vt/56 = 0 # Currents into node Vt
(Vt-Vz)/50 - Vd/75 - Vz/Z = 0 # Currents into node Vz

The correction factors get a bit more complicated as well.


Solving for the output impedance, I've modeled the nanoVNA as follows.
Note that node Vb is labeled as Vz-Vd in the diagram above.

TX o----+---50------------+---43.53---GND
| |
+---75---+---50---+ Vt
|
Vb +---50---GND

That 43.53 ohms is the effect of resistors R13,14,16,17, plus the 2*1.5k input impedance of U6
and the 50 ohm output impedance of the Si5351:
56 || (150 + (50 || (470 + (27 || 3000)))) = 51.53 ohms to ground
As I recall, the Si5351 datasheet does claim an output impedance of 50 ohms,
though the CMOS drivers can only be programmed for a max of 8ma of drive.

Solving this for the output impedance is again a matter of Kirchoff's current law,
this time with equations for nodes Vt and Vb.
I arrive at a number of 53.85 ohms.


The Si5351's CLK1 port is loaded down about as much as we dare.
Assume the TX port is grounded, assume the 82 ohm R9 is actually 75 ohms
due to R18,19,20,21, and U7, and assume the Si5351 is a 50 ohm source.
Since the Si5351 is driving a ~3v square wave through that 100nF cap,
it will be driving the load as if it were a plus/minus 1.5v voltage source.
The current out of the Si5351 will be 8.5ma, calculated from this:
1.5v / (50 + (470+27 || (150 + (56 || 50 || (50 + (75 || 50))))))
The maximum that the Si5351 outputs are spec'd to give is 8.0ma.

Jerry, KE7ER

Hi, new to this group,
I recently purchased a NANO VNA from eBay. Unfortunately it was before this series on the bad clones....I got a bad one. They showed pictures of the good ones and sent me a clone. What are the operational differences between the original and the clone? I can see where the lack of shielding would pose a problem at higher frequencies, but other than that it seems to be working correctly.

Tnx de Gary, KF9CM

On Aug 22, 2019, at 21:18, Larry Rothman <ac293@...> wrote:

Carsten,
From your statement, it sounds like you only like OS s/w.

This is less of a matter of liking, but more a matter of some 40 years of experiencing the “illusion of vendor support” (RFC 873, September 1982).

Did you compile the compiler you use to create s/w exe’s?

Yes.

If you do, what did you use - another compiler?

Yes.

Who compiled that one?

Apple.

How can you trust your compiler - or its compiler - or its assembler if you can’t see back to point zero?

I can’t prove anything here (see Ken Thomson’s 1984 Turing award acceptance speech “Reflections On Trusting Trust”, ).
On the other hand, a lot of people look at the compilation results from Apple’s compilers.

Did you inspect the firmware that was already installed in the NanoVNA?
Did you verify that it only provides a serial interface, and that it does not have a hidden HID interface that can inject keystrokes so fast, you’d never notice?

My NanoVNA is stuck in customs at the moment.
But I’m happy that I get to examine and compile its firmware on my own, should I choose to.

I do computer security and if you’re going to be paranoid and not trust just about everything 'out there' you won't get anything done.

I teach information security (of which computer security is a part). Paranoia is a medical condition that is often confused by medical laymen with the mental state someone gets into who starts to understand information security.

How about all the Hams that have created closed source software for the great test equipment
they’ve come up with?

It is their prerogative to do so. It’s just not very bright.

What we have learned in the field of computer science is that open source software leads to standing on the shoulders of giants, closed source software leads to others standing on your feet.

(Besides, if you want to, it turns out it is easier these days to monetize the visibility you get from a good open source project than to turn a closed source software project into money. Except for very few situations that need full attention and lots, *lots* of random luck.)

Your statement “Being trusted always creates a liability." just doesn't cut it.

Not sure you understood it. My point is that you can always choose to “trust" someone.
In daily personal interactions, trust may seem like a good thing for the trusted person.

In reality, it also creates a liability for the trusted person: to actually be trustworthy.
And that is not just about acting conscientiously, but also about being skilled enough not to get compromised.

Here we don’t know how that liability is being filled, which makes the trust questionable.
With open source software, more people can share that liability, which can (but need not) make the trust less questionable.

Grü?e, Carsten

I've observed that it is very common for custom C# applications to trigger
a false positive on a virus scan.

That doesn't help provide a definitive answer I'm afraid, but FWIW I have
zero concerns about this application.

Roger

If you're paranoid, back up your machine regularly.
Actually, you should do that if you aren't paranoid.

And if you think somebody else should have released the fruits of their efforts as open source,
then perhaps your best path forward is to spend a few months creating an open source replacement for it.

That said, I'm truly grateful that the firmware on the nanoVNA itself is open source.
Gives us an excellent platform for developing it further.
Kudos to edy555 and Hugen.

Jerry, KE7ER

As discussed in prior posts the NanoVNA uses a resistive bridge.
They are wide band and easier to construct then the alternative
bridge utilizing a directional coupler.
There are excellent references on the resistive bridge. The RF
bridge being a variation of the well know Wheatstone Bridge.

An excellent treatment and discussion is presented in
"The Wheatstone Bridge:How Does It Impact VNA Measurements"
Author: Brian Walker. Presented in RF and MIcrowave Magazine.
Further work in this area is presented by Joel Dunsmore and appears
in RF design Magazine Nov 91, "Simple SMT Bridge Circuit Mimics
Ultra Boradband Coupler".

I decided to calculate the port impedance of the NanoVNA R bridge.
It is 51.27 ohms and within the specified SWR source match.

Using this bridge as a template and based on a OPAMP version
of a similar bridge, I constructed a simple Spice Version. This schematic
is attached. The simulation is run in LTSpice. A free down load of this
simulator is available from Analog Devices web. A nice treatment of this
bridge and running it in simulation is instructive! You can plot Gamma as well return loss.
See a paper by Dunbar published in Electronic Design May 29, 2000, "Build A Vector Network Analyzer".
There the discussion presents VNA's which can span audio through RF and
includes functions like metal detectors, microphone analysis, etc... This
circuit operated in Spice simulation is handy to check the understanding of what the
NanoVNA returns in measurements.

I'll see if I can place the LT Spice circuit file in a folder for those interested.

has a great list of articles on tests they've performed on the most popular AV s/w.
It's a nice site for research - they publish their test methods as well.
For 2018, Microsoft had the most false +'s but that is to be expected from them as they want to safeguard the family jewels.

In any case, this is becoming noise in this forum - back to the NANO!!!

Anyone look at my housing design?

FWIW both a fresh download of NanoVNAsharp and one from July come up clean with BitDefender, which doesn't seem to be included in virustotal scans.

Quite a few antivirus companies share technology ag Avast = AVG, so it is possible your 'different' security programs use the same code base.

Mike

Carsten,
From your statement, it sounds like you only like OS s/w. Did you compile the compiler you use to create s/w exe's? If you do, what did you use - another compiler? Who compiled that one? How can you trust your compiler - or its compiler - or its assembler if you can't see back to point zero?

Did you inspect the firmware that was already installed in the NanoVNA?
Did you verify that it only provides a serial interface, and that it does not have a hidden HID interface that can inject keystrokes so fast, you'd never notice?

I do computer security and if you're going to be paranoid and not trust just about everything 'out there' you won't get anything done.
How about all the Hams that have created closed source software for the great test equipment they've come up with?

Your statement "Being trusted always creates a liability." just doesn't cut it.

Regards,
Larry

On Aug 22, 2019, at 16:46, Larry Rothman <ac293@...> wrote:

Hugen wrote the application spcifically for the nanoVNA and it is closed source.

Which doesn’t protect us from him catching something that then is compiled into all executables he makes. Yes, that has happened to others.

Being trusted always creates a liability.

(And the closed source nature heavily increases the amount of trust required.)

Grü?e, Carsten

Larry,

This is a good test site. It’s known that some specific AV programs have false positives for miscellaneous programs that other AV test approaches do not. It’s why, for instance, Wes Hayward doesn’t distribute the software to his popular RF circuit design book any longer ;and the ARRL has removed the CD from the book, too).

I sent out a PDF newsletter to my club that I got monthly from a club in Texas and all hell broke loose because one member used Avast which had a know false positive whereas all other AV software didn’t. So your use of the above website is a great tool to check these issues.

73,

Frank
K4FMH

I don't know what you're using for AV but virustotal gets 0 hits:


The nanoVNA Wiki has the google drive link for the software:


Hugen wrote the application spcifically for the nanoVNA and it is closed source.

I wanted to download the PC software, nanovnasharp from Hugen's Google
drive. Tried it on 2 pcs, and both using different security programs
flagged it as a potential ransomware program. Can anyone provide a site
from which the files are clean?

I'm assuming that this is the only software that is available?

Roger, I had a thought...

I think I might try your same technique of external calibration for my original test (per the first post in this message thread).

In other words, using both the NanoVNA and the 8753C, like you, I would capture s2p (or s1p?) files for each of my SOL loads, and then capture, uncorrected, the S11 data for my tuner. And then use this data and the correction formulas to create corrected s1p files for each VNA.

I haven't used Python (well, not for a very long time), but I'm thinking I could do it with a Matlab script. I know that one of Dick Benson's files on the Matlab Central website has a 10-point correction script (it's in the 3577A utility -- and it is only 10 points, rather than 12, because he skips the isolation measurements).

I'd need to figure out what the correction formula is for S11 -- I'm assuming it's a 3-point correction. But if I can do that, then I should be able to accomplish this goal. I'm very interested to see what the results would be, and how they differ from measurements made on the VNAs after they've been cal'd.

Anyway -- this won't happen until early next week, as the wife and I are off to Seattle to see her son.

- Jeff