开云体育

Here is some more stuff to contemplate.

I "modeled" so to speak the inductor under test, the DUT, as a perfect inductor. No SRF.

I added the measured fixture parasitic obtained from the "perfectly calibrated VNA". I cascaded these elements into the simulator and as expected, looking into the VNA port, port 1, we return a S11 and an equivalent parallel inductance of 1.3 uH. As measured our 1 uH inductor appears as a 1.3 uH inductor and reasonably constant at low frequencies as long as we stay well below the SRF... Huh! What SRF? I stated the SRF of the inductor was NOT present. Well, it is not, but the fixture has created its own! Namely the 10 pF fixture C shunts the 1 uH and there is a MEASUREMENT SRF introduced. Proper de-emdedding if you keep your wits together will permit you to get the proper answer right through and past the resonate point introduced by the FIXTURE! This example exemplifies how KEY fixture details are and if not considered, will render measurement error.

Alan

Take a look at this:



I've ordered a $18 SOL kit from ebay. I'll be adding comparisons to the Anritsu to the thread when they come in. I'll drag the 8753B to where I can power it up for that. So I'll have TDR to 20 GHz and VNA to 3 GHz.

The more I use my Tek 11801 the more I wish that there was a cheap sampling scope that would work with one of Leo Bodnar's pulsers. While the 20 GHz of an SD-24 or SD-26 would be very difficult to match, one might be able to get to 5-6 GHz with a clever design which was cheap to make using modern parts.

The great thing about the time domain is you can tell which part in a series of connectors is doing what.

Have Fun!
Reg

On Fri, 4 Oct 2019 at 00:49, sala nimi <sala.nimi@...> wrote:

I had some time and tested some terminators for fun. I thought that
commercial terminators are nearly perfect, but not mine. I have a N
terminator perhaps from Digikey, and It was worse than my very old self
made terminator from UKW berichte or VHF communication. Homemade terminator
was dot or comma in the Smith chart but commercial one had a clear arc.

But so much depends on what you use to perform the calibration with. To
take an extreme example

* 50 ohm system
* Load used for calibration has an impedance of 30 ohms
* DUT #1 is a resistor of 49 ohms
* DUT #2 a resistor of 31 ohms.

Although the 49 ohms resistor has a much better return loss in a 50 ohm
system than the 31 ohm resistor, the VNA would indicate the opposite
because it was calibrated with a load close to DUT #2.

It is very easy to fool yourself with VNA measurements.

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

On Fri, Oct 4, 2019 at 02:49 AM, sala nimi wrote:

Homemade terminator was dot or comma in the Smith chart but commercial one had
a clear arc.

probably it just means that you calibrated your VNA with bad terminator. So, now it shows good terminators as bad, because they are not bad as one which is used for calibration :)

I had some time and tested some terminators for fun. I thought that commercial terminators are nearly perfect, but not mine. I have a N terminator perhaps from Digikey, and It was worse than my very old self made terminator from UKW berichte or VHF communication. Homemade terminator was dot or comma in the Smith chart but commercial one had a clear arc.

Thank you all for helpfull posts and links.

Do you have any gnuplot code which can import a Touchstone file and plot it on the Smith Chart?

Nope, I always reformat data before gnuplot so that it reads in
traditional columns.

Attached should be the gnuplot Smith Chart command files

Thanks for all the feedback.

First to clarify, the correction process is 2 -step for the BNC-alligator clip lash up. One is the removal of the shunt C of the OPEN clips and Two is the total series inductance of the clips and wire attached to the BNC-SMA adapter. Of course without mention, that BNC-SMA adapter adds its own parasitic but were lumping that into the alligator clip leads and wires parasitic.

Now, I just did the "Mickey-Mouse" cal process. That is to say, OSL cal with no regard to what this OSL is really. The open, the dangling alligator clips, the short, just short them together and the load, a 1/8 W carbon film resistor. As far as the VNA firmware algorithm is concerned, these are ideal elements. A measurement at 3 MHz revealed the inductance at 1.05 uH. Wow!

Well, before you go off with jubilation, we need to take a close look at the S11 reflection sweep for both cases. And I suppose I should take pictures and record. However, I will try to describe because it is quite simple but an eye opener. Although the MM-cal routine returned a reasonable value of the L at one frequency, the trajectory of the S11, reflection coefficient was totally off base. S11 and the corresponding Z showed a marked decrease in reactance as the frequency lowered and at frequency approaching 30 MHz the S11 value spiraled inward demonstrating a horrible reduction in Q. All of this is WRONG. The prior measurement after corrections provided a S11 and a subsequent Z vs frequency that made perfect sense. The Q and hence resistance component were high/low through out the measurement range as it should since we are well below SRF. The S11 hugged the unit circle as it should over the entire sweep. The MM cal routine would be hard pressed to permit us to investigate where SRF occurs.

I think it is key to recognize that the identification of the cal pieces EXACTLY is paramount in obtaining a corrected VNA measurement response for ALL frequencies involved. The same is required of the test fixture, it must be properly characterized at all frequencies of interest so the proper un-raveling of the DUT (so called de-embedding) from the measurement system is possible.

Alan

Thanks!

WB2UAQ,

Between the HP8721 directional coupler and HP 85032 calibration standard you've got some pretty expensive toys there. The NanoVNA must feel outclassed when used with them :)


Herb

Hi Alan
Yes the type J pots are good. Years ago I built a K2BT noise bridge and also a Wilford Caron bridge. The pot called out for one side of the bridge was a 250 ohm type J linear pot. I still have these bridges as I don't throw anything away:) The K2BT bridge was published in Ham Radio Magazine in the 80's when money for ham radio projects was hard to come by. It was the best bridge I homebrewed and made the old Palomar noise bridge look pretty bad. If a type J pot would fit in the Palomar I would have installed one.

$ python3.7 -m pip install --upgrade --force-reinstall numpy

That solved the problem. NanoVNASaver now runs on my Mint workstation.

Hi, I'm looking for drivers for windows 7 nanovna usb port. thanks

/g/nanovna-users/search?q=windows+7+usb

From Nigel ZS6RN:
"I have updated with the newer version based on the file name (en.stsw-stm 32080 vs en.stsw-stm 32102)
and after a re-boot and changing device driver from ChibOS/RT Virtual Com port
to the one from st.com, the nanovna now has a port number"
/g/nanovna-users/message/3660

Oristo, I forgot to mention that I use the NanoVna to measure return loss with good dynamic range. First I tested the S21 performance of the NanoVna with an HP 355D and found that I could measure the 10 dB steps nicely down to about 60 dB. Then I attached an HP 8721 directional coupler with the Ch 1 watching the reflected port. With a BNC short across the 8721's test port I did a through cal. I now have a pretty good RL measuring system with the 44 dB directivity that the HP8721 offers. The 50 ohm load from an HP 85032 shows the S21 going into the mud as it should.

Here, I experienced that a USB 5 GHz Wi-Fi dongle was substantially slowed in
speed by having a USB mouse plugged into the same hub.
So some USB appears to "work that way".

Yes, slow devices >>on the same hub<< can impact thruput of faster devices.
USB is, after all, "Universal Serial Bus". so faster devices must toe-tap
while slow ones (e.g. HID keyboards and mice) clutter the bus with short slow packets.
Wi-Fi has a comparable situation..

Windows users can launch Device Manager and View Devices by connection,
since multiple case ports may share the same internal hub.

When it does get control, a USB device sends packets at negotiated speed.
I suspect, with no proof, that nanoVNA may negotiate a faster speed that it can
maintain under worst case conditions, so I insert a USB 2.0 hub to limit speed.

Thanks for the explanation! I haven't tried this yet but will do what the manuals for the old Gen Rad bridges suggest. They suggest "shunting down" the unknown and out of range impedance with a capacitance that will drop the impedance across the bridge to within the range of the bridge and then the mathematically remove the admittance of the known shunt capacitance to reveal the impedance that was out of range. I have done this with an old MFJ 259 (type that can measure Z's R and X but has a limit of about 1.6 k) to get an idea of the common mode impedance of a current balun......it worked well enough.

Hi, I'm looking for drivers for windows 7 nanovna usb port. thanks
Gaetano

Hi Herb -

My reading of Alan's post is that he performed a normal OSL calibration at
port 0. After that he connected his shorted clips to port 0 and noted the
difference between the OSL short and the fixture (test lead) short which was
300 nH. He then measured the inductor and subtracted the 300 nH fixture
inductance for his final value.

I understood that.

No "Mickey Mouse" calibration involve

Exactly, but (lacking another impedance bridge) I would like to know how far off
is nanoVNA measurement using >>MY<< "Mickey Mouse calibration"

they are mixed to 5kHz to make it easier for DSP and so on

They are mixed down for the analog-to-digital converter (ADC),
which arguably makes DSP harder, since it has to sort opposite side band suppression.

Here is a collection of articles that I find helpful:


.. specifically:

* Kurt Poulsen's 30 minute video of Michael Knitter's presentation for FA-VA 5,
which has architecture similar to nanoVNA


* DF8OE's opposite side band suppression discussion

Oristo,
My reading of Alan's post is that he performed a normal OSL calibration at port 0. After that he connected his shorted clips to port 0 and noted the difference between the OSL short and the fixture (test lead) short which was 300 nH. He then measured the inductor and subtracted the 300 nH fixture inductance for his final value. No "Mickey Mouse" calibration involve. I'm sure Alan will correct me if I misread.

Herb