开云体育

On Wed, 2 Oct 2019 at 20:51, alan victor <avictor73@...> wrote:

Good day,

When I began this process of working with the NanoVNA, I was not certain
how the cal standards were being treated. Since the completed cal showed a
perfect open, short and termination, I was pretty certain they were NOT
defined as REAL SMA connector devices! The SMA connectors are NOT that
perfect!

Alan

Ignoring random errors (noise, repeatedly of connectors etc), the standards
should look perfect opens and shorts, i*f that is how they are defined. *You
could use an SMA plug with a few cm of wire and leave it open to make your
open, and put your short on the end of a few cm of wire. If the calibration
standards are defined to be perfect text-book opens and shorts, they would
look like that on the Smith Chart. If you then put the supplied short on
the VNA, that would look wrong.

The female standards supplied with the kit are not bad at all. The problems
are going to occur when using most other connectors - everything except
APC7 is going to have some problems.

To check if I understand this correctly I ask, There is also a phase difference between reference and difference voltage from a bridge. I should have remembered that. And now they are mixed to 5kHz to make it easier for DSP and so on.

I may start this as a new thread titled Measurement of Components... again.

In any case, I put together the nice n' easy alligator BNC connector interface which is easy to deal out.

The part under investigation is a 1 uH inductor whose SRF is above 150 MHz. Nice. I can measure it down at 3 MHz or so with no worry about SRF complications.

I do a shunt mode measurement, one port only and a SOL calibration directly at the VNA CH0 port.
The frequency range is 1-30 MHz, 101 data points. I then add on the BNC alligator clip interface to the VNA.

My first observation is that this test set system with the alligator clips open and laying on the table has a shunt C of 10 pF at 3 MHz. I then short the clips together and find that the series inductance of this test system is 300 nH.

Next I connect my 1 uH inductor and note the CW movement of the chart with a readout of 1.37 uH at 3 MHz. I have at this point an inductor under measure with a shunt C of 10 pF and a series L of 300 nH.

I can remove the shunt C first by noting that the measured reactance must be of the form of the product over sum of the DUT and the shunt C of the fixture. Hence, the actual reactance measured, Xa must be:

Xa=(Xc*Xm)/(Xc+Xm) where Xa is the actual reactance of the DUT, Xc is the shunt C of the fixture and Xm is the measured reactance. You will note, if the Xc value is significantly LARGE then Xa=Xm and the only fixture component that I would have to remove is the stray series inductance.

The measured value of above experiment returns an Xa equivalent to 1.37 uH. Now I must remove the stray 300 nH and I arrive at a corrected value of 1.07 uH. This value correlates with an independent vector Z meter measurement within 50 nH measured at the same operating frequency.

Key here, you must calibrate with know standards to remove the aberrations-imperfections of the VNA instrument hardware, this is level one. Then level two must address the inherent parasitic values of the test set. This is essential. Then apply the appropriate arithmetic to find the unknown DUT itself.

Alan

Mike,
see attached for a first sneak peak of what I'm working on.

I think I'm currently under-estimating the roll-off by measuring it from
the -6dB point and to the end of the data. Is there a norm for how to do
these measurements? Or should I just make it up as I go along?

[image: image.png]

--
Rune / 5Q5R

On Thu, 3 Oct 2019 at 15:45, mike watts via Groups.Io <wy6k=
[email protected]> wrote:

Yes, all of those filter measurements would be useful for things I do.
Those functions would elevate you from "hero" to "demi-god" in my world.
:-)
Mike WY6K


"... somewhere in the distance, there's a tower and a light, broadcastin'
the resistance, through the rain and through the night..."

On Thursday, October 3, 2019, 4:20:50 AM CDT, Rune Broberg <
mihtjel@...> wrote:

Hi Norbert,
the application can run calibrations at far more than 401 points - I have
certainly done 5050 points at one point. ;-) But it does this in the
application, and not on the device, due to the hardware limitations.

I agree that a function to find cut-off frequencies for filters etc. would
be interesting, and I've put it on the list of features I would like to
implement for the next version, ie. for next week. I don't make any
promises though :-)

What measurements would be relevant to provide? Cut-off frequency, span for
band-pass, dB/octave and dB/decade fall-off, pass-band ripple?

--
Rune / 5Q5R

On Thu, 3 Oct 2019 at 11:12, <norbert.kohns@...> wrote:

Hi Rune,
thank you for fixing this issue!
I have a question for you in respect to the required number of data

points

to measure a band pass filter. With 101 points it is most likely not
possible to find the -3 dB points because the sample point is way off

from

-3 dB on both sides of the band pass. With 401 points it would work. An
automatic feature of measuring the bandwidth would be absolutely cool!

The

TAPR software is able to perform a calibration at 401 points or even more
than that, so I am certain that you could do that as well. I have no

glue

on how difficult that would be, but more data points for calibration

would

be a huge improvement.

Kind regards
Norbert, DG1KPN

Rune,

Thanks for your kind comments.

Here are some suggestions regarding your 4 questions above.

1) What window do I recommend?

A Blackman window has very low peak sidelobes (< -58 dB) which is unnecessary in a TDR application using a nanoVNA. The Blackman window achieves this at the expense of impulse response width. The pulse we see on the screen is wider than it needs to be. You could narrow the impulse response a bit by choosing a window with higher peak sidelobes such as a Hamming window (-43 dB peak sidelobe). The peak sidelobe of a Hanning window is <-31 dB which is a littler higher than I would recommend; however, it would be an acceptable choice for many applications. Make sure that the length of the window and the length of the samples in the frequency domain from the VNA are the same length. And, you window both the real and imaginary parts of the data in the exact same way.

2) Should you zero pad the data?

I would recommend zero-padding the data. I can't be sure but it appears you are already doing this based on what I see in your TDR plots. What I am seeing could be, in part, an artifact of the Blackman window you have chosen. Zero padding has the advantage of interpolating the time-domain data for the user. This enables more precisely estimating time delay for cable length if the cable is open on the far end from the nanoVNA.

3) How long should the IFFT be?

I would multiply the number of samples in the S11 data by at least 2 (4 would be better for frequency data of less than 500 points) and then take the next higher power of 2 for the IFFT length. If you are not familiar with how to do this. Here goes. If n is the number of samples (say 505), you use the equation IFFTlength = 2^(integer(log10(4*n)/log10(2)+1)). Basically, the log10(x)/log10(2) is the log base 2 of x. So for 505 samples from the VNA you would use a 2048 point IFFT of the windowed (see above) 505 samples from the nanoVNA. An even longer FFT won't hurt (n*8 in the equation above); however, it is probably not important to any measurement that comes to mind using the nanoVNA.

Changing the IFFT length does change the time-domain scaling of the samples being plotted compared to a non-zero-padded version. The relationship is inversely proportional to the IFFT length. In-other-words: the zero padding gives better time-domain sampling in the data. The window function chosen dictates the 3-dB width of the impulse; however, the length of the IFFT determines the time-domain sampling of the output.

4) What part of the output should be displayed?

Most IFFT algorithms will place the data we are interested in within the first half of the IFFT output. That appears to be the case for what is shown in version 0.1.0 of your great software. Therefore, you can just plot the data from the first FFTlength/2 samples. Of course, you need to plot the absolute value of the complex numbers out of the IFFT. I am pretty sure you are already plotting the absolute value.

I hope this helps with your continued development of nanoVNASaver. It is already a great tool.

Let me know if you have any questions.
--
Bryan, WA5VAH

On Thu, 3 Oct 2019 at 18:03, Henning Weddig via Groups.Io <hweddig=
[email protected]> wrote:

according to ebay my order is just shipped. It was marked shipped with a
a tracking number, so I can track the status of the shipment.

Henning Weddig

Same here, but clicking on the tracking number says that the label has been
created, but not actually shipped yet.

If this was $200 I would be a bit more concerned, but $20 is not much
money. If the VNA doesn’t show up, and it takes me a couple of months to
get my money back, I am not bothered.

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

No problem. There are many names around for the same file.That is why I uploaded them in the files section and dated each one.
Forum member ORISTO has created a WIKI page that describes the various releases:
Firmware Versions (with links)?by Oristo Po

More Forum member should be browsing through the WIKI /g/nanovna-users/wiki and FILES? /g/nanovna-users/files sections of this group.
There is a TON if info in there folks !!
And don't forget to print a copy of the Oct 2nd User Guide (29 pg) for reference while you're at it.
/g/nanovna-users/files/NanoVNA%20User%20Guide-English-reformat-Oct-2-19.pdf

Thanks !

Yup - been telling folks here for months about how they can get 'free parts' from all the old stuff they have lying around.
That's what this hobby is all about - hardware hacking and having fun doing it!

Really?
Read the latest Oct 2nd version of the User Guide in the files section on how to convert HEX or BIN files to DFU so you can update the flash on your Nano.

What is the update procedure?
Bob

Look for?? edy555_nanovna_v2.2-sept-30-19.dfu? or similar.
That's the bugfix version.
You can always go to the TTRFTECH github releases page:

Read the latest Oct 2nd version of the User Guide in the files section on how to convert HEX or BIN files to DFU so you can update the flash on your Nano.

Okay - snooped around the wiki, yet I still couldn't find edy55_bugfix (10-1-19), for example.

I got the switches off a scrap set top box PCB, but you can buy them almost anywhere. They are 6mm x 6mm and come in a variety of button heights.

4.3mm is flush with the PCB, 7mm sticks out slightly, take your pick.



--
Regards,

Martin - G8JNJ

Wow Bryan! Thanks for that in-depth piece of work!

I use a Blackman window, which is created by putting in "np.blackman()" ;-)
So I can use whatever window function is the most relevant. These ones are
available:

I've had suggested that I should pad the input data with as many zeroes as
I have data values, I think in order to increase either resolution or
range? I haven't done it yet, but I was thinking about having some
experiments.

If the second half the values shown are useless, I might as well just
discard them! :-)

You're right that the graph can currently only exist in one place. It's not
particularly difficult to fix, but it involves re-doing a tiny bit of the
application architecture; which I really should do at some point anyway. I
don't promise when I will get it done, though. ;-)

Thank you very much for your feedback, and I hope you can help me select
things such as:
1) The window function to use,
2) Whether to pad the input,
3) How many points to use for the FFT,
4) What part of the output should be displayed (half, like you suggested?)

--
Rune / 5Q5R

On Thu, 3 Oct 2019 at 17:16, bryburns via Groups.Io <bryburns=
[email protected]> wrote:

Rune,

I am taking a deeper look at your TDR plot. Thanks for implementing this.

I think you are plotting all of the samples of the IFFT of the S11 data.
If so, I would recommend that you chop off the last half of the plot. With
5 segments (505 total samples) from 50kHz to 900 MHz I am seeing the right
side of the plot indicating 55.4 m as the longest distance one could
measure. I set the the cable type to RG58/U with a velocity factor of
0.66. I think 55.4 m is too large by a factor of 2. The last half of the
bins in the IFFT are time delays that we won't measure in this way with
real cables. As it is, you give the impression to someone that they could
measure a 55.4 m cable for the settings indicated above which is not
correct.

Are you applying a frequency domain window function to the data prior to
doing the IFFT? If so, it would be useful to let folks know what window
you are applying when it comes time for documentation. If not, there are
an infinite number of possible windows. I would suggest something fairly
easy to implement such as a Hanning window or a Hamming window. For cable
length measurement, what you have is great.

I am interested in using the TDR to look for impedance deviations along
the transmission line which can get easily obscured by time-domain
sidelobes if no window is applied. I am using the TDR capability to look
for coax damage or breaks in the protective jacket caused by gardening. My
coax cables run through a garden area to my antennas which are above the
garden. Outright breaks in the coax are fairly easy to find; however, some
types of damage are a bit more difficult to observe. The TDR is especially
helpful for this application.

I do like the separate plotting capability you have implemented for the
TDR data which is in the same window as the coax definition. I have
noticed that once you put it into the main window as one of the graphs, you
cannot ever see the TDR data again as a separate plot unless you restart
the program. It seems like it would be a good idea to restore the plot to
the TDR window if it is removed from the plots in the main window or just
permanently leave it in the separate window that the user can bring up as
desired?

Another alternative is to only plot the TDR data in the main window. When
the TDR plot is a graph in the main window you should show at least the
velocity factor you have assumed on the TDR plot and perhaps the cable type
the user selected. This makes the plot self-contained when it is saved as
an image. I prefer a separate window where I can make the TDR plot as big
as my screen; however, that is not in any way essential and probably makes
more work for you.

Thanks, again, for your great work on this software. I greatly appreciate
it.

73,
Bryan, WA5VAH

where did you find the switches?

--
Dave, WB5NHL
5BDXCC, 311 countries confirmed
@dkjellqu

Always takes at least a day before it appears in the stracking

Actually, if you check, you will see that ONLY the shipping INFO was received by China Post - not the actual item.
This is what my eBay info says: ?? Shipping status Shipping info received

29 : ann : our nanovna will be evaluated tonight

according to ebay my order is just shipped. It was marked shipped with a a tracking number, so I can track the status of the shipment.

Henning Weddig


Am 03.10.2019 um 18:56 schrieb Larry Rothman: