HPAK has an informative application note on TDR.

See pages 27 - 30 of link:

https://www.testunlimited.com/pdf/an/5989-5723en.pdf

for discussion of TDR range and resolution.

Please read page 11, paragraph 4.1

"The measurement frequencies must be set so that the stop frequency
is equal to the product of the start frequency and the number of points (which would
yield harmonically related frequencies). The VNA has a function that performs this
automatically. From this, the rise time is determined by the maximum slope of the
highest frequency measured . . . "

Does the LabView NanoVNA system perform the automatic function ? [probably automatic only for 60G, $ 60K VNA]

On Mon, Aug 30, 2021 at 11:31 PM, John Galbreath wrote:

>
>
> Does the LabView NanoVNA system perform the automatic function ?

No.

While I implemented TDR for the 8754A, I had made some improvements based on the following article:

https://dspguru.com/dsp/howtos/how-to-interpolate-in-time-domain-by-zero-padding-in-frequency-domain/

Of course, if you want to see it in action:

https://youtu.be/XaYBpPCo1qk?t=4258

>From the same V2Plus4 review, showing my poor man's Beatty standard (very loose use of the word)

https://youtu.be/XaYBpPCo1qk?t=3122

Attached is the XL spreadsheet I use to help me find optimal freqency span and # of points to use for T D R with LabView NanoVNA.  Don't know what the "Samples/Dec" does (same row as # of points).  The TDR step plot is a "home-made" Beatty standard consisting of 4-foot RG-58 coax cables:  1 in series with 2 in parallel ( mismatch ) in series with 1 open at end.

On Wed, Sep 15, 2021 at 03:14 AM, John Galbreath wrote:

>
> Don't know what the "Samples/Dec" does (same row as # of points).

Covered in the manual.

Paragraph 15.1.1 (page 70) covers this, however I still do not understand it and I think it would be beneficial if I could understand it.
I am also not sure about the decimal point in the resolution calculation in my spreadsheet - please verify.

Also pg 142, section 23.7.  I suspect though you are not understanding what a log sweep or maybe what a decade is.  These terms are very common in the industry and you have the internet if you would like to learn more about them.

No, I do know what a log or linear sweep is, and I know about octaves and decades, but I do not understand what effect the number of samples per decade has on the inverse FFT to convert the frequency domain operation of the VNA to the time domain operation of the resulting reflectometry information.

I did not look at section 23 in detail since I am not yet interested in PDN.  The TDR display is linear in time/distance so I doubt if I would ever need to utilize the log sweep for TDR.

Actually it was very clever for you to augment a logarithmic sweep when the NanoVNA lacks this capability.

On Thu, Sep 16, 2021 at 10:10 AM, John Galbreath wrote:

>
> No, I do know what a log or linear sweep is, and I know about octaves and
> decades, but I do not understand what effect the number of samples per
> decade has on the inverse FFT to convert the frequency domain operation of
> the VNA to the time domain operation of the resulting reflectometry
> information.

>From your original comment of "...however I still do not understand it and I think it would be beneficial if I could understand it."  you leave it up to readers to guess at what you are not understanding.  I suggest in the future, think about what it is you are wanting to ask, then provide the details up front.

When I came up with the idea of a quasi log and segmented sweeps, I was very limited with what I could do with the old HP8754A.  You have no remote control over the frequency range.  The HP8501A nomalizer had a very limited number of points it could collect.  The only way to make narrow band measurement was to use an external RF generator.   I learned a lot of the basics from that system.

Compared with my first VNA, the original NanoVNA offers some advantages when trying to control it with a PC.  Still, it had a lot of the same problems.  A very limited number of data points and no log sweep.  Something I had worked around 15 plus years ago, still hunting me...  Then again, it's only $50.

PDNs are also feature rich at lower frequencies and having a log sweep can really improve the measurement by reducing the time and processing required.

You won't find any mention of log or linear segmented sweeps in section 18 where the manual covers the TDR measurements.   While the paper you linked and I assume read, mentions the bandpass mode not requiring frequencies that are harmonically related.  I'm assuming there's something you are wanting to exploit.  Just keep in mind as I mentioned earlier, the log isn't really a log but rather a sequence of segments just like with the linear mode.  The difference being that each segments start frequency is based on a log rather than a linear function.

Joe,

Appreciate a link to the "manual" commented on in this thread.  There's nothing I could locate in the Wiki or Files (just Nano User manuals) that had the number of pages or chapters being referred to !
Thanks,

Kit

The manual may be found on my Github account.   Just download the latest PDF.    The software as well as links to various videos I have made are also there.  Start with the README.

https://github.com/joeqsmith/NanoVNA_Software

A spreadsheet to calculate sweep parameters (Frequency span, number of points) and TDR range in accordance with guidelines discussed in HP-A-K application note, found at following link:
https://www.testunlimited.com/pdf/an/5989-5723en.pdf

is attached.  The resolution is close to, but not exactly equal to range shown in LabView NanoVNA program.  Would appreciate comments on corrections to get closer.

Rohde und Schwarz has an excellent 13 minute You Tube explanation of TDR including animation which displays relation between range, resolution and sweep parameters - link is at

https://www.youtube.com/watch?v=cabqRlQdNSY

1.  How does your LabView NanoVNA software calculate the handy "electrical length in meters" at the upper right hand corner ?  As shown in attached screen-shot, the range is 0.12 meters and the range calculated from HPAK application note is 0.18m.

2.  Do you think that the thru-line part of Bird model 43 directional wattmeter (portion contain slug and in/out connectors) could function as an airline ?

On Tue, Nov 2, 2021 at 04:30 AM, John Galbreath wrote:

>
> 1.  How does your LabView NanoVNA software calculate the handy "electrical
> length in meters" at the upper right hand corner ?  As shown in attached
> screen-shot, the range is 0.12 meters and the range calculated from HPAK
> application note is 0.18m.
>
> 2.  Do you think that the thru-line part of Bird model 43 directional
> wattmeter (portion contain slug and in/out connectors) could function as
> an airline ?

1) There are many papers available on-line that talk about the impulse wave.   Basically its just an inverse FFT that is filtered.
2) I have no clue what a Bird model 43 is so I did a quick search and downloaded the manual.   "The RF circuit of the Bird 43 is a length of uniform air line with Zo = 50 ohms."  So, yes, an airline would indeed make a very good airline!   The power of Google.

If you watch that video you linked, they talk about it starting here:

https://youtu.be/cabqRlQdNSY?t=279

The window function is covered here:

https://youtu.be/cabqRlQdNSY?t=637

I'm actually surprised the you would get anything meaningful.   Again, I can't stress it enough that the software does not typically limit what the user does.  Sweeping from 50kHz to 3GHz with only 18 points may cause a problem.  The software will certainly allow you to do it but your VNA may provide unstable results.

My first NanoVNA had stability problems with normal use.  When I pickup up an H4, the same problems were still present after well over a year.  This is why I never show the H4 in any demos.  The firmware has been the biggest problem.   Using the software just adds one more layer to the problem but I assume you tested it stand alone first.  If not, I strongly suggest you start there (no PC) and make sure you are getting correct results.   If you find your VNA is unstable (very possible), you may be able to increase the number of data points or bump up the start frequency.   You would just have to play with it.

Because you seem to stumble a lot and are just starting to learn the basics of TDR, to help you understand the link shows a short video of my V2+4 with a similar transmission line using your settings.  The firmware is 20201010-86c7055.   I would start by running a similar test just to make sure your setup is stable.

https://www.youtube.com/watch?v=iwFgnvfYN8Q

Guessing from your lack of any response, you are finding instability problems with your device.  The two plots below further show the effects of range and the number of data points on the stability.    The first is sweeping 50kHz to 3GHz starting with  18pts, then bumping it to 23pts.  Notice however in the second plot I have increased the span to 50kHz to 4.4GHz.  Even with 32 points, the V2+4 with firmware 20201010-86c7055 is unstable.  Once the number of data points is increased to 33, it becomes stable.

Again, you do not need my software (or even a PC) to try these experiments out.  All you are doing by adding the software is adding to your confusion and increasing your frustrations.  IMO, you would be far better served by learning some of the basics about your device first.  My software is not going to make fix any problems with the firmware.  Nor will it try to keep beginners in some safe operating region.  It's up to you to understand your test equipment and how you can leverage my software.

You most likely will find some versions of firmware more stable than others.  The reason I am using 2-year old firmware in my original NanoVNA is because it actually works well enough to get some decent results.  I have not had any luck with the newer firmware.  That's true for the V2Plus as well.  I have only found one version of firmware for that device that was stable enough.

Because I run these devices headless, I really don't care about all the features that are continually being added.  I just need the firmware to provide me with stable data.

I tried your suggestion and I had no problems with NanoVNA.  The attachment gives configuration details

The only reason I run with low number of points is to increase resolution.  Then the mis-match plots have right angles where they belong.

I am amazed what can be accomplished with free software and inexpensive hardware, which all works well.  The only problem is my slow PC with insufficient RAM.  It hangs up on 16 points and below, but works at 18 points up.  The mis-match standard is only 4 inches long.

On Mon, Nov 8, 2021 at 02:36 AM, John Galbreath wrote:

>
> I tried your suggestion and I had no problems with NanoVNA.  The
> attachment gives configuration details
>
> The only reason I run with low number of points is to increase resolution.
> Then the mis-match plots have right angles where they belong.
>
> I am amazed what can be accomplished with free software and inexpensive
> hardware, which all works well.  The only problem is my slow PC with
> insufficient RAM.  It hangs up on 16 points and below, but works at 18
> points up.  The mis-match standard is only 4 inches long.

Why are you amazed when you are seeing a FWHH of 51 with what should be according to you, 4"?    If you are using the meter scale, wouldn't you expect something close to 100mm?   With it being so square, it looks more like you have several feet, not inches.    My guess is there's a bit of hand waving going on.

You would need to define what you mean when you talk about resolution.   I suspect that you are looking at that graph you asked about in the upper right and as you decrease the number of data points, you start to zoom in.  So you feel this is somehow increasing the resolution.  That's just my guess.  If that is the case, I have another VERY simple experiment you could run that may help you better understand what you are seeing.

Sorry I created confusion.

The attached  plot is from my home-made mis-match consisting of 1-four foot coax in series with 2 parallel four foot cables in series with a four foot coax, all RG-58.  Reading should be 48 inches.  This plot has 300 points.

In the last sentence I state the reason for running a low # of points.  My Beatty Standard is 4 inches which requires lower points,

I am interested in learning more about a VERy simple experiment.

I am still not sure what you mean when you talk about resolution and why does your 4" Beatty Standard require less points?

you are right - it does not.

I still don't know what you meant by resolution.   Why did you feel your 4" Beatty Standard would require less points?   What made you now change your mind?

the facts

Another basic document, filled with facts on TDR.

https://mpd.southwestmicrowave.com/wp-content/uploads/2018/07/Utilizing-Time-Domain-TDR-Test-Methods-For-Maximizing-Microwave-Board-Performance.pdf

Thanks for the great reference - had not seen that one.

You asked about my resolution obsession - - what I was really thinking was getting a plot with square-wave type transitions at the mismatch points like I got with my home-made mismatch consisting of 4 each 4-foot Pomona RG-58 coax with BNC connectors.  The 51-inch result with 48-inch cables may be accounted for by considering the two BNC-T, one at each end of the center 25-ohm section.

I am enclosing TDR plots of terminated Beatty Standard comparing versions 2.04 and 2.07.  Same HP 85051, different results.  I suppose I will not see right-angle traces at the mis-match point unless I use a 50-GHz, $50K VNA.  If I show you my Beatty Standard plots, would you show me yours ?

So your idea of resolution was what I would have expected.   Read the following from the paper I linked for you:

*The resolution of a TDR is frequency dependant; the higher the frequency the more resolution. Resolution can also be compromised if there are not enough data points in the sweep.*

>
> I suppose I will not see right-angle traces at the mis-match point unless
> I use a 50-GHz, $50K VNA.

Which is why I mentioned the 8720 series to you some time ago and why I told you my software would not overcome this problem.   These can be found at a fairly low cost.

As I mentioned, I suspected you were looking at the graph in the upper right and seeing it zooming in as you reduced the number of data points and this led you to the idea that less points meant higher resolution.  That was my guess, you never answered.   The fact of the matter is that this graph defaults to auto scale is all.   I recommend you learn to turn these features on and off.

Indeed, the newer software's TDR calculations are much different than the older software.   BTW, you are running old software which is known to have a few problems.  The change to the TDR was explained during this video.  It also shows my home made 70mm air-line Beatty standard.

https://youtu.be/dI2LjW_joXA?t=525

This video shows the old software using point slope to look at another crude stepped transmission line.
https://youtu.be/XaYBpPCo1qk?t=3296

*joe smith wrote:
*
BTW, you are running old software which is known to have a few problems.

*john galbreath reply:*

Screen shots are from v2.04 and v2.07.  Improved version 2.07 has different distance values compared to version 2.04.  Is there a newer version I should download ?

You should always run the latest software.   I also suggest staying up on the videos and manual, otherwise you are back to what ever you used before deciding to base your choices on facts.

The video clearly talks about the point slope vs using S-parameters and the need to perform the SOL cal with the new software.  We can see you chose not to calibrate.  Personally, I would have calibrated it before asking why things don't work.  I would have no way of knowing how your VNA performs without calibration.  Maybe you can get away with it.  PERFORM an SOL CALIBRATION!!   Again, as the video covers, SOL calibration is not used with the older software (2.04) as I am using point slope based on a known Beatty standard.

The videos show tools that are built into the software that may have helped you/me understand where the problem is but you chose not to use them.

You measure 38mm and 77mm on a 10cm standard. You are comparing two bad measurements and asking why they are different from one another.  That's a whole special kind of problem.    For fun, here is my home made 70mm airline being checked with my V2Plus4 using both point-slope and S-parameters.   For the S-parameter, no calibration was performed (just like you).  My V2Plus4 can put up some decent data without it in some cases.   I then made the same measurement with the 2.04 software.  Notice, there's less than a mm of difference.  Had I taken the time to calibrate, these would tighten up.  Still both are reading a few mm longer.  Using 3GHz is clearly not the best choice.

Funny how you have moved from 18 pts to 600 for a 10cm standard.  There's no need for it but that's fine.   Also shown is the same Beatty standard attached to one of my vintage VNAs using 401 points.  This software uses S-parameters to calculate Z (like the new software).  The VNA was calibrated before taking this measurement.  Notice how it reads a little long.  This standard is made from some hobby brass.  It's an airline but  the connectors are not.  The brass sleeve is soldered on and so there is not a clean transition in the step.  It's not a metrology grade standard like what you are playing around with and I am not at all surprised it comes in a bit long.

As you can see, all three readings are within a couple of mm.   You're using a 100mm metrology grade standard and are 30mm off.  Don't you think there's a problem? Like usual, you leave out a lot of details.  For all I know, you could have slapped that standard on a 10' hunk of RG58u.   My guess, the problem is still between the keyboard and chair.  I would expect you to be well within 2mm, not 30mm, certainly not 62mm!!

Thanks for your informative reply - I have a lot to do.

My NanoVNA clone has an upper limit of 3 GHz - My long-range plan is to acquire the second version of V3, and do more study.

If I go ahead and calibrate the V2Plus4, you can see the plot now looks very similar to my vintage Agilent.

At 77.015 mm on a 10 cm metrology grade standard,  you have a 23% error and that's your good measurement.   Consider I have ran my home made standard on two different VNAs, using two different sets of standards, using two different techniques to derive Z and not even calibrating the V2+4 in one case.

Max: 73.352 mm
Min: 70.629 mm
Mean: 71.97 mm
P-P: 2.723 mm
Std Dev: 1.26 mm

If we assume my home made standard is 70.000000 mm (we know it's not), the error is still under 5% even using the point slope.

With the 2.04 software, you did not need to run the SOL.  So something else is going on to cause a 23% error.  Maybe you are using some of those 50 cent connector kits, don't believe in torque wrenches, like cheap cables,  Beatty standard was something from eBay that was dropped on the floor.   Maybe a bad VNA?  Never underestimate the human factor.

Odd, it stripped my attachment.  Let's try again.

Is your version for the Agilent available anywhere? I have an 8753E 75 Ohm that might benefit.

On Thu, Nov 11, 2021 at 02:05 AM, Joe Smith wrote:

>
> So your idea of resolution was what I would have expected.   Read the
> following from the paper I linked for you:
>
> *The resolution of a TDR is frequency dependant; the higher the frequency
> the more resolution. Resolution can also be compromised if there are not
> enough data points in the sweep.*
>
>> “I suppose I will not see right-angle traces at the mis-match point unless
>> I use a 50-GHz, $50K VNA.”
>
> “Which is why I mentioned the 8720 series to you some time ago and why I
> told you my software would not overcome this problem.   These can be found
> at a fairly low cost.  “

Dear all,

Might I just mention that second hand 8270- series analysers still cost about two orders of magnitude more than the device this forum is intended for.

I do like reading about some of the more extreme uses of the nano devices, even if I cannot understand all the details, and I am unlikely to ever be attempting such advanced measurements.

Steve L. G7PSZ

On Fri, Nov 12, 2021 at 12:12 PM, demianm_1 wrote:

>
> Is your version for the Agilent available anywhere? I have an 8753E 75 Ohm
> that might benefit.

10 years ago I would have offered to port it for the cost of an 8753E and accessories.  However, there was a person who had told me about software they had written for the 8753.   You may want to ask about it in some of the mail lists and forums.

On Fri, Nov 12, 2021 at 05:34 PM, Stephen Laurence wrote:

>
> Might I just mention that second hand 8270- series analysers still cost
> about two orders of magnitude more than the device this forum is intended
> for.
>
> I do like reading about some of the more extreme uses of the nano devices,
> even if I cannot understand all the details, and I am unlikely to ever be
> attempting such advanced measurements.
>

While it's mostly been focused on the V2,  as the OP is slowing starting to understand, the rise time they are interested in is directly tied to frequency not data points.  I see no reason to limit the discussion because of cost.  I see these low cost VNAs as a great training tool and eventually a few of us will require something better.

The V2Plus4 currently lists for $200 on Tindie (price has gone way up).   2 orders or about $20,000 USD seems a bit high.   They are commonly found for under $6,000.   I'm not aware of other devices regularly in this price range.  Still you would need to consider the accessories which could cost you more than the VNA.    I doubt that's any big surprise.   Price out a new 20GHz system from Keysight.