开云体育

The smith chart shows around the centre but you've got nothing connected to port 0. It should be at the right side if port 0 is open??
Are you sure your 50 ohm calibration part is good?


On Sat, 5 Sep 2020 at 9:17 PM, Vince Rooney<vince@...> wrote: Hi Larry,

Yea that’s the first thing I did. Ran through it a couple of times. The screen calibration seems to be a little wonky now too.

Is there such a thing as a hard reset? A real switch somewhere, not the screen? Maybe remove battery and reinstall?
My nanovna-h with ver 0.4.5-1 gfbbceca
was working ok, occasional hang but on off had cleared it before.
Now what happens is there is a very long boot Eventually the screen comes up and I get the display,marker, stim, etc bar on right but it is inconsistent responding to screen touches, one time it got me into cal mode but would not take any input.
maybe time for a firmware reload? I have not done that at all. I've been using it just like it came from the factory. build date dec 26, 2019.

on the left I have
c4
D
R
S

Hi Larry,

Yea that’s the first thing I did. Ran through it a couple of times. The screen calibration seems to be a little wonky now too.

That should be a stable release. Did you press the Reset menu button immediately before calibration?


On Sat, 5 Sep 2020 at 8:40 PM, Vince Rooney<vince@...> wrote: Good Evening Everyone,

I just uploaded the latest release from 6-19-2020 (2020-06-19T23:08:21.000Z 9f940ce 202) and the traces are jumping all over the place. I tried recalibrating it and it didn't make a difference. Is there a recommended firmware (stable) release that has been burned and tested that everyone is happy with? I think this release is too much for my unit?

Thanks,
Vince

Good Evening Everyone,

I just uploaded the latest release from 6-19-2020 (2020-06-19T23:08:21.000Z 9f940ce 202) and the traces are jumping all over the place. I tried recalibrating it and it didn't make a difference. Is there a recommended firmware (stable) release that has been burned and tested that everyone is happy with? I think this release is too much for my unit?

Thanks,
Vince

In that respect, who remembers the articulated test ports for use with the
8510 VNA (now lusted after, discontinued, but still expensive - originally
over $100k). Those were EXPENSIVE "cables"!!!!!!

Dave - W?LEV

--
*Dave - W?LEV*
*Just Let Darwin Work*

Clearly the negative real part in this case is in error.

As has been pointed out, calibration error is a very likely
culprit.

Here's one form of calibration error that does not seem
to be widely known: If, while doing the OSL part of the
calibration, the electrical delay in the test cable changes
a bit between when measuring the short and measuring
the open, wild things can happen that lead to measurements
on actual passive devices falsely showing negative real
part of impedance. And it doesn't take much. In my early
days of VNA usage I tended to use cheap test cables (RG58
for example), and the electrical length change of that cable
during handling was sufficient to cause the problem.
Also, if the standards definitions for the open and/or short
are off a small amount from what your CAL kit thinks, the
same thing happens.

Making good standards and stable cables are both difficult
matters, which is why these items tend to be so expensive.

Dana

I not see any reason use in SAA2 1024 not calibrated point mode for CPU send (this need a lot of ram)
I test speed on H4 (on V2 result should be some) (thanks OneOfEleven for add points/sec measure):
1 segment 401 uncalibrated point - 570 points/sec
2 segments 802 uncalibrated point - 560 points/sec
4 segments 1601 uncalibrated point - 560 points/sec
8 segments 3201 uncalibrated point - 560 points/sec
This points/sec points received from device (in device speed more beeg, not need send data, i measure 401 points on CH0 and CH1 in 620ms so speed ~650 points/second, for only one channel ~930)

So no matter how many you measure points in one time (yes if enable calibration in device 1/3 time lost on calc interpolation for segment, but V2 not do this) in segment scan

So 1024 points not give result, but grab a lot of memory, this memory can be used for increase device points count

On Sat, Sep 5, 2020 at 07:51 AM, OneOfEleven wrote:

Looks nice!

Does it use the same serial commands as the v1 I wonder ?
========================================================

OneOfEleven,
Since the new changes are only in display size and packaging, it brings along with it the same V2 serial commands - meaning it is not backwards compatible with programs written for the NanoVNA and NanoVNA-H4.

- Herb

On 9/5/20 8:13 AM, Jerry Gaffke via groups.io wrote:

Agreed that his values are very close to 1.0
and that re-calibrating may not solve it.
The relationship between S11 and the network impedance ZL is:
S11 = (ZL-Z0)/(ZL+Z0)
where Z0=50ohms resistive, S11 and ZL are both complex numbers.
For an S11 magnitude greater than 1.0, ZL must be either less than zero or greater than infinity,
and it's understandable that nanovna-saver decides not to report an impedance in this case.
But it probably should report something.
Would be good if somebody could write up the various causes of uncertainty
in these measurements and give some idea of how to estimate it.
But that's not an easy task.
And not something that is typically displayed with every measurement,
certainly not on any instrument that I own.

Unfortunately, for VNAs, the uncertainty of the displayed number depends on not only the value being measured, but also the calibration uncertainty. It's a non-trivial thing to write a simple guide for.

For the hard core, there are software packages that calculate it, propagating the uncertainty the calibration measurement, through the equations, and combining that with the uncertainty of the raw measurement (typically related to the SNR).





A simple example with ohm meters...

Imagine you have a box that puts a current through an unknown resistor and measures the voltage across the resistor. You don't know the current, and all you know is that the voltage measurement has 1% error.

So you have a 1% error resistor that's say, 100 ohms, and another that's 1000 ohms.

You measure each of them and get, say, 1.15 volts and 9.9 volts. From that you can get a scale factor and an "uncertain" estimate of the current (9.57 mA) from the slope, and other uncertain estimates of 11.5 mA (constraining it to go through zero, with the 100 ohm test R), or 9.9 mA (constraining it to go through zero, with the 1000 ohm test R). The latter two have an uncertainty of 1% in the resistance and 1% in the measurement. The first slope cal has an uncertainty of 1% for each resistor plus 1% for each measurement.

Which one do you choose? Well, you can check zero on the voltmeter by shorting or opening the input (or both). Maybe your 100 ohm resistor was high maybe the 1k was low? I'd use the 1k, just for grins. So I'm going to call the measurement current 9.9 mA +/- 1.4% (1 sigma) assuming the volt meter error is random and uncorrelated with the resistor value uncertainty, so I can "root sum square" combine them: 1.4 = sqrt(1^2 + 1^2)



Now you measure your unknown resistor (say it's around 470 ohms). You get a voltage measurement of 4.67 V, and that has a uncertainty of 1% as well. Let's use 9.9 as the "cal factor, so 4.67V works out to 472 ohms, but the uncertainty is (RSS again) 1.7 % = sqrt(1.4^2 + 1^2)

That's with one end of the calibration pinned to zero. Let's say we had to use the slope calibration (maybe the voltmeter doesn't always go through zero) - so we solve for y=mx+b for the two measurements (100 ohm, 1.15V and 1000 ohm, 9.9V)

Now we have two uncertain estimates of m and b (for this example, they are 0.0097 and 0.1778, with uncertainties of approximately 1.5% and 15% repectively)

So we make a voltage measurement of our unknown and get 4.74. We convert it to resistance by the formula:

R = (V-b)/m and we get 469.7 (pretty close), but the uncertainty is much larger:

Breaking it down.. the voltage measurement has an uncertainty of 1%, so 4.74V has an uncertainty of 0.0474 V. b has a huge uncertainty of 15% (0.0194) but the absolute value is small. Assuming they're uncorrelated we can RSS so uncertainty of V-b is 0.0541V (or 1.17%)

Now we have to divide by m (0.0098) which has a 1.9% uncertainty - when working with multiplication and division, you work with percentages, not absolute.
Uncertainty = sqrt(1.17^2 + 1.5^2) which is about 1.9% (that dominates over the 1.17)

So your resistance measurement is 469.7 +/- 8.9(1 sigma) ohms.


For a VNA, the calculation is a LOT more complex than just a linear slope/intercept, so the propagation of the uncertainties is more complex.

The measured reflection magnitude is greater than 1.0 by 0.002 at most. To put this into perspective, on a Keysight PNA, in the best case (calibrated with an ecal), the reflection magnitude uncertainty is specified to be ±0.02 (@50MHz and S11=1.0).
( , p.23)

I am working on a measurement model for the SAA2 based on the EURAMET VNA guide in order to characterize the uncertainties (). I still have some issues to solve though, mainly because the single-receiver architecture potentially introduces additional error sources not usually considered. But, if you use the included cal-kit (particularly if you don't calibrate out the cables), I think it is safe to say that this will be the largest error source unless you get down into the noise.

It contains much more than just HPAK information. Much, much more. Navigate up a couple of directory levels and one will see directories for all kinds of test equipment and other electronic topics.

DaveD

This is dedicated to ALL HP. Quite an excellent source. Looks like it has test gear and manuals dated back to the 50 and 60's. As well many of the bench briefs. Excellent source for the repair of hp boat anchors. Good to bookmark.

Alan

From the looks of it that's a site about HP and Agilent test equipment. Does it have NanoVNA information also? If not, I would inquire about it in a group dedicated to those brands of test equipment.

Agreed that his values are very close to 1.0
and that re-calibrating may not solve it.

The relationship between S11 and the network impedance ZL is:
S11 = (ZL-Z0)/(ZL+Z0)
where Z0=50ohms resistive, S11 and ZL are both complex numbers.
For an S11 magnitude greater than 1.0, ZL must be either less than zero or greater than infinity,
and it's understandable that nanovna-saver decides not to report an impedance in this case.
But it probably should report something.

Would be good if somebody could write up the various causes of uncertainty
in these measurements and give some idea of how to estimate it.
But that's not an easy task.
And not something that is typically displayed with every measurement,
certainly not on any instrument that I own.

Jerry, KE7ER

Hi,

In the Calibration menu of the NanoVna-H4? I have to start with the Open.
When I use the Calibration Assistant of the NanoVna Saver 0.3.6-rc2 to calibrate the NanoVna-H4 then I have to start with the SHORT.
Perhaps for consideration if it should be useful to change the order in the Calibration Assistant, choosing for the Open first also?
This is to avoid taking the wrong Cal component by accident if you was used to the order of the NanoVna-H4 Cal menu.

Kees, PE0CWK

Looks nice!

Does it use the same serial commands as the v1 I wonder ?

One trick with this kind of measurement (where the reflection coefficient is very close to 1, because the DUT is close to being a short) is to put a resistor in series with the inductor you are measuring. Then subtract the resistor from the measured R+jX


As noted by an earlier poster, what you might be seeing here is a numerical precision effect when the NanoVNA is converting from raw measurements to your mag/phase.

I think you're pretty good. It shows under 2:1 on all bands.

73, Zack W9SZ

On Fri, Sep 4, 2020 at 11:59 PM Darrell Carothers <rescuemedic1@...>
wrote:

Working with my H4 Nano. Just put up a random wire antenna. Trying to see
what frequencies it is resonant on. If I understand low is good. Can some
look at the attached photos and see if I am doing something wrong or if I
am good. Thanks Note the different frequencies at the bottom of the screen.





























Sent from my over-rated IPhone 7 Plus. Any Mis-spellings or grammar
errors are due to my IPhone auto correct feature.

<>
Virus-free.
www.avast.com
<>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

On Sat, Sep 5, 2020 at 04:08 AM, amirb wrote:

Hi Hugen
can you show the exact hardware inside this new unit? I mean the schematic and a picture of the board?
the original SAA-2, which has a 2.8" display, has a completely different hardware/schematic than all other nanovna.
=============================================
As hugen posted, there have been no changes to the original SAA-2 design, only layout changes to accommodate the addition of a 4" display. The schematics are the same as the original open source project.

I just received the SAA-2N that I ordered and will be posting a review on the V2 user group after I have verified its performance. So far the un-boxing experience has been positive.

- Herb