开云体育

I am confused. A friend of mine and I are trying to figure out why *my
nano H4 and his very early nano vna with the 2.8" screen will both run on
his nano saver 0.3.7.* His 2.8 nano runs on my computer on saver 0.3.8 but
mine does not. My w10 system complains about

*Error during sweep*

*Stopped*

*Failed to read data 1 10 times*
*Data outside expected valid ranges, or in unexpected format.*

*You can disable validation on the device settings screen.*

This just happened in the last couple of days. I have both 0.3.7 and 0.3.8
on my W10 system and this problem exists on both saver versions.

Help
Fred - N4CLA

On 6/26/23 11:12 AM, Fran?ois wrote:

It's possible, that with some active systems, you could get a reflection
that is bigger than the incident wave (i.e. if the Zload were negative).
But I think that for an entirely passive load, the reflected wave
cannot be greater than the incident wave.

We have a little trouble understanding each other. It is not a question of measurement but of calculation.
I have an adapter which for a given input impedance returns the impedance
Zout = 1.96593624236642 -j 5601.37911926951)}
The impedance I'm targeting is
Ztarget = 50 +j 100
the results are recalculated in the Excel sheet attached
S11 = 1.03617959022359 -j 0.0185190120640808
SWR = -56.028101163433028 negative
|S11| = 1.036345067 -> |S11| > 1

Indeed..
For instance if Z1 = 2 - 1000j and Z2 is 50 + 100j, then
gamma = (Z1-Z2)/(Z1+Z2)
gamma = 1.215 - 0.124j

|gamma| > 1 (and it's not some sort of numerical precision thing)

OK, we must have some interpretation problem. And it's not obviously "unphysical" - I can see a source with Z 50 + 100 j (A series inductor) and a load of 2-1000j (a series capacitor). Sure, it's not a conjugate match, but obviously power will be dissipated in the 50 ohm resistor.

So this must be a case where "reflection coefficient" as defined by impedances isn't correct. There's no real transmission line here, so I'm not sure that the concept of a "traveling wave" (where gamma is reflected/incident) is valid.

Here it is, thanks for taking a minute to look into it.

Or is the best path forward just to
remove the outliers manually when processing the data.

can you send as attachment your .s2p file

I wrote a little software to smooth out the errors. I would see what happens.
--
F1AMM
Fran?ois

De la part de astech119
Envoyé : lundi 26 juin 2023 18:32

It's possible, that with some active systems, you could get a reflection
that is bigger than the incident wave (i.e. if the Zload were negative).
But I think that for an entirely passive load, the reflected wave
cannot be greater than the incident wave.

We have a little trouble understanding each other. It is not a question of measurement but of calculation.

I have an adapter which for a given input impedance returns the impedance

Zout = 1.96593624236642 -j 5601.37911926951)}

The impedance I'm targeting is
Ztarget = 50 +j 100

the results are recalculated in the Excel sheet attached

S11 = 1.03617959022359 -j 0.0185190120640808
SWR = -56.028101163433028 negative

|S11| = 1.036345067 -> |S11| > 1
--
F1AMM
Fran?ois

De la part de Jim Lux
lundi 26 juin 2023 16:58

I'm measuring the group delay of some 60in cables with a NanoVNAH4 and I am getting outliers in the data. I am using nano-vna saver and I have tried both with on board calibration disabled and calibrated from 50kHz to 1.5GHz, the outliers are present still. I also have another H4 and I had the same issue just at a lower frequency ~150MHz. In the attached plot the points go to about +- 20 ns at just under 700MHz. I have also changed the sweep to have a lower df width but the same outliers were there at the same frequency. My sweep here is 1Mhz to 701Mhz 1212 pts 3avg df=578kHz

Is there any explanation or way to minimize this? Or is the best path forward just to remove the outliers manually when processing the data.

Thanks!

On Mon, Jun 26, 2023 at 01:23 AM, Mike wrote:

This is my test fixture, calibrated at the croc clips. It's not ideal but I'm
limited by the length of the coil. Should be OK at my measurement frequency of
70kHz though!

Calibrating with the crocodile clips will not give you a good reference plane. The reason I say this is if you keep them the same distance apart when you cal with an open, short and load you will have considerable inductance in the short and 50 ohm "cal loads". If you calibrate with the clips close together and then spread them the reference plane has changed. Neither is a good option.

I suggest you calibrate right at the screw terminations on the green block with the alligator clip leads removed. Then attach the leads and make your measurement. From the photo it looks like the leads are about 4" long and each one will add about 100 nH of inductance (total 200 nH or 0.2 uH). That extra .2 uH when you are measuring 110 uH is not significant. However you should get a better estimate of the SRF and be able to calculate the parasitic capacitance to more accuracy.

Try it and see what you find...

Roger

On 6/26/23 7:26 AM, Fran?ois wrote:

I knew about the phone lines. My problem is that if we calculate the S11 with complex values, it leads, for example, to a ROS which can be negative. S11 is no longer in a circle of radius 1.
Is this normal or am I mistaken?

It's possible, that with some active systems, you could get a reflection that is bigger than the incident wave (i.e. if the Zload were negative). But I think that for an entirely passive load, the reflected wave cannot be greater than the incident wave.

My study (amateur) concerns the cascading of 2 adapters in 'L' one high pass and the other low pass to carry out two adaptations of impedance at two different frequencies.

To generalize the brought back impedance can be different for the two frequencies and complex.

It works easily with a purely real impedance but when the impedance is complex, I tried to optimize the ROS by looking for its minimum value which I thought was ONE it does not work because the ROS varies from - infinity to + infinity.

I had to change the search criteria to subtraction
double optimizer = Complex.Subtract(Zouth_, Application.Zcibleh).Magnitude;

And now it works. What is curious is that when the solution adapter is found, the ROS is indeed equal to ONE. A crazy story.
--
F1AMM
Fran?ois

I knew about the phone lines. My problem is that if we calculate the S11 with complex values, it leads, for example, to a ROS which can be negative. S11 is no longer in a circle of radius 1.

Is this normal or am I mistaken?
--
F1AMM
Fran?ois

On 6/26/23 06:32, Jim Lux wrote:

On 6/26/23 2:22 AM, Fran?ois wrote:

Hello

In the attached formula (ROS.png), as long as Zo is real, even if Zl is complex, everything is fine and S11 remains in a circle of radius 1. and the ROS is between 1 and infinity.

If Zo is complex, this is no longer the case.

While writing C# code looking for a (complex) adaptation by dichotomy, I was looking for a ROS of 1. It doesn't work. I had to use the modulus of the difference between my target and the current value.

double optimiseur = Complex.Subtract(Zouth_, Application.Zcibleh).Magnitude;

Is it legitimate to talk about S11 when Zo is complex?

73

Sure it is: S11 is just a representation of the reflection coefficient on port 1 of the UUT.? Generally people use just the log magnitude in dB, but it has a phase, too.
Nothing prevents the Z0 being complex.

Definitely so. When working with telephone cable pairs at voice frequencies, Zo is always complex, generally with a phase angle approaching -45 degrees. Various graphical aids were published by the Bell System for solving transmission line problems with complex Zo, but today it's more common to simply solve the appropriate expressions using complex math in software.

Various schemes were developed for "loading" cable pairs for voice transmission, the most common being the addition of inductors at intervals of a few thousand feet. The most common loading scheme in the Bell System added 88 mH every 6000 feet, but there were quite a few other loading schemes in use. Loading a pair made it's Zo close to, but not exactly, real across the frequency span of the load scheme, reduced the loss of the pair within that frequency span, and acted as a sharp cutoff low-pass filter, reducing significantly transmission at frequencies above the cutoff frequency.

73,

Maynard
W6PAP

On 6/26/23 2:22 AM, Fran?ois wrote:

Hello
In the attached formula (ROS.png), as long as Zo is real, even if Zl is complex, everything is fine and S11 remains in a circle of radius 1. and the ROS is between 1 and infinity.
If Zo is complex, this is no longer the case.
While writing C# code looking for a (complex) adaptation by dichotomy, I was looking for a ROS of 1. It doesn't work. I had to use the modulus of the difference between my target and the current value.
double optimiseur = Complex.Subtract(Zouth_, Application.Zcibleh).Magnitude;
Is it legitimate to talk about S11 when Zo is complex?
73

Sure it is: S11 is just a representation of the reflection coefficient on port 1 of the UUT. Generally people use just the log magnitude in dB, but it has a phase, too.

Nothing prevents the Z0 being complex.

Hello

In the attached formula (ROS.png), as long as Zo is real, even if Zl is complex, everything is fine and S11 remains in a circle of radius 1. and the ROS is between 1 and infinity.

If Zo is complex, this is no longer the case.

While writing C# code looking for a (complex) adaptation by dichotomy, I was looking for a ROS of 1. It doesn't work. I had to use the modulus of the difference between my target and the current value.

double optimiseur = Complex.Subtract(Zouth_, Application.Zcibleh).Magnitude;

Is it legitimate to talk about S11 when Zo is complex?

73
--
F1AMM
Fran?ois

This is my test fixture, calibrated at the croc clips. It's not ideal but I'm limited by the length of the coil. Should be OK at my measurement frequency of 70kHz though!

--
Mike

Hey guys,

I have NO affiliation with this company but Battery Hookup has absolutely the best prices when it comes to the 18650 lipo cells. They also have flat cells with power levels of 4.5Ahr @ 3.7v

Mike C. Sand Mtn GA

On Sun, Jun 25, 2023 at 03:15 PM, Roger Need wrote:

The VNA can only measure R + jX or R//jX (with later firmware versions). It
calculates inductance by dividing X by 2*pi*frequency and this ONLY gives an
estimate of the true coil L if the coil is air wound and the frequency is low
enough that the skin effect is not having much effect on underlying
inductance.

Roger, the method I suggested requires inductance calculated this way. It yields an accurate coil model over a narrow frequency range. To my surprise, it seemed good enough over the whole 3.5-4 MHz band. But the wideband model suggested in the writeup noticeably improved accuracy over the somewhat wider 88-108 MHz band.

Brian

On Sun, Jun 25, 2023 at 11:04 PM, Roger Need wrote:

On Sun, Jun 25, 2023 at 02:27 PM, Mike wrote:

Thank you Roger. The coil is 95 turns of 0.9mm enamelled copper wire close
wound on a 36mm PVC former.

Mike,

Attached is an analysis of your coil using Coil64. The numbers are close to
what you measured. For an air wound coil the actual L will not vary much in
the frequency range of up to 12 MHz. Note the following:

--> ESR is increasing with frequency and simulation shows .299 ohms at DC,
.422 at 1 MHz. and 9.233 ohms at 5 MHz.
--> Self capacitance is calculated at 1.61 pF which is very small. You
estimated 2 pF based on your SRF measurement. Any stray capacitance in your
test setup will significantly affect your self resonant frequency so you need
a good test jig if this is an area of concern.

Note: For air wound coils assuming that apparent inductance at low frequencies
is equal to actual inductance L at higher frequencies is a reasonable
approximation. Therefore the method of calculating parasitic capacitance
based on using this value of L and the SRF to calculate parasitic capacitance
gives a decent estimate. BUT this method does not work if the inductor is a
ferrite core design.

Roger

Perfect! Thanks for that explanation Roger.

--
Mike

On Sun, Jun 25, 2023 at 02:55 PM, Brian Beezley wrote:

Mike, measure the inductance and series resistance at the model frequency. I
believe recent VNA firmware versions can provide these values directly.
Otherwise calculate them from R and X.

The VNA can only measure R + jX or R//jX (with later firmware versions). It calculates inductance by dividing X by 2*pi*frequency and this ONLY gives an estimate of the true coil L if the coil is air wound and the frequency is low enough that the skin effect is not having much effect on underlying inductance. If the coil is wound on a ferrite core you can't use this method to estimate L at higher frequencies.

Roger

On Sun, Jun 25, 2023 at 02:27 PM, Mike wrote:

Thank you Roger. The coil is 95 turns of 0.9mm enamelled copper wire close
wound on a 36mm PVC former.

Mike,

Attached is an analysis of your coil using Coil64. The numbers are close to what you measured. For an air wound coil the actual L will not vary much in the frequency range of up to 12 MHz. Note the following:

--> ESR is increasing with frequency and simulation shows .299 ohms at DC, .422 at 1 MHz. and 9.233 ohms at 5 MHz.
--> Self capacitance is calculated at 1.61 pF which is very small. You estimated 2 pF based on your SRF measurement. Any stray capacitance in your test setup will significantly affect your self resonant frequency so you need a good test jig if this is an area of concern.

Note: For air wound coils assuming that apparent inductance at low frequencies is equal to actual inductance L at higher frequencies is a reasonable approximation. Therefore the method of calculating parasitic capacitance based on using this value of L and the SRF to calculate parasitic capacitance gives a decent estimate. BUT this method does not work if the inductor is a ferrite core design.

Roger

On Sun, Jun 25, 2023 at 10:55 PM, Brian Beezley wrote:

Mike, measure the inductance and series resistance at the model frequency. I
believe recent VNA firmware versions can provide these values directly.
Otherwise calculate them from R and X. Then create a simple load with the
inductance and resistance in series. This works fine over a single ham band.
To create a wideband model, see this:



Brian

Thanks for the link Brian.

--
Mike