开云体育

On Fri, Nov 15, 2019 at 11:00 PM, QRP RX wrote:

" You can use complex pieces of cables with different impedance for testing it.

For example: 1 meter 50 ohm + 1 meter 75 ohm + 1 meter 50 ohm + 25 ohm load. "

-------------------------------------------------------------------------------------------------------

QRP,
A number of people, including myself, have intentionally added mismatches along the length of the cable and made the same observation as you. I haven't tried to read your MATLAB code, I'm having enough trouble learning python and C# without adding MATLAB to the mix.

Thanks for linking to your code. Hopefully someone familiar with both languages will take up the challenge and convert it to python. Rune is overworked these days and could use some contributions from the rest of us.

Do you expect to translate the MATLAB code to C# and add it to your NanoVNA-Sharp MOD?

- Herb

On Fri, Nov 15, 2019 at 09:34 PM, Gabriel Tenma White wrote:

" … It's the absolute accuracy and confidence in the measurements that you care about, and if you look at the smith chart that hugen posted, it looks the same as the image erik posted earlier in his "chasing ghosts" thread, namely big unexplained dips in S11 above 600MHz that is unphysical when measuring a stub …."
-------------------------------------------------------------------------------------------------

Thanks Gabriel. Its much easier for me to read the logmag plots and it takes more of a trained eye, like yours, to read the smith chart. hugen commented in the attached copy of the smith chart that reliable operation is limited to 1300 MHz, I suppose that is related to your comments as values after that become exceedingly noisy and eventually exit the circle of the smith chart (loss of directivity?).

Most of my own measurements above 1 GHz are S21 through measurements. I would more than happy with 40 - 50 dB of head room above 1 GHz. When I first purchased the NanoVNA I was only expecting it to be useable to 300 MHz. All this additional measurement capability is icing on the cake for me.

Regards,

- Herb

You can use complex pieces of cables with different impedance for testing it.

For example: 1 meter 50 ohm + 1 meter 75 ohm + 1 meter 50 ohm + 25 ohm load.

The problem with this IFFT approach is that result looks like truth, but it is not.
If you test it on different datasets, you will find incorrect results.

On Fri, Nov 15, 2019 at 11:01 PM, <erik@...> wrote:

extend the s11 with the the complex conjugate of the s11 as input to the fft

conjugate also cannot help. This is wrong approach.
You're needs to interpolate S11, then use interpolation for integration step response.
In such way it works ok. See my matlab script.

On Sat, Nov 16, 2019 at 12:41 AM, hwalker wrote:

I haven't actually looked at Rune's Python code but I believe he credits
nuclear_rambo for the tdr and the step impedance functions. Nuclear_rambo
calculates the initial impedance step correctly but then sign information is
lost for additional impedance steps along the length of the cable.

This approach is incorrect and leads to wrong results. It will give you even worse results with signed value.
I wanted to add this into NanoVNASharp MOD, but it shows incorrect results and there is no way to fix it.
Because approach is incorrect.

You can find correct approach in my MATLAB scripts:

This is because SWR meters have high measurement error.

@Ron, K7UV: When you lay your hand on the SWR meter enclosure, do the meter needles move?

Hi Ron,
Just to understand:
You are disconnecting the cable from the back of the radio and connecting
it to the vna? Is that correct? You are not using a different jumper or
removing any of the existing items that are inline when connected to the
radio, correct?

If you are, that might change what you are reading.

Thanks,

Marco

On Fri, Nov 15, 2019, 9:38 PM Ron - An Old Ham in Utah <k7uv@...>
wrote:

I'm using an external full-range tuner with a straight through (direct
without using tuner function) reading to use the cross-needle SWR/PWR
reading, so no, I'm not using the rig meter for SWR measurements. The
frequency for lowest SWR as well as the actual SWR level are off. Comparing
the NANOVNA readings versus the SWR measurement show 1.3:1 on the nano
device and 2.6:1 with the SWR meter when I'm transmitting. Perhaps this
discrepancy is to be expected between devices and exceeds measurable
accuracy? I would think the nano device reading would be more accurate, but
while transmitting, the SWR is high enough to cut back the power level from
that with a lower SWR. Usually, my transmitter self-protects when the SWR
is 2.0:1 or higher. Thanks and 73.

--
Ron, K7UV

I'm using an external full-range tuner with a straight through (direct without using tuner function) reading to use the cross-needle SWR/PWR reading, so no, I'm not using the rig meter for SWR measurements. The frequency for lowest SWR as well as the actual SWR level are off. Comparing the NANOVNA readings versus the SWR measurement show 1.3:1 on the nano device and 2.6:1 with the SWR meter when I'm transmitting. Perhaps this discrepancy is to be expected between devices and exceeds measurable accuracy? I would think the nano device reading would be more accurate, but while transmitting, the SWR is high enough to cut back the power level from that with a lower SWR. Usually, my transmitter self-protects when the SWR is 2.0:1 or higher. Thanks and 73.

--
Ron, K7UV

On Wed, Nov 13, 2019 at 07:16 AM, Larry Rothman wrote:

"There is an interesting discussion on hugen79's NanoVNA Issues page on Github

regarding circuit changes/mods to reduce overall noise levels."

-------------------------------------------------------------------------------

Larry,
Hugen posted a chart a few hours ago showing the results of a new PCB with some of the recent improvements installed (see attachment). S11 looks to be <= 40 dB to 1500 MHz and S21 dynamic range is >= 50 dB to 1500 MHz.

- Herb

On Fri, Nov 15, 2019 at 06:33 AM, Kurt Poulsen wrote:

" I have done a lot of such measurements in the past and eliminated the inductance of the wire thru the core.

In your setup I think you get even better result by doing a SOL calibration at the end of the BNC adaptor using the Shorting wire thru the core without the core as short and place a leaded 50 ohm resistor as load and nothing for the open. The you get eliminated the "funny" impedances of the adaptors in the calibration process "

-------------------------------------------------------------------------------------------------------------------------------
Kurt,
The one turn loop gave me a method of sorting my ferrites by visual signature, but I could never get a good SOL calibration, per your suggestion, that extended beyond a few MHz.

Since most of my unidentified ferrites are clamp-on style, I decided to "go down the rabbit hole" and try to build a fixture using some of the references you provided and material that I had on hand.

Attached is a photo of the fixture I built. I tried to construct a coaxial type fixture where I could easily get access to the center conductor. I performed an OSL calibration from 1 - 500 MHz with the fixture cover in place and no ferrite installed.

I made a reference plot with the fixture terminated in 50 ohms. The S11 impedance plot showed a straight 50 ohms with a slight bump at the 300 MHz band switch over point (see red trace in attached plot).

I then replaced the 50 ohm load with the short standard used during calibration, installed a Fair-rite 0443164151 ferrite on the inner conductor (see attached photo), replaced the fixture cover and made a test run. The S11 impedance plot showed a rise in impedance with at least 100 ohms of impedance between ~17 MHz and 400 MHz (See black trace in attached plot) .

The slight bump at 300 MHz during the reference run turned into huge outlier during the test run. I would typically remove it from my data set. The typical impedance data in the chart provided by Fair-rite does not closely match the data I obtained (see embedded chart data).

I discovered that the coaxial fixture could also be used to measure the transfer impedance of an rf current probe that I had constructed from ferrites, so there was a secondary benefit to its construction.

- Herb

Thanks. I'm using a cross-needle meter without any tuner, so I'm matching
the same antenna as measured with the Nano.

On Fri, Nov 15, 2019 at 01:01 PM, <erik@...> wrote:
" Rune,

extend the s11 with the the complex conjugate of the s11 as input to the fft

Then you can use only the real values as impulse response with the sign being correct.

All imag values will be zero
No need to apply any windowing so you do not lose any information."

--------------------------------------------------------------------------------------------------------------

Eric,
I haven't actually looked at Rune's Python code but I believe he credits nuclear_rambo for the tdr and the step impedance functions. Nuclear_rambo calculates the initial impedance step correctly but then sign information is lost for additional impedance steps along the length of the cable. I've examined his Python script before but don't have the background to suggest corrective changes to the code. The script is as follows:

# Create step waveform and compute step response
step = np.ones(NFFT)
step_response = np.convolve(td, step)
step_response_Z = Zo * (1 + step_response) / (1 - step_response)
step_response_Z = step_response_Z[:16384]

# Calculate maximum time axis
t_axis = np.linspace(0, 1 / cable.frequency.step, NFFT)
d_axis = constants.speed_of_light * _prop_speed * t_axis

# find the peak and distance
pk = np.max(td)
idx_pk = np.where(td == pk)[0]
print(d_axis[idx_pk[0]] / 2)

# Plot time response
fig, ax1 = plt.subplots()
ax2 = ax1.twinx()
ax2.set_ylim([0, 500])
ax2.yaxis.set_ticks(np.arange(0, 500, 50))
ax1.plot(d_axis, td, 'g-')
ax2.plot(d_axis, step_response_Z, 'r-')
ax1.set_xlabel("Distance (m)")
ax1.set_ylabel("Reflection Magnitude")
ax2.set_ylabel("Impedance (Ohms)")
ax1.set_title("Return loss Time domain")
plt.show()

- Herb

Never mind; I found them cheap on ebay.? Will order soon.
Bob

Herman,
Thank you very much for the information.

Marco

Good afternoon!

Depending on how everything in your rig is connected and/or grounded, switching from a rig that is connected to the powerline ground to the NanoVNA that is floating could change things so that both of your readings are correct. If that sounds possible, you could tie the shield of the transmission line to the ground/chassis/etc of your xmtr while the NanoVNA is connected. That would eliminate that as a possible issue.

Just a thought...

Tom
AE5I

Rune,
extend the s11 with the the complex conjugate of the s11 as input to the fft
Then you can use only the real values as impulse response with the sign being correct.
All imag values will be zero
No need to apply any windowing so you do not lose any information.




--
NanoVNA Wiki: /g/nanovna-users/wiki/home
NanoVNA Files: /g/nanovna-users/files
Erik, PD0EK

Rune,

I think that the TDR function is missing a trick in computing the observed impedance along a line. It looks like you are integrating the return impulses to get the impedance values, but ignoring the sign of the impulses. As a result, the impedance value always goes up, even if the cable is shorted at the end. It would probably be a good idea to include polarity in the impulse plot anyway, since that can help show the nature of the discontinuities along the line.

--John Gord

I did measure both with another vna. Ch1 is indeed worse but I found both to be closer to 50 ohm. Currently traveling so don't access to the exact numbers
Next to the resistance difference ch1 has a relevant reactive component. Best to use a 10dB attenuator at ch1
Calibrate the thru with the 10dB to remove the calibration impact
--
NanoVNA Wiki: /g/nanovna-users/wiki/home
NanoVNA Files: /g/nanovna-users/files
Erik, PD0EK