Re: Analyzing Noise versus Leakage on CH1
Erik, The FFT assumes that you have a perfect fit of the (co)sinus in the sample range.
At higher frequencies the 5kHz start to deviate so the samples may no
longer fit perfectly in the fft and you may get spectral bleeding.
This bleeding will vary with the error of the 5kHz signal
Can you apply a window to the input of the fft and see if that clears out
the spectral bleeding?
Spot on! I applied the blackman windowing function to the 1.29GHz output (the worst frequency), and the result is much improved - see the 10dB improvement between 48 samples and 480 samples in the figures below. The "spectral bleed" is replaced by a peak that is spread across just 2-3 fft "buckets". I ran a simple simulation and found that the spectral bleed I saw earlier could be reproduced with a frequency of 5020 Hz, a frequency error of just 20 Hz. It appears that at some output frequencies it's not possible to generate the LO frequency at exactly 5000Hz offset. This very nicely explains why I got widely varying results earlier - at those frequencies when the offset is exact, filtering over multiple frames lifts the peak above the noise floor and improves the dynamic range, and at other frequencies the increased filtering has no effect. We may now have ways to move this forward and get a genuine benefit from increased filtering. Two possible alternatives are: Option 1. Alter the si5351 frequency setting algorithm to guarantee the Ref and LO are exactly 5000Hz apart. This may mean having to adjust Ref slightly from that selected, but that should not be an issue for the user. Option 2. The dsp implements a windowing function. This option isn't as attractive as option 1, because I believe windowing degrades the noise margin improvement. Implementing the blackman, or many of the other windowing functions would be too onerous for this poor little microprocessor. However, I've simulated a simple triangle windowing function and it shows very similar results to blackman at the peak, which is the only area we're interested in. A triangle window would be easy to implement. Rgds, Dave [image: 1_1.29GHz_48.png] [image: 2_1.29GHz_480.png]
|
Re: RX-Port Input Impedance
To get a better CH1 input impedance I recommend to use a 10dB attenuator possibly with a 10dB amplifier when above 300MHz
|
Re: Analyzing Noise versus Leakage on CH1
Indeed, that is what I plan to do with my own build VNA using (almost) the same HW as the nanoVNA
Will let you know once done
|
Hi Rune, thank you very much for the improvements made to the NanoVNAsaver application. If not too demanding to accomplish, you would like to be able to define multiple segments, for which you can choose StartFreq StopFreq StepSize.
Your application makes nanoVNA a truly powerful tool.
Best Regards
Maurizio IZ1MDJ
|
Re: RX-Port Input Impedance
I measured the CH1 RL with a professional VNA and got this values:
100MHz 36dB
300MHz 28dB
500MHz 23dB
700MHz 20dB
900MHz 17dB
Measuring UHF filters would benefit from having a better termination impedance.
I tried to understand the CH1 input attenuator but i don't find any explanation to that topology. Using the SA612 input as single-ended, connecting IN_B to ground through a capacitor as indicated in the datasheet and simplifying the attenuator to a PI or just L arrangement could lead to a better RL with same loss. But i'm sure i'm missing something, as VNWA3 uses the same circuit although with different resistor values. It's input match is much better BTW. Maybe SA612 noise?
Carlos
|
Hi Tom,
good suggestion! I'll have a look, and there should be plenty of room for
it! :-)
--
Rune / 5Q5R
toggle quoted message
Show quoted text
On Mon, 23 Sep 2019 at 07:31, Tom VA7TA <tma.7ta@...> wrote: Hi Rune,
I have made a large number of routine impedance measurements using
your V0.0.10 exe binary under Windows 10 Pro 64b. I have not encountered
any stability problems and in general think it works wonderfully well -
thank you!
One part of the GUI that I didn't understand intuitively the first
few times I used it was the purpose of the "Segments" box. I changed it and
nothing seemed to happen. I of course noticed immediately that it is a
multiplier for the number of steps in the sweep once I changed it and then
happened to do a calibration which shows the number of steps in the sweep.
I decided I would like to make a suggestion that I hope is worthy of
consideration.
It occurred to me that it would be obvious to the user that the
segments setting increases the sweep frequency resolution if the number of
Hz per step could be shown. Additionally I think the provision of the
"Hz/Step" information would be useful in some situations. For example when
sweeping a network with narrow high Q responses the user may wish to select
a segments setting that provides a sufficiently small step size to ensure a
narrow peak or null is not missed without increasing the sweep time any
more than necessary. A second advantage for providing the "Hz/Step" data is
that it would provide immediate indication of how a change of the segments
setting changes the sweep resolution. This feedback would make the setting
more user intuitive I think.
Considering that the number of segments would never exceed two
digits and the number of Hz per step would never exceed 7 digits there
might be room to fit the two boxes on the existing segments line within the
GUI. If for example the "Segments" label were abbreviated to "Segs." with a
two digit box then possibly there would be room for another label something
like "Hz/Step" followed by a 7 (or 9 if 2 comma delimiters) digit box on
the same line. Just might be one of many possible ways to implement it
without impacting the existing GUI layout very much.
Thanks again for your great contribution - "NanoVNA-Saver" really
enhances the usefulness of my nanoVNA for what I intend to use it for!
Enjoy!
Tom
VA7TA
|
Hi Pete,
you don't need to reset the NanoVNA calibration - it probably even works
better if you don't, in that the data received from the NanoVNA will not be
rejected as mangled during data transmission if it "looks like" real data.
;-)
--
Rune / 5Q5R
toggle quoted message
Show quoted text
On Mon, 23 Sep 2019 at 03:19, Peter Mulhare <peter@...> wrote: Hello Rune,
Again thank you for your perseverance with the nanoVNA PC Software you are
bringing us all! It is very much appreciated.
However I have a question regarding Calibration using your software, I
presume that before I set about doing a Calibration in nanoVNA-Saver, and
reset it, I should also do a reset of the VNA Hardware itself? Or is this
taken care of by nanoVNA-Saver? As I still see the "0" come up alongside
the "C" in the left hand sidebar.
73's
Pete
ZL2iK
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf
Of Rune Broberg
Sent: Monday, 23 September 2019 07:37
To: [email protected]
Subject: Re: [nanovna-users] NanoVNA Saver
For those waiting for new binaries: I just published 0.0.11 :-)
This release brings averaging as a new feature: Press "Sweep settings" to
select between running a single sweep (of optionally multiple segments), a
continuous (live) sweeping of the range, or to average several sweeps from
the NanoVNA. You can configure how many averages to make, and optionally,
how many of the sample points to discard, based on which deviate the most
from the rest, and are least likely to contribute signal.
Also added is a new Resistance/Reactance chart, which shows both the R and
X component of R+jX. This brings the challenge of showing two traces for
the same data - added is therefore the option of picking a secondary colour
for sweeps under "Display settings".
There are further improvements to UI sizing, meaning the interface now fits
- tightly - on a 1366x768 screen, at least on Windows.
A few quality of live improvements made it in: Press escape in any of the
pop out windows to close them instantly. The calibration window now shows
more clearly when the source of calibration data is loading from a file,
and also the number of points loaded. A few crash bugs were fixed.
I look forward to hearing feedback from all of you!
--
Rune / 5Q5R
On Sun, 22 Sep 2019 at 17:14, Mario Vano <mvano@...> wrote:
On Sun, Sep 22, 2019 at 06:09 AM, Mario Vano wrote:
CORRECTION: (I mistyped)
- The multi-band dipole is my 20ft high 40/20 meter trap dipole that has
been augmented with parallel dipoles for 15 and 10. The feedline has a
toroid common mode filter.
I've been happily using the NanoVNA and various versions of "Saver" for
a lot
of projects lately, but some were just motivated by curiosity about the
device.
Unfortunately I have no other analyzer to compare to. In any case here are
some sweeps from earlier tests that might be useful for reference or
amusement. In most cases, they were done with my very early attempts at full
bandwidth calibration.
- The fm filter is a commercial product from the rtl-sdr.com people.
- The LW filter is a homebrewed design I use to augment VLF beacon hunting.
- The multiband dipole is my 20ft high 80/40 meter trap dipole that has
been
augmented with parallel dipoles for 15 and 10. The feedline has a toroid
common mode filer.
- The 6 meter antenna is an attic mounted dipole with a toroid common mode
filter.
I've also used happily used the device and "saver" to design and build
a
6
meter "squalo", but forgot to save the files from the latest testing run. I'm
currently working on a 400-700 mhz indor LPA design and plan to try to use the
device for relative pattern and gain testing.
Earlier tests with several types of 500-900mhz 1/4 wave mag-mount antennas and
TV rabbit ears showed useful relative gain results at 2-3 wavelength spacings.
73, AE0GL
|
Re: Firmware with arbitrary scan length
Hi Rune,
Thanks a lot for your work on NanoVNA-Saver. However it would be great to have averaging option in the standalone firmware too. It should be easy to do, i just need to find some time to get my hands on the code. Maybe next century :(
Carlos
|
Re: Analyzing Noise versus Leakage on CH1
Very interesting work, Erik. Maybe what we need is to use two separated Si5351 to avoid LO to stimulus leackage. They are cheap!
|
Here is the CH1 ADC input signal analyzed with a FFT for both 50MHz and 2.099GHz, both with cable connecting CH0 and CH1 (thru) and without any connection (isolation)
The phase noise of the SI5351 is visible in the 50MHz thru graph.
Dynamic range for 50MHz is 64dB due to leakage.
The signal level for 2.1GHz is 40dB down due to being the result of a 7/9 harmonic mixing.
The "leakage" for 2.1GHz is 10dB up compared to 50MHz
Net result is a loss of 50dB dynamic range leaving 14dB at 2.1GHz
|
Hi Rune,
I have made a large number of routine impedance measurements using your V0.0.10 exe binary under Windows 10 Pro 64b. I have not encountered any stability problems and in general think it works wonderfully well - thank you!
One part of the GUI that I didn't understand intuitively the first few times I used it was the purpose of the "Segments" box. I changed it and nothing seemed to happen. I of course noticed immediately that it is a multiplier for the number of steps in the sweep once I changed it and then happened to do a calibration which shows the number of steps in the sweep. I decided I would like to make a suggestion that I hope is worthy of consideration.
It occurred to me that it would be obvious to the user that the segments setting increases the sweep frequency resolution if the number of Hz per step could be shown. Additionally I think the provision of the "Hz/Step" information would be useful in some situations. For example when sweeping a network with narrow high Q responses the user may wish to select a segments setting that provides a sufficiently small step size to ensure a narrow peak or null is not missed without increasing the sweep time any more than necessary. A second advantage for providing the "Hz/Step" data is that it would provide immediate indication of how a change of the segments setting changes the sweep resolution. This feedback would make the setting more user intuitive I think.
Considering that the number of segments would never exceed two digits and the number of Hz per step would never exceed 7 digits there might be room to fit the two boxes on the existing segments line within the GUI. If for example the "Segments" label were abbreviated to "Segs." with a two digit box then possibly there would be room for another label something like "Hz/Step" followed by a 7 (or 9 if 2 comma delimiters) digit box on the same line. Just might be one of many possible ways to implement it without impacting the existing GUI layout very much.
Thanks again for your great contribution - "NanoVNA-Saver" really enhances the usefulness of my nanoVNA for what I intend to use it for!
Enjoy!
Tom
VA7TA
|
Re: Evaluating clamp on ferrite chokes
A |S21| measurement is reasonable. However, the signature for a lossy choke should also be apparent in a S11 measurement. So take a short length of #22 AWG enamel wire and tie it across the CH0 port of the VNA. The inductance measured if it is reasonable length of wire, say 100 nH should follow a CW trajectory on the chart, along the 50 ohm circle (blue). When the lossy ferrite material is added, I would expect an increase in L as well increase in series R. The result is an increase CW rotation over the same frequency swept range as well a shift of the circle contour to the right. See figure (red) . Try it, see what you get.
Alan
|
Re: Does anyone know how sensitive the nanovna is to electrostatic discharge?
Dave,
The minimum energy to cause damage has gone way down over the years as transistor geometries have shrunk, but I don't know specific values.
When I started doing chip design, the state of the art was with NMOS FET gates around 9 microns by 9 microns in minimum area. Present CMOS transistor gates (in high density digital circuits) are around 7 nanometers by 7 nanometers minimum. In area, that is more than a million times smaller.
RF devices are not normally that small, but you get the idea. Even in a fairly large device, gate oxide damage is usually concentrated in a small weak spot of the overall device, so it doesn't necessarily help a lot if the gate area is large. The usual plan in protecting sensitive circuits is to use things like (relatively) large junction devices to direct ESD currents to supply rails instead of FET gates. These junctions are reverse biased in normal operation, but they still can contribute to distortion problems.
--John
toggle quoted message
Show quoted text
On Sun, Sep 22, 2019 at 08:25 PM, Dr. David Kirkby from Kirkby Microwave Ltd wrote:
On Mon, 23 Sep 2019 at 02:36, johncharlesgord via Groups.Io <johngord=
[email protected]> wrote:
Dave,
50 ohms to ground provides some protection, but often not enough.
The standard human body model (HBM) for ESD testing is 1500 ohms in
series with 100pF. The 1500 ohms is supposed to represent the impedance of
the discharge path through the body, while the 100pF represents the
capacitance of the body including something like rubber soled shoes to a
grounded surface.
Even at a relatively low static voltage like 1500v, one ampere can
(briefly) flow, causing a drop of 50 volts across the 50 ohm resistor.
Small geometry semiconductors like those used in RF mixers can be damaged
by such a hit. Some situations can generate static voltages that are much
higher.
--John Gord
Thank you. I was unaware of that. Perhaps that¡¯s why my 8720 says no ESD on
the front panel. However, the test ports on that are no 50 ohns at DC, as
there are bias-T fitted too.
I assume that there must be some minimum amount of energy required to
damage a semiconductor.
Dave.
|
Re: Evaluating clamp on ferrite chokes
Coil coax around a ferrite and connect the outer shield for a S21 measurement. Loss in S21 = choking :) typical ferrite for line noise choking is #31 (also for top bands 160m - 80m). #43 does a great all-round job on higher bands. #52 is a newer kid on the block. Higher curie temperature, which translates to higher power rating + it's quite capable on all ham bands. All of this from perspective of an HF ham radio operator ;)
|
Evaluating clamp on ferrite chokes
Wondering if anyone has tried to use the nanoVNA to evaluate the properties of the clamp on ferrite chokes that are employed to reduce radio frequency interference (RFI) being emitted by wall wart switching supplies etc. I am not sure how one would do that...use some kind of standard pair of wires like an AC line or adapter cord and measure S11 (single end attached) or S21 (both ends attached) with and without the ferrite clamped on? A similar question would be how to evaluate a common mode choke made by coiling coax around a cylinder form of some kind.
|
Re: Does anyone know how sensitive the nanovna is to electrostatic discharge?
Dr. David Kirkby from Kirkby Microwave Ltd
On Mon, 23 Sep 2019 at 02:36, johncharlesgord via Groups.Io <johngord= [email protected]> wrote: Dave,
50 ohms to ground provides some protection, but often not enough.
The standard human body model (HBM) for ESD testing is 1500 ohms in
series with 100pF. The 1500 ohms is supposed to represent the impedance of
the discharge path through the body, while the 100pF represents the
capacitance of the body including something like rubber soled shoes to a
grounded surface.
Even at a relatively low static voltage like 1500v, one ampere can
(briefly) flow, causing a drop of 50 volts across the 50 ohm resistor.
Small geometry semiconductors like those used in RF mixers can be damaged
by such a hit. Some situations can generate static voltages that are much
higher.
--John Gord
Thank you. I was unaware of that. Perhaps that¡¯s why my 8720 says no ESD on the front panel. However, the test ports on that are no 50 ohns at DC, as there are bias-T fitted too. I assume that there must be some minimum amount of energy required to damage a semiconductor. 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
|
Re: errors of "error" models
Hello,
We just uploaded the currently available version of /F/L/O/S/S/ Maxima code:
as well as, the currently available version of its documentation:
for the Uncertainty Estimation of Full One-Port VNA Measurements.
Next to come : the currently available version of /F/L/O/S/S/ FORTRAN code.
Sincerely,
yin@pez@arg
3
|
Why is a rubidium standard at 10 kHz offset from the carrier at 10 MHz that noisy? Is the phase noise floor shown limited by the test set?
Regards, Alan
|
Re: Does anyone know how sensitive the nanovna is to electrostatic discharge?
Dave,
50 ohms to ground provides some protection, but often not enough.
The standard human body model (HBM) for ESD testing is 1500 ohms in series with 100pF. The 1500 ohms is supposed to represent the impedance of the discharge path through the body, while the 100pF represents the capacitance of the body including something like rubber soled shoes to a grounded surface.
Even at a relatively low static voltage like 1500v, one ampere can (briefly) flow, causing a drop of 50 volts across the 50 ohm resistor. Small geometry semiconductors like those used in RF mixers can be damaged by such a hit. Some situations can generate static voltages that are much higher.
--John Gord
toggle quoted message
Show quoted text
On Sun, Sep 22, 2019 at 12:10 AM, Dr. David Kirkby from Kirkby Microwave Ltd wrote:
I fail to see the need to worry if the active devices is connected across a
50 ohm resistor. 50 ohms is effectively a short circuit as far as static
is concerned.
|
Hello Rune,
Again thank you for your perseverance with the nanoVNA PC Software you are bringing us all! It is very much appreciated.
However I have a question regarding Calibration using your software, I presume that before I set about doing a Calibration in nanoVNA-Saver, and reset it, I should also do a reset of the VNA Hardware itself? Or is this taken care of by nanoVNA-Saver? As I still see the "0" come up alongside the "C" in the left hand sidebar.
73's
Pete
ZL2iK
toggle quoted message
Show quoted text
-----Original Message----- From: [email protected] [mailto: [email protected]] On Behalf Of Rune Broberg Sent: Monday, 23 September 2019 07:37 To: [email protected]Subject: Re: [nanovna-users] NanoVNA Saver For those waiting for new binaries: I just published 0.0.11 :-) This release brings averaging as a new feature: Press "Sweep settings" to select between running a single sweep (of optionally multiple segments), a continuous (live) sweeping of the range, or to average several sweeps from the NanoVNA. You can configure how many averages to make, and optionally, how many of the sample points to discard, based on which deviate the most from the rest, and are least likely to contribute signal. Also added is a new Resistance/Reactance chart, which shows both the R and X component of R+jX. This brings the challenge of showing two traces for the same data - added is therefore the option of picking a secondary colour for sweeps under "Display settings". There are further improvements to UI sizing, meaning the interface now fits - tightly - on a 1366x768 screen, at least on Windows. A few quality of live improvements made it in: Press escape in any of the pop out windows to close them instantly. The calibration window now shows more clearly when the source of calibration data is loading from a file, and also the number of points loaded. A few crash bugs were fixed. I look forward to hearing feedback from all of you! -- Rune / 5Q5R On Sun, 22 Sep 2019 at 17:14, Mario Vano <mvano@...> wrote: On Sun, Sep 22, 2019 at 06:09 AM, Mario Vano wrote:
CORRECTION: (I mistyped)
- The multi-band dipole is my 20ft high 40/20 meter trap dipole that has
been augmented with parallel dipoles for 15 and 10. The feedline has a
toroid common mode filter.
I've been happily using the NanoVNA and various versions of "Saver" for
a lot
of projects lately, but some were just motivated by curiosity about the
device.
Unfortunately I have no other analyzer to compare to. In any case here are
some sweeps from earlier tests that might be useful for reference or
amusement. In most cases, they were done with my very early attempts at full
bandwidth calibration.
- The fm filter is a commercial product from the rtl-sdr.com people.
- The LW filter is a homebrewed design I use to augment VLF beacon hunting.
- The multiband dipole is my 20ft high 80/40 meter trap dipole that has
been
augmented with parallel dipoles for 15 and 10. The feedline has a toroid
common mode filer.
- The 6 meter antenna is an attic mounted dipole with a toroid common mode
filter.
I've also used happily used the device and "saver" to design and build a 6
meter "squalo", but forgot to save the files from the latest testing run. I'm
currently working on a 400-700 mhz indor LPA design and plan to try to use the
device for relative pattern and gain testing.
Earlier tests with several types of 500-900mhz 1/4 wave mag-mount antennas and
TV rabbit ears showed useful relative gain results at 2-3 wavelength spacings.
73, AE0GL
|
gin&pez@arg
Peter Mulhare