|
It perhaps should have been obvious, but I recently realized that Electrical Delay can be used with the firmware TDR to allow high time resolution even on longer cables. A wide frequency sweep provides good time resolution, but limited time display range on the screen. Setting the Electrical Delay parameter can move that limited range window out in time to provide good resolution at the end of a longer cable.
Here is an updated version of my earlier setup procedure:
The following setup for the NanaVNA-H, -H4, or -V2 will give a display much
like that on a conventional TDR such as the Tektronix 1502:
Set stimulus for a wide sweep, say 50kHz - 900MHz or more.
Do an SOLT calibration at the desired measurement point (if not already done)
Set TRACE 0 for Real format
Turn TRACE 1 off
Set TRACE 2 for Smith,
Turn TRACE 3 off
Turn TRANSFORM ON
Select Transform LOW PASS STEP
Adjust ELECTRICAL DELAY to move the displayed window to the desired location along the cable
The resulting display should be similar to that on a step-type TDR like the
Tek 1502. With 900MHz max, the display width is about 43ns, with 3GHz it is about
8ns.
Impedance along any connected cable can be read by moving the marker to the
desired time (distance) and looking at the Smith chart marker values. You need to mentally add the
ELECTRICAL DELAY to the marker time to get the actual delay.
Save via the CAL menu when you have it all set up.
An open cable will show as an initial middle value depending on the cable
impedance followed by a step up. A shorted cable will have a step down. A
properly terminated cable will just be a steady middle value.
|
John, thanks for your original suggestion and the followup. I have found it very useful. While experimenting with it, I noted the extremely limited range of the original setup and found a method that allows viewing of the entire length of the line. I described it in another thread at /g/nanovna-users/message/17561Combining the two methods would allow the user to identify the general location of the defect and then zoom in on it using your updated method. I would like to understand the reasons for each setting in your original post and a bit more detail of the use of the Electrical Delay feature, though. Would you mind giving a more detailed explanation? Thanks, Neil W0NRP.
|
Here's a simplification of Neil's formula for stop frequency.
Length is in meters or in feet, velocity factor is a decimal number such as 0.66, result is in mhz.
fmhz = 38376*vf/feet
fmhz = 11697*vf/meters
From Neil's example where he somewhat arbitrarily increased the scan beyond 50 feet
by a factor of 100ns/77ns:
feet = 50*100/77
vf = 0.66
fmhz = 38376*vf/feet
That gives a result of 390.05 MHz, which agrees with Neil's example.
I haven't looked into it yet, but I assume John's "Transform, Low Pass Step"
involves a Fourier Transform or similar, and thus some rather nasty math.
So exactly how this works and where Neil's magic value of 39 comes from
may never be terribly obvious for most of us.
Jerry, KE7ER
|
I would normally think that Neil's upper frequency limit only sets the resolution
of the time domain response that gets displayed.
And that it is the lower frequency limit that determines the maximum length
of the cable that can be analyzed.
However, the nanovna only stores 100 measurement points in stand-alone mode.
So to get higher resolution we must reduce the length of the transmission line to be inspected.
Well, that's my guess as to what's going on here.
From
#########################################
Time domain operation
NanoVNA can simulate time domain measurements by signal processing frequency domain data.
Select DISPLAY TRANSOFRM TRANSFORM ON to convert measurement data to the time domain. TRANSFORM ON is enabled, the measurement data is immediately converted to the time domain and displayed.
The relationship between the time domain and the frequency domain is as follows.
Increasing the maximum frequency increases the time resolution
The shorter the measurement frequency interval (ie, the lower the maximum frequency), the longer the maximum time length
For this reason, the maximum time length and time resolution are in a trade-off relationship.
In other words, the time length is the distance.
If you want to increase the maximum measurement distance, you need to lower the maximum frequency.
If you want to specify the distance accurately, you need to increase the maximum frequency.
##############################
Never quite says it, but that sounds like a Fourier Transform to me.
Below is a brief excerpt from the above passage.
I suspect the first two statements are only true only because we can store a max of 100 points of data when standalone.
And I'm not convinced the final statement logically follows from the previous statements, though it is true
since it takes time for a signal to go down a length of transmission line.
# The shorter the measurement frequency interval (ie, the lower the maximum frequency), the longer the maximum time length
# For this reason, the maximum time length and time resolution are in a trade-off relationship.
# In other words, the time length is the distance.
Would be great if somebody who fully understands this stuff could expand on the material
posted on nanovna.com regarding "Time Domain Operation".
Jerry, KE7ER
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 09:29 AM, Jerry Gaffke wrote:
Here's a simplification of Neil's formula for stop frequency.
Length is in meters or in feet, velocity factor is a decimal number such as
0.66, result is in mhz.
fmhz = 38376*vf/feet
fmhz = 11697*vf/meters
From Neil's example where he somewhat arbitrarily increased the scan beyond 50
feet
by a factor of 100ns/77ns:
feet = 50*100/77
vf = 0.66
fmhz = 38376*vf/feet
That gives a result of 390.05 MHz, which agrees with Neil's example.
I haven't looked into it yet, but I assume John's "Transform, Low Pass Step"
involves a Fourier Transform or similar, and thus some rather nasty math.
So exactly how this works and where Neil's magic value of 39 comes from
may never be terribly obvious for most of us.
Jerry, KE7ER
|
Jerry,
I think the -spacing- between the test frequencies is what sets the maximum length.
--John Gord
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 09:59 AM, Jerry Gaffke wrote:
I would normally think that Neil's upper frequency limit only sets the
resolution
of the time domain response that gets displayed.
And that it is the lower frequency limit that determines the maximum length
of the cable that can be analyzed.
However, the nanovna only stores 100 measurement points in stand-alone mode.
So to get higher resolution we must reduce the length of the transmission line
to be inspected.
Well, that's my guess as to what's going on here.
From
#########################################
Time domain operation
NanoVNA can simulate time domain measurements by signal processing frequency
domain data.
Select DISPLAY TRANSOFRM TRANSFORM ON to convert measurement data to the time
domain. TRANSFORM ON is enabled, the measurement data is immediately converted
to the time domain and displayed.
The relationship between the time domain and the frequency domain is as
follows.
Increasing the maximum frequency increases the time resolution
The shorter the measurement frequency interval (ie, the lower the maximum
frequency), the longer the maximum time length
For this reason, the maximum time length and time resolution are in a
trade-off relationship.
In other words, the time length is the distance.
If you want to increase the maximum measurement distance, you need to lower
the maximum frequency.
If you want to specify the distance accurately, you need to increase the
maximum frequency.
##############################
Never quite says it, but that sounds like a Fourier Transform to me.
Below is a brief excerpt from the above passage.
I suspect the first two statements are only true only because we can store a
max of 100 points of data when standalone.
And I'm not convinced the final statement logically follows from the previous
statements, though it is true
since it takes time for a signal to go down a length of transmission line.
# The shorter the measurement frequency interval (ie, the lower the maximum
frequency), the longer the maximum time length
# For this reason, the maximum time length and time resolution are in a
trade-off relationship.
# In other words, the time length is the distance.
Would be great if somebody who fully understands this stuff could expand on
the material
posted on nanovna.com regarding "Time Domain Operation".
Jerry, KE7ER
On Mon, Sep 21, 2020 at 09:29 AM, Jerry Gaffke wrote:
Here's a simplification of Neil's formula for stop frequency.
Length is in meters or in feet, velocity factor is a decimal number such as
0.66, result is in mhz.
fmhz = 38376*vf/feet
fmhz = 11697*vf/meters
From Neil's example where he somewhat arbitrarily increased the scan beyond
50
feet
by a factor of 100ns/77ns:
feet = 50*100/77
vf = 0.66
fmhz = 38376*vf/feet
That gives a result of 390.05 MHz, which agrees with Neil's example.
I haven't looked into it yet, but I assume John's "Transform, Low Pass Step"
involves a Fourier Transform or similar, and thus some rather nasty math.
So exactly how this works and where Neil's magic value of 39 comes from
may never be terribly obvious for most of us.
Jerry, KE7ER
|
To be clear, by -spacing- I mean the step size between each of the (101 or so) test frequencies, not the min to max frequency difference.
--John Gord
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 01:37 PM, John Gord wrote:
Jerry,
I think the -spacing- between the test frequencies is what sets the maximum
length.
--John Gord
On Mon, Sep 21, 2020 at 09:59 AM, Jerry Gaffke wrote:
I would normally think that Neil's upper frequency limit only sets the
resolution
of the time domain response that gets displayed.
And that it is the lower frequency limit that determines the maximum length
of the cable that can be analyzed.
However, the nanovna only stores 100 measurement points in stand-alone mode.
So to get higher resolution we must reduce the length of the transmission
line
to be inspected.
Well, that's my guess as to what's going on here.
From
#########################################
Time domain operation
NanoVNA can simulate time domain measurements by signal processing frequency
domain data.
Select DISPLAY TRANSOFRM TRANSFORM ON to convert measurement data to the time
domain. TRANSFORM ON is enabled, the measurement data is immediately converted
to the time domain and displayed.
The relationship between the time domain and the frequency domain is as
follows.
Increasing the maximum frequency increases the time resolution
The shorter the measurement frequency interval (ie, the lower the maximum
frequency), the longer the maximum time length
For this reason, the maximum time length and time resolution are in a
trade-off relationship.
In other words, the time length is the distance.
If you want to increase the maximum measurement distance, you need to lower
the maximum frequency.
If you want to specify the distance accurately, you need to increase the
maximum frequency.
##############################
Never quite says it, but that sounds like a Fourier Transform to me.
Below is a brief excerpt from the above passage.
I suspect the first two statements are only true only because we can store a
max of 100 points of data when standalone.
And I'm not convinced the final statement logically follows from the previous
statements, though it is true
since it takes time for a signal to go down a length of transmission line.
# The shorter the measurement frequency interval (ie, the lower the maximum
frequency), the longer the maximum time length
# For this reason, the maximum time length and time resolution are in a
trade-off relationship.
# In other words, the time length is the distance.
Would be great if somebody who fully understands this stuff could expand on
the material
posted on nanovna.com regarding "Time Domain Operation".
Jerry, KE7ER
On Mon, Sep 21, 2020 at 09:29 AM, Jerry Gaffke wrote:
Here's a simplification of Neil's formula for stop frequency.
Length is in meters or in feet, velocity factor is a decimal number such
as
0.66, result is in mhz.
fmhz = 38376*vf/feet
fmhz = 11697*vf/meters
From Neil's example where he somewhat arbitrarily increased the scan
beyond
50
feet
by a factor of 100ns/77ns:
feet = 50*100/77
vf = 0.66
fmhz = 38376*vf/feet
That gives a result of 390.05 MHz, which agrees with Neil's example.
I haven't looked into it yet, but I assume John's "Transform, Low Pass Step"
involves a Fourier Transform or similar, and thus some rather nasty math.
So exactly how this works and where Neil's magic value of 39 comes from
may never be terribly obvious for most of us.
Jerry, KE7ER
|
Neil,
Here is some more information on how the TDR setup works. It is a bit "hand wavy", but its been a long time since I really worked with this kind of calculation.
First, the Fourier transform provides a connection between the time response and frequency response of a system. The time response is often described by the impulse response, which is the response of the system to a very narrow pulse. In theory the pulse is infinitely narrow, infinitely high, and has finite area under the amplitude/time curve. The theoretical impulse has infinitely high frequency components. In a practical system, we can use a reasonably narrow pulse as an input to a system, but for say, a 1GHz response, you need something like a 350ps wide pulse to stimulate it. Similarly, you need a 1GHz bandwidth system to look at a 350ps pulse. A 1GHz system (filter or whatever) can be described by how it passes a 350ps pulse or by how it passes (amplitude and phase) a range of sine waves from 0 to 1GHz. The Fourier transform relates the time description and the frequency description. If you know one, you can compute the other with the Fourier transform. In practice, we use something called the Discrete Fourier Transform (DFT) because we have discrete samples spaced out in frequency or time. Further, the DFT is implemented as the Fast Fourier Transform (FFT) because it is computationally efficient. (The FFT was figured out when I was in high school, so it is a recent development, although some folks say it was anticipated by Gauss.)
Conventional TDR:
In a conventional TDR system, an impulse (Tek 1503) or step (Tek 1502) is transmitted into one end of a cable, and the system looks for reflections. A perfectly terminated cable has no reflection, a cable open at the end reflects a positive signal, and a shorted end reflects a negative signal. An improperly terminated cable reflects something between the short and open signals; the termination value can be determined by the sign and amplitude. A piece of 75 ohm cable spliced into the middle of a 50 ohm cable produces reflections, and in the case of step input (Tek 1502) the reflections can be viewed as showing the impedance versus distance along the cable. For impulse input (Tek 1503) the interpretation is not so simple, but it does clearly show where discontinuities are located. Integrating the reflected impulses can produce a display equivalent to the reflections you would get with a step input. The distance resolution is something like the step risetime or impulse width times the speed of light. If you know what you are looking at, a cut cable end, say, you can get much better resolution, but it is hard to figure out arbitrary faults with dimensions less than risetime*(speed-of-light).
VNA TDR:
With VNA TDR, a series of sine waves are applied to the cable end and the amplitude and phase of the reflected signals are measured. Higher frequency applied signals get bigger changes in phase for a given distance to a reflecting fault and thereby allow better resolution. Wide frequency spacing of signals shortens the maximum unambiguous measurement range. Signals spaced, say, every 30MHz (3000MHz/100 steps) cannot distinguish reflections at 33.3ns and 66.6ns. For VNA TDR, the FFT converts the amplitude and phase of the reflected frequency signals into the impulse response of the reflecting cable. Integrating that impulse response converts the display to the reflection response to a step input, which is the Tek 1502 style of display.
Setup reasoning:
"Set stimulus for a wide sweep, say 50kHz - 900MHz or more."
High frequency gives good resolution, but shortens maximum length
"Do an SOLT calibration at the desired measurement point (if not already done)"
VNA works best if calibrated
"Set TRACE 0 for Real format"
We want linear (not log) input to the FFT. But "LINEAR" format looses polarity (short vs open) info
"Turn TRACE 1 off"
Reduce clutter
"Set TRACE 2 for Smith "
Smith markers let us read off cable impedances directly
"Turn TRACE 3 off"
Reduce clutter
"Turn TRANSFORM ON"
Convert from frequency to time representation
"Select Transform LOW PASS STEP"
I like this format, gives cable impedance directly, similar to Tektronix 1502 TDR
"Adjust ELECTRICAL DELAY to move the displayed window to the desired location along the cable"
This extends the good resolution to greater lengths, still subject to the ( 1 / (frequency-step-size)) limitation
The resulting display should be similar to that on a step-type TDR like the
Tek 1502. With 900MHz max, the display width is about 43ns, with 3GHz it is about
8ns.
Impedance along any connected cable can be read by moving the marker to the
desired time (distance) and looking at the Smith chart marker values. You need to mentally add the
ELECTRICAL DELAY to the marker time to get the actual delay.
" Save via the CAL menu when you have it all set up."
Saving is a good idea
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 08:31 AM, Neil Preston W0NRP wrote:
John, thanks for your original suggestion and the followup. I have found it
very useful.
.......
I would like to understand the reasons for each setting in your original post
and a bit more detail of the use of the Electrical Delay feature, though.
Would you mind giving a more detailed explanation?
Thanks,
Neil W0NRP.
|
John,
Good stuff!
Thanks for the write-up.
# Higher frequency applied signals get bigger changes in phase for a given distance
# to a reflecting fault and thereby allow better resolution. Wide frequency spacing
# of signals shortens the maximum unambiguous measurement range. Signals spaced,
# say, every 30MHz (3000MHz/100 steps) cannot distinguish reflections at 33.3ns and 66.6ns.
...
# "Adjust ELECTRICAL DELAY to move the displayed window to the desired location along the cable"
# This extends the good resolution to greater lengths, still subject to the ( 1 / (frequency-step-size)) limitation
So if we do at 3000mhz/100=30mhz steps, then window into the region around 66.6ns,
I believe you are saying that our reading will be confused by stuff that happens at 33.3ns.
Likewise, if we are looking at 33.3ns but our cable is greater than 66.6ns long, we could get
confused by stuff that is happening at 66.6ns. Correct?
John seems to be saying that tmax (the maxi delay through the coax in seconds) is 1/fstep,
where fstep is the step size in Hz. With 100 steps, that is the fmax (the max frequency)
divided by 100. tmax=1/(fmax/100)=100/fmax.
With 100 steps and a minimum frequency of close to zero, the step size is fmax/100,
and tmax=1/(fmax/100) = 100/fmax
In post 17561, Neil said that he found the relationship between max frequency and
the max delay through the coax was tmax=39/fmax.
Curious that there is a greater than 2:1 discrepancy.
I may have to play with it and see.
Jerry, KE7ER
|
Jerry,
I think the discrepancy in maximum delay range is confusion between the range shown on one screen and the range available when including Electrical Delay.
As you surmised, with 3GHz and 101 points, a discontinuity with a delay of 1.5ns also shows up at (1.5ns +33.3ns) and (1.5ns + 66.6ns), etc.
--John Gord
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 08:20 PM, Jerry Gaffke wrote:
John,
Good stuff!
Thanks for the write-up.
# Higher frequency applied signals get bigger changes in phase for a given
distance
# to a reflecting fault and thereby allow better resolution. Wide frequency
spacing
# of signals shortens the maximum unambiguous measurement range. Signals
spaced,
# say, every 30MHz (3000MHz/100 steps) cannot distinguish reflections at
33.3ns and 66.6ns.
...
# "Adjust ELECTRICAL DELAY to move the displayed window to the desired
location along the cable"
# This extends the good resolution to greater lengths, still subject to the (
1 / (frequency-step-size)) limitation
So if we do at 3000mhz/100=30mhz steps, then window into the region around
66.6ns,
I believe you are saying that our reading will be confused by stuff that
happens at 33.3ns.
Likewise, if we are looking at 33.3ns but our cable is greater than 66.6ns
long, we could get
confused by stuff that is happening at 66.6ns. Correct?
John seems to be saying that tmax (the maxi delay through the coax in seconds)
is 1/fstep,
where fstep is the step size in Hz. With 100 steps, that is the fmax (the max
frequency)
divided by 100. tmax=1/(fmax/100)=100/fmax.
With 100 steps and a minimum frequency of close to zero, the step size is
fmax/100,
and tmax=1/(fmax/100) = 100/fmax
In post 17561, Neil said that he found the relationship between max frequency
and
the max delay through the coax was tmax=39/fmax.
Curious that there is a greater than 2:1 discrepancy.
I may have to play with it and see.
Jerry, KE7ER
|
Jerry,
I noticed an interesting effect: If you set the Start frequency at half of the step size (say 15MHz to 3000MHz, 101 points), the impedance steps at the first (and every odd) "delay overflow" are inverted!! Probably makes perfect sense to someone (but not to me).
--John Gord
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 09:39 PM, John Gord wrote:
Jerry,
I think the discrepancy in maximum delay range is confusion between the range
shown on one screen and the range available when including Electrical Delay.
As you surmised, with 3GHz and 101 points, a discontinuity with a delay of
1.5ns also shows up at (1.5ns +33.3ns) and (1.5ns + 66.6ns), etc.
--John Gord
On Mon, Sep 21, 2020 at 08:20 PM, Jerry Gaffke wrote:
John,
Good stuff!
Thanks for the write-up.
# Higher frequency applied signals get bigger changes in phase for a given
distance
# to a reflecting fault and thereby allow better resolution. Wide
frequency
spacing
# of signals shortens the maximum unambiguous measurement range. Signals
spaced,
# say, every 30MHz (3000MHz/100 steps) cannot distinguish reflections at
33.3ns and 66.6ns.
...
# "Adjust ELECTRICAL DELAY to move the displayed window to the desired
location along the cable"
# This extends the good resolution to greater lengths, still subject to the (
1 / (frequency-step-size)) limitation
So if we do at 3000mhz/100=30mhz steps, then window into the region around
66.6ns,
I believe you are saying that our reading will be confused by stuff that
happens at 33.3ns.
Likewise, if we are looking at 33.3ns but our cable is greater than 66.6ns
long, we could get
confused by stuff that is happening at 66.6ns. Correct?
John seems to be saying that tmax (the maxi delay through the coax in seconds)
is 1/fstep,
where fstep is the step size in Hz. With 100 steps, that is the fmax (the max
frequency)
divided by 100. tmax=1/(fmax/100)=100/fmax.
With 100 steps and a minimum frequency of close to zero, the step size is
fmax/100,
and tmax=1/(fmax/100) = 100/fmax
In post 17561, Neil said that he found the relationship between max frequency
and
the max delay through the coax was tmax=39/fmax.
Curious that there is a greater than 2:1 discrepancy.
I may have to play with it and see.
Jerry, KE7ER
|
John wrote:
# I noticed an interesting effect: If you set the Start frequency at half of the step size
# (say 15MHz to 3000MHz, 101 points), the impedance steps at the first (and every odd)
# "delay overflow" are inverted!! Probably makes perfect sense to someone (but not to me).
That's very curious indeed!
Perhaps with some head scratching, one might use this effect to distinguish between
ambiguous regions on long runs of cables.
Though of course, far better to figure out how to process more data points.
Perhaps with nanovna-saver, or a nanovna-SAA2
At the top of Bryan's document, Step 2 states:
"I changed the nanoVNA stop frequency to 130 MHz which gives a maximum length that can be observed in the nanoVNA of about 31.5 m. "
Speed down the cable is 3e8*0.66 meters/second, assuming a velocity factor in his cable of 0.66.
So Bryan's 31.5m might represent a delay of 31.5/(3e8*0.66) = 160 ns.
If we assume Bryan did not take the velocity factor into account, that's 31.5/3e8 = 105 ns.
Neil's formula gives a delay of tmax = 39/fmax = 39/130e6 = 300 ns
John's formula gives us a delay of tmax = 100/fmax = 100/130e6 = 769 ns
I prefer the looks of John's formula, the others appear to have been arrived at empirically.
But that's a rather large spread.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 10:33 PM, John Gord wrote:
Jerry,
I noticed an interesting effect: If you set the Start frequency at half of the
step size (say 15MHz to 3000MHz, 101 points), the impedance steps at the first
(and every odd) "delay overflow" are inverted!! Probably makes perfect sense
to someone (but not to me).
--John Gord
Hide quoted text ( #quoted-172641791 )
On Mon, Sep 21, 2020 at 09:39 PM, John Gord wrote:
Jerry,
I think the discrepancy in maximum delay range is confusion between the
range
shown on one screen and the range available when including Electrical
Delay.
As you surmised, with 3GHz and 101 points, a discontinuity with a delay of
1.5ns also shows up at (1.5ns +33.3ns) and (1.5ns + 66.6ns), etc.
--John Gord
On Mon, Sep 21, 2020 at 08:20 PM, Jerry Gaffke wrote:
John,
Good stuff!
Thanks for the write-up.
# Higher frequency applied signals get bigger changes in phase for a given
distance
# to a reflecting fault and thereby allow better resolution. Wide
frequency
spacing
# of signals shortens the maximum unambiguous measurement range. Signals
spaced,
# say, every 30MHz (3000MHz/100 steps) cannot distinguish reflections at
33.3ns and 66.6ns.
...
# "Adjust ELECTRICAL DELAY to move the displayed window to the desired
location along the cable"
# This extends the good resolution to greater lengths, still subject to
the (
1 / (frequency-step-size)) limitation
So if we do at 3000mhz/100=30mhz steps, then window into the region around
66.6ns,
I believe you are saying that our reading will be confused by stuff that
happens at 33.3ns.
Likewise, if we are looking at 33.3ns but our cable is greater than 66.6ns
long, we could get
confused by stuff that is happening at 66.6ns. Correct?
John seems to be saying that tmax (the maxi delay through the coax in seconds)
is 1/fstep,
where fstep is the step size in Hz. With 100 steps, that is the fmax (the max
frequency)
divided by 100. tmax=1/(fmax/100)=100/fmax.
With 100 steps and a minimum frequency of close to zero, the step size is
fmax/100,
and tmax=1/(fmax/100) = 100/fmax
In post 17561, Neil said that he found the relationship between max frequency
and
the max delay through the coax was tmax=39/fmax.
Curious that there is a greater than 2:1 discrepancy.
I may have to play with it and see.
Jerry, KE7ER
|
Thank you, John. The founding concepts you described I was already aware of, but I'm sure they were beneficial to anyone following the post. I was specifically interested in the reasons for the settings you chose (Such as REAL format and TRANSFORM), as I am not yet familiar with all the features of the VNA. Understanding the f-to-T function makes it clearer.
My adjustment of the STOP frequency allows an overall view (albeit at low resolution) of the entire length of cable which hopefully will catch some indication of any discontinuities along the line. If desired, using a higher STOP frequency along with the ELECTRICAL DELAY can then provide a more detailed indication of the nature of the discontinuity.
Thanks again!
|
On 9/22/20 8:00 AM, Jerry Gaffke via groups.io wrote: John wrote:
# I noticed an interesting effect: If you set the Start frequency at half of the step size
# (say 15MHz to 3000MHz, 101 points), the impedance steps at the first (and every odd)
# "delay overflow" are inverted!! Probably makes perfect sense to someone (but not to me).
That's very curious indeed!
Perhaps with some head scratching, one might use this effect to distinguish between
ambiguous regions on long runs of cables.
Though of course, far better to figure out how to process more data points.
Perhaps with nanovna-saver, or a nanovna-SAA2
At the top of Bryan's document, Step 2 states:
"I changed the nanoVNA stop frequency to 130 MHz which gives a maximum length that can be observed in the nanoVNA of about 31.5 m. "
Speed down the cable is 3e8*0.66 meters/second, assuming a velocity factor in his cable of 0.66.
So Bryan's 31.5m might represent a delay of 31.5/(3e8*0.66) = 160 ns.
If we assume Bryan did not take the velocity factor into account, that's 31.5/3e8 = 105 ns.
Neil's formula gives a delay of tmax = 39/fmax = 39/130e6 = 300 ns
John's formula gives us a delay of tmax = 100/fmax = 100/130e6 = 769 ns
I prefer the looks of John's formula, the others appear to have been arrived at empirically.
But that's a rather large spread.
Jerry, KE7ER
The longest cable is more a function of the step size than the highest frequency. It has to do with ambiguity.
|
Jim wrote:
# The longest cable is more a function of the step size than the highest frequency. It has to do with ambiguity.
Understood.
All three of those formulas assume we only have the 101 datapoints of a standalone nanovna.
As I stated:
# Though of course, far better to figure out how to process more data points.
# Perhaps with nanovna-saver, or a nanovna-SAA2
Jerry, KE7ER
toggle quoted message
Show quoted text
On Tue, Sep 22, 2020 at 08:29 AM, Jim Lux wrote:
The longest cable is more a function of the step size than the highest
frequency. It has to do with ambiguity.
|
Scanning a cable from 1 to 101mhz, I see a horizontal scale on my nanovna-H4
of 390ns, which agrees with Neil's tmax = 39/fmax
I'll have to try John's Electrical Delay now, this is all starting to make sense.
Fun stuff!
Like everything nanovna, it's something you have to play with for awhile
before going out into the real world looking for trouble.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 09:39 PM, John Gord wrote: I think the discrepancy in maximum delay range is confusion between the range
shown on one screen and the range available when including Electrical Delay.
|
Scanning from 50khz to 100mhz works much better than scanning 1mhz to 101mhz.
The 50khz to 100mhz scan showed a long stretch of 50 ohm cable,
it then rose to 100 ohms when it encountered a 100 ohm termination at the end
and stayed there for the remainder of the display as expected.
The 1mhz to 101mhz scan gave some rather confusing results.
The 100 ohm termination at the end of the 50 ohm cable got displayed as
a dip to 25 ohms and then back up to 50 ohms.
Also, the transitions were more rounded than the 50khz to 100mhz scan.
I guess it really needs to see that first low freq sample.
Perhaps this is related to John's observation of post 17593:
# If you set the Start frequency at half of the step size (say 15MHz to 3000MHz, 101 points),
# the impedance steps at the first (and every odd) "delay overflow" are inverted!!"
Jerry, KE7ER
toggle quoted message
Show quoted text
On Tue, Sep 22, 2020 at 09:33 PM, Jerry Gaffke wrote:
Scanning a cable from 1 to 101mhz,
|
The post included below from John on the procedure for using stock nanovna-H firmware
as a TDR is my gold standard, it answered almost all my questions.
One point it does not explicitly state: The times reported indicate round trip times,
out to that part of the cable and back to the VNA.
I'm sure that's obvious to some. Not so obvious to others such as myself.
I have a 100 foot coil of RG8X, a web source said RG8X had a velocity factor of 0.66.
I measured 240ns on the display where the reflection at the end was displayed.
But 240e-9 * 3e8*.66 * 39.37/12 = 155 feet. What?
(The 3e8 term is the speed of light in meters per second, the 39.37/12 converts meters to inches to feet)
Reading the fine print on the cable jacket, in addition to "RG8X", it says "218XATC".
Googling that, I find the velocity factor is actually 0.84, not 0.66.
240e-9 * 3e8*.84 * 39.37/12 = 198.4 feet
Ahh, 100 feet of cable, and the nanovna tells me it's almost 200 feet long.
Makes sense the nanovna is telling me the round trip time.
The nanovna is just doing a Fourier Transform, and doesn't know that I am
looking at a length of cable.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Mon, Sep 21, 2020 at 03:29 PM, John Gord wrote:
Neil,
Here is some more information on how the TDR setup works. It is a bit "hand
wavy", but its been a long time since I really worked with this kind of
calculation.
First, the Fourier transform provides a connection between the time response
and frequency response of a system. The time response is often described by
the impulse response, which is the response of the system to a very narrow
pulse. In theory the pulse is infinitely narrow, infinitely high, and has
finite area under the amplitude/time curve. The theoretical impulse has
infinitely high frequency components. In a practical system, we can use a
reasonably narrow pulse as an input to a system, but for say, a 1GHz response,
you need something like a 350ps wide pulse to stimulate it. Similarly, you
need a 1GHz bandwidth system to look at a 350ps pulse. A 1GHz system (filter
or whatever) can be described by how it passes a 350ps pulse or by how it
passes (amplitude and phase) a range of sine waves from 0 to 1GHz. The Fourier
transform relates the time description and the frequency description. If you
know one, you can compute the other with the Fourier transform. In practice,
we use something called the Discrete Fourier Transform (DFT) because we have
discrete samples spaced out in frequency or time. Further, the DFT is
implemented as the Fast Fourier Transform (FFT) because it is computationally
efficient. (The FFT was figured out when I was in high school, so it is a
recent development, although some folks say it was anticipated by Gauss.)
Conventional TDR:
In a conventional TDR system, an impulse (Tek 1503) or step (Tek 1502) is
transmitted into one end of a cable, and the system looks for reflections. A
perfectly terminated cable has no reflection, a cable open at the end reflects
a positive signal, and a shorted end reflects a negative signal. An improperly
terminated cable reflects something between the short and open signals; the
termination value can be determined by the sign and amplitude. A piece of 75
ohm cable spliced into the middle of a 50 ohm cable produces reflections, and
in the case of step input (Tek 1502) the reflections can be viewed as showing
the impedance versus distance along the cable. For impulse input (Tek 1503)
the interpretation is not so simple, but it does clearly show where
discontinuities are located. Integrating the reflected impulses can produce a
display equivalent to the reflections you would get with a step input. The
distance resolution is something like the step risetime or impulse width times
the speed of light. If you know what you are looking at, a cut cable end, say,
you can get much better resolution, but it is hard to figure out arbitrary
faults with dimensions less than risetime*(speed-of-light).
VNA TDR:
With VNA TDR, a series of sine waves are applied to the cable end and the
amplitude and phase of the reflected signals are measured. Higher frequency
applied signals get bigger changes in phase for a given distance to a
reflecting fault and thereby allow better resolution. Wide frequency spacing
of signals shortens the maximum unambiguous measurement range. Signals spaced,
say, every 30MHz (3000MHz/100 steps) cannot distinguish reflections at 33.3ns
and 66.6ns. For VNA TDR, the FFT converts the amplitude and phase of the
reflected frequency signals into the impulse response of the reflecting cable.
Integrating that impulse response converts the display to the reflection
response to a step input, which is the Tek 1502 style of display.
Setup reasoning:
"Set stimulus for a wide sweep, say 50kHz - 900MHz or more."
High frequency gives good resolution, but shortens maximum length
"Do an SOLT calibration at the desired measurement point (if not already
done)"
VNA works best if calibrated
"Set TRACE 0 for Real format"
We want linear (not log) input to the FFT. But "LINEAR" format looses polarity
(short vs open) info
"Turn TRACE 1 off"
Reduce clutter
"Set TRACE 2 for Smith "
Smith markers let us read off cable impedances directly
"Turn TRACE 3 off"
Reduce clutter
"Turn TRANSFORM ON"
Convert from frequency to time representation
"Select Transform LOW PASS STEP"
I like this format, gives cable impedance directly, similar to Tektronix 1502
TDR
"Adjust ELECTRICAL DELAY to move the displayed window to the desired location
along the cable"
This extends the good resolution to greater lengths, still subject to the ( 1
/ (frequency-step-size)) limitation
The resulting display should be similar to that on a step-type TDR like the
Tek 1502. With 900MHz max, the display width is about 43ns, with 3GHz it is
about
8ns.
Impedance along any connected cable can be read by moving the marker to the
desired time (distance) and looking at the Smith chart marker values. You need
to mentally add the
ELECTRICAL DELAY to the marker time to get the actual delay.
" Save via the CAL menu when you have it all set up."
Saving is a good idea
|
Jerry, something is wrong here. The displayed values should show the one-way delay, and directly read the cable length if you set the velocity factor via the menu. How did you calibrate for this measurement, and what frequency start/stop are you using? On Wed, Sep 23, 2020, 12:58 PM Jerry Gaffke via groups.io <jgaffke= [email protected]> wrote: The post included below from John on the procedure for using stock
nanovna-H firmware
as a TDR is my gold standard, it answered almost all my questions.
One point it does not explicitly state: The times reported indicate round
trip times,
out to that part of the cable and back to the VNA.
I'm sure that's obvious to some. Not so obvious to others such as myself.
I have a 100 foot coil of RG8X, a web source said RG8X had a velocity
factor of 0.66.
I measured 240ns on the display where the reflection at the end was
displayed.
But 240e-9 * 3e8*.66 * 39.37/12 = 155 feet. What?
(The 3e8 term is the speed of light in meters per second, the 39.37/12
converts meters to inches to feet)
Reading the fine print on the cable jacket, in addition to "RG8X", it says
"218XATC".
Googling that, I find the velocity factor is actually 0.84, not 0.66.
240e-9 * 3e8*.84 * 39.37/12 = 198.4 feet
Ahh, 100 feet of cable, and the nanovna tells me it's almost 200 feet long.
Makes sense the nanovna is telling me the round trip time.
The nanovna is just doing a Fourier Transform, and doesn't know that I am
looking at a length of cable.
Jerry, KE7ER
On Mon, Sep 21, 2020 at 03:29 PM, John Gord wrote:
Neil,
Here is some more information on how the TDR setup works. It is a bit
"hand
wavy", but its been a long time since I really worked with this kind of
calculation.
First, the Fourier transform provides a connection between the time response
and frequency response of a system. The time response is often described by
the impulse response, which is the response of the system to a very narrow
pulse. In theory the pulse is infinitely narrow, infinitely high, and has
finite area under the amplitude/time curve. The theoretical impulse has
infinitely high frequency components. In a practical system, we can use a
reasonably narrow pulse as an input to a system, but for say, a 1GHz response,
you need something like a 350ps wide pulse to stimulate it. Similarly, you
need a 1GHz bandwidth system to look at a 350ps pulse. A 1GHz system (filter
or whatever) can be described by how it passes a 350ps pulse or by how it
passes (amplitude and phase) a range of sine waves from 0 to 1GHz. The Fourier
transform relates the time description and the frequency description. If you
know one, you can compute the other with the Fourier transform. In practice,
we use something called the Discrete Fourier Transform (DFT) because we have
discrete samples spaced out in frequency or time. Further, the DFT is
implemented as the Fast Fourier Transform (FFT) because it is computationally
efficient. (The FFT was figured out when I was in high school, so it is a
recent development, although some folks say it was anticipated by Gauss.)
Conventional TDR:
In a conventional TDR system, an impulse (Tek 1503) or step (Tek 1502) is
transmitted into one end of a cable, and the system looks for reflections. A
perfectly terminated cable has no reflection, a cable open at the end reflects
a positive signal, and a shorted end reflects a negative signal. An improperly
terminated cable reflects something between the short and open signals; the
termination value can be determined by the sign and amplitude. A piece of 75
ohm cable spliced into the middle of a 50 ohm cable produces reflections, and
in the case of step input (Tek 1502) the reflections can be viewed as showing
the impedance versus distance along the cable. For impulse input (Tek 1503)
the interpretation is not so simple, but it does clearly show where
discontinuities are located. Integrating the reflected impulses can produce a
display equivalent to the reflections you would get with a step input. The
distance resolution is something like the step risetime or impulse width times
the speed of light. If you know what you are looking at, a cut cable end, say,
you can get much better resolution, but it is hard to figure out arbitrary
faults with dimensions less than risetime*(speed-of-light).
VNA TDR:
With VNA TDR, a series of sine waves are applied to the cable end and the
amplitude and phase of the reflected signals are measured. Higher frequency
applied signals get bigger changes in phase for a given distance to a
reflecting fault and thereby allow better resolution. Wide frequency spacing
of signals shortens the maximum unambiguous measurement range. Signals spaced,
say, every 30MHz (3000MHz/100 steps) cannot distinguish reflections at 33.3ns
and 66.6ns. For VNA TDR, the FFT converts the amplitude and phase of the
reflected frequency signals into the impulse response of the reflecting cable.
Integrating that impulse response converts the display to the reflection
response to a step input, which is the Tek 1502 style of display.
Setup reasoning:
"Set stimulus for a wide sweep, say 50kHz - 900MHz or more."
High frequency gives good resolution, but shortens maximum length
"Do an SOLT calibration at the desired measurement point (if not already
done)"
VNA works best if calibrated
"Set TRACE 0 for Real format"
We want linear (not log) input to the FFT. But "LINEAR" format looses polarity
(short vs open) info
"Turn TRACE 1 off"
Reduce clutter
"Set TRACE 2 for Smith "
Smith markers let us read off cable impedances directly
"Turn TRACE 3 off"
Reduce clutter
"Turn TRANSFORM ON"
Convert from frequency to time representation
"Select Transform LOW PASS STEP"
I like this format, gives cable impedance directly, similar to Tektronix 1502
TDR
"Adjust ELECTRICAL DELAY to move the displayed window to the desired location
along the cable"
This extends the good resolution to greater lengths, still subject to the ( 1
/ (frequency-step-size)) limitation
The resulting display should be similar to that on a step-type TDR like the
Tek 1502. With 900MHz max, the display width is about 43ns, with 3GHz it is
about
8ns.
Impedance along any connected cable can be read by moving the marker to the
desired time (distance) and looking at the Smith chart marker values. You need
to mentally add the
ELECTRICAL DELAY to the marker time to get the actual delay.
" Save via the CAL menu when you have it all set up."
Saving is a good idea
|
Stan,
I was following John Gord's instructions as included in my previous email
This procedure uses the stock transform mode of the nanovna firmware,
it does not rely on any kind of special purpose TDR firmware.
There is no provision for setting a velocity factor.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Wed, Sep 23, 2020 at 02:42 PM, Stan Dye wrote:
Jerry, something is wrong here. The displayed values should show the
one-way delay, and directly read the cable length if you set the velocity
factor via the menu. How did you calibrate for this measurement, and what
frequency start/stop are you using?
|
Jerry,
I didn't mean to imply that one should not set the velocity factor. I'm usually content with just the delay, but that is partly because I am usually looking for things like cable impedance rather than fault location on a particular type of cable.
--John Gord
toggle quoted message
Show quoted text
On Wed, Sep 23, 2020 at 02:56 PM, Jerry Gaffke wrote:
Stan,
I was following John Gord's instructions as included in my previous email
This procedure uses the stock transform mode of the nanovna firmware,
it does not rely on any kind of special purpose TDR firmware.
There is no provision for setting a velocity factor.
Jerry, KE7ER
On Wed, Sep 23, 2020 at 02:42 PM, Stan Dye wrote:
Jerry, something is wrong here. The displayed values should show the
one-way delay, and directly read the cable length if you set the velocity
factor via the menu. How did you calibrate for this measurement, and what
frequency start/stop are you using?
|
