Empathy List Archives

DD

Dr. David Kirkby

Mon, Jun 2, 2014 5:24 PM

On 2 Jun 2014 15:50, "Attila Kinali" attila@kinali.ch wrote:

I recently got introduced into the usefullness of a VNA. But these
things are horribly expensive for home use, even if bought from ebay
(before you say anything, remember i live in europe, where every
boat anchor hast to travel a long way). But given that most of the
designs that are on ebay are from the 80s and early 90s, i thought that
with todays ICs it should be easy to come up with a design that does
the same thing but can be build on a kitchen table.

There are a few designs around. The early version of the VNWA was
described in QEX. There is the N2PK design too. The latter has a limited
frequency range, but a high dynamic range.

All modern professional systems with two ports use 4 receivers. Earlier
designs use 3 receivers which is not good for TRL calibration.

I think it would be a huge task. I think that the main issue would be the
software.
I have been considering adding the "unknown thru" calibration method to my
HP 8720D. That in itself would be quite a task, but writing all the
software for a VNA would be a huge task.

I thought TAPR had a similar project but I don't recall it producing
anything close to workable.

BTW the software options for the HP 8753's are now easily available, so if
you do buy an 8753 (probably the best choice), don't worry about what
software options it has.

I would really like to see an open hardware and software VNA, but it would
be a lot of work.

If you do it, think about having 3 or 4 ports with independent sources for
optimal balanced measurements.

I don't think that there's much point producing a design with just a TR
test set.

Dave.

On 2 Jun 2014 15:50, "Attila Kinali" <attila@kinali.ch> wrote: > > I recently got introduced into the usefullness of a VNA. But these > things are horribly expensive for home use, even if bought from ebay > (before you say anything, remember i live in europe, where every > boat anchor hast to travel a long way). But given that most of the > designs that are on ebay are from the 80s and early 90s, i thought that > with todays ICs it should be easy to come up with a design that does > the same thing but can be build on a kitchen table. There are a few designs around. The early version of the VNWA was described in QEX. There is the N2PK design too. The latter has a limited frequency range, but a high dynamic range. All modern professional systems with two ports use 4 receivers. Earlier designs use 3 receivers which is not good for TRL calibration. I think it would be a huge task. I think that the main issue would be the software. I have been considering adding the "unknown thru" calibration method to my HP 8720D. That in itself would be quite a task, but writing all the software for a VNA would be a huge task. I thought TAPR had a similar project but I don't recall it producing anything close to workable. BTW the software options for the HP 8753's are now easily available, so if you do buy an 8753 (probably the best choice), don't worry about what software options it has. I would really like to see an open hardware and software VNA, but it would be a lot of work. If you do it, think about having 3 or 4 ports with independent sources for optimal balanced measurements. I don't think that there's much point producing a design with just a TR test set. Dave.

AK

Attila Kinali

Mon, Jun 2, 2014 5:36 PM

On Mon, 02 Jun 2014 09:35:41 -0700
"Richard (Rick) Karlquist" richard@karlquist.com wrote:

Julius said that he designed it before
the era of cheap calibration. Now that everyone has calibration,
you don't need a good directional coupler. You can get away
with a MiniCircuits coupler.

How about impedance matching issues? Can those be calibrated away?

But in fact it is
even easier to just use a resistive bridge. Four ordinary
resistors will easily go to 3 GHz. Use a differential amplifier
at the output. Lots of info on this in the literature.

I somewhat fear that the parasitics of resistors will give me
garbage when going to >1GHz. Maybe i should look into those and
calculate how much performance they would cost.

Another way to make in effect a directional coupler is to use
a 180 degree hybrid.

My understanding is, that those are rather narrow band.
Those i've seen were at most one octave in range.

Don't even think about trying this at home.

You mention that for both Podell's and Botkas design. May i ask what
the reason is? Why do you think it cannot be done at home?

Also, i quickly tried to search for both Botka and Podell. While
Podell gives a couple of matches on ieeexplore, Botka comes back
almost empty. Any hints to for finding references to their work?

BTW, you didn't mention software, but that's a big part of
the job.

Because i have not had a look into that yet. My guess is that
phase detection of a known frequency (i ignore multi-tone detection
due to harmonics for the moment) is known and should be easy.
So the big part of the softare is getting thins into the PC and doing
a good job of presenting the data, ie GUI stuff. And GUI is, once you
have a graphical design, just implementation work.

Please correct me if i'm mistaken.

		Attila Kinali

--
I pity people who can't find laughter or at least some bit of amusement in
the little doings of the day. I believe I could find something ridiculous
even in the saddest moment, if necessary. It has nothing to do with being
superficial. It's a matter of joy in life.
-- Sophie Scholl

On Mon, 02 Jun 2014 09:35:41 -0700 "Richard (Rick) Karlquist" <richard@karlquist.com> wrote: > Julius said that he designed it before > the era of cheap calibration. Now that everyone has calibration, > you don't need a good directional coupler. You can get away > with a MiniCircuits coupler. How about impedance matching issues? Can those be calibrated away? > But in fact it is > even easier to just use a resistive bridge. Four ordinary > resistors will easily go to 3 GHz. Use a differential amplifier > at the output. Lots of info on this in the literature. I somewhat fear that the parasitics of resistors will give me garbage when going to >1GHz. Maybe i should look into those and calculate how much performance they would cost. > Another way to make in effect a directional coupler is to use > a 180 degree hybrid. My understanding is, that those are rather narrow band. Those i've seen were at most one octave in range. > Don't even think about trying this at home. You mention that for both Podell's and Botkas design. May i ask what the reason is? Why do you think it cannot be done at home? Also, i quickly tried to search for both Botka and Podell. While Podell gives a couple of matches on ieeexplore, Botka comes back almost empty. Any hints to for finding references to their work? > BTW, you didn't mention software, but that's a big part of > the job. Because i have not had a look into that yet. My guess is that phase detection of a known frequency (i ignore multi-tone detection due to harmonics for the moment) is known and should be easy. So the big part of the softare is getting thins into the PC and doing a good job of presenting the data, ie GUI stuff. And GUI is, once you have a graphical design, just implementation work. Please correct me if i'm mistaken. Attila Kinali -- I pity people who can't find laughter or at least some bit of amusement in the little doings of the day. I believe I could find something ridiculous even in the saddest moment, if necessary. It has nothing to do with being superficial. It's a matter of joy in life. -- Sophie Scholl

DD

Dr. David Kirkby

Mon, Jun 2, 2014 5:38 PM

On 2 Jun 2014 17:33, "Ed Palmer" ed_palmer@sasktel.net wrote:

There's a DIY project to build a spectrum analyzer at

http://scottyspectrumanalyzer.com . Since it's modular, one version of the
project is to add a couple of modules that change it into a network
analyzer.

But I think a VNA is an order of magnitude more complex than a SA - pun
intended!

The software is probably where the biggest work is.

A modular design for a VNA would be interesting, as different sources,
couplers etc could be used depending on frequency range.

Dave.

On 2 Jun 2014 17:33, "Ed Palmer" <ed_palmer@sasktel.net> wrote: > > There's a DIY project to build a spectrum analyzer at http://scottyspectrumanalyzer.com . Since it's modular, one version of the project is to add a couple of modules that change it into a network analyzer. But I think a VNA is an order of magnitude more complex than a SA - pun intended! The software is probably where the biggest work is. A modular design for a VNA would be interesting, as different sources, couplers etc could be used depending on frequency range. Dave.

DD

Dr. David Kirkby

Mon, Jun 2, 2014 5:45 PM

On 2 Jun 2014 18:14, "Thomas S. Knutsen" la3pna@gmail.com wrote:

The design of an VNA is an interesting thing. It requires quite high focus
on good RF practices and screening.

In the range 0-3GHz there is no low cost devices avaible, not counting the
copper mountain tech boxes ( http://www.coppermountaintech.com/ ). Up to
1.3GHz there is the DG8SAQ VNWA avaible from sdrkits, these can also be
used with mixers to extend the max frequency. The VNWA is an 2 detector

VNA

that needs an S-parameter testset in order to get all the 4 S-parameters.

An homebrew alternative would be fun to do, but its a lot of work, both in
getting reproduceable data from the hardware and in programming. Building
somthing that is connected to the PC simplifies things a lot.
Couplers and detectors are not the hardest thing to make, some small SMD
resistors, an balanced amplifier - detector and things should work to 6GHz
or higher with some care in the layout.
Signal generation is perhaps the hardest part, there is AFAIK no single
solution working from LF to high UHF, one cool alternative is to build an
generator with an YIG and mixing down, but that requires a lot of work to
get stable over the range 0-3GHz. In addition you need to keep the signal
from the generator out of the detector in order to keep the dynamic range
high.
If you are building your own VNA, I would build it with 4 detectors and

the

posibility to re-configure those. It opens for several of the more

advanced

calibration methods and eliminates some of the errors in the VNA.

If I were to build something, I think I would base it on the N2PK design,
as there is documentation and programs avaible that makes for some part of
the work.

There are some IC's avaible that do the detection of the power levels,
AD8302 comes to mind, the common denominator for these are that they don't
solve for the phase sign, and thereby are not true vector. In addition,
those I have tested don't behave to well with

As an alternative, the HP8410 series are avaible here in the EU, sometimes
quite cheap

I assume you mean 8510.

The accuracy of the VNA is determined by the calkit used to calibrate it.
There is no way around obtaining an good calkit, learning how to use it
without destroying it, and do repeatable calibrations. The calkit is the
single most important part of the VNA. Do use an calkit for the connectors
you are going to measure, don't add adaptors or worse, coaxial cable after
the calibration plane.

I had at one point an HP 8753A (3 GHz) VNA with a full S-parameter test
set. It cost me about 50% of what my HP 85054B 18 GHz N calibration kit
cost me and I think I got the 85054B cheap at about $3800.

I do sell low cost calibration kits for N, SMA and X-band waveguide.

Dave

On 2 Jun 2014 18:14, "Thomas S. Knutsen" <la3pna@gmail.com> wrote: > > The design of an VNA is an interesting thing. It requires quite high focus > on good RF practices and screening. > > In the range 0-3GHz there is no low cost devices avaible, not counting the > copper mountain tech boxes ( http://www.coppermountaintech.com/ ). Up to > 1.3GHz there is the DG8SAQ VNWA avaible from sdrkits, these can also be > used with mixers to extend the max frequency. The VNWA is an 2 detector VNA > that needs an S-parameter testset in order to get all the 4 S-parameters. > > An homebrew alternative would be fun to do, but its a lot of work, both in > getting reproduceable data from the hardware and in programming. Building > somthing that is connected to the PC simplifies things a lot. > Couplers and detectors are not the hardest thing to make, some small SMD > resistors, an balanced amplifier - detector and things should work to 6GHz > or higher with some care in the layout. > Signal generation is perhaps the hardest part, there is AFAIK no single > solution working from LF to high UHF, one cool alternative is to build an > generator with an YIG and mixing down, but that requires a lot of work to > get stable over the range 0-3GHz. In addition you need to keep the signal > from the generator out of the detector in order to keep the dynamic range > high. > If you are building your own VNA, I would build it with 4 detectors and the > posibility to re-configure those. It opens for several of the more advanced > calibration methods and eliminates some of the errors in the VNA. > > If I were to build something, I think I would base it on the N2PK design, > as there is documentation and programs avaible that makes for some part of > the work. > > There are some IC's avaible that do the detection of the power levels, > AD8302 comes to mind, the common denominator for these are that they don't > solve for the phase sign, and thereby are not true vector. In addition, > those I have tested don't behave to well with > > As an alternative, the HP8410 series are avaible here in the EU, sometimes > quite cheap I assume you mean 8510. > The accuracy of the VNA is determined by the calkit used to calibrate it. > There is no way around obtaining an good calkit, learning how to use it > without destroying it, and do repeatable calibrations. The calkit is the > single most important part of the VNA. Do use an calkit for the connectors > you are going to measure, don't add adaptors or worse, coaxial cable after > the calibration plane. I had at one point an HP 8753A (3 GHz) VNA with a full S-parameter test set. It cost me about 50% of what my HP 85054B 18 GHz N calibration kit cost me and I think I got the 85054B cheap at about $3800. I do sell low cost calibration kits for N, SMA and X-band waveguide. Dave

DL

Don Latham

Mon, Jun 2, 2014 5:49 PM

Those interested might look at RedPitaya.com for a new piece of hardware that
might be used. Less than $500 without a "box"
Don