VNA advice

[G0HZU]

Playing about with a smith chart can only take you so far.

The main function of a modern n port VNA is to extract n port model data from it and then to make best use of this data (usually in a simulator).

If you just use it to look at S11 and then decide to go off and play with a smith chart and a pencil (as in classic impedance matching tutorials as found in books like Bowick) then you aren't making best use of the instrument's full potential or the potential of modern RF simulation packages.

It's 2014 not 1965

[eurofox]

F@$#g Windows explorer
Spend an hour typing, typed a few backspace and the browser leaves this page and away is my text. This was the last time, Now I'm sure, I will replace it by Linux on this laptop too. 

Use Chrome, is not perfect but better.

http://www.pa4tim.nl/?p=1594 I made a series of tutorials because there was/is not much literature aboutt NA for beginners. I had to learn it by myself the hard way.

I already download all of them  Thanks

Calkits are the most important together with cables. My best cables are 279 euro a piece (I got them for free from a friend), besides that mil grade RG58 (the see through version), home made extra shielded belden/radial cables for bnc and GR cables.

I use calkits from Amphenol, rosenberger (sma), homemade (N and bnc) and from GR (GR874 and GR900)

I have already very good cables with the ultra fast pulses it is mandatory, today I scored 27ps rising edge.

My interest in VNA's is component behaviour related.

I have the same interest with respect to component testing, I'm not an RF guy but I do have an RF generator, a microwave generator that I build myself, a Rigol 815 + TG and a 21Ghz HP SA

I like the VNWA a lot, it has a lot of functions. OK, it is limmited to 1500 MHz (only with my best cables and call kit) but if you want to play at 6 GHz the calkit you need for that will cost a lot more as a complete VNWA.

How did you manage to get 1,5Ghz since it is supposed to be limited at 1,3Ghz?

[G0HZU]

I still use a smith chart when matching in microwave office or ADS. It makes it easier to see what's happening as you build up a matching circuit whether it's broad or narrow band. I can't imagine trying to match to the input of an RF power transistor, where the input impedance may be 0.5ohms with a chunk of series inductance without the aid of a smith chart.

Yes but you ARE using a simulator If we stick to looking at small signal (linear) analysis then the crude oldschool way would be to just measure the S11 (and S22) of the whole circuit with a VNA and then go off and try matching to it using a smith chart and a pencil. This would run into problems if matching changes to the output side of the network affected the input S11. So at best this would require an iterative approach with lots of soldering/desoldering of components and lots of VNA time.

[KJDS]

I still use a smith chart when matching in microwave office or ADS. It makes it easier to see what's happening as you build up a matching circuit whether it's broad or narrow band. I can't imagine trying to match to the input of an RF power transistor, where the input impedance may be 0.5ohms with a chunk of series inductance without the aid of a smith chart.

If we stick to looking at small signal (linear) analysis then the crude oldschool way would be to just measure the S11 (and S22) of the whole circuit with a VNA and then go off and try matching to it using a smith chart and a pencil. This would run into problems if matching changes to the output side of the network affected the input S11. So at best this would require an iterative approach with lots of soldering/desoldering of components and lots of VNA time.

I agree, I can't remember that last time I used a paper smith chart. I just like looking at one in MWO or ADS as I'm sorting out the matching.

My typical approach for an LNA, where we can think only about the small signal case, is to start with the device s-pars and add some source inductance to improve the stability. Then I'll start working the match outwards, first with the output match and use that to push the optimum noise figure point towards the input impedance. Depending on frequency I'll either use lumped or distributed components. If I'm using lumped then I'll also add in some microstrip lines into the models between the components  for anything above VHF. I'll then work the input and output matching circuits until I'm happy. Sometimes I'll leave the optimizer to run if I'm fairly close, however it's always worthwhile checking just what the optimizer is doing. Nothing worse than getting great performance after optimizing only to find out it's put in something impractical.

Too often I've seen engineers run the optimizer, build the circuit and miss something fundamental, like a tuning cap sitting in series with a bias line. The S parameters will still say that it will work.

I started out in the 80's when VNA time was hard to come by, much hair pulling and too many iterations. A lot of time on the bench chasing issues that today you'd sort out before the layout started. I have no idea how people managed PA design then, except that mostly what they did wasn't that good.

[Howardlong]

... So at best this would require an iterative approach with lots of soldering/desoldering of components and lots of VNA time.

Been there, done that. It's worth it for the learning experience, I found it very educational. At the end of the day, we all need to get the hot pointy thing out anyway!

[KJDS]

... So at best this would require an iterative approach with lots of soldering/desoldering of components and lots of VNA time.

Been there, done that. It's worth it for the learning experience, I found it very educational. At the end of the day, we all need to get the hot pointy thing out anyway!

The last discrete LNA I did was very very close without the soldering iron. If the model is that close, then tweak the model a little to give the change you want, then get the soldering iron out. It's far far quicker to try five different variations in MWO or ADS than it is with a soldering iron, especially when you're working with 0201 or smaller.

[Howardlong]

My standard for production boards for the past four years has been 0402, I can at least reasonably hand solder them under the microscope, or if I'm doing several boards in a panel of ten or so, I can hand place them on pasted boards for reflow with just an illuminated magnifier which speeds things up. You can reasonably squeeze a track between on 0402, not sure you can get away with that on an 0201. 0201 is a rarity for me, I've only used them twice, and it has to be under a microscope. My board assembler only recently acquired a pick and place machine good enough for 0201: they tell me it'll go to 01005, I know I won't be doing that soon.

[thewyliestcoyote]

Haha discrete components, its all about the minimum number of components and doing everything in physical structure of the board, material so on, that is the real design magic is. Say hello to HFSS, FDTD, FEM, or MoM. Ok joking aside. For simple the using the smith chart is great because of the learning experience of it. But it all comes down to what you are doing, if you are making a lot of measurements or specialised measurements there is no way around it VNA's and good software. If you are a little computer savvy you can save your self thousands of dollars in software, hardware, and the kind of stupid software options for it. Again I don't know what is your frequency range but a Keysight fieldfox is hard to beat, they can function as 2 port VNA and spectrum analyzer with signal generator. If you need more ports then 2 then invest in a good switching network. Have a play with python and scikit-rf. I wrote some python to control fieldfoxes, I will try and find it and post it. I am a bit biased to Agilent/Keysight.

[G0HZU]

I have already very good cables with the ultra fast pulses it is mandatory, today I scored 27ps rising edge.

Is this called pulsenutting?
Not sure how a basic low cost USB VNA is going to prove very useful to you if you are already trying for that level of performance?

[thewyliestcoyote]

I have already very good cables with the ultra fast pulses it is mandatory, today I scored 27ps rising edge.

Is this called pulsenutting?
Not sure how a basic low cost USB VNA is going to prove very useful to you if you are already trying for that level of performance?

The only way to get smaller and smaller equivalent requirements is to measure higher and higher frequencies and good phase resolution. There are ways to get equivalent response by generating a approximate circuit (LRC and transmission lines). This can be a dangerous game to play as it extrapolation not interpolation. None of these are direct measurement of pulse response, but given some basic assumptions(trap for young players as Dave would say) like linearity and complexity or circuit. I do not know the measurement performance of the USB VNA in question. 

[eurofox]

For people who want learn the smithchart.

http://www.fourier-series.com/rf-concepts/smithchart.html

[thewyliestcoyote]

For people who want learn the smithchart.

http://www.fourier-series.com/rf-concepts/smithchart.html

There was a couple of YouTube videos that really helped me when I was just learning about the smith chart in school.

[barnacle2k]

Since we have a nice VNA thread going here:   

I have the opportunity to buy a E-CAL module (With USB and Parallel port) for my 8753E.
But i would need a 85097 interface module to connect it to my VNA, now i wondered if anyone ever tried to build one.
It seem to be only a power injector / level shifter as of this:
http://www.keysight.com/owc_discussions/message.jspa?messageID=54151#54151
Or is there software to control it from a normal PC over USB?

[cncjerry]

I bought one of those boards asian boards, I think it is generically called a winnwt?  or a nwt500?  It does a decent job for the money from 0-500Mhz.  It can also be used as a simple SA as well as a DDS signal source.  Gives impedance with phase, has some built in attenuators, etc.  The RLB that came with it works ok though I built my own directional couplers using Minicircuits transformers that seem to work better.  The NWT500, just search on ebay, only runs like $175 but it maps perfectly against my SA with TG.

If you just want scalar output, all you need is a noise source and a scope with FFT and long persistence.  That will give you a curve but you have to work to get impedance and no phase.

I am also playing around with the ADL5519 from Analog devices.  A pair of them with I/Q signal source will give you forward/reflected as well as phase though I haven't worked out the phase part.  Those chips aren't cheap, but with a 20db directional coupler you can also use it as a fairly accurate power meter across 60db or so without switching.  The app note and data sheet have good examples. 

So if you had a pair of the ADL5519 for like $9 per = 18, plus an I/Q DDS for $50 plus two DCs with connectors at $20 per, etc and a microprocessor, you would throw together a simple VNA with power measurement for under $200.  I haven't done it yet, but plan to and I will add external ADC 16 bit or more to get higher resolution than the typical micro at 12bit with averaging.

I've been looking at VNAs and commercial high-function products are very expensive.  But using the NWT500 coupled with a computer  should give you just about everything you would want.  You would have to write software to do in channel power measurement and other things my spectrum analyzer does but hey, it's a project.

Jerry

[Marco]

The Chinese "clone" uses the software for the original NWT-500, but the hardware is very different (and the firmware is slightly buggy and closed source). So yeah, it might work ... but it's very shifty.

[vk6zgo]

Again, you're not going to get Pulse response information off of a VNA, pulse measurement specifically will be done with a Communications Sampling Scope, not any old "Oscilloscope" but a ridiculous piece of kit (http://www.tek.com/datasheet/csa8000b). Pulses are a "time-domain" kind of thing. VNAs are, in a terrible simplification, a highly glorified multi-channel Spec-An with phase measurement; a frequency domain device.

Of course you can get time domain information from a VNA!

http://cp.literature.agilent.com/litweb/pdf/5989-5723EN.pdf

Interesting article--HP had a tutorial back in the early '90s describing this,

They don't teach young EEs much history,though!---

"Then, in the 70’s, it was shown that the relationship between the frequency domain and the time domain could be described using the Fourier Transform"

1970s, my bum!!

[XFDDesign]

Again, you're not going to get Pulse response information off of a VNA, pulse measurement specifically will be done with a Communications Sampling Scope, not any old "Oscilloscope" but a ridiculous piece of kit (http://www.tek.com/datasheet/csa8000b). Pulses are a "time-domain" kind of thing. VNAs are, in a terrible simplification, a highly glorified multi-channel Spec-An with phase measurement; a frequency domain device.

Of course you can get time domain information from a VNA!

http://cp.literature.agilent.com/litweb/pdf/5989-5723EN.pdf

Interesting article--HP had a tutorial back in the early '90s describing this,

They don't teach young EEs much history,though!---

"Then, in the 70’s, it was shown that the relationship between the frequency domain and the time domain could be described using the Fourier Transform"

1970s, my bum!!

I got plenty of theory (re FFTs), but in this case we have the guy who is trying to use a hammer for things that are not nails. I can also apply the FFT route on my scope to get a pseudo Spec-an, but that doesn't mean it does the job well. Unlike a spec-an, the VNA gets phase information too, so sure, you can back-interpolate, but that doesn't mean the end result is good. It's one of my beefs with test gear today: little is made where it's primary purpose is done well, but instead it's made to do too many things and does them poorly.

[tggzzz]

It's one of my beefs with test gear today: little is made where it's primary purpose is done well, but instead it's made to do too many things and does them poorly.

It is one of my beefs with many things nowadays - too many "marketing festures"[sic] and tick boxes.

[Alexei.Polkhanov]

BTW the 8753A have an infamous CRT that more than often fails. HP did trade in our A for an E, as they really wanted the A to disappear, and not show up as 2:d hand, I kept the s-parameter test set but this did not affect the trade-in Now what to do with that box??

You can buy color LCD kit for it on Ebay.

[KJDS]

Again, you're not going to get Pulse response information off of a VNA, pulse measurement specifically will be done with a Communications Sampling Scope, not any old "Oscilloscope" but a ridiculous piece of kit (http://www.tek.com/datasheet/csa8000b). Pulses are a "time-domain" kind of thing. VNAs are, in a terrible simplification, a highly glorified multi-channel Spec-An with phase measurement; a frequency domain device.

Of course you can get time domain information from a VNA!

http://cp.literature.agilent.com/litweb/pdf/5989-5723EN.pdf

Interesting article--HP had a tutorial back in the early '90s describing this,

They don't teach young EEs much history,though!---

"Then, in the 70’s, it was shown that the relationship between the frequency domain and the time domain could be described using the Fourier Transform"

1970s, my bum!!

I got plenty of theory (re FFTs), but in this case we have the guy who is trying to use a hammer for things that are not nails. I can also apply the FFT route on my scope to get a pseudo Spec-an, but that doesn't mean it does the job well. Unlike a spec-an, the VNA gets phase information too, so sure, you can back-interpolate, but that doesn't mean the end result is good. It's one of my beefs with test gear today: little is made where it's primary purpose is done well, but instead it's made to do too many things and does them poorly.

Using time domain with a VNA is a very powerful technique that many RF engineers have used to improve products. If you want a decent coupled line directional coupler then the lines alone need to be 50* ohm, and that includes everything after the lines stop running parallel, so corners and transitions to coax or a load. Being able to sort the discontinuities and measure the overall performance with the same piece of equipment is so useful when optimizing a design.

[Alexei.Polkhanov]

"Then, in the 70’s, it was shown that the relationship between the frequency domain and the time domain could be described using the Fourier Transform"
1970s, my bum!!

FFT (Fast Fourier Transform) algorithm was popularized in 1965 according to Wikipedia. I believe that application of FFT became feasible in around 1970 when algorithm and computing power were at  levels when such application became practical. I don't think they were referring to Fourier transform itself.

[tggzzz]

"Then, in the 70’s, it was shown that the relationship between the frequency domain and the time domain could be described using the Fourier Transform"
1970s, my bum!!

FFT (Fast Fourier Transform) algorithm was popularized in 1965 according to Wikipedia. I believe that application of FFT became feasible in around 1970 when algorithm and computing power were at  levels when such application became practical. I don't think they were referring to Fourier transform itself.

Yes, but the modern form of the Fourier Transform was first stated 150 years earlier, in 1820.

Hells teeth, I was introduced to Fourier Transforms in school (not university) in the early 1970s!

Youngsters indeed.

[XFDDesign]

words

Ok, and what's the sticker price on the VNA you use to do this with? Here at work, we just dropped $500,000 on an Anritsu 70GHz VNA which does almost everything under the sun (including a few methods of measuring NF). I suppose I could suggest such a beast to the OP.

[eurofox]

I suppose I could suggest such a beast to the OP.

I will give you my address and you will be very kind to ship it to me 

[XFDDesign]

Sure, if you can provide a $600,000 payment.