Can a SigGen+SA+Directional Coupler be used as a Poor Man's Scalar Analyzer?

[niconiconi]

For some simple experiments, I'd like to test the level of impedance mismatch caused by component footprints on PCB microstrip layouts, up to 3 GHz. I don't have a VNA for that, but I do have a spectrum analyzer and signal generator. I believe all I need is scalar data, namely S11 magnitude in decibel, and low frequency data below 1 GHz doesn't matter, is it possible to make a poor man's Scalar Network Analyzer with an inexpensive directional coupler?

[radiolistener]

No

[hendorog]

Yes you can make an SNA that way, but you would need some sort of cunning plan to have some confidence that you were measuring the footprint instead of everything leading up to it.

[joeqsmith]

For some simple experiments, I'd like to test the level of impedance mismatch caused by component footprints on PCB microstrip layouts, up to 3 GHz. I don't have a VNA for that, but I do have a spectrum analyzer and signal generator. I believe all I need is scalar data, namely S11 magnitude in decibel, and low frequency data below 1 GHz doesn't matter, is it possible to make a poor man's Scalar Network Analyzer with an inexpensive directional coupler?

Yes, a scalar network analyzer can be made with a tracking generator and spectrum analyzer.  There is no need for a coupler.   Impedance is a complex number and you're missing half the data.   Maybe I am not understanding what you are asking or how you plan to get the complex data. 

[radiolistener]

Yes, a scalar network analyzer can be made with a tracking generator and spectrum analyzer.

How? I think this is impossible, because there is needs at least measurement bridge.

I tried to use tracking generator with logarithmic power meter and measurement bridge to measure SWR. It can be used for SWR estimation. But it works terrible and needs a lot of movement to calibrate such installation. It's much easier to buy NanoVNA because it will give much better results and more easy to use.

And I would like to see how such DIY installation will be used to inspect a small impedance variations along the microstrip line on a PCB.

[joeqsmith]

Yes, a scalar network analyzer can be made with a tracking generator and spectrum analyzer.

How? I think this is impossible, because there is needs at least measurement bridge.

I tried to use tracking generator with logarithmic power meter and measurement bridge to measure SWR. It can be used for SWR estimation. But it works terrible and needs a lot of movement to calibrate such installation. It's much easier to buy NanoVNA because it will give much better results and more easy to use.

And I would like to see how such DIY installation will be used to inspect a small impedance variations along the microstrip line on a PCB.

I suspect you are reading more into my response than what was there.  I was not suggesting that the overall answer was yes.    A tracking generator and SA can certainly be used to get scalar data. 

https://www.everythingrf.com/community/what-is-a-scalar-network-analyzer

Impedance is a complex number and will require mag and phase.   

As far as looking at "mismatch caused by component footprints on PCB microstrip layouts, up to 3 GHz", that seems like a difficult problem for a hobbyist to pull off with any level of certainty.   I would start with what ever their requirements are and work backwards from there.  Cost of the probing station, standards, ....   Seems fun to try it but costly.

****

For some simple experiments,

  IMO it opens with an oxymoron.   

[joeqsmith]

Before I bought my HP141T (which included a tracking generator)  I was using an RF generator and diode detector attached to a DC meter with some LabView software to make a poor man's scalar network analyzer.    Before that, I was using a sweep generator, diode detector and analog scope.   I ran a lot of fun experiments using these methods and learned a few things along the way but even today I'm not sure I could tackle OPs experiment. 

It would be interesting to hear more about it.

[Kalvin]

In principle yes, but in practice it is up to how careful you are with your measurements. You need to perform the calibration of your setup very carefully in order to get useful, accurate and repeatable results. It would help if you had some reference circuits available who's impedance is known at the frequencies of your interest, so that you could check your measurements and measurement setup.

w2aew has a nice introduction video on directional couplers and measuring SWR / return loss / VSWR:

[niconiconi]

Yes, a scalar network analyzer can be made with a tracking generator and spectrum analyzer.

How? I think this is impossible, because there is needs at least measurement bridge.

I tried to use tracking generator with logarithmic power meter and measurement bridge to measure SWR. It can be used for SWR estimation. But it works terrible and needs a lot of movement to calibrate such installation. It's much easier to buy NanoVNA because it will give much better results and more easy to use.

And I would like to see how such DIY installation will be used to inspect a small impedance variations along the microstrip line on a PCB.

I suspect you are reading more into my response than what was there.  I was not suggesting that the overall answer was yes.    A tracking generator and SA can certainly be used to get scalar data. 

https://www.everythingrf.com/community/what-is-a-scalar-network-analyzer

Impedance is a complex number and will require mag and phase.   

As far as looking at "mismatch caused by component footprints on PCB microstrip layouts, up to 3 GHz", that seems like a difficult problem for a hobbyist to pull off with any level of certainty.   I would start with what ever their requirements are and work backwards from there.  Cost of the probing station, standards, ....   Seems fun to try it but costly.

****

For some simple experiments,

  IMO it opens with an oxymoron.

First, speaking of complex impedance, I don't need impedance itself, just a rough estimation of how bad the (mis)match is, from the scalar return loss or VSWR. The everythingrf.com link says,

They can be used to measure parameters like VSWR and Return loss, which only requires the measurement of the magnitude of a signal at a particular frequency or frequency range. In this case, we do not need to measure the phase.

Thanks for the comments. So the apparent conclusions are:

(1) It's at least theoretically possible, and my understanding of what a scalar network analyzer does is correct. Also from what I've read, I believe I can even theoretically export the data to a PC and perform a Open-Short-Load calibration in postprocessing to remove some errors (but true deembedding is not possible due to the missing phase data).

(2) Nevertheless, what I'll be doing is essentially putting an SNA from scratch, the pitfalls are just too numerous, unless I really know what I'm doing, the data I get will likely to be bogus, and completely meaningless.

So it can be a fun "SNA from scratch" experiment, but to be able to trust the result to even the slightest degree, I still have to get some 3 GHz test equipment for comparison, or at least some reference boards with known characteristics.

Case closed, thanks for the input.

P.S: My original plan was actually to get a Tektronix 11801. I found an offer and paid. A few days later, the seller told me "We have listed this machine for a long time without buyers, it has already been dismantled for parts before your order. Please cancel the order..."    So I thought whether it's possible to try something simpler, apparently I'd better to keep hunting...

[virtualparticles]

If you can measure some complex voltages (or mag and phase) you can determine raw Source Match, Directivity and Reflection tracking then you can apply a short, open and load to your coupler and measure the results on the couple port. The correction can then be done using the mathematics in the attachment. If you can't digitize these outputs then you could apply a load to the coupler and measure the leakage (raw directivity) on the couple port. If you can then put couple port offset in memory and then subtract it from your measurements then you can measure return loss down to within 10 dB of the raw directivity with +/- 3.3 dB of accuracy or +/- 1 dB of accuracy within 20 dB of the raw directivity.

Without doing the whole 1 port calibration, the raw directivity is the "floor" for your measurement where the error goes to infinity.

You could probably get a coupler with 40 dB of raw directivity over a narrow range and make reasonable measurements of return loss to 20 dB.

There are also some nice network diagram manipulations in the article.

Best,

Brian

[hendorog]

Creating a simple SNA or a VNA from scratch isn't that difficult and a great way of learning.
Brian's doc is excellent and there are others on the www too.

Of course, as mentioned, it is very cheap now to just buy a device such as the NanoVNA. A characterised cal kit is still going to be needed if you want to make more precise measurements.

As a random aside, there is an old vector reflectometer design which uses only scalar power meters to take vector measurements.
It is mentioned in this link I found, which has some other info which might help:

https://indico.cern.ch/event/69321/contributions/2069666/attachments/1028706/1464818/Caspers-Meas-2.pdf

[Marsupilami]

inexpensive directional coupler?

Define inexpensive please. With a good coupler that has high directivity and decent match it's super easy to do.
https://www.minicircuits.com/WebStore/dashboard.html?model=ZHDC-16-63-S%2B


Connect the DUT as close as possible to the coupler
The only calibration step that you have to do is to run it a sweep without the DUT (coupler port open (I short might be better but I think for what you're trying to do it will be fine)) and normalize the SA trace so that you see a flat 0dB line. That is going to be your 0dB return loss. (all power reflected)
Then connect the DUT and you should see a fairly reasonable return loss magnitude plot.

[neilhao]

I remember that Mr. Henrik Forstén  designed such a kind of VNA.
https://hforsten.com/cheap-homemade-30-mhz-6-ghz-vector-network-analyzer.html

[joeqsmith]

With a good coupler that has high directivity and decent match it's super easy to do.

If indeed it is super easy as suggested,  I would like to see the experiment ran. 

OP didn't provide a lot of background on package types they were interested in investigating.     

... mismatch caused by component footprints on PCB microstrip layouts, up to 3 GHz.

I assume you would start with some sort of custom controlled impedance test board with your various patterns.  You need to somehow isolate the effects of the footprint.  Run the entire test on a single PCB?  OP may be able to provide some insight as to what footprints looked like.   Maybe try and use all 50ohm parts.    It seems like you may be able to get some sort of relative measurement at best.     

It seems like a job for the simulator but if it really is supper easy,  please set it up.

Thanks.

[Marsupilami]

It seems like a job for the simulator but if it really is supper easy,  please set it up.

Salty are we today, ey?

The question was

is it possible to make a poor man's Scalar Network Analyzer with an inexpensive directional coupler?

To which the answer is yes, and it is easy, depending on what do they mean by inexpensive. They didn't specify the particular setup on the board which, you're right, can make things a lot more complicated or outright impossible. However I can easily imagine a scenario where you did your math, did your simulations yet you're interested in how good or bad the built thing turned out. No, they won't be able to tell how much the actual footprint contributes to poor results, you're right, but that's not necessarily needed and was not the question.

I doubt the skepticism was about this part, but I was nevertheless interested to set it up quickly. Here's a comparison measurement of some random device I dug out of the trash pile. The red trace is calibrated VNA measurement S11 magnitude while the yellow is the SA + tracking generator through the coupler. +-1.5dB difference, that's good for a lot of things and it could be further improved. The coupler I had on hand is not ideal for the purpose.

[joeqsmith]

It seems like a job for the simulator but if it really is supper easy,  please set it up.

Salty are we today, ey?

Always. 

The question was

is it possible to make a poor man's Scalar Network Analyzer with an inexpensive directional coupler?

To which the answer is yes, and it is easy, depending on what do they mean by inexpensive. They didn't specify the particular setup on the board which, you're right, can make things a lot more complicated or outright impossible. However I can easily imagine a scenario where you did your math, did your simulations yet you're interested in how good or bad the built thing turned out. No, they won't be able to tell how much the actual footprint contributes to poor results, you're right, but that's not necessarily needed and was not the question.

I doubt the skepticism was about this part, but I was nevertheless interested to set it up quickly. Here's a comparison measurement of some random device I dug out of the trash pile. The red trace is calibrated VNA measurement S11 magnitude while the yellow is the SA + tracking generator through the coupler. +-1.5dB difference, that's good for a lot of things and it could be further improved. The coupler I had on hand is not ideal for the purpose.

Pretty much what I had responded to Radiolistener.  I read more into your post. 

[Marsupilami]

Pretty much what I had responded to Radiolistener.  I read more into your post.

Story of my life. Many people see a lot of things in me then they all leave disappointed.

[joeqsmith]

Pretty much what I had responded to Radiolistener.  I read more into your post.

Story of my life. Many people see a lot of things in me then they all leave disappointed.

Same!   

[coppercone2]

I think its going to be hard to use it with PCB but I think it might be fine for testing full modules (i.e. things you screw in line to coaxial cable)

I.e. pcb card + attenuator #1, PCB card + attenuator 2, etc.. thinking big

things need to warm up and you would benefit from good cable, connectors, etc.. big changes. detecting a package change needs really stable equipment (comparing 2 components in different footprint). i remember a thread here with a high speed 20GHz TDR that showed changes with footprint nicely. that trace 2 posts above would be as far as I go with it, otherwise you get dubious quality.. as you see its already hard to interpret because the change is small.. I imagine you wanted to test something like a 0602 vs 0402 part on a board with no other changes, thats hard, its down in the inaccuracy of such a setup

I did this alot but it was along the lines of see if inline filter or 20db attenuator component is working correctly , or if some thing like a TVS block is not doing something bad, when you are looking at a sub db change it turns into a all day affair where you are not sure of anything

but I highly recommend doing it because then you will not feel bad spending VNA money on something because you know exactly what you are avoiding, if you don't try you will never get the 'why did i buy this expensive thing' out of your head


Unless you are really short on time, performing a cheap tool investigation prior to a $$$ purchase is never a bad idea. thats called being smart and learning, and it will help you communicate with people that only have 1 method of doing things, you will know what you are saying once you do both. in a nutshell the stability and sensitivity of a integrated VNA is extreme and you will learn to appreciate that, it is essentially like buying a microscope. if you have a SA and SG, buying a directional coupler and playing with it is the next step to learning.

[niconiconi]

With a good coupler that has high directivity and decent match it's super easy to do.

If indeed it is super easy as suggested,  I would like to see the experiment ran. 

OP didn't provide a lot of background on package types they were interested in investigating.     

... mismatch caused by component footprints on PCB microstrip layouts, up to 3 GHz.

I assume you would start with some sort of custom controlled impedance test board with your various patterns.  You need to somehow isolate the effects of the footprint.  Run the entire test on a single PCB?  OP may be able to provide some insight as to what footprints looked like.   Maybe try and use all 50ohm parts.    It seems like you may be able to get some sort of relative measurement at best.     

It seems like a job for the simulator but if it really is supper easy,  please set it up.

Thanks.

To be honest, I already did some simulations and the results even looked real ( )! But there are many ways to fool oneself in a simulation. For example, in one of my simulations (I don't remember the data now, but if I recall correctly), a capacitor footprint's S11dB crosses the 20 dB mark at 1 GHz or so. But with a ground plane cut-out,  S11dB is still below 20 dB at 3 GHz. But this simulation is done using lumped capacitance, not real capacitance, with perfectly matched lumped ports, lossless metal, with possibly additional misconfiguration. So it's dubious at best.

So my original motivation is to get a reality check - can I trust my simulation setup? I want to start at something really basic, something like...

1. A 50-ohm microstrip with two SMA connectors, far-end terminated by 50-ohm coax load. One board has solid ground plane and another has cut out under the center conductor for impedance compensation.
2. A 0603 AC coupling capacitor footprint between a 50-ohm microstrip, far-end terminated by a 50-ohm coax load, One has solid ground plane and another has cut out.

Compare experiments and simulation results for different cut off sizes. If I can even detect a difference at all, I'll call it a success.

But after reading your criticisms, I think it might not be a good idea. The measurements I was trying to do will possibly be drowned into the noise floor. Looks like my goal is too ambitious, perhaps I should start from some easier experiments that are easier to measure, such as a microstrip couplers, filters, etc, 800 MHz.

[joeqsmith]

To be honest, I already did some simulations and the results even looked real ( )! But there are many ways to fool oneself in a simulation. For example, in one of my simulations (I don't remember the data now, but if I recall correctly), a capacitor footprint's S11dB crosses the 20 dB mark at 1 GHz or so. But with a ground plane cut-out,  S11dB is still below 20 dB at 3 GHz. But this simulation is done using lumped capacitance, not real capacitance, with perfectly matched lumped ports, lossless metal, with possibly additional misconfiguration. So it's dubious at best.

So my original motivation is to get a reality check - can I trust my simulation setup? I want to start at something really basic, something like...

1. A 50-ohm microstrip with two SMA connectors, far-end terminated by 50-ohm coax load. One board has solid ground plane and another has cut out under the center conductor for impedance compensation.
2. A 0603 AC coupling capacitor footprint between a 50-ohm microstrip, far-end terminated by a 50-ohm coax load, One has solid ground plane and another has cut out.

Compare experiments and simulation results for different cut off sizes. If I can even detect a difference at all, I'll call it a success.

But after reading your criticisms, I think it might not be a good idea. The measurements I was trying to do will possibly be drowned into the noise floor. Looks like my goal is too ambitious, perhaps I should start from some easier experiments that are easier to measure, such as a microstrip couplers, filters, etc, 800 MHz.

I would like to see a write up of your simulation. 

I am not criticizing the experiment as much as I have a healthy skepticism.    I wouldn't let it deter your efforts. 

For the experiment,  it seems like you would want to get a high quality board.  If you  made everything on one board, at least you could get some sort of relative data.  Or, talk with the board house and see how tight they can control your test boards.  I would imagine the differences you are looking for could be very small.  You may have to invest in good connectors, cables....  Then again, it's fairly low in frequency.

For fun, I could setup a simple experiment using a VNA and crude test board.  The boards I have are not good for this frequency.  We could run an SOL with them, then  remove say a 2mm square from under the trace on the short, or open and see what difference it makes.  Totally uncontrolled mess (the way I like to work  ).    We could run the experiment using OWOs low cost VNA.

***
Thinking about it, it would make more sense to try it with a part mounted to the board.  I had made up a test board using some ATC (prior to the AVX debacle) 100B RF jelly been parts.  Three caps were mounted to a test board configured as a shunt thru, series and shunt.  I could mount a fourth capacitor to the same type PCB, series configuration, using the same connectors.  We could then compare the return loss for the two PCBs.  It's not much of a test but may give you some confidence in the variance.   I could then remove  2mm sq from the ground beneath the cap and remeasure it.    I could provide you with details about the build that you could then attempt to simulate and see how the results compare. 

If this seems worthwhile to you, let me know and I will set it up.  Again, it would be very crude.

[eb4fbz]

Yes but your S11 or reflection measurement will be very limited due to directional coupler directivity. Read about RL uncertainty 

Most broadband directional couplers have directivity figures arround 20dB. That means you can't measure any better, and all measurements below -10dB will have a big uncertainty.

VNAs are able to cancel these sources of error.

[hendorog]

Yes but your S11 or reflection measurement will be very limited due to directional coupler directivity. Read about RL uncertainty 

Most broadband directional couplers have directivity figures arround 20dB. That means you can't measure any better, and all measurements below -10dB will have a big uncertainty.

True but its not that uncommon to find one which performs much better than the specs. I've got a few here which can get to 30dB or a bit better in this <3Ghz area.
Also it is frequency dependant so sometimes the frequency range of interest falls into a range with higher directivity.

A bridge could be used instead of a coupler for better directivity.

[DrNefario]

I would also recommend using a bridge.  Here is a paper detailing a simple one you can build.

[joeqsmith]

I would also recommend using a bridge.  Here is a paper detailing a simple one you can build.

146MHz is a bit shy of the 3GHz OP had asked about.   

I would have liked to have heard more about their simulation.