VNA measurements of pcb transmission line

[mik4el]

I’m an absolute beginner using VNAs but have started playing around with measuring antennas. I’m using the PocketVNA. Now I’d like to understand the characteristic impedance of a transmission line on a pcb.

Measuring S11 on antennas have been working well for me but I’m a bit lost on measuring transmission lines on a pcb.

As I understand the method:
1. Make sure the VNA is warmed up.
2. Calibrate the VNA with open, short and load on port 1.
3. Connect DUT with one end of the PCB to port 1 and the other end terminated by a 50R load.
4. Measure S11 for desired frequency.

If the transmission line is working correctly (50R Z0) then shouldn’t the |Z| on a Smith chart be 50R? It is not for me.

The system I have setup is a U.FL contact soldered to one end of the transmission line and GND. In the other end there is an SMA contact. The SMA contact is directly connected to port 1 on the VNA. On the U.FL contact I have a U.FL to SMA coax. On the end of that SMA I have a 50R load. The VNA is calibrated directly at port 1. The connection is:

50R load -> sma to ufl coax -> ufl contact -> pcb transmission line -> sma contact -> VNA port 1

This is how the system looks connected up:



For 800-900 MHz I measure |Z| at 22R in this system. See the smith chart below that also shows the calibration. The calibration is not perfect and could be improved I guess but looks like a bad calibration alone doesn’t explain my results. Either my transmission line is way of (which it shouldn’t be) or I’m doing something wrong when measuring (more likely).


smithchart_50R_on_ufl.png

Another question, if I compare gain on the 50R load placed directly at the VNA compared to placing the 50R load at the U.FL coax after the transmission line I see a difference in S11 gain at 15dB (-10dB vs -25dB). What conclusions can I make from this? Does this mean that there is 15dB of loss in the system?


50R_gain.png

Also, if I look at the Smith chart for all the measurements I see a big difference in comparing measurements done directly at the VNA and measurements done with the DUT in between. How should I interpret this? If the transmission line was ideal the measurements would be similar right? I suspect the coax from U.FL to SMA has a big impact here but I can’t confirm since I don’t have any means to only measure that cable.


Smithchart.png

Right now my plan is to proceed with finding ways to measure the ufl to sma coax cable to understand its impact and also to calibrate with that cable attached. But anyway I would appreciate advice on how to proceed and what I’m missing Are there any good guides on how to measure pcb transmission line characteristic impedance with a VNA? Is it meaningful at all to use a device as the pocketvna and my cheap cables and connectors or do I need professional grade stuff for this?

Thanks!

[xaxaxa]

It's pretty obvious that either the u.fl cable or your 50ohm load is bad; I would guess it's the former. Btw you can make a really decent 50ohm reference by soldering 2 100ohm resistors to a sma female connector.

[Bud]

I am not familiar with the VNA gadget that you are using so I do not know why they use Gain term for S11. Nobody is using Gain term for S11, for S11 one port measurements you use Return Loss term. Gain is used for 2 port measurements such as S21.

To measure transmission line impedance using a VNA a few conditions must be met, read here
https://chemandy.com/technical-articles/measuring-track-characteristic-impedance/measuring-track-characteristic-impedance-article1.htm

But it appears you need to sort out your calibration and regular 50 Ohm S11 measurements first.

[TheUnnamedNewbie]

I agree that gain is a strange (incorrect) name for that graph. Return loss would make more sense, or just 'S11'. That aside.

What you are seeing in the graph is the ratio of reflected power to the transmitter power. Ideally, no power is reflected and the S11 should be -infinity. An S11 of -25 dB means that the reflected power is -25 dB weaker than the transmitted power. That is pretty standard for a simple SMA connector.

S11 of -10 dB thus means that your UFL+PCB (and from what I can gather from the image, a male-male SMA adapter) is reflecting a lot more power (15 dB more, in fact). You are running a quite narrow frequency sweep, so it is hard to tell where this power is reflected. If you run a larger sweep, and you see a slow pattern over the band, it is probably due to the cable. If there is no pattern you can see, it is likely happening at one of the adapters before that cable, or the connector of the UFL line. Now, looking up the UF.L connector, the maximum VSRW is around 1.6, which corresponds to the ballpark of -10 dB S11. In other words, what you could be seeing is just the performance of your UF.L connector.

[hagster]

Your PCB tx line looks very short ~5mm
 That is very short compared to the wavelength you are using. The Zo will be dominated by effects of the connectors and solder joints.

My advice to test it is to build another PCB of the same spec, but just have one long tx line(at least lamda/2). Put an SMA at either end. You can even have multiple track widths and select the best.

[mik4el]

Thanks everyone for great advice! I'm learning a lot from this.

Good point in the "Gain" discussion. I was a bit confused by that term since I've only learned about return loss before. Thanks for clarifying.

Very interesting to learn about the U.FL connector. I'll try to do another measurement with just the U.FL coax later next week. Maybe that can make this clearer.

For my next run of PCBs I'll definitely try to include a better test PCB since that will isolate the transmission line measurement better.

[michael2]

Take the time and measure S11 of Port 2... I promise you will roll your eyes...   

Port 2 is broken on the Pocket VNA. The box is not worth its money... 

[mik4el]

Take the time and measure S11 of Port 2... I promise you will roll your eyes...   

Port 2 is broken on the Pocket VNA. The box is not worth its money... 

So you mean taking a male to male coax and put from port 1 to port 2 and measure S11? What should one expect to see?

[Mechatrommer]

50R load -> sma to ufl coax -> ufl contact -> pcb transmission line -> sma contact -> VNA port 1
This is how the system looks connected up: ... <picture>

i'm not rf/vna expert but if you are me, i would not do such a way in the picture for S11 measurement. imo you are characterizing your board combined with the long coax and 50 ohm terminator, not your board alone. i suspect there will be alot thats going on in the long coax. i would prepare a smd resistor pad on the board and just put 1 x 50 ohm resistor (or 2 x 100 ohm) so ambiguity due to long coax ("add on" transmission line) will be lessen. but thats is still half of the story, the real S11 measurement should be made when everything (IC, passives etc) are populated and in real operation, my 2cnts. another thing is S21, but since your subject is S11, i wouldnt go further, ymmv.

[mik4el]

i would prepare a smd resistor pad on the board and just put 1 x 50 ohm resistor (or 2 x 100 ohm) so ambiguity due to long coax ("add on" transmission line) will be lessen. but thats is still half of the story, the real S11 measurement should be made when everything (IC, passives etc) are populated and in real operation, my 2cnts. another thing is S21, but since your subject is S11, i wouldnt go further, ymmv.

Yes and that is also more inline with other guides I've seen now. Will try!

[mik4el]

I got a UFL-SMA adapter today so I could do a few more measurements. My measurements shows clearly that there is a big difference in S11 measurements on the cable vs directly on the VNA (no surprise). I also noticed an improvement covering the UFL coax with ferrites so I used that for my new measurements and recalibrated with a setup as shown below:



Plotting the S11 measurements for different setups shows clearly that there is a big difference from measuring on the UFL coax and measuring on the VNA. There are for example clear differences around 1.3GHz and 2.6GHz:



For a new measurement I also tried the method of soldering 2X100R 0402 resistors connected between the end of the transmission line and GND:



Plotting all the measurements 800-900MHz on a smithchart showed that the calibrated and ferrite covered U.FL coax at least provided more reasonable results. The 50R pcb measurement also looked more stable than the U.FL coax measurements:



If the 50R pcb measurement is the most reasonable this however shows that impedance is around |Z|=25-30R for 800-900MHz.

At this point I'm not sure what to do with these results. I think for the future I have higher trust in the method of putting a resistor on the PCB. What do you think? How could I improve my methods and measurements?

[LukeW]

For your PCB the line is so short it's unlikely to be critical anyway.

You've got the DUT, then a uFL connector (I always hate these) then some unknown uFL to SMA conector, then a SMA terminator. All these unknown variables between the DUT and the terminator.

It's hard to measure PCB characteristic impedance of your CPWG/microstrip on your "real" PCB layout because it's hard to connect to.

Make a test PCB with a test microstrip width, from the same fab with the same PCB stackup.

Make a transmission line between two SMAs, and/or a transmission line terminated to ground with 2x parallel 100R 1% 0603.

If you design your PCB with a pi network before RF reaches the IC, you can use it for many things like filtering, impedance matching, attenuation, as well as easily attaching a 50R termination to the end of the PCB transmission line.

[radioactive]

I can't tell from the images of the pcb, but it appears that you have a hand-soldered bare wire from the u.fl connector ground to a distant pad for ground?  If so, that is probably most of the problem.  That is a pretty large inductance as you go up in frequency.  The other thing is it appears that you just have one via on each of the SMA connector grounds with a short trace stub.  That is also inductive and probably why you are seeing a resonance around 3.5 GHz.