How to measure VSWR of tiny antennas - how to eliminate effect of VNA cables?

[rf-fil]

I find myself working on small devices these days, sometimes only few centimeters across. And in the same context, I would use a PIFA or a chip antenna. Now.. how do I actually connect my VNA in a way that the test cable doesn't become a part of the antenna system? With lower frequencies, say, below 1GHz, I have used ferrite sleeves in the past. I'm kind of at loss (no pun intended) at what to do for 2.4GHz, for example..

[coppercone2]

maybe try a thinner coaxial cable if the size is annoying  if your frequency is only a few GHz then you can use a very thin cable.


If your VNA is close to the antenna then it will effect the antenna (loading), so a long cable is like a requirement towards accurate tests of the antenna, but it might be less realistic of a result then you get if the antenna is mounted inside of a small enclosure or whatever

[selcuk]

The important thing is to calibrate the VNA on the antenna feeding point. If you calibrate it with typical SMA kits, it will be inaccurate. If you calibrate at the VNA port and add offset time for the cable length, it won't help either. So I recommend using a 0 ohm resistor (or short with solder), a 50 ohm resistor and open circuit and keeping the VNA and cable as it is while calibrating.

I normally add 0402 resistors to circuit for antenna matching network and so I use a low tolerance 0402 50 ohm resistor (FC0402E50R0BST1) as a calibration load. You can find alternatives for your PCB. You can use ferrites around the cable to decrease it's reception from the surrounding.

I've attached an image. As far as I remember the cable is RG-405 having about 28 pieces of 74270033 around it. The cable is rigid so that I can keep the orientation and bends between calibration and measurements. I keep the styrofoam and VNA fixed as well.

[Bud]

You can use VNA Port Extension feature, but mikkie mouse gadgets like nanovna may not have it. In which case the best way would be to calibrate right on the PCB at the antenna terminals, using a SMT 50 Ohm resistor, as mentioned above.

[jfet]

classic way is use attenuators like 3 or 6 db on the cables at each end.  it removes the effects of the cable by matching.

if you doing field strength measurements, you could use a very large ground plane and drill a hole just big enought for the antenna connection.  Use a attenuator again to connect the antenna thru the hole to the VNA output.

[Stringwinder]

Many  of the antenna designers I have met have used "Carbon-Based Foam Absorbers" to stop
power travelling along coaxial cables (and interfering with measurements). Far better than
ferrites that are narrowband in comparison. Resistive losses can also come from carbon fibres
applied correctly. Apply over the last wavelength(s) or so along the coax.

https://www.cumingmicrowave.com/products/carbon-based-foam-absorbers.html

The "320-2 C-RAM MT" data sheet has lots of information.

[E Kafeman]

If it is a real small DUT with integrated antenna such as a BT earphone and antenna relies on a ground as part of antenna, monopoles, F-antennas, which mirrors itself in  PCB ground => adding a measurement cable will change measured antenna impedance as cable braids will extend PCB ground size.
Correct impedance matching is then not possible if not some action are taken.
   
Even if PCB ground length is long relative actual wavelength will there be hot places along ground where a coaxial cable leaving PCB more heavily will affect measured impedance. Depending on situation can cable effect be reduced in a number of ways.
Sleeves and ferrite tubes to "cut off" cable braid works relative well at 1-3 GHz even if ferrite tubes material not are totally cutting or absorbing the braid current. Select ferrite material intended for as high frequencies as possible.
Never let test cable leave PCB ground at a ground corner or near antenna location.Especially if PCB ground is lossy from RF view or very short, such as split in small isles and traces, is it hard to avoid that coax cable braid becomes a vital part of antenna, which will result in a very different and probably poor impedance matching when cable is removed.
Sometimes must several ways be tested how to attach test cable without affect too much how RF current behaves in PCB ground. Hard to explain but I use a finger to search for places along ground that reacts more heavily at Smith chart, to find places to avoid letting cable leave PCB.
   
 Do below show how a cable is attached at a medium small PCB using just two small ferrite tubes around measurement cable. It is not ideal high impedance blockers or absorbers but it is just a marginal antenna affect by test cable remaining. https://youtu.be/RyMFun_KhAc?si=PkSXL6TAsHgfyyEe
 
 As the PCB in above video will be used close to a human head must also that be taken in account in how ground behaves and how body nearness can move around ground hot spots. If above video had been a customer project  had I probably done more detailed testing in different environments to find best impedance matching. Now was it a fictive job just to show principles how to do a 10 minutes quick impedance matching using AnTune software. 
 
>Many  of the antenna designers I have met have used "Carbon-Based Foam Absorbers"
None of my antenna design  colleges can use such material to avoid that cable braid  affects antenna impedance as its absorption effect is very low for short cable lengths, close to PCB.
 
 >ferrites that are narrowband in comparison 
   
 High freq. ferrites can very hardly be said to be narrow band but do mostly peak at 1-1.5 GHz in absorption efficiency and at 5 GHz are effect a bit too low to result in avoiding test cable very measurable acts as an low loss ground extension.
 So fare have I been lucky with antenna design and matching  at 5-6 GHz as in such cases have PCB ground been low loss and several lambdas in size which reduces problem that braid adds to ground size and hot spots can easier be avoided.
 A sleeve as alternative will always work. It is more narrow band but mostly wide enough to cover bands of interest at 5-6 GHz.

[EE-digger]

Generally, I allow for a U.FL jack on the PCB if possible.  By moving one short (capacitor) the jack looks toward the radio or toward the antenna.  I've been using short U.fl jumpers cut in half so only about 3" from my VNA cable to the DUT.  I cal to the end of the VNA cable, attach the U.FL stub then use port extension on a short applied at the PCB.  I've had good results from 0.8mm to 1mm coax, no smaller.  Most work at 2.4GHZ but some cellular antennas from 600MHz to 1.8GHz.

I will position the cable to minimize its impact as a ground plane extension.  Sometimes I'll use small ferrite tubes if the application is more critical.

[virtualparticles]

For the problem of RF current on the shield affecting the measurement: As you know, ferrites fall apart above several hundred MHz, but you can simply wrap your cable around a dowel and tape it with duct tape to keep in place. Arrange for the reactance to be 1 k-ohm or better at your frequency of interest. I use (R^2*N^2)/(9*R+10*L), R being radius in inches, L being the length of the coil in inches, N being the number of turns, and the result is in micro-Henrys.

[tszaboo]

In my experience it hardly matters at 2.4GHz. I'm not chasing after 0.1dBs of return loss, and the part accuracies are going to dominate measurement errors anyway. If you use a 1pF+/-0.1pF capacitor to do the matching, that will matter more than the placement of your coax. Of course try to get your cable out at the opposite end and do the calibration on the PCB itself.
And what I found that the quality of the 50 Ohm resistor is very important. FC series is a good suggestion. I learned it the hard way after wasting a lot of time soldering random components because the 50 Ohm standard was subpar. And use the final enclosure. But once you get ~ -15dB return loss, you are done, because you hold it a different way and your measurement changes. And there is very little you can do because the steps and accuracy of the actual components is low.
So do that, then do range testing. And realize that you are already better than 95% of commercial products.

[Randy222]

The important thing is to calibrate the VNA on the antenna feeding point. If you calibrate it with typical SMA kits, it will be inaccurate. If you calibrate at the VNA port and add offset time for the cable length, it won't help either. So I recommend using a 0 ohm resistor (or short with solder), a 50 ohm resistor and open circuit and keeping the VNA and cable as it is while calibrating.

I normally add 0402 resistors to circuit for antenna matching network and so I use a low tolerance 0402 50 ohm resistor (FC0402E50R0BST1) as a calibration load. You can find alternatives for your PCB. You can use ferrites around the cable to decrease it's reception from the surrounding.

I've attached an image. As far as I remember the cable is RG-405 having about 28 pieces of 74270033 around it. The cable is rigid so that I can keep the orientation and bends between calibration and measurements. I keep the styrofoam and VNA fixed as well.

You can cal the vna on it's S11 port, then use it to characterize the cable (transmission line), then attach the antenna. There should be enough numbers for the math to make sense. If the antenna is small then some sort of test jig would make sense.