RF metrology - measuring ultra-low dielectric loss tangents

[TheUnnamedNewbie]

As some may know, I am a millimeter wave\$^1\$ researcher in my day job.

In said job, I have to work with very low loss (\$ \tan{\delta} < 10^{-3}\$) materials. And these low losses still have a big impact on our applications.
One of the things we work with are foams, but we don't really know how well the foam works - say we have a foam made of PTFE (ptfe has a loss tangent in the ballpark of 0.0002-0.0007, depending on who you ask). Can we just scale the loss tangent by the density of PTFE?

So, I was hoping to do some measurements, but this is where problems come in. Most methods I come up with (measuring transmission loss of a slab between horns, measuring resonators, etc) all end up giving you losses well below the measurement capabilities of most VNAs (which tend to be at best 0.1 dB transmission uncertainty). So how do you measure such low loss tangents?

My frequency of operation is around Ka band on this job, but more general approaches would also be welcome.

\$^1\$ EDIT: for those who don't know, millimeter waves are signals from 30-300 GHz. Above that is usually considered 'sub-millimeter wave' which I also work in, but to a lesser extent.

[Conrad Hoffman]

Measuring those losses is tough at low frequencies, certainly crazy difficult at RF. Maybe the answer is in your question- "these low losses still have a big impact..." If you have any system where the losses have an impact, that's the place to test.

At low frequencies I use traditional bridges, but I don't know if they could be practical at RF. A Schering bridge can be done almost entirely with air capacitors, so might be a candidate. Though very good for loss measurements, I don't know if it could resolve down at the PTFE level. Transformer bridges can, but again, I don't know if a design is possible for RF.

[branadic]

What about relative measurement of a capacitive test fixture with first air (measure ambient parameters such as temperature and humidity) in between and the PTFE foam afterwards? We have Agilent 4285A with 6 digit resolution, basic |Z| measurement accuracy is 0.1%, 75 kHz to 30 MHz measurement range. Depending on what you're after you should specify what is rf in your case, could help a lot.

-branadic-

[TheUnnamedNewbie]

Depending on what you're after you should specify what is rf in your case, could help a lot.

In my original post I already said the band in question was Ka band, which is somewhere in the ballpark of 26-30 GHz. The remainder of my job is above that, this is the lowest I've ever gone down to (usually we work around 140, 280 or 540 GHz)

Maybe the answer is in your question- "these low losses still have a big impact..." If you have any system where the losses have an impact, that's the place to test.

Yes, that is something we are trying to do, but the problem is isolating the loss due to the dielectric and the loss due to manufacturing tolerances. In order to figure out where our problem lies, we need to be able to separate these two, and that was what I was hoping to achieve with a dedicated measurement of the foam.

If our assumptions about the loss of the foam being simply the weighted average of the loss of the material and the loss of air (weighed by their volumetric ratios), then we should have less loss, but as I already said, we don't know if this is because our assumptions on the loss of foam was not valid (as it does seem to give accurate results for permittivity) or some other loss factor we are not considering in our simulations.

[chuckb]

About 20 years ago I set up an open air loss range at 10 GHz with better that 0.01db repeatability using a scalar analyzer. It may even have been 0.001dB but it was a long time ago.

We had two 27dBi flat plate antennas about 10m apart to try and be in the far field of the antennas. There were directional couplers directly attached to the waveguide antennas. This removed temperature dependent coax losses and told us exactly what the forward power was. The test sample was about 6 inches from the transmit antenna and it was larger than the antenna so all of the main beam energy passed through the sample.

We would average the data for a minute then move the sample (or the Tx antenna) exactly 1/4 wavelength. This caused any reflected power to change by 1/2 wavelength and the power measured at the directional coupler would change. If you have any loss (coax) between the Tx antenna and the directional coupler you can't detect the true Forward power. Theoretically you should move the sample slowly over that 1/4 wavelength to find the peak and null, but a fixed step worked fine for us.

The other main issue was side lobes from the antenna. The side lobe beam energy would reflect off the metal walls of the building, the floor, the ceiling or people walking in the area. We used flat sheets of pyramid foam absorber placed where the side lobes would be to absorb that energy from the Tx antenna. Then we place absorber around the Rx antenna to block reflection coming in the side lobe.

Make sure the sample is perpendicular to the beam path. Otherwise beam steering may reduce the energy at the Rx antenna.

You start with a pathloss test, including the 1/4 wave shift, average those numbers for your baseline pathloss. Then inset the sample and retest.

Good Luck!

[TheUnnamedNewbie]

Quick update - haven't been able to read through your reply properly yet chuckb.

I found this document from DuPont while doing some research, and it seems like an good introduction to the topic:

Low-Loss Materials in High Frequency Electronics and the Challenges of Measurement by Glenn Oliver

[Conrad Hoffman]

Very clever using the interferometer!

[branadic]

Maybe this is of interest too and worth to contact them?

http://www.polymer-engineering.de/en/research/fields/functional-polymers/thermoplastic-circuit-board.html

-branadic-

[pwlps]

If you don't want to build (and calibrate) a resonator and prefer to directly measure  the  transmission loss of a slab between horns you can increase the sensitivity by orders of magnitude using a simple mechanical chopper and lock-in detection. I mean a slab in form of rotating disk with holes (or a similar arrangement) so that the beam goes alternatively through air and the medium you are testing (don't laugh please  , you can find experiments working on the same principle in physics labs, see e.g. https://www.ophiropt.com/laser--measurement/knowledge-center/article/10378).  Then you don't need a VNA, just an amplifier and an RMS power sensor. You will need a sort of position encoder (a simple slot-type photodetector will do) to provide a reference signal for the lock-in amplifier.   With rotation frequencies of a few tens of Hz you should suppress any 1/f noise enough to be able to get a gain of 10^4 or more with reasonable time constants.  As the response should be very well linear you can calibrate it easily with similar shaped disk made of a known dielectric.

[RoGeorge]

...simple mechanical chopper and lock-in detection...don't laugh please 

That was exactly my first thought, too.  A lock-in amplifier combined with a VNA.  Yet, I have no practical experience at the frequencies used here.

My guess is one of the big problems will be to distinguish between the energy absorbed vs reflected by the slab of the tested material.  I think the VNA is still needed, and the lock-in detection should be implemented later, by software post-processing the logged measurements over long enough periods of time.