I have had a lot of time recently to work on personal projects. To aid in this, I picked up a NanoVNAv2 from Tindie. One month after placing my name on the waitlist I was contacted that the hardware was back in stock. Less than a week later, it arrived to my lab bench.
I found a measurement candidate in a variable air capacitor I picked up at a ham swap. I had used it with brief success on a homebrew antenna, so I figured it may prove useful as a test load (device under test, or DUT) in a network measurement.
Opening up the enclosure and looking over the connections within the project box I found a loose wire. After refastening, this capacitor was ready to measure. See the circuit below to illustrate the electrical connection within the project enclosure.
Ok, great, I have a DUT… I want to measure its capacitance, but how? Looking over the specifications of the NanoVNA, I determined I could get the most accurate measurement by performing a shunt two-port measurement of the capacitor and deriving an impedance and ultimately a capacitance. A single port measurement can provide usable accuracy for impedance measurements near system impedance but the two port measurement will be more accurate in general as it is essentially a high frequency 4 wire measurement.
Now, for the logistics- I didn’t have the correct adapters, but I did have some coax and an assortment of connector ends, so I decided to make my own test fixture for this measurement. I made a two-port SMA-to-UHF connector so I can measure the (capacitive) load shunt to ground. I also floated solder into the end of a UHF connector until it created a short. The new parts and a schematic representation for the circuit for measurement:
I left the connecting coax length a bit long, but a quick calculation assured me I should not expect to see the impact of this line length in the data at the measurement frequency (at max sweep frequency, the coax is one-quarter wavelength), so I reasoned the inclusion of the transmission line in the measured data would not prevent an accurate evaluation of the NanoVNA’s functionality. As the soldering iron cooled I performed a T/R calibration (this NanoVNA has a single transmit port) and began measurement.
Following measurement I began plotting data. I have plotted the open circuit parameters (also known as Z parameters) to determine the amount of capacitance in the circuit:
Now I can pick a frequency where the negative-going portion of the curve is approximately linear, say 10 MHz. At this frequency the impedance (magnitude) values yield an approximated capacitance of 79 pF, 52 pF, 25 pF, and 51 pF.
This is a good start. However, there is still the length of coax that is embedded in the measurement. An easy way to approximate the capacitive contribution of the coax is to use the capacitance per foot value from the cable datasheet (30 picoFarad/foot) and, realizing it is in parallel to our DUT, subtract the length-adjusted value from the measured capacitance to account for the extra capacitance of the line within the measurement setup (30pF/ft * 0.833 feet = 25pF). I’ve tablulated the data to summarize the measurements, including deembeded capacitor data using the method above to estimate the contribution of the fixture:
|Knob Position||Measured C||Estimated DUT|
|0||79 pF||54 pF|
|90||51 pF||26 pF|
|180||25 pF||0 pF (not measurable)|
|270||52 pF||27 pF|
Overall, I am satisfied with my new measurement tool and expect it will serve to improve my future homebrew designs. I’m similarly pleased to have a measured capacitor around the lab to open up antenna configuration possibilities. Although this post is intended to serve as a summary of my evaluation of the NanoVNA, the electrical model for the capacitor I have created from these results can be improved upon and verified. If you have any suggested changes to my measurement, drop it in the comment section- I am always open to improving my test methodology.
One feature I’d like to see implemented in a future NanoVNA software update is a logarithmic sweep option to make these power-type measurements easier. As you can see in the plot, the capacitive portion of the response has a non-linearity at low frequency. Another area for improvement in the current software is the uncorrected error introduced in the return loss data due to the non-characteristic termination at the port 2 reference plane. An additional measurement with ports 1 and 2 reversed at the DUT (plus some math) will include all error terms needed to remove the cable length terminating port 2.
I should acknowledge there is another way to solve this problem: by treating the measured data as a transmission line, the DUT can be transformed the electrical distance of my transmission line number of Wavelengths Towards Generator (recall WTG from the smith chart) to extract the admittance of the DUT. Usually this can be done by using port extensions on the VNA, which mathematically subtracts phase from the scattering parameter measurement, based on a specified time or electrical distance. However, due to the shunt nature of the measurement the math isn’t as straightforward; the s-data will be converted to admittance parameters before the phase is subtracted. This will require some python trickery, since the nanoVNA software doesn’t support admittance parameters. After I’ve wrapped my head around the math, I’ll plot the data in python.
Have ideas? Leave a comment on how you’d approach this problem.