Practical Impedance Matching for the WISP
The following tutorial describes methods for tuning the RF front end of the WISP
You should have some knowledge or background in RF engineering. A good reference is: “Microwave Engineering” - David M. Pozar
You should know how to use and calibrate a vector network analyzer. I suggest making friends with somebody that uses VNA regularly. There are many pitfalls and nuances to making good RF measurements that cannot be covered by this tutorial. It is best to leverage the knowledge and experience of others to help you. Remember, graduate students can usually be bribed with a six pack of beer (good beer)!
You fabricated your own WISPs. Even if your PCBs were made exactly to our specs you will need to measure and tune the matching network. There is quite a bit of a variation for one batch of WISPs to another due to the variations in component and PCB manufacturing.
Need to retune the WISP for your region's frequency band
You want to add a custom antenna
You do not need to tune every WISP manually. Once you have the correct matching network values for your practical batch of boards, matching should stay consistent form one WISP to another.
First read the following paper on impedance measurements of UHF RFID Chips. It is a good place to get orientated with the issues surrounding RF measurement and impedance matching. One thing to keep in mind is that the WISP has different advantages and tradeoffs compared to IC tags. For instance the WISP has a discrete matching network which we will use to tune the RF front end, and the user has access to internal signals such as the rectified voltage, which will be handy when determining sensitivity.
Nikitin, P.V.; Rao, K.V.S.; Martinez, R.; Lam, S.F.; , “Sensitivity and Impedance Measurements of UHF RFID Chips,” IEEE Transactions on Microwave Theory and Techniques, vol.57, no.5, pp.1297-1302, May 2009
“A balun is a type of electrical transformer that can convert electrical signals that are balanced about ground (differential) to signals that are unbalanced (single-ended [i.e. referenced to earth ground]) and vice versa. They are also often used to connect lines of differing impedance.” http://en.wikipedia.org/wiki/Balun
For example:
A dipole antenna has a balanced output
Patch antennas, like the ones used by the RFID readers, are unbalanced
A Vector Network Analyzer (VNA) is unbalanced
Thus if you want to measure the input impedance of a dipole antenna with a VNA you will need to use a balun to convert from the balanced VNA to the unbalanced antenna. If a balun is not used the shield, or reference, of the coaxial cable leading from the VNA to the antenna will begin to radiate energy. If the user touches the coaxial cable in difference places the impedance of the antenna will change. Clearly this is not a valid measurement. As a side note it is important to remember that accurate measurement of an antenna’s input impedance is not just a matter of using a balun. It is necessary to consider the RF environment around the antenna.
One of the biggest misperceptions regarding RFID tags is the need for a balun between the antenna and IC. WISPs, like almost all RFID tags, can be driven in a balanced mode. In this case the differential RF signal is injected into the (+) antenna node and the (-) GND node of the rectifier. The result is that the GND plane is floating with respect to earth ground.
Now you may think that you must have a balun to convert for the balanced antenna to the un-balanced WISP analog front end. However, this is not necessary and would be a waste of precious RF power. The reasoning for the lack of a balun is that RFID tags have very, very low capacitive coupling to earth ground, thus their ground plane (or silicon substrate) can truly float with respect to earth ground. The same argument is true for the WISP. The only caveat is the ground plane on the WISP is oscillating at 915MHz, and if an antenna designer is going to do detailed simulation of an dipole-like antenna they should include this extra piece of metal into their design.
Refer to the WISP firmware tutorial for the corresponding WISP model: WISP 5, WISP 4.1
For this specific application, there are two files that need to be loaded. These files are saved as pinDefWISP4.1DL.h and pinDirectionsAtStartUp.c Click the links to download the files. These files will force the WISP to operate in sleep mode only. The reason why this is important is because, when the WISP is operating at full capacity, the impedance of it varies significantly based on what operation it is performing and the input voltage. When the WISP is in sleep mode, the only determining factor on input impedance is the input voltage. Also, since the WISP consumes so much power during full operation, it is impossible to feed enough power to indefinitely power the WISP, so the impedance of concern is that which occurs in sleep mode as this is the mode at which the WISP storage capacitor can be charged. If you want to write your own code to do this, or adapt existing code, that's of course fine. The key is to minimize the time that the micro spends awake after power-on-reset (POR). The code should set the port pins to their correct states to minimize power consumption in the external circuitry and then go immediately into LPM4.
In order to connect the WISP to the VNA a SMA connector must be mounted to the WISP. Take a through hole, female SMA connector and cut three of the four ground pins off. Then, solder the modified SMA to the WISP as shown below. The addition of this SMA, while necessary, will change the apparent input impedance of the WISP unless it is calibrated out. The quality of your soldering job will make a significant difference in the reliability and repeatability of your measurements. Beware of cold solder joints, too much solder / globs of solder, and mechanically weak joints that shift when you connect a SMA cable.
The remaining ground pin of the SMA should be soldered to the gnd antenna port on the WISP. Like wise the center signal pin on the SMA should be connected to the positive antenna port.
WISP with SMA connector
WISP with SMA (front view)
WISP with one antenna cut and trace ready to be removed
WISP with both legs of the antenna cutoff, SMA on back side
Now that we have added the SMA connecter to the WISP we need to calibrate it out since it has an unknown amount of electrical length and reactance.
The figure below shows a custom calibration kit. This is the same technique use in Nikitin et al's paper (see the primer above). It is important to use the exactly same type of SMA connectors as was used on the WISP. Remove the extra gnd legs on there SMA connectors as before. Next simply solder a short across one of the SMA connects to create a “Short”. Next solder a 50 Ohms chip resistor to the second SMA connector to create a “Load”. Make sure the buy a tight-tolerance resistor. For the ‘Open” just leave the signal and one remaining ground pin of the SMA open. Now use your custom “cal kit” to calibrate the VNA using its internal calibration process. Make sure that the VNA is warmed up before calibrating, otherwise the results will drift as the unit warms up. The warm-up time will vary between different models of VNA, but 30 minutes is a good starting point.
Custom Calibration Kit (must use same SMA connectors as used on the WISP)
Everybody makes the same mistake when they go to measure the input impedance of the WISP. They think …
“I’ll just short out the inductor so that I can directly measure the input impedance of the rectifier. Then I’ll use my favor impedance matching calculator to find the L-match values. Slap down a matching network and I’ll be done, right?”
That would work if you were matching to a passive load. Diodes are non-linear devices. Seriously, they are non-linear!
If you measure the rectifier without a matching network you will basically see a short. This means most of the RF energy will be reflected off the diodes. Since it is a short the incident voltage at the input node will be small and thus the diodes will not turn on. During the matching process the voltage incident on the rectifier increase and the diodes will begin to turn on. More and more current will pass through the diodes and the equivalent input impedance will change. Therefore you will need to add a matching network that is about correct. Measure the rectified voltage (Vout) to insure the diodes are actually turning on. Then modify the matching network to get you closer to your desired input impedance.
Should we adjust the VNA output power such that the WISP VOUT = 2.0V. Then take a measurement. Then adjust match. Then repeat. ??
You will need an inductor kit, a variable capacitor or two, and a capacitor kit in order to achieve a well-tuned WISP front end. The kits must be RF-grade, not conventional inductors and capacitors. A good starting point for component values in the WISP L-match circuit are given in the tutorial below, but be aware that your final values will likely differ significantly from those shown once you're done with tuning. This is why it's important to have a full inductor kit at hand while tuning, as well as a few choices for variable capacitors. We recommend using Coilcraft inductors and Murata TZY2-series variable capacitors.
This application allows for easier selection of values for the inductor and capacitor in the L-match circuit. It will produce expected input impedance in the form of a Smith chart when given a matching network and load impedance, and can thus be used to determine a good starting point for L and C values. However, it will not accurately model all aspects of the system and therefore you must still use an iterative process to find the best component values.
Open the above URL and click on Downloads. Use the latest version of the “Smith” application.
See above for a tutorial on putting the MSP430 into a low power mode.
Because the impedance of the WISP front end varies greatly with input power, the test power used while tuning should be chosen carefully. We recommend setting your VNA test power to -9dBm to ensure a good match at the proper power level. This is based on previous experience, and is based on aiming for a Vout of 2.0 V. Your application may have unique requirements for which a different Vout level is more appropriate. This may mean that you need to use a different power level. Be sure to recalibrate the VNA whenever you select a different power level.
The following figure illustrates the impedance measured across a harvester with a shorted inductor in the L-match circuit. The numbers you determine using this method can be used to select an inductor and capacitor value. NOTE: The circuit under test in this example is NOT a WISP front end, but is a similar RF harvester. It is aiming for a 50-ohm match, not the 70-ohm match of the WISP. Your impedance values and component values will be different.
Shorted inductor, No capacitor
Open up the Smith chart application described above, and put in a starting point for load impedance using the “Keyboard” button at the top. This will be the value reported by the VNA when the inductor is shorted. Next, insert a series inductor and then parallel capacitor using the buttons at the upper right. Each component will have some effect on the circuit. Play around with the values until the Smith chart shows a good match. NOTE: The circuit under test in this example is NOT a WISP front end.
First approximation of component values using Smith v3.10.
As a first guess, you may choose to use the component values reported by the Smith chart application. Alternatively, you can use the values we recommend below. Adjust the variable capacitor to its minimum value (0 degrees). Reconnect the WISP to the VNA and determine what effect the components had on the impedance. Adjust the variable capacitor to determine its effect, but bring it back to a minimum value before further iteration.
Example starting point for L: 6.8 nH Coilcraft inductor (0402CS-6N8XGL)
Example starting point for C: 4.5-20pF Murata variable capacitor
MFG PN: Murata TZY2R200A001R00
Digi-key PN: 490-2003-1-ND
The following figure illustrates the effect of too little inductance in the L-match network. If one imagines drawing an arc of constant-admittance, it would intersect the mid-line at too low an impedance. The variable capacitor is set to its minimum value. NOTE: The circuit under test in this example is NOT a WISP front end.
Inductor and Capacitor added, variable cap set to 0 degrees. This network has too little inductance.
The following figure illustrates the effect of too much inductance in the L-match network, as indicated by the impedance being much too high. The variable capacitor is set to its minimum value. NOTE: The circuit under test in this example is NOT a WISP front end.
Inductor value increased slightly, variable cap still set to 0 degrees. This network has too much inductance.
The following figure shows a circuit with the right amount of inductance. The variable capacitor is set to its minimum value. NOTE: The circuit under test in this example is NOT a WISP front end.
Inductor value decreased slightly. Variable cap still at 0 degrees. This circuit has the right amount of inductance!
The following figure illustrates the effect of adjusting the variable capacitor in a circuit with the right amount of inductance. In this plot, the circuit is best matched (to 50 + 0j Ω) when the capacitor position is 75 degrees. NOTE: The circuit under test in this example is NOT a WISP front end.
Assuming your first guess didn't result in a perfect match, you now begin the iterative tuning process. Choose a new inductor value based on the impedance measured with the capacitor at a minimum (0 degrees), and repeat the measurement until you're able to achieve a good match.
If you're planning to make more WISPs with the same characteristics, de-solder the variable capacitor and measure its value. Now you can use (much cheaper) chip capacitors with this value on future WISPs. Please note that a discrete capacitor will not behave exactly the same as a variable capacitor of the same value, so you'll need a little trial and error. The value of the variable capacitor is to enable you to quickly adjust the reactance to zero and see if the impedance is close to your target and thus determine if you have the correct inductor or not. Once you're sure of the inductor, replace the variable cap with a discrete cap and re-measure. If the resulting impedance is inductive, increase the value of the capacitor; decrease it if the impedance is capacitive.
There is still a several key topics that need to be present for this tutorial. Feel free to bug me if you need more info and I have not yet posted it.
measuring sensitivity
known issues with VNAs
Jeff Braun - kennebraun and Alanson Sample - alanson maintains this section
Below is a link to a PDF that describes how to measure the input impedance and the power consumption of the WISP using a balun.
This procedure is not as simple as the method described above and is generally more appropriate for very accurate WISP impedance measurements or for measuring the input impedance of the antenna.
How To Measure the Input Impedance and Power Requirements of the WISP.pdf
Altium CAD Files form the PCB mentioned in the above document: Calibration PCBs With Baluns.zip
The above document mentions a MatLab file titled OnePortErrorModel.m This matlab file can downloaded using the following link: OnePortErrorModel.m