I’ve been considering adding a warmup exercise to the class-D amplifier lab, having the students plot Ron for the FETs vs. Vgs. The problem is not completely trivial, since we want to measure small resistances around 100mΩ. To get sufficient resolution with the ADC on the Teensy boards, we want a fairly high current at those low resistances, which means using a small load resistance. If I use 0.25W resistors and a 3.3V supply, I can’t go less than 43.6Ω without exceeding the rating for the resistor.
My first attempt is to measure just my nFETs. I generated Vgs with my FG085 function generator doing a slow triangular wave (24s period) set to 4Vpp and a 3.1V offset (so Vgs is swept from about 1.1V to about 5.1V). I also put a 470µF capacitor across the function generator outputs, to smooth out the steps of the low-quality DAC a bit (which wouldn’t be needed on a better function generator with a higher-resolution DAC). To measure the gate voltage, I used a pair of 6.8kΩ resistors to cut the voltage in half, to keep it in the range of the Teensy ADC (0–3.3V).
To provide large currents, I used the 5V USB supply (instead of the 3.3V regulated supply, which cannot supply much current on the Teensy 3.1 board and even less on the Teensy LC). To measure the supply voltage (Vdd) and the drain voltage (Vdd), I used two voltage dividers, each with a pair of 8.2kΩ resistors. It turned out I had two different batches of 8.2kΩ resistors, which were slightly different, so I had to redo the measurements after rearranging the resistors to make sure that each voltage divider used resistors from the same batch.
I used a nominally 16Ω (measured at 15.3Ω) 10W resistor as a load. At 5V this would dissipate 1.6W with a current of 327mA (within the USB limits).
I first made measurements with no FET—both with and open circuit between drain and source test points and with a short circuit between the drain and source test points. The open-circuit test gave me an estimate of the maximum on-resistance I could measure (and is how I detected the mismatches of the 8.2kΩ resistors). The short-circuit test gave me an estimate of the test fixture wiring resistance, which needed to be subtracted from the measurements. I got about 134mΩ of wiring resistance, which is large compared to some of the power-transistor on-resistances, but fairly small for long breadboard wiring.
I tested all 6 nFET transistors that I have: 4 power nFETs and 2 low-power ones in TO92 cases:
I was not very impressed with these measurements and decided to do them over using the more obvious technique of using a 47Ω resistor to the 3.3V supply as the load and recording Vds and Vdd directly, without voltage dividers (still using a voltage divider for Vgs, though). The wiring resistance came out smaller this time (77mΩ), but it varied a bit when I reran the same measuring process, so I’m a bit dubious about the Ron measurements for the power transistors—changes in the wiring resistance estimate were larger than the final estimated Ron for AOI514.
The results were somewhat better with the larger load resistor and smaller drain currents.
The gate voltages at which the resistance dropped abruptly seems to have shifted lower when using the 15.3Ω load to 5V. I’m not sure why this is, as the same voltage divider was used for the gate voltage in both cases. For the smallest transistors, the problem may have been overheating, as they were only rated for 200mA or 300mA of current and I was delivering about 325mA. But the bigger transistors should not have been heating much, and even on the small transistors I’m not seeing the hysteresis you would expect from slowly heating and cooling the parts.
I next need to look at whether a similarly simple pFET circuit will work for them to measure Ron for pFETs.
Testing all the nFETs I have on hand should also have been useful for deciding which ones to use next year, but the AOI514, AOI518, and NTD5867NL-1G are no longer available. In any event, the AOI thresholds are a bit high for using with 3.3V logic, but the PSMN022-30PL,127 (39.06¢ in 100s) looks perfectly reasonable with an on-resistance of about 47mΩ at 5V and a threshold voltage around 2.08V (where Ron=1Ω). I should probably get and test the NTD4858N-35G nFETs (38.56¢ each in 100s) that the students used this year, as well as the IPU50R950CEAKMA1-ND 500V nFETs (47.2¢ in 10s) that my son and I are thinking of using for the theater-light LED drivers.
I think that doing nFET measurements for a single nFET should be fairly quick for students, though convincing them of the need for “drain-source short” test to subtract off wiring resistance may be difficult. I next need to look at whether a similarly simple pFET circuit will work for them to measure Ron vs. Vgs for pFETs.