I’m running out of the AOI518 nFETs, and they’ve been discontinued, so I decided today to try characterizing replacement parts that I had bought some time ago: PSMN022-30PL nFETs. Since I now have a function generator (the FG085) that can do slow triangle waves with reasonably good control over the voltage range, I decided to use that to drive the test and PteroDAQ to record the data. The use of PteroDAQ limits me to 3.3V on the analog-to-digital converter, but I was mainly interested in determining where the nFET turned on, not measuring how it behaves with high voltages or currents, so that’s ok.
The circuit I used started out very simple, with just the function generator, the nFET, a current-sensing resistor, and the KL25Z board for PteroDAQ, but I later added a pair of op amps as unity gain buffers, because I thought I was having some trouble with high-impedance inputs to the ADC:
I set the function generator to create a 3.3V triangle wave with an offset of 1.7V and a period of 15s, then recorded E21, E20, and E20–E21 at a 60Hz sampling rate. I tried a wide range of resistor sizes, to plot over several decades of current:
To get the straight lines to fit so nicely, I had to do two things: throw away any points where E20–E21 was less than 2mV and add a tiny offset to the E20–E21 voltage (1.2mV). I don’t know whether this offset was coming from the op amps or from the differential amplifier in the KL25. I believe it is from the KL25, since I have seen it in other measurement contexts, even though the offset exceeds the worst-case spec for error on the ADC. I might have been able to get the lines to superimpose better if I had been able to measure the resistors accurately, but I had to rely on the color code, because my Fluke multimeter is broken and my cheap multimeters aren’t accurate to 1% on measuring resistance. I think that the 1kΩ resistor is a bit off (it was a 5% part, not a 1% part).
The interesting thing for me was the negative dynamic resistance around 5.8mA to 68mA. As the current increased, the voltage across the diode-connected nFET dropped. I’ve seen this behavior in nFETs before (see More mess in the FET modeling lab and Negative resistance oscillator), though only in the discontinued NTD5867NL. In that experiment I was able to get a high enough current (around 100mA) to start getting positive resistance again, but I was not able to achieve that with this setup—with smaller resistors to get higher currents I lost the negative resistance and got a nearly constant 1.8V for all currents.
I still have no idea what causes the negative dynamic resistance, nor why some nFETs have it and others don’t (I never observed it for the AOI518, for example). I’ll have to ask some experts on device modeling if this is a commonly observed and modeled phenomenon. It doesn’t appear in the intro books that are I’ve looked at in the field.
(Update 2015 July 8: the I-vs-V curve for this transistor and even more for the NTD5867NL in the More mess in the FET modeling lab post looks to me like a thyristor effect, but I have no idea where in the FET structure a 4-layer thyristor NPNP would be formed.)