Some results from the MOSFET tester project!

The 120 m 50 V BUZ71A that served as the crash test dummy while I got the thing working:



A detail of the interesting area near the origin:



The datasheet drain resistance values are the maximum values, so they’ll generally be higher than what I measure.

A plastic-encapsulated W7NB80 with a 1.9  (!)  drain resistance, due to its 800 V (!) rating:



Hold the gate voltage constant at 10.0 V and step the temperature from 0 °C to 50 °C:



I haven’t figured out how to get the actual temperatures from the Gnuplot input dataset to the graph without knowing them in advance. The “index” is simply the 0-origin block number, which conveniently (and coincidentally) lines up with the 0 °C to 50 °C temperature range.

An overview of a 400 m 200 V IRF630:



The juicy part:



And the variations with temperature:



A 1.5  200 V IRF610, another high-resistance transistor:



The temperature variations:



The winning entry for high resistance, though, is the 500 Ω (!!!) BSS127 that emerged from a paper on current sensing using mirror FETs for temperature compensation. It has a 600 V rating, but I have no idea why such a high drain resistance makes any sense in a SOT-23 package. They’re obsolescent and I won’t buy any just to have ’em around.

Just for completeness, a 1  1% resistor:

Resistor - 1.0 ohm

Resistor - 1.0 ohm

And a 100 m 1% resistor:

Resistor - 0.1 ohm

Resistor - 0.1 ohm

It turns out that the wire leads I soldered on contributed 6 m to the total, so the tester actually reports the truth! I checked that by passing 1.000 A through the resistor, which put 100 mV at the base of the resistor pins, then measuring 106 mV at the end of the wire leads. One can quibble about voltmeter accuracy, but it’s pretty close and much better than the ohmmeter accuracy at that resistance.

The firmware forces 0.0  for drain current identically equal to 0.0 (it’s a floating point number cast from a 10-bit unsigned integer) to avoid numeric explosions. The next few points away from the origin show the effect of small errors on small measurements; the voltage resolution is 15 mV and the current resolution is 2.5 mA; you can actually see the steps near the origin.

All in all, a fun project…

Need the datasheets? Ask your favorite search engine for, say, IRF610 datasheet. That should do the trick.


  1. #1 by Bill Rutiser on 2012-03-23 - 09:53

    The omega symbol isn’t displaying in the text. It shows up as a small hollow square box. Google Chrome 17.0.963.83 m. As far as I know this should be a pretty default configuration.

    I spent far to much of my professional life trying to make funny symbols display correctly on a terminals and PC’s. It seems like with modern standards and bit mapped graphics displays, this should be a solved problem. Apparently, there is at least one computer/OS/browser combination that doesn’t have that character at that code point in that font.

    The more standards you have to depend on the more ways things can break.

    Please ask if I you need more info. BTW this isn’t a complaint. The biggest problem is that I may start telling font issue war stories.

    • #2 by Ed on 2012-03-23 - 10:13

      Google Chrome 17.0.963.83 m

      Of course, it works perfectly in Chromium 17.0.963.79 on Ubuntu 10.10…

      I have absolutely no control over anything other than the text & pictures: everything else comes directly from whatever does. For a while, I was using the text equivalent (“120 milli-ohm” and “50 micro-amp”), but I finally decided that just didn’t make any sense at all… we may not have MathML, but at least the fonts should work.

      Yeah, right. (A rare example of a double positive forming a negative.)

      Probably the best thing for you to do is fire off a comment to the tech support folks, who seem surprisingly good at handling stuff like that.

      The biggest problem is that I may start telling font issue war stories.

      Which will quickly degenerate into a catfight involving mobile phone formatting, then we’ll begin reminiscing about paper tape… best not to start with that. [grin]

      • #3 by madbodger on 2012-03-29 - 10:28

        Weird, it isn’t working on Safari either. I looked at the HTML source, and it isn’t an entity, it’s actually inserting a EE 82 B6 which is a UTF-8 encoded E0B6 character, in the Unicode “private use” area. I’d expect 03A9 (GREEK CAPITAL LETTER OMEGA, as the Unicode consortium refers to it in great runes), which would be encoded in UTF-8 as CC A9. The entity would be Ω or \u03a9. Looking at the HTML source, WordPress uses a hodgepodge of unicode shorthand, decimal encodings, and broken UTF-8.

        • #4 by Ed on 2012-03-29 - 11:20

          WordPress uses a hodgepodge of unicode shorthand, decimal encodings, and broken UTF-8.

          Well, now you have enough information to hand the problem back to them…

          I’m sure there’s a Good Reason for all that. Let us know how it turns out! [grin]

  2. #5 by JETGUY on 2012-03-23 - 12:08

    Nice work Ed. Sad to say that when going over this in my engineering classes, I think I was one of the few who actually understood the why behind this sort of testing and what it all really means. Buiding high powered EV controllers, this kind of info becomes extremely relevant. Especially the thermal tests which show how quickly you can go into thermal runaway with a bank of transistors.

    • #6 by Ed on 2012-03-23 - 15:49

      MOSFETs have an increasing resistance-vs-temperature curve, so they’re harder to fry by accident, but it can be done. I’m always surprised at how much juice those tiny SMD MOSFETs can handle, even if I have had some, uh, accidents…

  3. #7 by JETGUY on 2012-03-23 - 12:27

    Just a question, how big was the peltier in watts? Just got finished replacing several in a wine fridge and the originals failed from condensation (the seals failed on the thermal package) which caused galvanic corrosion to the copper buss between each junction and left a green copper slime down the heatsink. DC+disimiliar metals+water=destruction and fast! But this means I have a spare peltier left for this project. So a tip for any peltier, seal that module with silicone or keep the cold side above the dew point.

    • #8 by Ed on 2012-03-23 - 16:08

      how big was the peltier in watts?

      It’s surplus sans documentation. All I know is that it has a (cold) resistance of about 1.5 ohm: 17 W at 5 V, 216 W at 12 V. That suggests it’s really not intended for operation at the usual 12 V; I don’t know how you’d dump 200+ W from the poor thing without a water-chilled cold plate.

      I have another batch of the same size (42 mm) with a hand-scrawled notation: 15 V 4 A 33 W. Given that 15×4 = 60 W, it seems those allegedly pumped 33 W under the usual ideal conditions. The resistance works out to 3.8 ohm, which is pretty close to what I measure.

      On the in-plan: measure no-load duty cycle, apply known power through the MOSFET, measure loaded duty cycle. That should give the additional power required to pump the applied power, thus the efficiency.

  1. Hall Effect LED Current Control: First Light | The Smell of Molten Projects in the Morning