It’s been a chicken and egg race for the last couple months – I was working on power supply projects, but at the same time needed an electronic load so I could test the supplies at known current draws. I have had a working electronic load for a few months, but only got around to making it pretty and boxing it up in the last week. The design that I came up with is not very sophisticated, but it uses common parts and it seems pretty solid. The picture at right shows my recently built variable power supply working into the electronic load.
My parameters for this project were practical: I wanted to built the load using parts on hand – as designed, the load conservatively accommodates up to about 60V and current to 3A, but not both extremes at the same time – I don’t want to push total power beyond 45W. The voltage range is determined by the Vds of the mosfet that I use to dissipate power. The lowly IRFZ24Z that I chose is rated up to 60V. I know some projects have used logic-level mosfets here, but the Z24’s higher gate threshold comes along with the higher voltage tolerance and also smaller gate area, hence less gate capacitance. From some other electronic load projects on the web, I was concerned that the load might oscillate if the driving op amp were pushed beyond its phase margin operating into a capacitive load. That turned out not to be the case with this design, even when implemented on perf board.
Regarding the limit of 3 amps, I picked it because it falls within the safe operating area for the mosfet at the rated temperature, even with voltages up to the limit (which I would avoid, since 60v * 3A would be way over the total power rating). I do not usually work with voltages about about 18V or higher current, so these limits are not likely to come into play on most days, but I needed to have them in mind while designing the load. The max current draw determined the size of the shunt resistor, in 0.333 ohms. The shunt was implemented as three parallel 1 ohm surface mount current sense resistors, each capable of handling 2W. At a maximum current of 3A, these resistors would need to be capable of handling 3*3*.333 = about 3W, so their 6W aggregate rating leaves some headroom. Both the resistors and the mosfet are mounted on the heat sink.
Finally, the power rating is a little arbitrary. I started with the spec sheet’s max of 60W and derated it comfortably. Since all of these ratings are junction-temperature dependent, I also took some pains to limit the temperature of the mosfet. The mosfet is affixed to a finned-aluminium heat sink with screw hardware, a silastic insultator and some heat sink compound. I would have really preferred to have built the project in a metal case with the heat sink bolted onto the back, but all I had on hand from the last trip to Austria was a plastic one. So, the heat sink is mounted on a bit of perf board, with spacers that elevate it above the base of the case; it also has clearance on every side. Air vents have been cut in the side of the case and the fan is mounted on top of the unit so that air is drawn from both sides, parallel to the heat sink fins and then exhausted through the top-mounted fan.
The most likely way for this load to fail would be overheating, so I added some protection circuitry. A thermistor is tacked to the mosfet using a Permatex Muffler and Tailpipe Sealer (see below, $6 on ebay):
This paste is used to secure the thermistor against the mosfet. When dispensed from the tube, the grey paste is somewhat watery. I found that it took a couple applications to build up the coating that I wanted. Allow at least a day for the paste to harden as its water content evaporates. When moist, the paste is conductive (in the hundreds of kilohms per cm range for some I smeared on a bit of plastic), but when dried it is an insulator. The product is made for automobile exhaust systems and it conducts heat well and is not harmed by it. In fact, the paste is meant to dry in two phases – first, by evaporation and second by baking. So, once the product dried, I ran the load at high current for a while to try to “set” the paste.
The thermistor drives two thermal safety features: fan speed and a fail-safe voltage cut-out. When power is initially applied, the exhaust fan runs at low speed to keep it quiet. The fan speed is controlled by a 555-timer that is set to 187.5Hz and 71% duty cycle, but because the fan is driven through inverted logic, it operates at 29% duty cycle. However, when the thermistor gets above about 45C, the fan is switched fully on. If the thermistor approached 60C, it will drop out the voltage that is an input to the first op-amp; as a result, there will be no voltage across the shunt resistor and no current draw.
I designed the fan to be on at baseline because I thought that it would be easier to get ahead of the heating than to play catch-up, and because the longer the mosfet stays cool, the better its longevity. In practical terms, operating with 15V load, the fan doesn’t kick into high speed with a constant load of 1A. At 1.5A, it will go back and forth between low and high-speed. One improvement that could be made on this project would be to add some hysteresis so that the fan makes fewer transitions from low to high speed, but this is a minor annoyance and actually somewhat comforting because it lets the user know where the temperature is. At 2 amps continuous, the fan is in high speed mode almost all the time. I haven’t run the load long enough and at high enough voltage and current to trigger shutdown (although I did in earlier testing, with a smaller heat sink).
The project is based around an LM324 op amp; two subsections are devoted to control of the load mosfet, while the other two are concerned with temperature control. A multi-turn potentiometer inputs into the first op amp subsection, which is configured as a voltage follower. That section is followed by resistive divider that limits the output to a range of 0 to 1 volts on the positive input of the next op amp section. Feedback above the shunt resistor forces the output of that section to drive the load mosfet as needed to keep the two op amp inputs at the same potential — i.e., to set a voltage between 0 and 1 volt above the shunt resistor. The output of the shunt resistor returns to ground through an LED volt/amp meter (which adds negligible additional load).
The 15k thermistor output voltage drives the other two op-amp sections, configured as comparators and biased to reflect the temperature set points mentioned above.
A few comments about construction. First, I have to admit that the first iteration was built Manhattan-style on copper clad board, and that this was the version used to get the power supply going. I started with one design, made changes, added bits — by the end, it wasn’t pretty but it worked fine. However, it was too jury-rigged and exposed to serve as a long term equipment, hence the revision with nicer layout and packaging.
In both designs, I used scrounge LM324’s that I had on hand, but had to adapt them from SOIC to 0.1mm spacing to make connections. In the second version, I used an adapter board that makes the pin out similar to a DIP package, i.e., 0.4mm wide. However, I much preferred the wider package from QRPme.com shown in the original copper-clad version. The traces in the narrower package are thinner and spacing between solder points and pins are just too close.
I chose to make the load operate from 12V since a lot of my equipment is set up to operate from 12V and I have 12V power distribution on the work bench, partly because I work on a lot of radio equipment standardized around 12V, and partly because this makes it easier to move back and forth between different mains voltages depending where I am living. As usual, I used Anderson power pole connectors on the rear of the unit.