The dummy load returns and I do some cooking
I do very much enjoy watching youtube channels on electronics. It has probably been the single most instructive resource for me and led to real understanding of what I am actually doing.
Some time ago, I watched Dave Jone’s excellent episode on building a constant current sink – a “dummy load” which is an essential piece of test gear for testing out power supplies. So much less fiddly than messing with loosely spec’d power resistors, I had always had the intention of building one but never quite got around to it.
Apparently, many others had the same idea (after Dave showed just how easy it is) and expanded on his original idea. Not least is Martin Lorton’s excellent series on his own take on the dummy load which is similar, but capable of double the current of Dave’s design. His project turned out exceedingly well and is even housed in an attractive enclosure.
I then forgot about it for a while, hoping to return to it once I sourced parts and figured out how the MOSFET actually works so I can build it correctly and choose my own MOSFET for my needs.
About a week ago, I stumbled upon a new channel, one by Peter Oakes and he needed to build his own dummy load also. I rather like his approach, his designs tend to be simple but very effective and he happily goes through all sorts of testing and debugging with the viewer to trim it up and make it work well.
I thought, right, it’s about time I built the thing.
I have always had a defecit of N-channel mosfets for some reason. This is due to me mostly using power transistors for my projects. My recent study of switchmodes and of course the dummy load led to me needing a good supply of MOSFETs. I had a couple of IRF740s but one I fried (by trying to drive it with an NE5532 opamp instead of a 555 timer lol) the other lives happily as part of my clock project in the boost converter for the nixie tubes.
Crestfallen at once again having to go to Chinatown or order online just for a few parts, it then dawned upon me to check my junk box.
I’ve always been a big believer in salvaging old electronics and harvesting them for parts. At least the higher ticket items like power transistors, ICs, large caps, MOVs, relays and of course MOSFETs. I had a junked UPS whose battery was beyond hope and I had saved the board. As it happened, I was able to desolder four N-Channel MOSFETs: two IRFZ34Ns and two IRF3205s.
Looking the datasheet, I was delighted to discover the IRF3205 is wonderfully suited to my dummy load project. Hooray!
Thanks to the three fellows listed above (Dave, Martin and Peter) I was easily able to cobble together a design using (I think) the best of each of the three designs. I watched all three channels in detail again to really come to grips with it. A good word of advice when doing something like this is really pay attention to their testing, don’t just copy their schematic and modify it blindly. These fellows really are a wealth of information and it will help you immensely especially if you have to modify your design to suit your specific purposes.
Anyway, in addition to the salvaged MOSFETs, it turns out I had pretty much everything I needed to get going on this, and I adapted their designs to suit my needs and the parts I actually had. I chose the IRF3205 for it’s 200W power dissipation, logic-level gate, and high voltage and current ratings.
Here is the result:
I opted to use a 9V power supply in case I needed headroom (which I do as it turns out) and also because I literally have a pile of unused 9V AC adaptors. The circuit (as you can see) only draws a few milliamps so a small one would be just dandy. All the high current stuff passes through the MOSFET and sense resistor and out to the ground of the power supply under test.
I decided to use a simple LM7805 voltage regulator to create my 5V control voltage, instead of the REF02 Peter Oakes used (since I didn’t have one) and not wanting to simply use the unregulated power rail like Dave’s design, or the LM7806 like Martin Lorton’s. The 9V supply provided more than enough headroom for this. Drawing only a few milliamps, I needn’t worry about it’s heat.
Like Dave’s design (and Martin’s) I’m using the venerable LM324 opamp. Using only two (buffer and comparator) of its four opamps. The LM324, though not without it’s problems being an older IC, is quite sufficient for this application. It will run happily off a single supply and get down very close to ground. The volt or two lost at the top towards the positive rail is not a concern for me as I wish to limit it’s output power to a sane level anyway.
Doing the actual sinking is the IRF3205 as mentioned, and the sense resistor is using one of my Dale 0.1Ω 1% 5W resistors I managed to get in a junk shop. Though I had mentioned these might be problematic in my milliohmeter posts, I highly suspect it is my DMM or my meter design that is causing this problem in the first place.
The theory is quite simple. The LM7805 provides power to the opamps and a reference voltage that is limited at the top by the 5.6kΩ resistor and can be adjusted using a 10kΩ ten-turn pot (which I have thanks to my power supply, need more of these expensive buggers). This set voltage, which will be in the range of 0~3.2V is fed into the first opamp of the LM324 which acts as a buffer to prevent loading of the circuit, then out to another 10kΩ trimpot which is adjusted to divide the output by 10, leading to a range of 0~320mV. The second opamp is wired as a comparator which compares the voltage drop of the 0.1Ω sense resistor (which is a function of the current running through it, 100mV/1A or 100µV/mA) with the set voltage from the previous stage. The output of this opamp then drives the gate of the MOSFET. The negative feedback and comparison with the set voltage allows the opamp/MOSFET combination to actively stabilize a constant current draw from whatever source you have it connected with.
Pretty simple concept and very effective. I did chose to go the Peter Oakes route mostly. His influence can be seen in the choice of the 0.1Ω sense resistor over the 1Ω used in Dave’s design due to my inability to find 1Ω (or 10x 10Ω) precision power resistors. This has the unfortunate effect of being less accurate for the LM324 since the voltage levels are 1/10th those of Dave’s design. Dave did warn about this in his power supply project, noting that cheap opamps will not “cut the muster” for such a precision setup thus his favouring the 1Ω sense resistor.
Breadboard it up
This is my favourite part, where I take my schematic and actually assemble it for testing. As expected, I had all the parts I needed on hand and gleefully assembled it.
The first test was fantastic. It worked exactly as expected. I had predicted it would, considering I watched Dave, Martin and Peter demonstrate theirs in detail. Very little could go wrong. But of course, it did.
A stupid mistake leads to magic smoke
Being a little too eager to test it it out, I started playing with the control pot and cranking it up to ridiculous levels. What I had neglected to do (despite the picture above) is actually mount the MOSFET on a heatsink! D’oh! Sure, it *can* dissipate 200W, but with a heatsink! I was testing my power supply mains board with it, which puts out an unregulated 40ish volts. I was cranking it in excess of 2A which of course lead to the MOSFET having to dissipate 80W of heat with no heatsink! The results were dramatic, it actually melted the solder on the drain pin and the wire came undone.
It must have caused a short somewhere, as it blew the 4A fuse in my (thankfully constructed) safety outlet.
A bit confused
So it was obvious I melted the MOSFET. Just out of curiosity (and not wanting to pitch a part unecessarily) I decided to test the MOSFET. I verified (using this simple method) function of the mosfet switching using my diode tester on my DMM and it seems to work fine (surprisingly) and I cannot detect a gate short to drain or source. In powering up the circuit again, I end up with more blown fuses however, and when I get it to work at all, I can only draw between 11 and 35mA regardless of my setting on the control pot.
This haphazard testing has resulted in a MOSFET that is most probably melted yet deceptively still kinda works, and several blown cartridge fuses which I will have to replace. I’m hoping against hope that nothing else is fried in my circuit, though it seems the opamp is still doing it’s job so that looks good.
I’m a bit frightened to whack in the other IRF3205, though this is most likely the solution I need, simply because it is my last one. I shouldn’t be such a wuss, I can always buy more. Cheap as chips as Dave says. Next time I WILL be mounting it on the largest heatsink I can find and use a lower voltage power supply for my testing. The 12V mastercrap battery charger I salvaged would probably be ideal since it can put out 6A and even has a rather coarse ammeter on it. In opening that however, I discovered it has no capacitance decoupling it, just a transformer and a couple of diodes for rectification. Ripple-y! Maybe I’ll whack in a 1000µF cap across it to smooth it.
With every electronics project you (or I or anyone) does, even and especially in failure valuable lessons can be learned. This is part of the reason I’m documenting my projects. It acts as my lab notebook (those of you who were in a science program know the value already) and revision history where I can reference all of the decisions, lessons, and whys of how I did things. Bonus I can share it with the public . Here are the take aways from this project:
- Always mount power transistors and MOSFETs (and voltage regulators) on heatsinks if you expect to draw any significant power. Or better yet, just always do it. With thermal paste. I cooked my MOSFET hot enough to melt solder, which is almost certainly hot enough to kill it. It will save you burned out parts and possible fire hazards.
- When designing any project, but particularly power electronics, read all you can and watch all videos in detail. Learn everything you can about each part of the circuit, its function and which parts to be careful of and most especially be mindful of limits you cannot exceed.
- Compare and contrast different designs that perform the same function, novel solutions can be found along with greater understanding of the subject and saving from possible pitfalls.
- Junk electronics are a free and wonderful source of parts, and you save the environment by re-using components and save yourself a trip to the shops or yet another online order. Get good at desoldering.
- Parts that hold up to simple function tests may in fact not be functional. Though the IRF3205 I melted still passed the diode check, it quite obviously cannot do it’s job and took several fuses along with it.
- Always grab several of each part you need for a project, frying them is somewhat inevitable
- Safety is always first. My recently built safety outlet for builds under test acquitted itself by blowing it’s fuse for me rather than passing an obscene amount of current through my project.
Well I spent a few minutes and wired up the other MOSFET, sadly to the same result. I can adjust the current up to about 30mA where it just stops. I’m a bit mystified. Assuming the MOSFET is working as it should, it means it’s gate is not getting the voltage it needs to turn on. It could be the opamp is damaged, but I checked that also. I replaced the opamp to the same result (unless I damaged a second opamp ARGH). I need to go through and check my wiring and apply the golden rule of troubleshooting – though shalt check voltages. Though mildly upsetting, I know I can probably isolate the issue, and all the bits and pieces that make up the dummy load are cheap, readily available and thus easily replaceable. Learning experience, right?
Success because I was just being foolish
Or as we know it here, “Tuesday”.
One thing I have learned in my experiments is most often when there is a problem, it’s usually the most foolish reason. In this case, I had simply forgot a couple of simple concepts that lead to me being mystified.
The difficulty was that the circuit seemed to be working very well, apart from the fact that the MOSFET was only opening a trickle to let about 30mA through which is, for my purposes, useless.
What I had forgotten is that I was using the opamp to drive the MOSFET gate and although the MOSFET threshold voltage was a mere 2-4V (logic level) the thing won’t be fully turned on until one approaches up to 10V. I had noted that full out the gate control was only driving 3.7V into the gate which (if the MOSFET was just a simple switch) should have turned it on for me. It did, just not as much as I expected or needed. I needed to pull that gate to a higher voltage to get my desired result.
But why? why wasn’t the opamp swinging right to the rail? Because, dumbass, the LM324 is a cheap opamp that is most definitely not rail to rail. In reviewing the datasheet (and countless warnings from the electronic gurus mentioned above) I discovered that up to 1.5V below the positive rail is simply unavailable to me! This makes perfect sense. I turned up the current and the gate voltage seemed to halt at 3.7V which, as it happens is 1.3V less than the 5V rail I was powering the damn opamp from!
I feel a right fool, after replacing basically every component in the circuit to try and find a faulty one, the solution was staring me in the face. Just connect the V+ of the opamp to the 12V supply I was powering it off of! It doesn’t care. The voltage I’m feeding its inputs is still the one I want it to be, but the difference is its output can swing much higher and really drive that gate and get some current flowing.
This simple modification led to the perfect solution. I’m now able to dial up right to just shy of 3A no problem. Delightedly, the heatsink gets warm, but not hot, which is good news for power dissipation.
Now it’s just a matter of figuring out how to achieve 3A over a wide variety of voltages but that is a question of trimming it up. I know with nothing connected my set voltage will reach 3.2V (which would mean 3.2A on the output) but it’s falling slightly short by the time it gets to the MOSFET. This could either mean there is a current drain somewhere that is robbing me of some control voltage, or the opamp needs a bit more voltage to really open up that MOSFET. A simple matter of testing my control voltage to see if it’s 2.9V with a power supply attached (as opposed to 3.2V without) or not. If it’s 3.2V the MOSFET needs a higher voltage, if it is 2.9V I need to find out where that 300mV is going to and block it.
Quickly this morning I figured I’d check voltages just to find that mysterious 300mV discrepancy. Normally I wouldn’t care for such a small amount, but I do want my dummy load to at least hit 3A with a bit of padding above it before I can say I’m satisfied.
To do so, I had to take my multimeter, which was acting as an ammeter, out of circuit to use it to check voltages. Lo and behold, that was the problem. It’s a well known fact that multimeters in ammeter mode have what’s called “burden voltage” which can be a serious problem especially at low currents. The cheaper the meter, generally, the worse the burden voltage.
Once I had it out of there, my load quite happily was able to just climb over 3A with all voltages checking out okay. I’m finding the result out by about 10-15mV which is more than acceptable.
The road to completion
This project looks like it might actually be completed in the near future. I’m excited! Since the design is reasonably close Peter’s (and Dave and Martin’s) I know that I have very little to actually worry about, still it pays to do my homework. I will run some more performance tests with different loads and see if anything needs tweaking. I might check with my scope also, though since it is not a fancy pants DSO I will be unable to capture the transient response and turn on/off characteristics of my load since they happen in a matter of microseconds and are non-repeating.
Unlike Peter Oakes, I did not include an input for a separate signal drive as I have nothing to drive it with yet (my function generator is still a ways off, though that is on my plate also).
Next up, I will hit the electronics shops and see if I can find a reasonable panel meter for it. A standard 200mV full scale LCD should do me fine, backlit preferably. Definitely need one that can use a common ground. Not hard to find methinks. A suitable enclosure for it would also be a good idea. I did find that the heatsink got warm at full output (as expected) but not hot. That being said, I didn’t run it for any considerable length of time so I will have to ensure it has adequate cooling for longer term operation. If I start testing higher voltage supplies this will also increase the heat dramatically so I may have to look into a larger heatsink and/or active cooling. I have a bunch of junk fans so this won’t be an issue.
It will, however, also influence my choice of power supply. I’ve been using 12V as the main power for the circuit mainly to give the opamps as much output swing as possible so my original idea of 9V will have to be revised for 12. The math says the current consumption is still very low (about 5mA) so a small wall wort which I probably have will do. Of course, I must factor in the display (whatever that will be) and optionally a fan into my current budget.
Other than that, I think a power switch, banana posts and it’s done!
Here it is on the breadboard: