Powerful electronic load based on the DL24 board and a new power module — updated

Not long ago, I featured a new power module in several of my videos. You may also have seen the backstory behind it in a few other reviews. I can now say that the project has reached a kind of logical conclusion. Though, to me, a conclusion is really an opportunity to improve on what came before and start something new, perhaps even more interesting.

To keep up with new topics on the blog, I recommend following me on Instagram, where I'll post notifications about all new articles and maybe write about something interesting now and then: link to my account

You can also subscribe to updates and new articles on the Telegram channel — https://t.me/KirichBlog

I'll say right away that this project isn't for me to keep; it's intended for sale, and there are a few reasons for that. The first is obvious: to be able to develop something new, you need money, so yes, it's about the money. I partly explained the second reason at the beginning of the video about transistors: I'm simply tired of watching people try to build a powerful load and get it completely wrong. Many of my readers know that I've published all the necessary information and explained exactly how it should be done.
Anyway, I'll keep things relatively brief today, without spelling out the obvious. There will still be useful information for those who can solder and have some experience.
And if you want a load like this but don't want to get into the weeds, order components, assemble it, cut, drill, solder, and swear at it, you can simply buy one ready-made. I'll add more details at the end of the review.

Let's start with the backstory. In 2015, I built myself a relatively powerful electronic load. I made a few improvements after assembling it, mostly software-related, with help from one of my readers. The current limit was raised from 15 A to 30 A, a temperature sensor based on an 18B20 chip was added, software thermal protection was implemented, along with a few cosmetic tweaks. Over the past eight years or so, it's clocked who knows how many hours, days, or weeks, and I've never had to repair it.


It was built in the same form factor as my main power supply at the time, which I found to be about as practical as it gets.


Inside, there's a bank of transistors with a current-balancing circuit between them. The principle is simple and reliable. The transistors are from China, probably even counterfeit, but during testing I sometimes pushed it to 600 W, albeit for relatively short periods, until the power module overheated and the protection shut it down.
Speaking of protection, you can see thermal switches on the heatsinks. Unfortunately, this hardware protection method didn't work out for several reasons:
1. Current: the relays simply couldn't handle it, especially after the upgrade to 30 A.
2. Voltage: interrupting the power supply at voltages above 30 V can destroy these thermal switches, just like on the Apollo (here is the story), exactly the same thing.
3. The basic idea: the thermal switch should shut down both power modules, not just one, because if one is switched off, the other will try to do the work of both.


But the basic design itself proved to work very well, so I decided to build the new loads the same way as the old one, while taking into account the experience I'd gained and the design mistakes I'd made before. I think eight years of active use is a pretty good learning experience.



Plus, I've already built loads using several different approaches, so this is essentially a compilation of the best solutions.



Naturally, we start by assembling the power stage. I've shown the schematic before; it hasn't changed much. The key difference is that it uses a quad op-amp instead of a dual op-amp. The PCB will later be available for download by my Patreon supporters (as a version for DIY fabrication using the toner transfer method), although it's really not that difficult to lay it out yourself. In the future, I plan to order double-sided boards with a solder mask from a manufacturer, so there's a good chance they'll also be available assembled.

To start with, I made a couple of test boards, one for each power module. The power module is rated for up to 40 A, but there’s nothing stopping you from increasing that, for example, by doubling it.



Assembled, it looks something like this. The only thing I don’t like is the capacitor, which has to be mounted off-board. I used to put it on the back of the board, but the heatsink warms it up there, so I put it on top instead.


I had a dilemma about where to mount the board. At first I wanted to put it in the center, but knowing that this would lead to uneven heat dissipation (the temperature would be higher farther from the fan), I decided to shift it slightly to the left. I was aiming for 5 mm, but settled on 2.5 mm for the first try.
I drilled the heatsink and mounted the board. At this stage, I didn’t even use thermal paste.


For the first test, I checked how the current depended on the control voltage. I expected the entire module to draw 40 A with 5 V at the input. That’s pretty much what happened: one half drew 2.04 A at 0.5 V and 20.85 A at 5 V. The imbalance is caused by the shunts warming up, but that’s normal, since their only job here is to balance the current; measurement is done using the common load shunt.
I was dissipating 300 W on one half in this mode.


In this mode, of course, everything overheats quickly, since I wasn’t using proper cooling or thermal paste. But you can see that the transistors heat up fairly evenly.


Next, I stack the heatsinks together and drill additional holes, both for mounting and for installing extra components.
That makes 38 threaded holes in total (plus one unthreaded hole). Some spots are very awkward to drill. On the second power module, I ended up breaking two taps: the first one tilted against a fin as it came out (I’d been using it for years), and the second broke because of my own stupidity—I tapped a blind hole with a tap meant for through-holes. No big deal, I’ve already bought new taps :)
The mounting-hole layout is standardized, i.e. it’s the same on all the power modules.


To improve heat dissipation, I bought ARCTIC COOLING MX-4 thermal paste locally (thermal conductivity: 8.5 W/(m·K)). I wanted to buy Honeywell PTM7950 phase-change material, but it costs three times as much, while its stated thermal conductivity is the same 8.5 W/(m·K). Maybe when I’m richer, I’ll buy some for testing and comparison.
The paste is genuine; the website and code check out without any problems. I didn’t risk buying it on AliExpress.


I applied the paste and screwed down the transistors. It might look like I used a lot, but I didn’t—only a tiny bit squeezed out.


With the rough assembly done, I could start testing. I decided to compare several fans at roughly the same dissipated power.
First up was a Delta. Testing wasn’t very convenient: a powerful power supply powered the load itself, the board, and the fan, so I set it to 15 V. A small power supply served as the reference voltage source.


The power dissipated by the power module was about 350 W. I was feeding it more, but some of that was dissipated in the wires, which heated up to over 80 degrees.


With the Delta, the transistors near the air outlet reached about 78 degrees. The first and second photos show both sides of the module. The transistors closer to the fan were about 7 degrees cooler.


I didn’t try the Sunon fans from AliExpress because they run very slowly. Instead, I tested the ones I’d bought locally, starting with the PMD1207PTV1-A.
The outer transistors reached 72–73 degrees, and the one closest to the fan reached 67.


Then I tried the less powerful PF70201V1-000U-A99. The temperature went up, of course: 77 degrees on the outer transistors and 71 on the one closest to the fan.


Then I got curious about what would happen if I installed both, as I’d been planning to. The more powerful PMD1207PTV1-A was set up as the intake fan, and the PF70201V1-000U-A99 as the exhaust fan. The cooling system drew more than 12 W in total, so slightly less than that went into heat on the module—though only a few watts less.


The noise is impossible to describe, of course, but I got 65–66°C on the outer transistors and 62°C on the one closer in.


Out of curiosity, I raised the voltage a little, to 15.5 V. The power went up, but this had little effect on the temperature, since the fans' power consumption increased proportionally.


I tested the AVC fan in two configurations. First, I simply screwed it onto the heatsink and got 76°C on the outer transistors and 71°C on the one closest to the fan—not bad, really.


But in my review of these fans, I mentioned that they have a temperature sensor. In this case, it was telling the fan to speed up, but perhaps the temperature wasn't high enough, since I was measuring the temperature of the fins being cooled by the fan itself.
In the end, I simply shorted it out.


The fan drew about 2.7 W before I shorted it out; afterward, that went up to 7.5 W.


On the other hand, this fan on its own kept the outer transistors at 69–70°C and the one closest to the fan at 65.6°C. In effect, the module is now running with a temperature rise of about 40°C (70−30=40), so at a transistor case temperature of 110°C it can dissipate twice the power, or around 650–700 W, and still be operating within its normal range.
Yes, it's very noisy, but as they say, do you want the fancy trim or do you want to get there? You can also build a quiet load at this power level, but it'll cost a lot more.



Right after that, I assembled a second set of boards. I made a few small changes to the component values, but as it turned out, they didn't really change anything. Let's call it an experiment.


I put an adapter board on top of the modules to make connections easier. It also looks noticeably neater this way.


Now let's move on to the brains. Connecting the power module is actually very straightforward. Desolder the original transistor, connect the power module's negative terminal to the transistor's source (on the load board) and its positive terminal to the drain. Take power from the load board, and connect the control input to the transistor's gate. Connect the fan to the original connector.

Yes, there are a few more changes to make afterward, but they mainly involve the shunts, diode, and so on.


Apply power, and you can already pull 300 W.


But 300 W isn't much to write home about; we need a more powerful power supply.


No problem: first we pull almost 600 W, then around 670 W. After that, I hit the power supply's limit. And remember, I was powering the board and fan from 12 V here, not 15 V as I originally planned.


So, everything seems to be working so far. That means it's time to grab the calipers, mark it out, get the scissors, and... cut the board in half. Some might call that vandalism; I'll say I need to fit the board into the enclosure, and making it work properly at full size is simply impossible.


But I decided to start with the front panel rather than soldering. I had to figure out how to mount the display. The panel is about 2 mm thick, so mounting it from the inside wasn't an option. Cutting an opening the full size of the display would be even worse—it would look completely makeshift. In the end, I had to mill a window so that only the visible part of the display would be on the outside, with the rest inside. I milled it about 1.5 mm deep, and it almost worked. At one point I accidentally entered the wrong coordinate, and the bit went off course, gouging a groove—fortunately, on the inside.
Yes, I had to cut it manually, controlling the machine directly from the software (very slowly and tediously), because the plastic is soft and the bit could catch and wind it around itself at any moment. I think I'll do it automatically next time.

But I got what I wanted. It's not industrial-grade, but it's definitely better than it could have been. I used a broken display as a test piece; for no apparent reason, it had started showing a solid white screen. A new one has already been ordered and is waiting at the post office for me to pick it up. Since I have several boards, this didn't hold up the assembly.



Another issue I had to think through was reverse-polarity protection. The board originally has a diode, I think rated for 100 V. But not only is it rated for just 100 V, it also gets hot. In the original setup, it gets some airflow from the fan, but still overheats.
In my case, the original diode simply wouldn't work at all, so I had to install external ones—either one on top or two on the sides to distribute the heat evenly.
What's more, after rummaging through my parts stash, I found I didn't have any suitable diodes. What to do? Then an idea suddenly came to me: I have a few power MOSFETs, and each one also contains a parasitic diode with the same ratings as the transistor itself.
So we take the transistors, short the gate to the source, and get a diode—in this case, rated for 200 V and 30 A. I had to place the diodes near the fan, because on the other side they would heat up the already hotter transistors even more. Not very elegant, but what can you do.
By the way, in this case it doesn't matter what the voltage drop across these diodes is, since we're turning it all into heat anyway.


I also modified the fan and moved the temperature sensor outside. I relocated the fan connector to a more convenient spot. By the way, it's controlled via the power rail here, not the negative line, which can sometimes be useful.


Next, I removed the original display from the board. It's held on with double-sided tape at the top and bottom. I used the original mounting holes for the board.


The second task was the shunts. These are complete junk and can only work with active cooling. Remember the converter review where the solder on the shunts melted at 25 A? Those shunts had half the resistance, so they dissipated half as much power.
Here, though, I plan to raise the current to 30 A, and there's no cooling.
So I desoldered the original shunts. For comparison, just look at how thin they are next to the much more suitable shunt I use.
But this time we're going a step further and installing a “shunt for a sensible person.” Yes, it's a two-terminal type, but it's a proper branded part. The stated accuracy is 5%, but there's no need to worry: according to the datasheet, they're available in 1%, 2%, and 5% versions, but they all have the same TCR. Absolute accuracy didn't matter to me anyway, since there's calibration.
I fitted a heatsink to the shunt here. In the second version, I left it off; after some changes and calculations, it became clear that it wasn't needed, and the shunt gets a little airflow anyway.


The prepared board is on the left, the unmodified one on the right.


Now we fit everything into the case. More cutting, drilling, trimming, and soldering—in short, a pile of shavings and debris all over the workbench. The heatsink at the back is covered by a grille, but there's a problem: the heatsink carries not only heat but also the input voltage potential. Unfortunately, there's no easy way around that, unless I use plastic screws for the grille. I'll think about it...


Now it's time for the front panel. I wanted it to look neat, if not exactly beautiful.
For that, I ordered self-adhesive paper. It's advertised as white, with a decent thickness, and so on. It came to about 350 UAH including shipping; the paper itself was 253 UAH for 100 sheets.


There's a scored pattern on the back. For a circuit board, I'd call it scoring; I have no idea what it's called when it's on paper.
By the way, this is a pretty important detail. Sometimes the back is printed on to show that it's the non-printable side; sometimes there's nothing (I've been caught out by that before). Here, there are score lines—more on those later.


Next came printing, tape, scrapbooking, and all that. To get ahead of myself, yes, tape isn't exactly an ideal solution, but experience has shown that even after 8 years the panel still looks almost like new, so this solution has stood the test of time.
As for those score lines on the back, I have to rant. I don't know who came up with this, but not only is the bottom layer hard to lift for peeling, it also comes off in pieces along the score lines. A request to the manufacturer: don't add score lines; just print something on the back, a logo or whatever.

The paper itself is great, by the way: opaque and a clean white. I once got some that was too thin, and the black panel ended up looking grayish through it.



On the first, “prototype” unit, I soldered the display on with wires. Though calling it “just wires” is a bit of an exaggeration: the pitch is 1 mm, and the wires aren't exactly thin. MGTf wire would have worked better, but I'd run out of the right kind at home.


I cleaned the solder joints to make things a little neater.
I connected the power negative using a through-hole jumper for reliability and rigidity. The sense wires went into the original holes; the power positive no longer goes to this board at all.


Let's test-fit it in the case. By the way, inside there’s a Sanmim 15 V power supply rated at 6 W. It actually draws up to 8 W at maximum, but I’ve tested these power supplies and they can really handle more. Plus, there’s airflow here.
The power supply is always connected to the mains; the output is switched instead. But I added a fuse, and there’s a varistor on the board, so everything’s fine.
I also considered switching the fan separately with a thermal switch: you turn off a hot load with a button, but the fan keeps running until the heatsink temperature drops below 35 degrees. For now, though, I’ve decided not to do that.


But I still haven’t managed to solve this problem. The board needs to know the transistor temperature to protect them from overheating.
You saw the original temperature sensor in my review of this load. I removed it from the board, soldered on wires, made a holder from an aluminum tube, filled it with Kafuter thermal adhesive, and installed it, but it reads about 15 degrees too low.
The protection will still kick in, and there’s a safety margin, since the transistors can operate normally up to 120–130 degrees, while the protection is set to 100. But I’d like it to be more accurate. On the second unit, I did a better job: I soldered thin wires to the thermistor and insulated it with special tape instead of heat-shrink tubing, but I didn’t notice much difference.

I’m still thinking about a solution. I don’t like it when something doesn’t work as intended, but I can’t see any alternatives.


That’s pretty much it. Next comes setting up the power module board, calibration, and testing.


And here it is with the case closed. It’s not perfect, but I think it looks quite neat.


A little testing.
First, power. I specify 500 W, which is the power it can dissipate continuously at a normal room temperature.
But the experiment showed that the load handles higher power quite well too—for example, 670 W.


I specified a current of 25 A. I wasn’t sure it would manage 30 A, but it works perfectly well at 30 A too.


The load can draw 24 A starting at an input voltage of 1.4 V at the terminals. Naturally, this threshold will rise at maximum current, but not by much. Beyond that, it all depends on the wire gauge between the load and the power source.
This makes it possible to fully test individual LTO cells, which have a minimum voltage of 1.8 V.



The second one was, of course, easier to build. I had fewer little details and nuances to think through, so it turned out a bit more optimized. I think the third, fourth, and so on will be a little more refined too.


For example, on the second one I used a ribbon cable. It has its drawbacks, of course, but also some advantages. I installed a connector on the load board. In the future, I plan to install two connectors, one on the board and one on the display, so they can be connected without soldering. Connecting the ribbon cable to the board is very awkward, after all.


The wire routing is also more optimized here. By the way, the USB board is connected with a plug on both versions, essentially making a short extension cable. It currently uses microUSB, but Type-C could be used too.
Also, both power module boards, as well as the top board, are coated with protective varnish to prevent dust-related arcing and improve operating stability.


The USB port is on the rear panel. On the second unit, I also put the external temperature sensor connector on the back. I wanted to put it on the front, but it looked really makeshift.
By the way, in my personal opinion, an external temperature sensor is completely pointless. At least, I haven’t needed one even once in all these years. Ideally, you should measure temperature without contact, and using a sensor like this to measure the temperature of the power supply under test is completely useless. So I’ll probably make it an optional extra.


And here are the first two units of what’s almost a production-ready device :)


A few other things.
I may shoot a video about these loads tomorrow or the day after. If you’re interested, ask your questions now; I’ll answer them in the video and may demonstrate anything you’d like to see.
I’ve shot the video; it might be useful


A model rated for up to 40 A has been added, with a power rating of 500 (600) W. Its only drawback is that it can’t be connected to a PC by cable—Bluetooth only.
By the way, someone once asked me why I specify 500 (600) W. The continuous power rating is 500 W, but the load can dissipate 600 W for a while, which is important for people testing batteries. As you know, battery voltage drops as it discharges, and so does its power. In other words, you can connect a fully charged battery and load it at 600 W; the voltage will drop fairly quickly, and the power will fall to the usual 400–500 W. The point is that this load lets you start discharging at more than 500 W.

This model is based on the DL24M board. Since this board is no longer being made, this will be the only unit in this configuration. Future units will be based on the DL24, DL24MP, or DL24EW, depending on the required configuration.
An encoder has been added alongside the buttons. It duplicates the “Mode,” “More,” and “Less” buttons, making the unit much more convenient to use.



I’ve also tested two more powerful versions. I’ll go into more detail later, but briefly:
A 1000 W model based on the DL24. This version has no enclosure; I may build at least one more in an enclosure.


And for now, the most “advanced” version. 1500 (1800) W, current up to 40 A, with the option to connect a charger supplying up to 40 A, allowing automatic testing. Two-speed fan control, plus separate control of an additional fan. Control based on the DL24EW, with additional hardware thermal protection.

Next on the agenda: building more powerful models, for example, 3000 W and up.

The photo shows an intermediate version. It’s already been tidied up since then. The customer suggested assembling it as shown in the photo, then installing it in an enclosure themselves, which is why everything has been put together to make installation easier.



I already know one question: what about noise and heat? It gets hot and noisy. If anyone knows how to dissipate 500–600 W in a compact enclosure without noise, I’d be glad to hear about it :) But it really only gets noisy at around 150–200 W of load; below that, the noise is moderate. I may optimize it in the future.
By the way, it can sometimes be useful for the fan to ramp up earlier. When testing at power levels above the maximum, this gives you extra time before overheating. I’ve tested this myself: I used my old load to test 600 W power supplies, even though the load can handle only 300–350 W at most.

Price. I’m planning on $150. I agree, it’s not cheap, but all the mechanical assembly, fitting everything together, and the other little details take a lot of time and effort. When I was finishing the first unit, at one point I was ready to give up. The second one went more smoothly. Plus, I test and calibrate everything, and provide technical support, etc. For now, the most I can offer is a small discount on the first unit (with the green buttons).

Plans. Lots. For example, optimizing the existing design—but doing it properly, not the “Chinese” way: improving what’s already there, rather than cutting costs.
There may also be more powerful versions, 1000 and 1200 W, naturally in a different enclosure, possibly metal. A 1500 W load is also being considered. And of course, I mean real power, not PMPO. I’m also thinking about 800–1600 W models with PC connectivity, current up to 60 A, and so on. These are just ideas for now, though it would be interesting to gauge demand.

By the way, I happened to find the manual for one of the devices a friend and I developed some time ago. Can anyone guess what year this manual was written? :)




Well, that’s about it for now. I’d be very glad to hear your comments—and even more so, your questions and opinions.