The story of the TEC-80K electronic load, or a year-long journey

Yes, today this isn't a review or an article, but a story. And it's about the TEC-80K electronic load, which is fairly popular and well known.

A reminder: to stay up to date with new topics on the blog, I recommend following me on Instagram. I'll post notifications about every new topic there, and maybe write about something interesting too - link to my account

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


This story began quite a while ago. In one of my articles, I showed a load that stood out from the others because it could handle currents of up to 60 A and power levels of up to 1600 W.
It comes in three versions: 400, 800, and 1600 W. In my photo, the second and third versions are actually swapped. Yes, the 1600 W model looks smaller than the 800 W one, although in reality the 1600 W model is noticeably larger.


All three loads have roughly the same design: a tube made of heatsinks, with transistors on the sides. The 1600 W version is noticeably different, though: the transistors on each side are arranged on two levels. So the 400 W model has 2 transistors, one on each side; the 800 W model has 4; and the 1600 W model has 8.
Since it has more transistors, the load's designer used a wider heatsink: 120 mm versus 80 mm on the lower-power models. With the same heatsink length, that makes it look smaller.



But I'll start the story somewhere else. About a year ago—or, more precisely, in August last year—I started building electronic loads not just for myself, but for other people too. From the outset, I knew that 30–40 A and 500–600 W weren't enough and wouldn't be interesting. I quickly solved the power problem, so I started making 1000–1200 W loads as standard models, with even more powerful ones available to order.

Unfortunately, increasing the current wasn't so simple. I do have a load that can draw up to 100 A and even plot a simple graph, but more about that later.

I knew about the TEC-80K load, but I wasn't happy with its design. It felt rather unfriendly. If I was going to build one, I wanted it to be something you'd be happy to pick up—a complete, finished product, not a semi-finished kit. On top of that, shipping a massive heatsink from China, even by sea, is quite expensive. So I decided to build everything around the cooling system I was already using successfully, and buy only the electronics.

So I wrote to the manufacturer, who turned out to be the designer as well. He also supports Ukraine, which immediately gave me the idea that the new version of the load should not only carry my brand, but also be localized.


Branding was sorted out very quickly: the load's splash screen got my website's name. I later tweaked the spelling a little, but that wasn't important. The point was to make sure buyers knew they'd bought it through me, and that I had also handled the final assembly, setup, and testing.


The next step was to localize it, meaning translate the interface into Ukrainian. The designer was enthusiastic about the idea, and we got to work on the translation. He sent me a file with all the menu items to translate into Ukrainian, but we ran into a few problems: some items had very strict character limits, so I had to figure out how to shorten the wording.

It looked something like this. Yes, it wasn't perfect, but it was the best I could come up with at the time. We decided to work out the details as we went—or, more precisely, after I received it, tried it out, and told him what needed adjusting. After all, it could be updated with new firmware (I didn't yet know how many complications that would involve).


Then came the font selection. At first, the designer (remember, he's Chinese; that matters) tried to find examples of the Cyrillic alphabet himself and drew the letters as best he could, even though I'd sent him pictures of fonts.
It looks pretty sad, of course. Keep in mind that in China, not only do they use characters, but they also have no concept of uppercase and lowercase letters, because they don't have letters at all.


Fortunately, the person I was talking to was an engineer, which made it easier for us to communicate to some extent. I described the characteristics of the fonts and explained how they worked, also pointing out that Cyrillic and Latin fonts share many characters with identical shapes. That made things a little easier.
As you can imagine, communicating meant more than just exchanging messages—we also had to send pictures and photos. So we moved from the Taobao chat to the more convenient Telegram.

It was something of a revelation to the designer that letters not only have different shapes, but also different heights, and that fonts themselves can vary too—and there are... a lot of them. To show him, I sent the same text written in different fonts.


I don't know what the designer was thinking, but he didn't give up, and I have a lot of respect for him for that. The only thing was that the process moved very slowly, since we were both busy with our own things—and the length of our correspondence is another story :)

Things got a lot easier after someone in one of the Telegram groups suggested using a font editor. After that, things moved along much faster.
Yes, I had to edit some characters manually to make them look neater. And thanks again to the people in our Telegram group who suggested not only the font editor, but also which fonts would work best with the character cell size we were using.



I had to check the results, make adjustments in the editor, send the developer pixel-by-pixel screenshots, and mark up his photos to show what needed correcting.
But we made progress, and it started looking much better and neater. The photo shows an almost-final version.


This is the final interface. The photo is mine, but the board belongs to a friend. He ordered it before I did, and it came to me for reflashing.
Unfortunately, there was some kind of bug with the board, so the developer sent a replacement display board along with my order.


As I mentioned, the firmware can be updated. But if you think that's easy, I'm afraid I have bad news. Not only are there hardware-level quirks, but the update program is, to put it mildly... not very user-friendly. And figuring out what it says when an error occurs is another challenge....


After a whole lot of iterations, we somehow got the software side sorted out. It was around March 2024, about six months after we first got in touch.

Next came the hardware. As a reminder, I wanted to order just the electronics, without heatsinks, a fan, and so on. I almost forgot: there are two versions of the control board. The standard one has a simple display on the main board; the newer version has two boards, a main board and a separate display board.
I'll show the two-board setup later. The single-board version with a basic indicator looks like this:


Since my enclosure is different and I'll be using two heatsinks, we discussed mounting options and wire lengths. We agreed that he would simply send the connectors and wires, and I would cut them to length and install the connectors myself.


But that wasn't all. It turned out the original display board wouldn't fit in my enclosure, since the internal height is only 84-85 mm.
I also didn't like the layout of the controls, and I wanted a universal board. By the way, besides the encoder, the display board has two buttons: a start button and an additional button for turning off the display.
So I got the display board drawings and routing files, sketched out what I needed in Sprint-Layout, which I'm used to, and sent it to the developer:


There was plenty of unused space on the board, so it turned out we could remove the excess without any problems.


After that, we discussed the board routing in detail, since I needed the controls and indicators in very specific locations.
I asked the developer to add a routed cut so I could snap off the lower-right corner of the board if I didn't need it, but either he didn't understand what I meant or it wasn't possible to do. In the end, we left it as it was.


We also added mounting holes, and the new display board was ready. The new one is on the left, the old one on the right.


Once everything was approved, it was time to buy. A friend placed the order; he already uses a couple of loads from this manufacturer, rated at 400 and 800 W.
The seller created a special listing for buyers from Ukraine, with several options:
1. 1600 W kit without a heatsink or fan—that's the one I bought.
2. 800 W kit, also without a heatsink or fan
3. Complete 1600 W unit
4. Complete 800 W unit
5. Complete 800 W unit with an LED display.

I don't remember the current rating of the 400 W version, but the 800 W model can handle up to 48 A. The most interesting one is the 1600 W version, though even that has changed. Originally, this version was rated at 60 A everywhere. At my request, the developer started making an 80 A version. I asked for 100 A, but he said he couldn't do that yet. We'll come back to the load current later.

I'll have to check, but as far as I remember, only the last listing comes with an LED display. All versions have a Ukrainian-language interface, but aren't branded.


Since the total order came to more than 150 euros, we decided to split it into two parts and ship it in two boxes to two recipients. The seller included different extras for each, and photographed the contents of each parcel before shipping.
That is, We discussed everything first, decided what would be included, and then placed the order. I received the extras because I plan to continue working with this person.


About a couple of weeks later, I picked up my boxes from the parcel locker. A quick aside: yes, this manufacturer does have some "rough edges" and issues, but one thing he definitely deserves credit for is being responsive, willing to help, and... his packaging.
My friends and I have received packages from him, and they have always been packed extremely well. It’s very difficult to damage anything.


The amount of "stuff" is certainly impressive: boards, bags, wires, and so on. Just to remind you, this is two kits plus the extras.


To avoid confusion, I’ll sort everything out so you can get at least a rough idea of what comes in one kit.


1. Connectors and wires for connecting the power boards; these are also used to connect the display.
2. Wires for connecting the measurement input and the 485 interface.
3. 50 kΩ temperature sensor
4. Cable for connecting to a PC, a standard USB-microUSB cable.


1. Diodes, connected in series in the power circuit to protect against reverse polarity
2. Power transistors
3. A set of clamping bars and terminals. They are soldered to the center leads of the diodes and transistors, then screwed onto the heatsink.

Note that there was one more diode and transistor than needed, but these weren’t extras. The seller usually includes at least one spare transistor. Later, though, I’ll show what comes in the full version of the kit.



1. Power boards. In my case, the transistors and diodes weren’t soldered onto them, but we had agreed on that in advance.
2. 15 mΩ current-sense shunts are installed on the back.
3. The main power board, where the load input itself is connected.
4. An auxiliary board that works together with the main board.
5. The board on the other side of the heatsink; the temperature sensor also connects to this board.
6. An auxiliary board that works together with the board on the other side of the heatsink.

There’s not much else to say, really. By the way, all the boards are coated with lacquer. Personally, though, I think it would be better to standardize them somehow: all four boards are different, but they could have been reduced to at least three types, or even two with a bit of effort.


The motherboard and the display and control board. Note that the plastic overlay was also cut to match the layout of the new display board. I hadn’t even thought of that when we were talking, but the overlay turned out to be very useful.


To me, there’s another oddity here: each board has its own microcontroller and a USB port. But on the motherboard, the port is for connecting to a PC, while on the display board it’s for updating that board’s firmware.
By the way, the PC connection is galvanically isolated. These boards are also coated with lacquer, but in my subjective opinion, the soldering could be better, especially around the USB connectors.


This is what definitely isn’t included in the standard kit. I received all of it only because I plan to continue working together with the developer, so some of these parts might come in handy while assembling my version of the devices.
Yes, the kit included a spare motherboard, and as it turned out, I needed it.


1. A pair of op-amps used on the power boards, a pair of microcontrollers used on the display boards, and a pair of USB connectors.
2. A button and an encoder knob
3. Temperature sensor
4. Firmware update cable. I haven’t used it yet, but it has a switch, probably to turn the power line on and off so you don’t have to keep plugging and unplugging the connector, since starting the firmware update requires disconnecting and restoring power. It’s clearly been put together rather haphazardly, but it’s only needed very rarely, and the developer offered it himself to make the process more convenient.



I received these two kits in May, but ended up putting them aside for a long time. I simply didn’t have time to work on them. As it turned out, that actually helped, but more on that later.

Now I’ll tell you about the finished version of this electronic load. It so happened that a friend needed to order one, and since the order was shipped through me, I got to "play around" with it a little.
I hadn't planned to add such a large section to this article, but since it's clearly on the same topic, I'll show it anyway.

Anyway, it was ordered in a hurry, so it was shipped by air—expensive, but fast. I think it took about 2–3 weeks.
I received a hefty box. Inside it was a large box and a small one. The small box contained circuit boards, wires, and various bits and pieces. As I mentioned, everything was packed really well.



The small box contained the fan, motherboard, display board, and assorted small parts.


The fan caught my attention: it's big and heavy, and subjectively, it looks like a quality unit.
It originally had four wires, but the tachometer wire has been cut, and the connector seems to have been replaced, although I personally used a standard "fan" connector.


The fan's behavior is interesting too; it feels like it has some kind of "brains".
At 5 V it doesn't spin, drawing only a small current. Raise the voltage to 6 V and it first gradually speeds up, then slows down a little and continues running at a fixed power level.
But what I found even more curious was that if you raise the voltage above 11–12 V, it reduces the current to maintain constant power. So there's simply no point in supplying it with a higher voltage; its power is fixed.


The fan is clearly designed to generate pressure rather than airflow: if you block its air outlet, its power rises from 18–20 W to 27 W, but if you block the intake, it stays almost the same, increasing only slightly.
Keep this in mind when choosing a power supply (one isn't included). You'll need a PSU rated for at least 3 A of actual output.


The motherboard is exactly the same as the one I bought. The label says 16 because the package also contained a replacement board for my friend's electronic load.
The microcontrollers on the motherboard and display board are different: the motherboard has an 8A8K48D4, while the display board has an 8H3K48S2. Honestly, this seems like overkill to me—two fairly capable controllers for a task that isn't all that complicated.
The display board, on the other hand, is the standard type for this electronic load: it's full-height, with a different arrangement of indicators and buttons, as well as a different interface connector. On mine, the connector is mounted at 90 degrees.



The package included a cable for connecting to a PC, mounting hardware for the motherboard and fan, wires, three power-board ribbon cables (one spare), a display-board ribbon cable, and a spare power transistor.


The heatsink and power modules.
The packaging gets a 5+, and so does the weight—but that's 5 kg. The heatsink certainly looks seriously substantial. Subjectively, with the included fan it should be able to dissipate far more than 1600 W, but that would require good thermal transfer from the transistors.
So in this case, the heatsink isn't being used very efficiently. My electronic loads dissipate around 600 W of heat with a heatsink weighing about 0.5 kg; this one weighs 5 kg (well, a little less, since that's the weight including the boards) and dissipates just 1600 W.



It has a full set of thick fins.


The power boards, transistors, and temperature sensor are already installed, so all you need to assemble it is a screwdriver and a pair of reasonably capable hands.


1. Install the motherboard so its connectors line up with those on the power boards.
2. Connect the power boards to the motherboard using the included ribbon cables.
3. Install and connect the fan.
4, 5. That's basically it; all that's left is to connect the display board.
6. There are six feet on the bottom, so you can use the whole assembly without any extra stand.


The whole assembly, put together. What's left:
1. PC cable
2. Spare ribbon cable
3. Spare transistor
4. Connector with wires for the measurement input
5. Connector with wires for the RS-485 interface.

The display board isn't mounted at all. I think I would have planned from the start to mount it next to the motherboard on the heatsink; that would be more convenient.



The firmware isn't branded, since the electronic load was bought directly from the seller. Otherwise, it's pretty much the same as mine, except perhaps for an additional connection settings item.
By the way, when ordering, you need to tell the seller which version of the load you want, USB or RS485. In this case I ordered the USB version, but the menu includes options for the 485 interface.


For a quick test, I checked the voltage measurement accuracy. No issues there, and the current calibration looks good too, at least at low values.
You can see in the photo that the voltage drops slightly under load, which is normal. A separate sense input is used to compensate for this.



It also works without any problems at a higher voltage.


Next, I wanted to see how effective the cooling is and how the load behaves at higher power.
To test this, I connected it to a more powerful power supply and let it heat up for a while.
At low power, the fan is off. Once the temperature reaches a certain point, it turns on, and its speed is automatically regulated. I should note that it moves a decent amount of air, even through the heatsink.

And remember to be careful: the heatsink is essentially connected to the load's positive input terminal.



I ran it for a while at 24 A. This was to check two things: the temperature and how much the current drifted as it warmed up.

There is some drift: as it heats up, the current drops slightly. I had the same issue with my load based on this board, but managed to reduce the effect somewhat by adding temperature compensation for the reference voltage to the circuit.



Now for the temperatures, which was what I was most interested in.
First, it turned out that the transistors on one side run hotter than those on the other. The difference isn't huge, but it's there. I not only know why, but also discussed it with the developer to see if we could fix it.

At around 1300 W, the transistors on the left reach 106–107°C, which is a bit high in my opinion. This is because there are only 8 transistors, so at the full 1600 W each one has to dissipate almost 200 W (some of the power is dissipated by the protection diodes).


On the right side, the transistors were at around 100–104°C.


When I was designing my power modules, I took a slightly different approach. Transistor temperature also depends on the area available for heat transfer. So if I use 16 transistors and this unit has 8, the heat-transfer area is almost twice as large (the transistors I use have a slightly smaller area). That makes it easier to transfer heat to the heatsinks.
I also use good thermal paste, and it's not cheap :(
And the heatsinks themselves are more effective (in my opinion).

As a result, in my load—which I rate for 1000 W continuous and 1200 W short-term (relatively speaking)—at around 1000–1250 W (almost the same as in the test above), I get nearly the same transistor temperatures while needing only half the fan power (12 versus about 24).
Looking at the hottest ones (those farthest from the fans), I got 100–106°C.

So in practice, the transistors in my power modules run at about the same temperature as those in the TEC-80K, even though I rate mine for one and a half times less power.


All right, let's move on to the software. I wouldn't say it has a huge number of features, but it covers the basics. It also has a very useful feature: the software can control several loads at once. I've been asked about this more than once, but unfortunately Atorch can do this only with the CL24 version, not the others.
You can also save and load a graph, but as far as I can tell, there's no way to overlay multiple graphs.
Personally, I think ZKEtech has the simplest and most convenient software. Sure, it's not perfect either, but it's very handy for battery testing.

For example, the TEC-80K software lets you save a graph and display it later, but it doesn't retrieve the actual test results (capacity in Ah and Wh). You also can't view the results at an arbitrary point in the test, as you can with ZKE. I'll have to discuss this with the developer.


The data is saved to a file with the .tec extension, and it turns out you can open it in Notepad. Why not CSV?
It includes test results in mAh and Wh, so maybe it would be worth putting together some kind of converter...




I’d like to pause here and explain why these two sets had been sitting around for so long, and why I hadn’t built a load based on them.
The thing is, I couldn’t figure out how to fit the TEC-80 power modules to my heatsinks. There simply wasn’t enough room to install them, especially with the wires and everything else.
But when I started looking more closely at the control module, I realized I could basically use my power modules as they were.
Yes, testing showed there were some quirks, but with a few small modifications I turned them into equivalents of the original modules. Thanks to the developer for providing me with the schematics for the original power boards.

Key points.
1. The load current isn’t measured at all—not in any way. We calibrate the load’s current, but essentially all we’re doing is mapping the reference voltage to a current value. If the shunts on the boards start to drift a little as they heat up, the current will drift too. Atorch-based boards have a separate common shunt, so this isn’t a critical issue. The current also drifts on my 100 A load, but it has a separate meter, so the drift is visible and taken into account. I spent almost a week wrestling with stability here before getting it to something reasonably acceptable. As it turned out, the result was at least as good as—and possibly better than—the original TEC-80K. I managed this by improving the shunts’ cooling, modifying the power module boards, and adding positive temperature compensation. In other words, as the unit warms up, the reference voltage rises slightly, reducing the current-setting error.
2. Voltage measurement is implemented in a rather unusual way. Normally, there’s just one “voltmeter” circuit; with a two-wire connection, you simply connect the power and sense inputs together. Here, you shouldn’t connect them, because there are two measurement circuits—and actually, even three.
It’s quite unusual, since calibration is done separately for the power board and the motherboard, and these inputs must not be connected during calibration. What’s more, the power board has two voltage dividers, for low and high voltages, so it has two measurement inputs, in addition to a separate input on the motherboard. And by the way, there’s also a fuse on the motherboard’s sense input!
One of the motherboards had a problem from the start: it wouldn’t calibrate for voltage. Luckily, I had a spare.
3. In practice, the motherboard always outputs the reference voltage; the load is turned on and off by applying or removing the inhibit signal. I wasn’t happy with that because it caused a current surge, so I added a reference-voltage inhibit circuit with a soft start. That made things better.

I discussed all of this with the developer and suggested some improvements. I also plan to make a small change to the current-control circuit using the common shunt, but that’s for later.

I also gained some useful experience, which showed that if I can use my own power boards, the limit on maximum current and power goes away. So I’m planning a version that can deliver not just 80 A and 1600 W, but, for example, 120–150 A and 3000–4000 W. The developer just needs to change the limits in the program.


Anyway, I got a bit carried away.
As I mentioned earlier, the display board was modified so I could mount it on the front panel. Cutting a section out of the board also made room for the power button. When we changed the board layout, we agreed right away that there shouldn’t be any other traces in that area, for example, from the encoder.

The only inconvenience was that the display is thinner than the front panel, so I had to make spacers to raise it a little. But I needed to try it first to figure out what to do and how to do it properly.



I didn’t take any photos during assembly; it just didn’t occur to me. I only have one photo from an intermediate stage, when I was struggling to get the current stable. I had to remove the power modules, change something, put them back, and so on—a real headache.


By the way, my loads now have professionally made PCBs, so they look better and neater. What’s more, I managed to order the power boards in purple, the same color as the Atorch boards :)
The photo shows another batch of power modules. Also worth noting: the 1000 W loads use longer modules, which lets me increase the power a little. They simply won’t fit properly in the 500 W version.
And here are a couple of loads (already shipped to the customer) with the modules shown above.


In the end, I solved some of the problems, worked around the rest, and got almost what I wanted: a load with current up to 80 A, power up to 1200 W, PC connectivity, and four-wire connections—all essentially based on my stock 1000 W load.
The power leads look more substantial than those on the 40 A version, which makes sense—the current is twice as high. The 40 A modules have been thoroughly tested for a long time, and in practice this is two 40 A loads in one enclosure, so there’s nothing unusual here.
The only thing is that, as with my 1000 W loads, I added thermal protection to the second power module, since the protection everywhere covers only one channel. Naturally, the fans are still regulated automatically according to temperature.



The end result is this unit. It’s the same TEC-80K, but rated for 1200 W. It can handle up to 1600 W for short periods, though I wouldn’t recommend doing that.


At the back are the fan grilles, power input, and a Type-C port for connecting to a PC. Yes, I decided to replace microUSB with the more modern USB-C; I think it’s better this way.


After writing the article, I tested it under real-world conditions: I connected a 12 V battery and started discharging it.


First I gradually increased the current from 20 to 60 A. The actual current was slightly higher, but that’s not critical, since it drops a little as the unit heats up.
The accuracy of the clamp meter itself is questionable, though. My main goal was to assess how much the reading drifted as it warmed up, since I don’t calibrate using a clamp meter.


I set it to the full 80 A. At first the reading was a little high, then it settled at exactly 80 A and stayed there for quite a while. Later the current did drop to 79.7–79.8 A, which I think is more than acceptable.

One thing to note: the cables to the battery must be good quality, with a cross-section of at least 10 mm², preferably 16 mm², and good-quality lugs, since the connection point can heat up.
Also, remember to connect the cables to the load first, then to the battery; disconnect them in the reverse order, never the other way around. If a cable accidentally falls and shorts something, it’ll be a real disaster.


Thermal image: overall view, the top of the load enclosure, the hot air outlet, and the input terminals. The terminals heated up not because of their quality—there’s nothing wrong with them. My cables are heavily used, and the lugs aren’t very good quality.



Group photo. On the left are the standard 500 (600) W and 1000 (1200) W models. I try to keep them in stock when I can, although for the past few months I haven’t managed to put anything in the “warehouse”—I simply haven’t had the time. These loads have been in production for a long time, but they’re still being improved. For example, they now have larger heatsinks, factory-made circuit boards (all of them now), larger terminals, and various other small upgrades, while the price remains the same.
On the right is the new model based on the TEC-80K, along with a 1300 W, 100 A version, which is also available to buy, although I originally built it for myself.



Although the 100 A model doesn’t have “smart” controllers, it can also be connected to a PC via Bluetooth. The software is very basic and rather clunky, but sometimes even that is better than nothing.




That’s all for now. I may add to this post later, or more likely write a new one with a more detailed description of the new model. That may take a while, especially since there will probably be some changes.
There will also quite possibly be a dual-unit version, so to speak, 2U, rated for 2400 W and 100 or 120 A.

I’d be very happy to hear your questions, suggestions, advice, or anything else related to this topic.