Atorch DL24M 600 W Electronic Load Kit

I'll start with a quick announcement: to keep up with new posts on the blog, I recommend following me on Instagram. I'll post notifications about every new article there, and maybe write about something interesting too - link to my account

Not long ago, I posted some information about the relatively new DL24M electronic load, a follow-up to the DL24, which I reviewed about a year ago.
When I posted about the DL24M, AliExpress was having a sale, so I bought one. Today, I want to tell you about it.


First, a couple of links: to the DL24M announcement and the DL24 review. They may be useful, especially since the second link includes details on how to build the power module. I'll also skip over some points in this review, since the DL24M shares some features with its predecessor.

Key DL24M specifications, depending on the configuration:
Input voltage - up to 200 V
Load current - 25, 30, 35, 40 A
Power - 150, 300, 450, 600 W
Operating modes - CC, CV, CP, CR
Connection - four-wire.
Power supply - 12 V
Communications - Bluetooth

It came in a fairly large cube-shaped box. Nothing was rattling around, and everything arrived intact.


The main load module came with a power supply, instructions for four-wire connections, and a bag of assorted bits and pieces.


The bag contained:
1. Temperature sensor
2. Adapter for the power supply
3. Adapter for connecting via USB ports
4. A pair of power leads with alligator clips
5. Mounting hardware

The power supply is the same as last time, more or less: very budget-friendly.


The product page lists four configuration options:
1. Main module with heatsink and control panel
2. With one, two, or three power modules.

The main module can operate on its own, with a maximum current of 25 A and power up to 150 W.
The power modules only work together with the main module, adding 5 A and 150 W of power each.


The first major difference from the previous model is the separate display and control unit. It's genuinely convenient. The only drawback is that this module doesn't snap into the enclosure of the finished device; it's attached with double-sided tape instead.
External dimensions: 85x62x16 mm; mounting cutout: 75x52.5
The two halves of the enclosure are held together by pins, with no clips or screws. Inside are the control board and display.


The whole thing is controlled by an HC32L170 microcontroller, which is unfamiliar to me. Next to it is a Bluetooth chip. I haven't seen such a simplified solution before: usually it's a multi-pin chip, but here it's just an SO-8.
There are no components on the back, but you can clearly see some contact pads. From what I can tell, the ones at the top left are probably for programming. I couldn't figure out the ones below and to the left, but on the right...

And on the right is the second major difference: there's now an option to connect an encoder. Hooray!!! There are 5 contacts, from left to right:
On/Off
M
encoder
encoder
GND.

I connected an encoder to try it out, and I can say without a doubt that it's much more convenient to control this way. It's also better to connect the M button to the encoder's push button and use a separate button for on/off. So, I recommend soldering a five-pin socket there, connecting an encoder, and setting up proper controls.


Now for the main module. Its design is noticeably more compact than the previous model, mostly because it no longer has a display.
Dimensions: 110x118x70 mm.


The front panel has:
1. A four-position terminal block for connecting the device under test or an adapter. Note: positive is on the left.
2. External temperature sensor connector
3. Positions for two connectors: power and a switch. Their contacts are wired in series.
4. Fan connector with three contacts, though the tachometer signal isn't used
5. Eight-pin connector for the separate control unit.
6. 5.5x2.1 jack for connecting the power supply.

The fan is controlled automatically. It runs briefly at startup, then runs under load when power exceeds 10 W, regardless of the number of modules, and also when the temperature exceeds 40-45 degrees. It's moderately noisy, and sometimes the noise can be a little uneven, but there's no way to adjust the speed.


Three of the four sides have pads for connecting the power modules, and the contact layout is mirrored. When connecting them, make sure they line up with the corresponding pads on the power module. For clarity, the manufacturer also made the cutouts on the boards different.


On the underside there’s a frame for securing the heatsink, along with an explanation of the board connector functions and other additional information—though it’s in Chinese.


The third important difference is that the manufacturer has significantly changed the circuit design. There are now two HLW8110 ADCs, each with two channels: one for measuring current and the other for measuring voltage. These are used in electricity meters.


But it turns out they really went to town with this. The full description of the ADC is at the link above; I’ll keep it brief.
Each HLW8110 ADC has two inputs, one differential and the other single-ended. The differential inputs are connected to the external and internal temperature sensors, while the single-ended inputs measure the load’s voltage and current.
Each ADC also has TX and RX pins for receiving and transmitting data, respectively, and there’s a pair of optocouplers nearby on the board.
As far as I can tell, the ADCs’ digital inputs and outputs are logically connected in parallel, but the first one (current measurement) is connected directly to the controller, while the second one (voltage measurement) is isolated by optocouplers.
The GND pins of both ADCs are connected directly, while the power input to the current-measurement ADC has a resistance of about 260 Ω.


1. Voltage is measured directly from the divider, while an additional op-amp is used for current measurement. It’s visible on the right and marked 333, but I couldn’t figure out exactly which op-amp it is. Current is measured across a 1 mΩ shunt. That would be fine, except in my opinion it’s a bit small for the claimed 40 A—not so much in terms of power rating as the cross-section of the resistive layer and the small contact area. It’s also about twice as thick as the others.
2. The current-balancing circuit for the power transistors, on the other hand, is done properly, with a separate op-amp for each transistor. Current is measured across 10 mΩ resistors connected in series, so each transistor’s circuit has a resistance of 20 mΩ.
There is, however, a small difference with the main transistor: it also has 20 mΩ, but made up of eight 10 mΩ resistors. That’s because the main model is rated for up to 25 A when used on its own; when additional modules are added, the current is shared between them.
The jumper to the left switches the divider resistors in the input circuits of these op-amps. The voltage from the controller output reaches the op-amp input through a resistor of about 10 kΩ. The 300 W jumper connects a 10 kΩ resistor to ground, creating a 2:1 divider; the 450 W jumper connects a 4.7 kΩ resistor, and the 600 W jumper connects a 2.7 kΩ resistor.

There’s also a small fuse visible in the photo, but it’s on the ground bus, quite far from the input. Judging by its location, it seems to protect the measurement circuits—but that’s just a guess.


Of course, I wanted to see what transistor was under the heatsink. They certainly didn’t skimp on the thermal paste :)
The heatsink base is just barely large enough to fit.


1. It uses an IRFP250M transistor, which is rather undersized: these transistors can normally dissipate up to 50 W, or 75 W at most, as in ZKEtech loads, for example. So 150 W is very optimistic. On the left is the STPS41H100CG protective diode assembly, rated for 40 A and 100 V, connected in series with the input power circuit. There’s no current headroom at all, and it’s rated for just 100 V, which isn’t enough to provide protection at the claimed 200 V.
2. I accidentally discovered that the thermal paste under the transistor is there for a reason: there’s a thermistor hidden underneath. I almost pried it off.
3, 4. Just out of curiosity, I measured the on-resistance (with 12 V on the gate) and gate capacitance. The measurements aren’t entirely accurate, but they give you a general idea: 33 mΩ and 5.5 nF.


I put it back together and tried turning it on.
To my surprise, it powered up just fine. The display lit up, so let’s see what we’ve got.


Basically, it’s almost the same as the DL24, but with a few changes:
1. The current setting now has more digits and can be set in 1 mA increments.
2. Current and voltage readings now have a resolution of 0.01 A and 0.01 V. Why set the current in 0.001 A increments, then, remains a mystery to me. The scale also goes up to 9999 V and 9999 A, despite the claimed limits of 200 V and 40 A.
3. The same goes for power: the range is larger, but the resolution is finer.
4. The display for measuring internal resistance has been replaced with an indicator showing the power transistor temperature; the temperature inside the control module is shown above.

Overall, it’s easier to compare the displays visually. On the left is how it looked before, and on the right is how it looks now. The current readings were already there straight out of the box; perhaps it counted them while I was calibrating.
By the way, there’s an LED on the front panel indicating a Bluetooth connection, but it’s hard to see. The Bluetooth symbol shown on the display is much easier to spot. There’s also a power indicator LED on the main board, but it’s barely lit.


Unfortunately, the screen is still dim, and the protective film makes it even harder to read. So, apologies in advance for the poor photo quality—I did my best to bring them out.

The main functions haven’t changed: the load can operate in CC, CV, CP, and CR modes. But a couple of new modes have been added—or, more precisely, one new mode was added, and another was improved and moved to a separate screen. Both modes are selected with the M button, followed by pressing +/-.
BRT — Battery internal resistance test. There are two modes: small and large battery. In this mode, the controller gradually increases the load current while measuring resistance at different current levels.
PT — Power supply test. The load current gradually increases while the output voltage is measured. It’s useful; I wrote that I missed this feature in the previous version. But it works so slowly... It’s more of a power supply endurance test than a test of its load capacity. Sure, it’s useful too, but they could at least have added some way to adjust the speed.



1. Main settings menu, accessed by holding the power button. Current and voltage calibration have been removed, and a choice of minimum current measurement resolution has been added: 10 or 1 mA. I’ll show this later.
2. Hold down the M button before powering on to open the menu for configuring the connected power modules. For example, with two modules, select 300 W, and so on. To switch the setting, keep holding M after powering on and release it when the desired line is highlighted.
3, 4. I stumbled into this mode by accident, then spent ages trying different button combinations to figure out how to get back into it. As far as I can tell, it’s a calibration mode, but it’s not entirely clear how to use it. I managed to get in by pressing M, then pressing -, M, + in sequence, or by pressing power, -, M, +.
The top line is highlighted first. If you apply power, the line below becomes highlighted and the parameters change. Pressing any button displays OK; I assume this saves the settings. The only way to exit is to turn off the power. I tried changing the calibration voltage (apparently you’re supposed to apply 0.5 V), but nothing changed.


First, I powered up one main module and applied a 100 W load. Everything was fine. I tried increasing it to 150 W, also without any problems, but as soon as I changed the current or voltage a little more, the load immediately reported an overload. Protection is good, but this one is extremely sensitive and trips at the slightest overload.
I reduced the power to 140 W and ran it for a while. Even at this point, I noticed that the temperature was rising very quickly. In a short time it reached 83 degrees and kept climbing, even though the room wasn’t particularly warm.


This is a test of the minimum input voltage at which the load can still regulate the current. At 15 A, it turned out that the current is still regulated at 1.58–1.65 V, depending on the temperature. Below that, the current starts to drop.
Naturally, this threshold will be even higher at higher currents, and adding power modules won’t fundamentally change things, since the voltage drop across the input diode and current-sense shunt will increase.


More tests, lots of photos, but there’s not much point in looking through them all; it’s easier to read the conclusions.

Setting the current from 1 mA to 5 A in a 1, 2, 5 sequence.
At low currents, operation is unstable. It settles into regulation at around 10 mA and above. The accuracy is quite acceptable, and at 100 mA and above it’s even good.


No problems at 10 A either. I could have measured higher, but that would have made things more complicated. In my opinion, the load behaves normally.
At low currents, though, there’s not only a large setting error but also a lot of fluctuation. For example, when set to 1 mA, the current can be as low as 0.02 mA, and at 5 mA it can drop to 3 mA.


This is what it looks like at 5, 50, 100, and 500 mA. It seems the current fluctuates by around +/-3–4 mA; it’s just more noticeable at low values.



The menu has a high-precision mode, which sets the current in 0.1 mA increments. I still don’t see the point, since we can only measure in 1 mA increments anyway. Seems like a bit of overengineering to me.


Here, the load lets you set low current values, but it only does so reasonably well at 5 mA and above. Current settings: 0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50, 100, 200, and 500 mA
I should note that the readings fluctuate noticeably less than in the previous mode. Achieving this kind of accuracy at low currents with a 1 mΩ shunt takes some doing.


In high-resolution mode, at currents of 50 and 500 mA.


There were no problems at currents up to 2 A either. I’d like to say it’s all great, but I can’t: the maximum current in this mode is only 2 A, and switching modes means going back into the menu. In my opinion, they could have provided two ranges that switch automatically.


Checking voltage measurement accuracy.
I applied voltages from 0.1 V to 61 V at the input, again using the 1, 2, 5 sequence, except for the highest value.
At 0.1 V, the load displayed nothing. The results were pretty good after that, but at almost every test voltage the reading was about 0.2 V too high. Overall, the result is decent and works for me. I think it could be improved, but it can’t be corrected in hardware; this is a software calibration error that needs investigating.


As with the previous load, CV mode has its quirks. Most likely, because the current ramps up gradually, it overshoots the set value, the power supply drops its output voltage, and the cycle repeats.


This is also CV mode, but at different settings: 10 V 2 A, 20 V 2 A, 20 V 5 A. Then I accidentally raised it even higher, to around 30 V, I think, and the load went into protection.



As I wrote at the very beginning, I bought the load in the “fully loaded” configuration, with three power modules.
The power module is as simple as it gets: a board with a transistor and a heatsink on top. Contact pads are brought out on all four sides of the board, in two pairs: one pair in the normal orientation and the other mirrored.
Each module came with five screws and nuts. The main module came with the same set, though there’s no real need for it there, so consider it a spare.


From here, it’s straightforward. To get a 600 W system, you need to:
1. Connect all three power modules, matching the markings on the boards. With fewer modules, you can position them however you like relative to the main module, but you still need to connect them correctly. With the full configuration, they can only be arranged as shown in the photo.
2. Set the jumper to 600 W.
3. Before powering it on, hold down the M button and, while holding it, select 600 W 40 A.
4. The 600 W mode will then be available in the regular menu.


I don’t remember which mode I started testing in, and it doesn’t really matter. For the next test, I decided to use 50 V and a load current of 10 A.
I set the power supply to 50 V with a 15 A current limit and connected it to the load input. I set the load to 10 A. After 1–2 seconds, I heard a very brief “zzzz” sound. At first I didn’t pay much attention to it, then I tried to figure out why the load was drawing 15 A instead of the set 10 A, with the voltage a little over 2 V...

In this case, I got off lightly: I was powering the load from an adjustable power supply, which simply went into current limiting. If I’d been testing a battery, things could have ended badly. By the way, you MUST install a fuse—and not just any fuse, but one rated for the voltage at which you plan to use the load.



I first tried to find the “burnt-out” module with a milliohmmeter, but ended up simply disconnecting the first power module I came across.


And I immediately picked the one that had failed, even though I initially suspected the main board. The transistor had nearly shorted. When checking MOSFETs, of course, you should first discharge the gate capacitance. The other three transistors were fine.

The problem was that the transistors were already operating beyond their limits, and higher voltage makes things even harder for them. Keep in mind that 2 V at 100 A, 20 V at 10 A, and 200 V at 1 A are all the same power, but they’re very different conditions for a MOSFET operating in linear mode.


I quickly replaced it with another one and continued testing. Along the way, I also decided to see how the heat was distributed between the modules and how well the current was balanced.


To check current sharing, I measured the voltage drop between the load’s negative power input and the transistor sources. On the main board, I measured directly at the transistor lead; on the additional modules, I measured at the connecting terminals.
The measurements were taken at currents of 1, 3, 6, 12, and 18 A.

There’s a noticeable imbalance at low currents, which then decreases considerably. This creates an additional problem when operating at high voltage. That is, When testing a source at 100 V and 3–5 A, we get not only a shift in the safe operating area, but also an uneven load distribution between the channels. As a result, one of the channels will leave the safe operating area sooner and become even more overloaded.

At low voltage, the situation is reversed: the transistors have an easier time, and the balance is better.


Board temperatures in different operating modes. Here, the whole assembly is lying with the fans facing down, so cooling is worse, but this lets us see how hot the transistors get by measuring the temperature of the board underneath them.
Modes:
40 V 1 A
40 V 2 A
10 V 6 A

In the last mode, I had to cover the area where the protection diode is located, because it was "glowing" brightly in the IR range.
Incidentally, it is precisely because of the additional components that the main board operates under a heavier load. The transistor, protection diode, and current-sense shunts all heat up. At 40 A, more than 30 W will be dissipated across all of these components!

Also keep in mind that the current-sense shunts in the balancing circuit will be under heavy load when using just one additional module. The main module is rated for 25 A and uses a 20 mΩ shunt made up of eight 10 mΩ resistors, so they will dissipate 12.5 W, or about 1.5 W per resistor. With two modules and a current of 30 A, each transistor carries 15 A. That is certainly fine for the main transistor's shunts, but the pair used for the additional module will dissipate 4.5 W, or 2.25 W per resistor. The situation is noticeably better with three or four modules; the most critical case is using just one power module.


Having learned my lesson the hard way, I decided to test it at 40 V this time, gradually increasing the load current. I also checked how accurately the current can be set using the power supply's ammeter, and I can say it works great here.
The photo also clearly shows how much brighter and sharper the power supply's display is than the load's. Without the flash, the load's display is certainly readable, but noticeably worse than the power supply's.


At around 560 W, this whole assembly quickly (in about a minute to a minute and a half) exceeded 100 degrees and shut down due to overheating. The device wasn't in its case; it was sitting freely on the table with the heatsinks facing up, and the room was around 20 degrees, so 600 W is out of the question...


Thermal image taken immediately after shutdown. By the way, I found a significant design flaw: when the load shuts down due to overheating, the fans shut off too. They only turn back on after you press any button, which also clears the overheat warning. I recommend installing a thermal switch to keep the fans running at high temperatures; even one rated for 60–80 degrees should do.


One more minor thing I noticed: the main module's transistor is hot, the load is off, the fans are running, and the temperature is dropping. But if you turn the load back on, the temperature reading on the display drops slightly at first, and only then starts rising again. It's not critical, just an observation.


Last time, I was asked to measure ripple in the load's power circuit. I somehow forgot, but this time I decided to run that test too.
Ripple was measured across the main module's current-sense shunt, with the power modules connected, at currents of 0.5, 1, 2, 5, 10, and 20 A. So the ripple was measured across a 20 mΩ resistor at currents of 0.125, 0.25, 0.5, 1.25, 2.5, and 5 A.


And also at 10 and 20 A (2.5 and 5 A across the 20 mΩ shunt), using a faster timebase. The short pulses visible here, and the noise in the screenshots above, are most likely coming from the oscilloscope's power supply. A battery pack was used as the source under test.


As was clear from the start, the load can only be connected to a PC or phone via Bluetooth, since there's no familiar USB port here. But in this case, that actually turned out to be more convenient, and there's galvanic isolation too.
I'm using Windows 10. I opened the BT connection settings, and during the search the DL24M device appeared. I selected it.


The device manager showed both the DL24M device and two virtual COM ports, COM12 and COM13.


I downloaded the software, selected COM12, and it connected. That was almost too easy—where's the usual ritual of waving a magic wand?
But the software isn't exactly "smart and capable"; nothing new appeared in it, unfortunately. Why not make decent software, at least something like ZKEtech's? Is it really that difficult? The software is practically empty...


Conclusions.
In my opinion, this load is interesting in its own right and has a number of improvements over the previous version, for example:
1. ability to connect three additional power transistors alongside the main one
2. added a power supply capacity test
3. an encoder can be connected
4. high-precision current setting and measurement mode, but only up to 2 A
5. the display and controller are on a separate module
6. Power transistor temperature is measured more accurately.
7. Measurement of the source's internal resistance at different load currents.

But there are some drawbacks:
1. The transistors clearly have little power headroom; this will be especially noticeable at voltages above 50 V
2. 600 W is only possible for a short time before it overheats
3. The current-sense shunts operate at their limits, especially the main shunt and the auxiliary ones when using only one power module.
4. The protection diode has no current headroom
5. Dim display
6. No information on load calibration
7. Dismal software
8. The power supply capacity test is too slow, and there is no speed adjustment.
9. CV mode problems again
10. Current “wobble” at low values.

Overall, I recommend buying it, but you may need to make some modifications:
For the best results, move the transistor off the board and mount it on a separate heatsink with a fan. In that case, though, you need to replace the diode (it’s better to replace it anyway) with a more powerful one and cool it as well. You could use the same heatsink as the transistor, especially since its flange is at the same potential as the transistor, so they can be mounted on one heatsink without insulation.
Next, the shunts. Ideally, replace the common shunt with a 1 mΩ one, preferably a four-terminal type, as I showed once before. The problem is calibration: you’ll either have to select it “by hardware” or figure out how to do it in software.
The balancing shunts can be left as they are, but I’d replace them with higher-power ones too, since they’ll also heat up at high currents.
Definitely add an encoder; it makes operation much more convenient.
Remove the terminal block; 40 A is too much for it. Solder the input wires directly to the board.

Also, if you’re buying one, I think it’s better value to get just the main board, since it has everything needed to work properly with external transistors. You can use your own heatsink and transistors. For operation at 600 W, it’s better to use different transistors and, as always, keep them cool—and then cool them some more!

That’s about it for now. I hope this was useful.