BW150, Atorch’s new electronic load

Not long ago, I spotted a new product from Atorch, got interested, contacted the manager, negotiated a discount (unfortunately, not a very big one), ordered it for review, received it, and now I’m sharing it with you.

To stay up to date with new blog posts, I recommend following me on Instagram, where I’ll post notifications about all new topics and might also write about something interesting from time to time - link to my account

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

Previous reviews of Atorch devices
1. DL24 - the original four-wire load - review
2. DL24M - a modular load, but with one transistor per module - review
3. DL24MP - a brief look-over; this one is my favorite - review, differences in the new version - link
4. DL24EW - a load with charging control, first revision (the third revision is currently on sale) - review
5. CL24 - a load with charging control, but rated for up to 150 W - review
6. Protocol trigger - review




In fact, this load is very similar to the previous CL24, which I’ve already reviewed, as well as its modified version, the CL24R. I didn’t see any point in reviewing that one, since its key difference was support for connecting more devices at the same time.

The store offers four package options:
1. Basic
2. Basic + protocol trigger
3. Basic + protocol trigger + adapter for connecting a charger + standard battery holder
4. Basic + trigger + switch + battery holder + new adjustable battery holder

I chose the second option because the protocol trigger is useful to have, I already have plenty of holders, and I don’t need the charger switch at all, since I use my own chargers rather than this workaround. It was also cheaper.


The package included:
1. Two wires for connecting a power source. Ah, there was a time when they included four...
2. Temperature sensor
3. USB-A to USB-C cable
4. Small board for connecting USB chargers
5. Protocol trigger.

There’s not much to say here: apart from the protocol trigger, it’s a standard package, and I showed the trigger recently. One important difference: the power supply is no longer included. You’ll understand why as you read on.


The overall appearance is almost identical to the CL24; we’ll compare them later. The package includes a sticker that, among other things, says the device can be powered in three different ways:
1. From 5 to 12 V through the DC5521 input
2. From a 5 V, 2 A power supply through the USB-C port
3. From 12 V through the same USB-C port, using the PD/QC protocols.

Most people already have power supplies like these at home, so there’s no point in including a separate one. That may be for the best, since the included power supplies not only had a “Chinese” plug, but were also pretty poor quality.



Like the CL24, this is essentially a continuation of the DL24 line, at least in terms of its design: a board with a display mounted on it and one transistor under the heatsink. As I’ve said, in my opinion this is either a step backward or, at the very least, a step in the wrong direction. The DL24EW already exists, even if its software is a bit rough, so why not develop that line instead?

You might ask what the difference is. The difference is that although this load is rated at 150 W, that figure comes with a lot of restrictions.
Up to 36 V - 150 W
From 36 to 80 V - only 60 W
And above 80 V - just 45 W.




I explained in this video why this happens, and why a transistor rated to dissipate 150-200 W may only be able to dissipate 45 W. I strongly recommend it to anyone who only looks at the maximum figures in datasheets. It may also simply be useful to watch.


The terminal block for connecting the device under test or a battery is on the left side. The contact pads for connecting the charger's switching module are also located here.
On the right side, there’s a USB port for connecting to a PC and a pair of mysterious contact pads. Interestingly, the CL24 also had a mysterious contact pad on this side. Now there are two; maybe the next version will have three, and we’ll finally find out what they’re for.
On the end opposite the display are a USB port and a DC5521 jack, which we use to power the load.
Nothing interesting on the bottom: the fan backplate and some additional information, in Chinese.


Near the power connectors, you can see a pair of chips and an inductor. I’d guess one chip is a step-up DC-DC converter and the other is a PD/QC controller.
I’ll also note right away that the DC5521 jack is good quality. It’s the same one used on the trigger boards—nice work.


The display area is almost identical to the one on the DL24. What disappoints me most is the lack of an encoder. Those who’ve used one will understand.
There is a catch, though: not everyone can use one properly, because the encoder on the DL24EW works downright badly, while the one on the DL24MP has to be installed by the user. It’s a real paradox: on the load that comes with an encoder, it works badly; on the one where you have to solder it in yourself, it works perfectly...

I just want to say: developers, get in touch—I’ll explain how to make your load perfect, or close to it.



The circuit design appears to be exactly the same as the CL24’s, at least in the analog section. That’s why I’ll be cutting the tests short; there isn’t much point in repeating them.
The one advantage is that the chip markings haven’t been sanded off here.


A few detailed photos.
1. They’ve finally replaced the shunt made from a pair of 2512 resistors with a proper one. I’ve complained about these cheap shunts so many times, and they’ve finally listened. By the way, on every load I build, I have to replace the stock shunts with higher-power ones for three reasons: reliability, power rating, and TCR. In other words, if you use a “beefier” shunt (they’re available in 1, 2, and 3 W ratings in the 2512 size) and improve its heat dissipation, it will run cooler and therefore drift less for the same TCR. Plus, it seems that higher-power resistors have a thicker resistive layer. In short, I fully approve of and support switching to a resistor like the one in the photo.
2. The second difference is that the wireless interfaces have been moved directly onto the main board: Bluetooth is handled by the controller, while WiFi is provided by a submodule. I approve of this too, although there’s a catch: you pay for everything upfront and have no choice, but the total cost of the fully equipped version will be lower.
3. There’s a small DC-DC converter above the submodule. I don’t know what it powers; most likely, it supplies power to the board’s electronics in general.
4. There’s another converter near the USB port. I’d guess the first one steps the voltage down to 5 V and the second to 3.3 V, but I haven’t checked.




And of course, let’s compare the two loads side by side. The CL24 is on the left, and the one under review is on the right. At a glance, the only differences seem to be the color of the circuit board and the fan, but that’s far from the whole story.


First of all, despite having the same analog section, it clearly uses a different microcontroller, and wireless interfaces have been added.


The previous model only has a DC5521 input, and it only accepts 12 V.


And it turns out even the fans are different. By the way, I’m building a pair of CL24 units into an enclosure, with my own modifications. It’s not just because an enclosure is more convenient: the stock cooling system also bends the board quite badly. Plus, I’ll be using four transistors instead of one, which should improve reliability, along with a few other modifications aimed at improving reliability and performance.



The undersides of the boards are almost identical; nothing interesting there.


When you turn it on, it first asks you to select a language, then download the software, then walks you through a full tutorial. After that, it lets you connect to WiFi, and only then takes you to the main screen, which will also switch automatically after a while to the default CC mode.
All of this appears every time (except for the self-test and WiFi connection screens), but you can disable it in the settings.


The main screen. First, it’s noticeably larger here; second, the description of the second section is missing from the store listing.


There’s a brief description on the product page, but apart from the last few modes, there’s nothing new here. It’s all exactly the same as the CL24 modes.


But the mysterious menu items with some kind of sine wave caught my interest. At first, I even thought: no way—did they really add dynamic tests? Yeah, right—just wait. The digital/graphical display mode used to be selected in the settings menu; now they’ve given each mode its own button. What’s the point?
The selected mode is shown in the upper left. If you try to enable the battery test mode with charging, it prompts you to connect a charger, and the Developing menu still has a placeholder in it.



The settings menu is almost identical to the previous model. At a glance, the display mode selection is gone—which makes sense—as is the option to choose the type of PC connection.


There’s some description on the product page, but it’s more of a demonstration of what the listed options do, since there’s very little useful information.


I’ll comment on a few of the options separately.
1, 2. Maximum power. It’s 150 by default; you can select 155, but it will tell you the maximum limits for each voltage. I feel like shouting: Bring back at least 1000 W like the DL24 had—or better yet, 2000–3000!!!
3. An extremely useful option: you can set the pause between the charging and discharging stages. Why this wasn’t added to the DL24EW is a big mystery. Please add it.
4. This is the most mysterious option. I don’t know who named it, but the name is definitely misleading. I thought it meant “discharge amp”—some kind of extra amplifier that kicks in when discharging at low currents. But after thinking about it a bit more, it turned out to be simpler: in Chinese shorthand, it means minimum discharge current, i.e. the lowest value you can set. It’s 5 mA by default, but you can set it to 2 mA. In any case, it’s not particularly useful, as I’ll show later.


First, let’s test voltage measurement accuracy. In short, it starts measuring at around 50 mV. Above that, it’s generally quite accurate. There is some measurement error, of course, but I’d say it’s very small and shows up mainly above 20 V.


Voltage is interesting, of course, but I was much more curious to see how it handled setting the current, especially at low values.
First, let’s test from 2 mA to 9 mA. I put this in a separate photo on purpose, because this is the most difficult range.

So, at 2 mA it tries to show something on its display, but the power supply doesn’t register that current—photos 1 and 2.
At 5 mA, things are a bit more promising: it sets the current, and the power supply even registers it, but it fluctuates, so the current can be either 5 mA or 2 mA.
Next, tests at 6, 7, 8, and 9 mA. Better, although there’s still noticeable fluctuation, so it’s not usable.

The point is, you can set the minimum current to either 5 mA or 2 mA in the settings; this range is still dead either way. On the other hand, what did you expect? The shunt here is 5 mΩ, and it measures both polarities. At 5 mA, the voltage across the shunt is just 25 µV :(

By the way, someone once asked me about electronic loads that can work properly at currents of 10 mA or even 1 mA. Yes, they exist, but they’re extremely rare because they’re very specialized. One option is to replace the 5 mΩ shunt with, say, a 50 or 500 mΩ one. You’ll have to mentally adjust the divider readings, of course, but at least it’ll be a factor of 10. And yes, with this setup, the maximum current will be 2.5 A or 250 mA, respectively.



Thinking about it afterward, I probably should have checked with a multimeter. But the point is that even the power supply measures the current quite reliably from 2 mA, although I was concerned not only with accuracy but also with stability—and the load has trouble with that in this range.


Next, tests at 10, 20, 50, 100, 200, and 500 mA, and at 1, 2, 5, 10, 20, and 25 A.
In my opinion, the load reaches a reasonably stable operating mode above 20 mA, which is more than enough for the vast majority of situations. There’s noticeable measurement error at the maximum current, but the power supply itself isn’t as accurate as the 6006P I used for testing currents up to 5 A.




After that, mostly out of curiosity, I tested CV mode. I set the power supply to 50 V and 0.5 A, then set the load to 5, 10, 20, and 48 V.
It basically works, but there’s something odd: with the current limit set to 0.5 A, the power supply showed 540 mA, and the load showed the same current. This is usually caused by ripple, and you can partly fix it by connecting a capacitor of around 1000 µF in parallel with the load input. The exact value will take some experimenting.
Since the analog section is identical to the CL24 load, you can check out the CV mode tests and related quirks there.



I didn’t run this test to measure temperature, but to check operating stability. First I set it to 150 W, then almost 155 W—154.90, to be exact, right up against the 155 W maximum. The load ran like this for quite a while, but apparently the current exceeded the set limit briefly, and it went into protection. It runs very quietly. The fairly decent fan speed control helps, too, though the CL24 had that as well.

Temperature-wise, things are about the same as with the CL24, although personally I think the new version’s heatsink is a little more effective thanks to its greater number of fins. For the most part, though, it all comes down to how well the transistor is pressed against it. When I sold a couple of DL24 units from my remaining stock, I immediately modified them by bolting the transistor through the board to the heatsink and applying a good thermal paste. I have to say, it made a noticeable difference to the transistor temperature. I did also install a more suitable transistor—the type I use in my electronic loads—but that’s a minor detail and has nothing to do with the temperature. Maybe when I decide to sell this load, I’ll make the same modification. The DL24 had its temperature sensor off to the side, though, while the new models most likely have it in the center of the transistor’s mounting hole.



Connecting to a PC and phone. I ran into a bit of a snag here: the E-test program is easy to find on Atorch’s website, and the program for connecting over Wi-Fi has nothing to do with Atorch, so it’s even easier to find. But for some reason I couldn’t find the PC software on the website. There are tons of manuals, but no software. In the end, I found an archive with the software in my backups. I’d downloaded it back when I reviewed the CL24, so I used that.

Oddly enough, even though the program is for the CL24, it found the new load without any trouble. There is one catch: the Chinese interface displays the model name correctly, but if you switch to English (the tiiiny button in the top right), it shows “Simulation” among the connected devices and CL24 in the title bar. Still, it lets you open the settings and works just fine.



You can configure and start it, then view both the operating graph and graphs for temperature and capacity/energy.



Connecting over Bluetooth was trouble-free, too, but the app is ancient and unbelievably buggy. On top of all that, there’s no way to switch it to English.


Tuya was trouble-free as well: it’s easy to find and add to the app. Damn, I’ve got quite a collection in there already :)


Everything works, but, as with the CL24, it’s not very convenient to use. Unfortunately, this app just isn’t really designed to work with devices like these :(



As an extra, I decided to run one more test and see how hot the transistor gets.
I set the voltage to 13.7 V—or rather, that was already set on the power supply, and it suited me just fine. The load current was 11 A, for about 147 W.


I let it run like that for about 10 minutes, although I noticed the temperature had already stabilized after roughly 5 minutes.
The load read 83.2 degrees, while the actual temperature was 90.8. I’d say the readings are fairly accurate. When the sensor was off to the side—or, even worse, somewhere who-knows-where, as on the 24MP and 24EW—the difference was much greater.


I removed the heatsink. I was curious to see what transistor they’re using now.


Let’s also take a look at the diode assembly: 40 A, 100 V, and judging by its appearance, used.
They didn’t skimp on the thermal paste. It looks like GD900, but too much thermal paste is actually not a good thing—it’s not oil in porridge...


I removed the transistor. As expected, the temperature sensor sits inside the transistor’s mounting hole, and the transistor is the right one now, as I wrote earlier.

I took a look, snapped a photo, applied some fresh, higher-quality thermal paste—the kind I usually use—and put everything back together.


I repeated the test with the same settings and for the same amount of time.
According to the load, the temperature dropped from 83.2 to 77.2; according to the thermal camera, it went from 90.8 to 85.2. Well, a 5.6-degree drop is pretty good, if you ask me—and all it took was some good thermal paste. By the way, the thermal camera now tried to focus on the transistor’s leads rather than the transistor itself. but with a very small difference.


Now for the conclusions.
The good:
Well, at least they didn't make it worse, which is already a relief.
They installed a proper shunt
They used a transistor rated for linear operation.
They put all the wireless interfaces on a single board.
They used a different controller, though probably because it has Bluetooth.
They made the power supply universal and stopped bundling their own cheap power supplies
They added a kit option with a four-wire battery holder
Unlike the current DL24EW and DL24MP models, this one still has a full four-wire connection.

The not-so-good.
Well, it's basically the same CL24 with a few tweaks. As I wrote, the analog section is identical, so practically all the shortcomings listed in that review have carried over here too.
The limit is still just 150 W, which isn't much
The power limit still depends on the input voltage: above 36 V, you can't draw more than 65 W. Worst of all, there's no way to disable these limits. These are the biggest shortcomings for me. And as I wrote before, I'll say it again: developers, get in touch with me. I think we could make this device much better, and I'm sure it would become a bestseller.
The encoder—how could you? Bring back the encoder!!!

And overall: subjectively, this load is better than the CL24—or rather, that's not quite right. It hasn't gotten better; it's become more convenient. Unfortunately, some of the shortcomings are still there. But I can now recommend this load, especially if you work with batteries up to 12–24 V and 150 W is enough for you.

By the way, a quick question. Do you think it would be worth making a proper equivalent of the original battery switch? Actually, one already exists, but we could make it fit in place of the original, using the same terminals, while providing lower voltage drop, handling more current, protecting against reverse polarity on both sides, and fully disconnecting the circuit rather than only on the load side.


That's all for now. I hope this was useful, and as always, comments are welcome.