12 V 3 A Power Supply for Access Control Systems

Lately, quite a few of my reviews have started with the words, “Some time ago I posted a review...”
This time it’s exactly the same story: I previously posted a review of an access control power supply, but that one was a 12 V, 5 A model with an uninterruptible power supply function, while this time it’s a simplified version.
But you can find out more in the review itself.


First, a few words about why you’d need such a specialized power supply, and why I used the abbreviation ACS in the introduction.
Many of you have probably seen doors with electronic locks in apartment building entrances, offices, private homes, and so on. In its simplest form, such a system consists of a lock (electromagnetic or electromechanical), a controller, a reader (TM, RFID, fingerprint, etc.), and a power supply.
Together, these make up an ACS—an Access Control System. You can also connect an output from an intercom to the system, so it can unlock the door on command.

A few words about the locks, and why some power supplies have an uninterruptible power supply function while others don’t.
If you use an electromagnetic lock, you need backup power to keep it holding the door when the electricity goes out. You can see this kind of lock on the entrance door of an apartment building. It’s usually mounted at the very top or near the handle, and is often concealed.
An electromechanical lock doesn’t need continuous power, since it’s essentially just a regular latch. People often think an electromagnet moves the latch bolt (although systems like that do exist), but usually it works a little differently. There’s a mechanism hidden inside the lock that is cocked when you close the door and stays cocked until the release mechanism is triggered. It can usually be released in three ways: turning a key, pressing the lock’s mechanical button, or activating an electromagnet with an external command. There is technically a fourth way—a sharp, hard blow near the lock—but it doesn’t work reliably.
The release mechanism shifts a small mechanism inside the lock, and then a powerful spring retracts the main latch. To reset everything, you need to open the door slightly and close it again.

The reason I’m telling you all this is that last time I showed a power supply with the option to connect a battery. Today’s model is a simplified version. I planned to modify it easily for use with a battery, but that didn’t work out.

As before, the kit comes in a bag, with another bag inside containing the remote controls, plus a cardboard box with the power supply.


As before, there are three package options: with one, two, or four remotes. I ordered the most complete version.
1. Power supply
2. Four radio remotes
3. Radio receiver
4. Manual.


The manual is in two languages, Chinese and English. It includes a wiring diagram, but, as with the previous model, the diagram is wrong. The lock is shown connected to the relay contacts, but it needs to be wired in series with the contacts, and a jumper is required. If needed, I’ll draw the correct diagram.


Sometimes you need to operate the lock from somewhere other than a fixed location. In that case, radio remotes are used. A quick warning: this solution significantly reduces the system’s overall resistance to unauthorized entry, so it should only be used where that isn’t a critical concern.
I think the remote design will look familiar to many of you. By the way, exactly the same remotes came with the previous power supply.


Powered by a single 12 V battery.


The transmitter and receiver. You can see the jumper positions used to set the remote’s code. The transmitter and receiver codes must match.


And now we come to the main part of the review: the power supply.
This time the enclosure is entirely metal, but it can still only be used indoors, since it has no moisture protection.


The basic specifications are printed on top: input voltage 100–260 V, output voltage 12 V, current up to 3 A. The connector pinout is also listed.
There’s also a power indicator LED on top.


For the mains connection, the manufacturer simply routed a pair of wires through a hole in the enclosure. At least they thought to protect the hole with a rubber grommet.
On the other side is the connector for external devices. The previous model had a terminal block, which was less convenient. Just a few days ago, I had the pleasure of switching a turnstile control board with a bunch of unlabeled wires, and it reminded me of connectors like these.


We remove the cover, which is held on by a couple of small screws. I was hoping the power wire would be connected inside with a plug, but unfortunately, it's soldered to the board.


The device is very densely packed; the circuit board is completely crammed. This came back to bite me during later experiments, when I had to desolder another component just to replace one.


On the power supply side, there's an improvised heatsink made from a roughly cut aluminum plate, which in turn rests against the case. There's no thermal paste.
On the other side, you can see a connector for the remote-control receiver and a trimmer resistor used to set the relay hold time.
The timer is needed for electromagnetic locks; it sets how long power will not be supplied to the lock. Electromechanical locks are much simpler: they only need a short pulse to open.


For a closer look, I removed the board from the case. There's a protective film underneath it. In my opinion, this design is clearly missing a ground terminal. It's strange that the manufacturer didn't provide one; the previous power supply had one.


There's a proper mains filter at the input. I once made a series of videos about the individual sections of power supplies. Nothing looks particularly problematic here: the input capacitor is 33 µF, and the power supply is rated at 36 W. That's barely enough capacitance for this power rating and our mains voltage; it won't be enough for the claimed 100 V input.
Interestingly, the previous power supply had the same input capacitance and was rated at 60 W.
The photo shows two blue capacitors. The one on the left connects the power supply case to the negative terminal of the input capacitor; the one on the right is an interwinding noise-suppression capacitor. But the first one is the correct type, Y2, while the second is just an ordinary high-voltage capacitor. This circuit is only acceptable when the case is grounded, and there's nowhere to connect ground :(


The PWM controller and output rectifier are identical to those used in the 60 W power supply. The controller is a KA5L0380R, rated for 75 W, and the rectifier is a YG902C2, rated for up to 10 A—plenty of headroom for the claimed 36 W and 3 A.
The power component soldering is a little unusual. Most likely, they first install the board in the case, then screw the components to the heatsink, and finally solder them.


The output section of the power supply also has an inductor to reduce output ripple. I'll check that later.
Everything looks neat on the outside, too. By the way, the board has no SMD components; everything is done the old-fashioned way. Though, as far as I'm concerned, the result is what matters.


1. This time, though, they skimped on the capacitors. Last time they were name-brand; here they're generic no-name parts. The output capacitors are rated for 16 V, which I consider unacceptable in devices like this, since they have to run around the clock.
2. The lock power control circuit is implemented almost the same way as in the previous model, using the same NE555. I'll show it in the schematic later.


The board is double-sided, but the back only has traces. There are protective slots and traces of protective coating, so I had no complaints about this part either.


Even during the initial inspection, it was clear that the circuit would be similar to the one in the previous power supply. Further investigation only confirmed that. The circuit in this power supply is about 95% identical to that of the 12 V, 5 A power supply.


Since it doesn't support backup power, the unit can be divided into two functional sections.
1. Red — 12 V, 3 A power supply
2. Blue — relay control circuit and release-delay timer.

There are also a few cosmetic differences. For example, the previous version had two inputs for connecting a button, while this one has only one. There are differences in the primary side as well; for instance, the negative terminal of the input capacitor is connected to the case (which brings us back to the missing ground terminal).


Unlike the previous version, the receiver module is installed more neatly, although the inductor still rests on the relay case. But there's a drawback: they forgot to make a hole for routing the antenna out. In the previous model, the case was partly plastic, so this wasn’t an issue. Here, though, it’s metal, which means the antenna will have to be brought outside. The only way to do that reasonably neatly is through the hole for the trimmer.


Let’s move on to the tests.
The first power-up went without a hitch: the LED lit up, and I got the specified 12 V at the output.


The trimmer lets you set the output voltage anywhere from 11.7 to 13.7 V.
The adjustment is very smooth, and I had no trouble setting it to exactly 12.000 V. Of course, that level of precision isn’t really necessary; even 12.5 V at the output wouldn’t have caused any problems. Besides, the voltage rises as the unit warms up, and ordinary resistors are used rather than precision ones.


I checked how well it maintains the voltage under load. No problems here either, although there is a slight drop.


Output ripple is excellent overall. A little 50 Hz mains ripple gets through at maximum current, but again, it’s nothing critical.
The probe was set to 1:1, with a resolution of 50 mV per division. Even in the worst case, the ripple was only about 10 mV—low, very low.


Now let’s look at the power supply’s efficiency, since that matters for a device that stays plugged in all the time.
Here’s what I got:
1. No load: 2.2–2.4 W. Essentially, this is the power supply’s own consumption.
2. Current 1 A, power 12 W — 69%
3. Current 2 A, power 24 W — 77%
4. Current 2.5 A, power 30 W — 75%
5. Current 3 A, power 35.5 W — 73%

The highest efficiency was around 24 W. At a load current of 3 A, the output voltage dropped, so the final power was 35.5 W rather than 36 W.


Next came the usual tests at 1, 2, and 3 A, with 20-minute intervals. At the end of each interval, I took thermal images to check the temperatures.
1. No-load mode. There’s a lot of heat on the right side. That’s where the 330 Ω load resistor is located (same as in the previous model), dissipating about 0.5 W. Even with no load on the power supply, the resistor reaches 90°C. I think the manufacturer was trying to eliminate the whine this way, since it’s still present in this mode.
2. Current 1 A. The transformer is at 55°C, while the resistor has heated up to almost 100°C due to the heat from nearby components.
3. Current 1 A, but from a different angle. Now you can see the thermistor at the input, and the peak-temperature cursor has shifted to it: over 100°C. That’s quite high, especially given how close it is to the input capacitor.
4. Current 2 A. The transformer has warmed up to 75°C, and the thermistor to 110°C. After all the tests, I bent it farther away from the input capacitor; it really doesn’t need a heater right next to it.


After warming up, efficiency at 2.5 A dropped noticeably, to 67% instead of the 77% measured when cold. But as it turned out, this was just the first warning sign.
At 1 A and 2 A, efficiency remained unchanged.


I think many of my readers are familiar with my method for testing power supplies: I increase the current in steps and test each level for a set amount of time. This usually lets me find the maximum power the PSU can deliver without overheating.
Above, I showed the temperatures at 1 A and 2 A. The next step was 3 A, which seemed perfectly logical.
But when I raised the current to 3 A, the electronic load shut off because the output voltage had dropped below its protection threshold. That was a big surprise, since PSUs usually deliver their rated current without any problems, and sometimes I’ve even pushed the output power to 150%.


All right, no problem. Let’s find the cutoff threshold. The PSU was already warmed up, which actually made it a better fit for the experiment.
I increased the load current from zero to 3.5 A in 100 mA steps. At 2.8 A, the output voltage started to drop. You can see it both on the graph and in the table on the right, which shows the results of the last few steps.


After finding that the maximum current at which the output voltage doesn’t drop is 2.7 A, I set it to 2.5 A and decided to continue the test. But then a completely unexpected problem cropped up. To start with, I had to stop the test manually after half a minute.


I stopped it because I had a feeling something was overheating. I think the electronics folks will understand.
Sure enough, one of the components reached 132°C. Here are a couple of thermal images from my “experiments.”
Based on the location of the hot spot (the thermal image is offset relative to the photo), I figured out that it was the snubber diode. The thing is, apart from this diode, there’s simply nothing else in there that could get hot.
The snubber is needed to absorb parasitic spikes in the primary winding, but it’s usually the absorbing component itself that heats up—the RC network or the TVS diode—not the diode. In this case, though, it was the diode that was getting hot.


I tried cooling the power supply down and then running the test again, but nothing much changed. At 3 A, the diode mentioned above quickly started overheating; with the PSU already warmed up, the same thing started happening at 2.5 A.


I found an indirect explanation when I checked the power meter’s display during testing. The PSU was still connected to it after the efficiency test, which helped me track down the problem.
Two tests, both at 2.5 A: the first before it started heating up, the second afterward. The clock on the multimeter at left shows the wrong time, but it does show that everything happened within a minute.
At startup, the power drawn from the mains is 40.4 W, but after 20–30 seconds it’s already 50 W. In other words, the PSU starts consuming about 10 W “by itself.” And this is where the drop in efficiency after warm-up comes to mind—I think the two are related.


There’s clearly a problem, and I wanted to figure out what might be causing it. Let’s go back a few months, to when I reviewed the 12 V, 5 A PSU. Back then, everything was fine and it handled its rated load without a problem. Here’s its schematic.


Now let’s open the schematic of the PSU under review. I removed the parts unrelated to the power supply and highlighted some components in color:
1. Blue — different from the previous version, but it doesn’t affect the problem.
2. Orange — the connection has changed, but it doesn’t affect the problem.
3. Red — different, and this one could affect what we saw above.

The key differences are in the primary-side snubber components and the RC network across the output diode. The power dissipated in the first circuit depends on the component values in the second. For example, increasing the value of the capacitor connected in parallel with the output diode will slightly reduce the load on the snubber. But that creates another problem: if you increase the capacitance significantly, the inverter’s output transistor has a harder time. Increase it too much and you risk a failure—I once had a TOP2xx-series PWM controller burn out this way.
Anyway, I digress. The manufacturer reduced the snubber resistor’s resistance and increased the capacitance in the RC network across the output diode. In other words, they were deliberately trying to prevent the snubber from overheating.


I decided not to give up and tried a small experiment. First, I desoldered the snubber diode because I suspected they’d used a regular 1N4007 instead of a “fast” diode, but it turned out to be a perfectly decent FR107. Just to be safe, I replaced it with the more familiar-to-me 1N4937. During testing, there was a suspicious click, and at first I thought the diode had cracked from overheating.
Then I reduced the capacitance of the RC network capacitor across the output diode to 2.2 nF, as in the previous PSU.
I ran the test, and the results were exactly the same as before the experiment. But when the PSU started going haywire and I routinely reduced the load current from 3 A to 2 A, the output voltage dropped even further, while the input power didn’t decrease.
My conclusion is that there was a design problem, most likely in the transformer calculations. You see this kind of behavior when you try to “overclock” a power supply by raising its output voltage. Another possibility is incorrect winding, resulting in weaker coupling between the windings, but that seems less likely.
While writing this review, I thought I could replace the transformer, but then decided it wasn’t worth it in the end. Replacing a resistor is one thing; replacing a transformer is quite another. Besides, I don’t really want to do the developers’ job for them when they’re getting paid for it.

Most likely, they designed the PSU, made a mistake with the transformer, and had already ordered the production run. They somehow managed to work around the issue, but not completely, and ended up ruining a pretty decent PSU.


That’s where I could end the review, but before wrapping up, I should mention that the review was originally supposed to be quite different. I planned to add an uninterruptible power supply function to this unit, similar to what I did in this review. The thing is, that modification requires raising the voltage from the stock 12 V to about 14.4 V. A decent power supply usually handles this without any problems. Its output power will drop a little, but otherwise it works normally.
But in this case, the power supply is already working hard as it is, and raising its output voltage by 20% would make things much worse.
One option is to add a few turns to the secondary winding. This could also be done as a modification to fix the problems, but that would go beyond the scope I set for modifications in these reviews.

I can say that I plan to come up with something for both the first power supply and, possibly, the one shown above, but the solution will be quite different.


Now I can sum things up.
In terms of build quality, I’d say it’s more good than bad. Everything is put together fairly neatly, although the input and output filter capacitors are unbranded. The way the radio module is installed is inconvenient, and there’s no hole for the antenna. If you feed the wire through the hole for the trimmer resistor, it works great—I tested it throughout my apartment.

The electrical characteristics are a mixed bag. Ripple, regulation, temperature rise, and efficiency are all acceptable for this type of power supply, but only at load currents up to 2 A. At higher currents, especially from 2.5 A upward, things go haywire, and it can’t be used at those currents for any length of time.

So, technically, this is a decent power supply rated at 12 V, 2 A, and it’s perfectly suitable for powering an access controller and a lock. For example, an electromagnetic lock rated for a load of up to 280 kg draws just 400 mA. Even with the controller’s power consumption, that leaves plenty of headroom, so you can easily power a couple of surveillance cameras as well. With an electromechanical lock, it’s even simpler: its power consumption is higher, but it’s intermittent.

For this application, there are only two things to flag. The output capacitors are rated for 16 V instead of 25 V, which could shorten their service life. If you replace them, it’s enough to change only the one before the filter inductor. However, since the capacitors have some headroom in their capacitance, the original ones may last just fine. Also, if the power supply isn’t grounded, one of the small high-voltage capacitors should be replaced with a Y-class capacitor.

That’s all. I hope this information was useful. As always, I’ll be glad to answer questions or read your comments.