Bakon BK950D, a compact soldering station with a T12 soldering iron

There's a sticker on top with a number that you're apparently supposed to enter on the website to check whether the product is genuine.


I tried it a couple of times, but got the same result each time. The screenshot shows a translation from Chinese.
Anyway, I don't know whether it's genuine or not. I was more interested in how it was put together and how it works.


Everything inside was tossed into a pile like this. Nothing was damaged in transit, though I think the packaging could have been sturdier.

Remove the four screws hidden under the missing feet and open the case.


The power supply board is screwed to the bottom of the case, while the control and temperature-control board is mounted in the top cover.


The power switch is connected via a connector, which is a plus. The downside is that it only breaks one wire. Normally, the power switch on a switching power supply should disconnect both wires.
The ground connection, on the other hand, is done properly: no connectors, just soldering. It would have been better to connect it directly to the connector, but this approach is acceptable too.


The first pleasant surprise was a proper input filter, a fuse, and a thermistor to limit the capacitor's inrush current. I was very pleased to see a 100 µF input capacitor; I expected something like 47–68 µF.
No complaints at all on this point; everything is excellent.


Now let's take a closer look at the main components installed on the circuit board.
1. The primary-side high-voltage switch is an IRF840. I was surprised to see a transistor rated for just 500 V in this circuit, since 600–650 V parts are more commonly used. Of course, some readers will ask why that's a problem when the input voltage is only a little over 300 V.
That's true: it's around 315–320 V. But this is a flyback power supply, and the primary winding has a voltage spike of at least Vout × N (the turns ratio). Judging by the 60:8:8 marking on the transformer, I'll assume that's the turns ratio—which makes sense for this topology. That means the transistor will see around 320 + 20 × (60/8) = 470 V, leaving very little headroom.
This is, of course, a very rough calculation, but it gives us a ballpark idea of the situation.
I did a quick search online for user reports and didn't find any mention of this transistor failing, but a 500 V transistor in the primary circuit is still not a good choice.
2. No complaints about the output diode: STPS20H100CT, rated at 20 A and 100 V. This diode assembly is used in 24 V power supplies, so no issues here.
3. The PWM controller is an OB2202CP. I'll probably come back to it later, but for now I'll just say I'm skeptical about placing the controller and power transistor far apart.
4. The interwinding noise-suppression capacitor is a standard type, not a Y1 or Y2 safety capacitor.
On the one hand, that doesn't matter if the unit is grounded. On the other hand, the manufacturer has skimped on a dirt-cheap capacitor, so I recommend replacing it—especially since it's rated for just 1 kV.
A 78M05 regulator is also visible nearby, powering the controller board.


The output capacitors have plenty of capacitance, but their voltage rating is just about adequate.
For an output voltage of 19–20 V, 35 V capacitors are recommended, although I think these should work fine too.
The thing is, the power supply doesn't have a particularly heavy load, and it's only used intermittently. The soldering iron draws around 15–20 W on average, with a current of about 1 A.
At that load, 25 V capacitors should also work fine. You might ask what capacitance and voltage rating have to do with output power.
The capacitance is straightforward: 2000 µF in total is more than enough for a maximum current of 2–2.5 A. The voltage rating is a little more subtle. Capacitors heat up when operating at high frequencies. The higher the voltage rating, the larger the capacitor. The larger the capacitor, the greater its case surface area. And the greater the surface area, the better the heat dissipation and the cooler the electrolyte inside stays.
Also, a switching power supply produces brief voltage spikes, which don't do the capacitors any favors.
Speaking of spikes, here's a nice little touch: there's an output noise-suppression inductor inside.


The circuit board layout is surprisingly good, especially around the output rectifier and filter. The board is also soldered properly and is clean, earning it a solid 4+. I took off half a point because putting the power transistor and PWM controller on opposite sides of the board isn't ideal; it should only be done if the layout is designed with that in mind.


I reconstructed the power supply schematic from the circuit board. There may be some errors, but I tried to include all component values (the capacitance of some capacitors is approximate) and reference designators.


Controller circuit board. This is the part that actually regulates the temperature and controls the operating modes.
While taking it apart and putting it back together, I stripped the threads cut by the self-tapping screws. It felt like they hadn't been screwed in at all, but simply forced into the plastic—a small mark against the manufacturer. It doesn't affect operation, but it can make disassembly and reassembly a bit of a pain.
By the way, the photo shows the connection to the power supply board using four wires, marked In, 20V, GND, 5V. In reality, the connection is 20V, GND, GND, 5V.


The controller board. Overall, I'd give it 4 out of 5: there are traces of flux and some rather untidy soldering, although most of it is fine.


On the back are the solder pads for the connection to the power supply, the soldering iron connection, a trimmer, and the contacts for reprogramming the microcontroller.


1. It uses one of the most popular microcontrollers, the Atmega8.
2. The thermocouple signal is amplified by a dual LM358 op-amp, but only one channel is used.
3. Power to the soldering iron's heating element is switched by an AOD425. This is an N-channel MOSFET rated for a maximum current of 50 A, with an on-resistance of 17 mΩ—more than enough for this application. It dissipates no more than 0.1 W :)
4. At the end of the teardown, I measured the power supply's output voltage: just over 19 V, although it's rated at 20 V. The tip is actually designed to run on a 24 V supply.


After I'd already drawn the power supply schematic, one of the readers sent me a link to the schematic for this station. I already had the power supply schematic, so here I'll include the linked schematic for the control board.
The power supply schematic for my unit is slightly different. In particular, the resistor values in the PWM controller's voltage divider differ.

In brief, calibration works like this:
1. Measure the actual temperature.
2. Write down or remember the reading (I simply kept the thermocouple pressed against the tip).
3. Press both buttons on the station at the same time and wait a couple of seconds for the first digit on the display to start flashing.
4. Use the buttons to enter the measured value (the left button changes the value, the right button selects the digit to change).
5. Press the right button once more, and the controller will save the setting.
6. If you're not happy with the result, repeat the process.

So, let's give it a try. First, I returned the trimmer potentiometer to roughly its original position (fortunately, I still had photos of the measurement results).
1. Measured value.
2. Press both buttons at once in the appropriate mode. In the photo, the first digit is off; in reality, it's just flashing.
3. Enter the measured value of 307 degrees, then release the buttons.
4. The temperature «jumped» upward, but the difference between the actual and set temperatures became smaller.


1–4. I repeated the process several more times. The result didn't improve any further, but it seems perfectly acceptable to me.


Now let's run through the entire range once more.
Before software calibration on the left, after on the right.


You can see that the result is better than when adjusting it by turning the trimmer potentiometer.
But the nonlinearity is still quite noticeable. I'll try again, but I suspect the trimmer potentiometer is actually what controls the calibration linearity.

First, let's figure out how much power the station actually delivers.
I measured the resistance of the soldering iron together with its cable, since that's how it will be used, and I'm modifying the power supply specifically for this setup.
I got 8.73 Ω. Knowing that the power supply output is about 19.36 V, we get a current of 2.22 A. The total power in this (cold “startup”) state is 42.9 W.


The heater accounts for (8.43/8.73) × 42.9 = 41.4 W. In my opinion, that's still not enough power for “heavy-duty” work.
While I was at it, I checked the resistance of the tip's grounding path: 0.24 Ω, which is perfectly fine.


There are many ways to modify it; here are a few:
1. Simply change the resistor values in the feedback divider and the IC protection divider.
There is a risk of the high-voltage transistor failing, since the voltage across it will increase, and it already has very little headroom.
2. The same thing, but replace the transistor as well. This is a reasonable option, but the power supply will operate less than optimally because the output voltage is increased by more than 20%.
3. Add a few turns to the transformer's secondary winding and change the feedback divider values. In my opinion, this is the best option. You may not even need to replace the high-voltage transistor, since its voltage stress will remain almost the same. The downside is that the modification is more complicated.

There were even some completely crazy ideas.
1. Just replace the power supply. That didn't work: the dimensions of the “generic” power supply are clearly larger than the original's.
2. Rework the high-voltage section, removing the PWM controller and all its associated components along with the high-voltage transistor, and replacing them with a TOP248 or an equivalent. Complicated and expensive.
3. Remove the transformer too, and build the power supply around an IR2151-2153. Since a soldering iron doesn't really need a regulated voltage, this is a good solution, especially in terms of heat. But winding a transformer and building almost everything from scratch? No, that's too long and complicated.

Anyway, I thought it over and decided to add turns to the transformer's secondary winding and change the feedback divider.
But first, I needed to figure out what the original power supply could deliver.

The photo below compares it with a 24 V, 100–120 W power supply. Our transformer is clearly smaller.


Since the PWM frequency is the same as that of the TOP2xx ICs I use, I tried simulating a similar transformer to see roughly how much power one of this size could handle at the same frequency.
My estimate was that this transformer can deliver about 62–63 W continuously in a closed enclosure.


In principle, I could have left the high-voltage transistor as it was for my modification, since I planned to add turns to the winding.
But for better reliability, I decided to replace the transistor too.
The output capacitors also have to be replaced as part of this modification.

The originals are 1000 µF and rated for 25 V. That's already cutting it close with a 19 V output, and 23–24 V is out of the question.
It's best to use low-impedance capacitors, since the switching frequency is 133 kHz.
I found a LowESR Capxon LZ 1000 µF, 35 V capacitor at home, but it was too long, as well as a 470 µF, 35 V KF-series capacitor. The KF-series capacitors would have been a better fit. They seem more reliable to me, but a 1000 µF, 35 V capacitor from this series would be too large in diameter, so I decided on a small “hybrid”: a 1000 µF, 35 V capacitor before the filter and a 470 µF, 35 V one after it.
I'll show later why I did it this way.


For the modification, I chose an SPP20N60C3 transistor. This transistor is used in a 25 V, 4–5 A power supply that I reviewed some time ago. The transistor cost just under two dollars offline.

For comparison, here are the key figures from the datasheets. You can see that the replacement is much better than the original; the only thing the new one loses on is gate capacitance, which is twice as high.
The higher gate capacitance increases switching losses and heat: the higher the capacitance, the more slowly the transistor turns on and off. But I figured 133 kHz wasn't high enough for this to be a major concern, and I could always see how it performs in real-world conditions.


First, I desolder the output capacitors and the high-voltage transistor. Perfectionists can straighten the heatsinks while they're at it. In my case, the output diode's heatsink was tilted toward the transformer, so I bent it back a little. Of course, I didn't just bend it—I heated the diode leads first, then bent it :)


I decided to add a few turns for a reason. There’s plenty of room on the transformer for an extra winding, so it would be a shame not to use it.


To add the winding correctly, you need to understand how the existing one is wound. I didn’t want to unwind the transformer, and couldn’t be bothered to even desolder it, so I simply took a close look at the direction the wires run from the terminals.
Transformers are often wound with the turns crossing over each other like this, but you still need to be careful: it’s easy to accidentally connect the winding out of phase and spend ages wondering why nothing works.


Next, cut the trace leading to the output diode as shown in the photo, and drill a couple of holes. The holes should be about 1.8–2 mm in diameter.
Clean the solder mask off the contact pad on the right and tin it.


Take two pieces of copper wire (insulated, of course), about 0.5–0.7 mm in diameter and twice the length of the board.
It’s better to use two wires rather than one thicker wire, since this reduces voltage spikes and puts less stress on the snubber, improving reliability.


Strip and tin one end of each wire, insert them into the hole to the left of the transformer terminal (above it in the photo), and solder them to the transformer terminal.
Then wind two turns using both wires. Try not to bunch the turns together, but don’t wind them too close to the edges either: turns from other windings protrude there, which could be unsafe.
The goal is to get the turns laid out evenly, at least as neatly as in the photo.


Experiment to determine how much wire you need to solder it into the board, trim off the excess, strip and tin the end, slide on a piece of insulating sleeve or heat-shrink tubing, and insert it into the hole on the right.


Solder it as shown in the photo: solder the contact on the left first, then wind the wire, and finally solder the one on the right.


If you feel like it, you can put the original insulating tape back in place.
I couldn’t be bothered to desolder the input capacitor, so I did everything with it still in place. But it’s easier to desolder it first, then wind the coil and put the tape back on.


Don't be alarmed, this isn't the output voltage :)
The photo shows the voltage across the PWM controller's supply capacitor. I'll explain why I measured it.
Since the output voltage is regulated, we can add or remove turns without changing the output voltage (unless, of course, we push the power supply completely out of regulation).
To check whether we did everything right or connected the winding backwards, we can use the IC's supply voltage as a guide.
If everything is correct, the voltage should drop by 3–4 V. If the winding is connected backwards, it should rise by the same amount.
Warning: this voltage is measured on the “hot” side of the power supply, which is directly connected to the mains. Be extremely careful!
Also, I recommend doing the first tests with the power supply connected through a light bulb, for example a 15 W bulb. It should flash briefly, then either go out completely or glow very dimly.

I made a mistake and didn't measure the voltage before the modification. But since I knew the PWM controller's supply voltage was above 11 V, and mine measured 12 V, I couldn't have raised it: if I had, the voltage before the modification would have been outside the normal range.
Still, it's better to measure it before making any changes.


Next, we need to change the feedback divider's resistance slightly to raise the voltage.
There are several ways to do this, too.
1. Solder a 20 kΩ resistor in parallel with the 4.42 kΩ lower resistor. I calculated this value for an output of 23 V (I arrived at 23 V based on a maximum power of 62 W).
2. Replace the divider's lower resistor with a 3.6 kΩ one.
3. Replace the divider's upper resistor with a 36 kΩ one.
4. More complex options, which I won't cover.

The values in options 2 and 3 won't give exactly 23 V, but they'll get you closest. I chose these values because they're from the common E24 series, rather than the less common, more precise series.

For a 24 V modification, however, you can't always stick to the E24 series, so the values are as follows (in the same order as above):
1. 16 kΩ.
2. 3.45 kΩ
3. 38 kΩ (you can use 39 kΩ, but the voltage will be higher than 24 V)


We soldered another resistor on top of one of them. Let's see how it turned out.
We got the calculated 23 V, actually a little more, but that's fine.
The voltage on the PWM controller's supply capacitor also went up. By the way, this made me suspect that the secondary windings may have had different numbers of turns to begin with, since in that case the PWM controller's supply voltage would have returned to 19 V. But I can't say for sure, because I didn't check it before the modification.


After that, I planned to put the board in the case and run it for an hour or two. I also wanted to use this photo to show why I chose these output capacitors.
The capacitor on the right needs to be shorter than the one on the left; otherwise, it will hit the case standoff. Of course, I could bend it sideways, but that wouldn't look very neat. Since the right-hand capacitor is after the inductor, reducing its capacitance to 470 µF will have almost no effect on operation in this case. In other words, 1000 µF before the inductor and 470 µF after it is perfectly acceptable. I wouldn't do it the other way around, though: the capacitor has a harder job before the inductor, and the lower capacitance would make things worse.


But unfortunately, the test didn't go as planned. The soldering iron powered up quite briskly, but when it reached 350, the station rebooted. The power supply's overload protection kicked in briefly, since we'd increased the power.
It was clear that the power supply was operating right at its current limit: everything seemed fine, but it was just a little short.
I decided to raise the protection trip current slightly, especially since I'd installed a more powerful transistor. But because the trip current was already almost equal to the actual current, I didn't want to increase it by much.

A 0.27 Ω resistor is connected in series with the high-voltage transistor. I had two options: replace it with a lower-value resistor or add another one in parallel.
A 2.4 Ω resistor happened to be at hand (although I'd originally planned to use a 2.7 Ω one). Connecting it in parallel would reduce the resistance (and increase the current) by about 12%, which was more than enough for me.
For this approach, you can use 2–3 Ω resistors. If replacing the resistor, use a value of 0.22–0.24 Ω.

I covered the resistor with heat-shrink tubing and soldered it in place on the trace side.


Now it starts up normally, with no more reboots.
Let's put the cover back on and move on to the tests.


I left the station on at 400 degrees for a while, then lowered the temperature to 350 and soldered a board. But since the station had worked perfectly with ordinary boards even before the modification, I finished up by trying the more “demanding” test again.
When I tried to desolder the same capacitor as last time, the improvement was obvious: the power had increased by almost 1.5 times. But I still had trouble desoldering small capacitors, mainly because their bodies were heating up.
Things should improve if you raise the temperature to 380–400 degrees and use a more suitable tip than a sharp conical one.


The temperature profile, however, remained unchanged.
How can that be, some readers will surely ask: it was 40 W, now it's 60 W, and the power supply is still just as hot?
Here's how. The higher voltage (and therefore higher power) affects how quickly the soldering iron heats up and how well it recovers. During normal use, its average power consumption may even go down.
Imagine this: a 50 W soldering iron is heating a huuuge copper pour for about 5 minutes. With a 100 W iron, you might heat it in 1.5 minutes, not 2.5. That's a saving.
With a 25 W iron, you might not be able to heat it at all, and you'd be there forever.

Of course, if you continuously draw heat from 50 W and 100 W soldering irons, the 100 W iron will dissipate exactly twice as much. But that calculation doesn't apply to normal use.

While I was at it, I checked the temperature inside with the cover off.
The transformer was around 60, so there's room to increase the power further.
The high-voltage transistor's heatsink was around 40–45.
The diode's heatsink was around 60, about the same as the transformer.
All those temperatures leave plenty of headroom, but the 5 V regulator heated up to nearly 100, which isn't good.


For the 5 V regulator, I made this improvised “heatsink” from a tinned piece of 2.5 mm² wire. Some people might find it funny, but this tiny heatsink is enough to bring the temperature down by 15–20 degrees, and I don't need any more than that. Besides, it's always close at hand :) If you're feeling fancy, you can use a tab from a high-power diode.
Since I was already tinkering with it, I also replaced the interwinding noise-suppression capacitor, just to be safe.
People often ask me what these look like in a shop and how to search for them.
The capacitance usually isn't critical; 100–2200 pF is enough, with 1000 pF being ideal.
The voltage rating is determined by the capacitor class, so any of them will work.
They look like this, or like this.
I used a capacitor from the circuit board of some monitor power supply.

Almost forgot. If you want to make it even safer, don't screw a self-tapping screw in near the 5 V regulator. The manufacturer laid out the circuit board correctly, with protective air gaps and proper routing.
But they overlooked one thing: if you screw in that self-tapping screw, the total distance between the components on the “hot” side and the regulator's flange will be very small. And if a breakdown does occur, it'll happen right there.
Of course, this is a rather hypothetical situation, but it's theoretically possible.


On the right is what I ended up with, and on the left is what's left over. Don't throw away the leftover components; they may well come in handy :)


And here's the schematic, updated to reflect the changes made.
Replaced components are shown in red, and added ones in blue.


Clean up the residue with a cotton pad and acetone. I thought about coating the circuit board with varnish, but then changed my mind.


Solder the power and ground wires back on, then put everything back together.
By the way, reassembly isn't very convenient—the power button keeps trying to pop out.