New version of the PSU-UPS adapter board: prototype

Not long ago, I posted an example of modifying a Mean Well UPS using my own board. That time I used an older version of the board, which had some drawbacks due to its simple circuit design. Today I’ll show you a new version that addresses the shortcomings of the previous design.

This is the first part of the article. The second part will show how it’s used in the target power supply.

A reminder: to keep up with new posts on the blog, I recommend following me on Instagram. I’ll post notifications about all new topics there, 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


So, as a reminder, last time I modified a Mean Well UPS, specifically an older model marked AD-55.


The circuit was greatly simplified and was essentially a higher-power version of a circuit from nine years ago that I wrote about before.
The main drawbacks of that circuit were:
1. A large voltage drop across the reverse-protection diode when running from the battery; about 0.4-0.5 V was simply "wasted".
2. Significant voltage overshoot during charging: at the end, the voltage would "run away" to 14.4 V.
3. It was difficult to set charging currents above 1.5-2 A because the current-sense resistor got very hot.


The new circuit fixes these issues, although it is more complex.
The main circuit itself is highlighted in black.
The charge-complete indicator circuit is in red.
The "cold" automatic startup circuit is in green.
The circuit for automatically starting the protection board, if one is fitted and it locks out after a full discharge until voltage appears at its input, is in blue.

I’ll say right away that the "blue" circuit hasn’t been tested, so it’s just a sketch. The green one was tested, but without the protection diode and resistor; I added those to protect the capacitors if the battery is left connected with reversed polarity for a long time.
A separate note about the red circuit: it’s included here only because I used a dual op-amp and had a "spare" channel. Since an op-amp can be used as a comparator, and the board has a voltage reference based on the VD1 zener diode, it could be used, for example, to generate a battery low-voltage warning. In my case, it’s populated on the board, but I didn’t use it.

As you can see, the current-measurement circuit is very similar to those on well-known Chinese boards; it’s hard to come up with anything new here.
It’s best to use precision resistors for R4 and R5, not so much for their accuracy as for their lower temperature coefficient.

The circuit and component values differ slightly from those shown in the article because the article shows the prototype, while this circuit is closer to the final version.


The PCB layout shown here also fixes some prototype errors and adds a fuse.


Then it’s the usual process: toner transfer, etching, and so on.
By the way, people have asked what flux I use. For making the board, I use our F3 flux. There’s also water-washable F5; once I run out of the first one, I’ll switch to the second. But I only use flux when making the board. After that, I wash it off and rely on the flux in the solder, which is more than enough and leaves things noticeably cleaner.

People have also asked why I use toner transfer instead of photoresist. The answer is simple: I’ve got the process so well dialed in that, compared with everything else, I barely even notice it. I hadn’t made a board for maybe a year and a half or two, but when I needed one, I made it right away without any problems.



I also made a board for another project at the same time, but "something went wrong" and it didn’t work as intended.


Since this was a prototype, I corrected routing errors along the way. One of them had been "hanging around" for about the last 10 stages of the routing process, and I hadn’t even noticed it was there—I just carried it from one stage to the next.


I needed something to test the board with, so I decided to use one of the 12 V power supplies I reviewed some time ago. These supplies are a bit underpowered for this job, of course, but they’re more than adequate for a prototype.


I did have to make two changes:
1. Cut the trace to the overvoltage protection (OVP) circuit.
2. Raise the output voltage to 13.8 V.

The power supply schematic is in the review, if you need it; the link is above.


The voltage ended up a little lower than required, but again, it’s good enough for a prototype. For future reference, I recommend using fixed-value resistors. If you use a trimmer, it MUST go in the lower leg of the divider!!!


The fourth improvement is that this board is easier to connect circuit-wise than the previous one. There’s no need to cut the trace in the power supply’s negative feedback circuit, and apart from the CC signal, the board simply connects between the power supply and the battery.


In practice, the wiring looks something like this: connect positive and negative directly to the power supply output, and connect the CC signal to the cathode of the power supply’s zener diode or to an optocoupler pin.


First test: adjusting the value of the feedback compensation capacitor on the board. I also wanted to increase the charging current.


The board originally had two 40 mΩ shunts in parallel, giving 20 mΩ. I added a third and adjusted the divider. The current went up, but was still somehow lower than expected.


After that, I thought I’d discharge the battery a little and, while I was at it, check how the board performs through a full cycle.
I connected a lamp as a load, waited a while, then plugged the power supply into the mains. The result was 2.35 A.
My divider was 10 kΩ on top and 150 Ω on the bottom, roughly 66:1, with a 13.3 mΩ shunt. So:
The reference is 2.5 V, which gives about 38 mV after the divider.
Calculated current: 0.0375/0.0133 = 2.8 A.

Why was the result lower? First, there are tolerances in the resistors themselves (both the divider and the shunt). Second, there’s the temperature coefficient: the lower resistor’s value drifted downward, so the current dropped too. That’s exactly why I recommend using precision resistors.
But really, what matters here is that the current doesn’t drift upward. The rest isn’t all that important: the current is limited correctly and charging proceeds as it should.

At an output current of 5 A, the calculated voltage drop across the circuit is 0.065 V, clearly less than the 0.4–0.5 V across a diode. There’s no voltage overshoot at all. The only caveat is that right at the end of charging, the circuit will take longer to top off the battery, since the 13 mΩ resistor is effectively in series with it. But what is 13 mΩ compared with the resistance of the wires, protection board, relay contacts, and so on?
Of course, you could also cut the negative feedback zener connection on the power supply to compensate for the shunt’s voltage drop during charging, but in my opinion, there’s really no point.


So, the goal has now been achieved. And to get ahead of myself, I’ll say that it was achieved with the target power supply too—but that’s for another time.

That’s all for today. I hope this was useful, and that you’ll need to make modifications like these as rarely as possible.