19 V 3.42 A power supply: cheap and dreary

Today I'm reviewing a typical Chinese power supply with a "laptop-style" design, i.e. one intended to replace a laptop's original power supply.
Inspection, tests, a small modification—but overall, everything was predictable even before I ordered it.

As usual, a reminder: to keep up with new posts on the blog, I recommend following me on Instagram. I'll post notifications about every new topic there, and might also write about something interesting from time to time: link to my account

I probably wouldn't have bought this power supply, but the stars aligned:
1. The power supply in my monitor burned out. I couldn't be bothered to repair it, so for now it's being replaced by another power supply, but I wanted something more permanent.
2. I had a 2/5 coupon, which brought the price down to $5.04.
3. I was curious to see what the Chinese are currently offering as laptop power supplies.
4. I don't have any 19 V power supplies around, or a suitable 24 V one, and overclocking a 12 V power supply to 19 V just seemed wrong.

So I ordered it, and some time later it arrived in a combined package with a few other small items I'd also bought with coupons.

It comes in a little bag, with some kind of QR code and the specifications printed on the back.


It looks quite neat. The dimensions are 115x53x30 mm, and there's a power indicator on top.
The model name and specifications are printed on the bottom.
It uses a standard three-pin connector for the mains cable and a fairly standard 5.5/2.5 mm plug for the load. The mains cable is not included!


Naturally, I took it apart before even switching it on for the first time. I was very pleased to find that the case isn't glued together; it's held by clips and came apart relatively easily and cleanly.
The case and board are clearly a universal combination: the board is noticeably smaller than the case. Manufacturers of decent power supplies can't afford that kind of luxury.


Well, what can I say? Everything is as expected: inside is a cheap board—not even one salvaged from some branded power supply, just the cheapest power supply board you can get.


1. There's no input filter, fuse, thermistor, or anything of the sort. Actually, that's not quite true: a resistor is doing double duty as a thermistor and fuse.
2. The input capacitor is 33 µF, the inverter transistor is a 2SK3868, and I was surprised to see a fairly large capacitor in the PWM controller's power supply circuit.
3. Y capacitor
4. On the output, there are two 680 µF, 25 V capacitors separated by a small inductor, and a YG865C10 diode assembly.


The transformer may look basic, but it can handle the rated power just fine. The secondary winding is made up of several wires wound together.


The underside is just as cheap, but still reasonably neat. At least the board is relatively clean. There's a protective slot under the optocoupler and Y capacitor.


The power supply is built around a UC3843—a classic, plain and simple.


All right, let's switch it on.
To my surprise, nothing blew up, and the output was at the rated 19 V—slightly over 19 V, but nothing significant.
But the no-load power consumption of 2.2 W seems a bit high to me. I'm used to modern power supplies drawing closer to 0.5–1 W.
When it's running 24/7, even an extra 1.5 W adds up.


I won't be measuring efficiency today. I accidentally fried my wattmeter almost immediately after the previous measurement :(
I wanted to buy one on OLX, but the seller insisted on payment in advance, so I gave up and just ordered one from AliExpress.

I ran the load test using an EBC-A10H electronic load with a four-wire connection.


I tested the maximum load several times, gradually increasing the current limit. It turned out that the power supply holds its output voltage very well, and the protection kicks in above 5.4 A. But it simply acts as a power limiter, so the voltage drops gradually.

In any case, I think the protection threshold is too high. It's easy to push the power supply into an overload condition, and it has no overtemperature protection.


Ripple was measured by connecting the probe directly across the second output capacitor.
The test conditions were no load, and 50%, 100%, and 150% power. In my opinion, the ripple is rather high: 250–300 mV, not counting the spikes.


The same test conditions, but with a different time base. Surprisingly, 100 Hz ripple only started showing up at 150% load.


I checked the temperatures at 50% and 100% load, running each stage for 20 minutes.
Thermal images were taken 20, 30 and 40 minutes into the test. I took the intermediate shot because, in my opinion, the output diode assembly was getting noticeably hot. After 40 minutes in the closed case, 20 of them at 100% load, the output diode assembly was almost 130°C. The transformer winding wires were at a similar temperature; everything else was fine.


During the test, I noticed that the output voltage gradually dropped. Yes, it sagged noticeably as soon as I connected the load, but the load was connected at the end of the output cable, so that was normal. The fact that it kept dropping afterward, though, was not good.


And yes, the output voltage does depend on temperature. With the power supply warmed up and unloaded, it was 19.2 V. After I removed the load, it rose to 19.4 V as the unit cooled down; when cold, it was about 19.5 V.
It's not exactly terrible, but it clearly shows that they skimped even on the feedback circuit resistors.


There wasn't much else to test, and since I wasn't very impressed with the test results, I decided to make a few improvements to the power supply. I picked out a small pile of parts for the job.
By the way, the fuses came in the same combined shipment as the power supply. I bought them on sale for next to nothing.


1. I decided to replace the Y capacitor so I could “sleep easy.” The original is on the left; the one on the right was desoldered from the circuit board of some branded monitor.
2. Inductors: the original is on the left, and the one on the right is from the same monitor as the capacitor.


I replaced just one output capacitor—the one immediately after the diode assembly.
The original is on the left, the new one on the right.

I also replaced the input capacitor, fitting a Samwha 47×400 V in place of the original.


1. The inductor originally had slightly higher inductance; I unwound some turns. I did this because “too much of a good thing is bad,” and also to reduce its DC resistance, which should lower the inductor's temperature.
2. The original current-sense shunt measured 0.26 Ω. If I hadn't been so lazy, I would have replaced it with a 0.33 Ω one.
3, 4. The input current-limiting fusible resistor was 1 Ω. I replaced it with a thermistor whose cold resistance is 10 Ω.


What was left after the modifications


This is how it turned out. I also bypassed the output capacitors with small 0.15 µF ceramic capacitors. I soldered the ground contact of the input connector as well. It isn't connected to anything, but it will help the connector stay in place.

I'll say up front that I wasn't trying to build a good power supply. I just wanted to make it a little more reliable and better quality with minimal expense and effort.


Output ripple:
1–4. The same conditions: no load, then 50%, 100% and 150% load.
5, 6. Low-frequency ripple at 100% and 150% load. This is cleaner now too.


And here's what is now happening at the capacitor immediately after the diode assembly. Increasing the inductor's inductance reduces ripple downstream of it, but increases ripple at the capacitor upstream. That means the first capacitor has to meet more demanding requirements.



As I wrote at the very beginning, this power supply is a pretty sorry sight, but it works and can be improved.
Without the modifications, I probably wouldn't use it to power a laptop. I'd rather pay more and look for a branded unit. In my case, though, I was able to improve it a little, and now it's perfectly usable—especially since the monitor actually draws around 2–2.5 A, so the power supply runs well below its limits.

That's about it. I hope this was useful.