Addendum to the previous power supply review

A follow-up to my simple, everyday-user reviews of power supplies (with useful links).
First, a few words on choosing a power supply—which is better or worse in terms of ripple. Out of the box and at the same price, power supplies based on the TL494 have less "ripple
amplitude" than power supplies using MOSFETs. These days, most are "quasi-resonant flyback" power supplies based on the NCP1216 and LD7575. Their circuit board layouts are identical.
I'll split the main AliExpress sellers of MOSFET-based power supplies into two groups. Each of these groups has power supplies from the same manufacturer.
The only difference is the batch: some are still selling the old batch, while others already have the new one.
I have nothing to do with power supplies or electricity; I'm a mechanic—I "turn wrenches." PSUs are simply an interesting topic; I got interested out of curiosity
to get a basic understanding.
Group 1 (3 main sellers, LD7575 PWM controller)
Seller 1 - CLICK.
Seller 2 - CLICK.
Group 2 of sellers (3 main sellers, NCP1216 PWM controller).
Seller 1 - CLICK.
Seller 2 - CLICK.
Seller 3 - CLICK.
If you don't want to do anything to the PSU, I recommend buying one from a seller in the second group.
As for the first group, You can find transistors and output diode assemblies in TO-220 packages in 500–600 W power supplies. In these MOSFET-based power supplies,
the diode assemblies run hotter than they do in power supplies based on the TL494. Diode assemblies in smaller packages should be replaced with proper ones in TO-247 packages.
If you need this power supply for a critical application, be sure to replace the capacitors with name-brand ones. The output capacitors are non-negotiable. It’s also a good idea to replace the fan with a reliable one. These power supplies can fit four 18 mm capacitors at the output, with no gap between them.
In the latest batches of power supplies based on the LD7575, only two capacitors are fitted at the output. For the input filter, there’s space for two common-mode chokes and two X capacitors. One common-mode choke and one X capacitor are installed. If you want, you can add a second choke (in place of the jumpers) and a second X capacitor. Search AliExpress for "X2 Safety capacitors" and look for 0.33 µF–0.47 µF (15–22 mm). There’s almost always more room for the first X capacitor, so install a 1 µF one (22 mm).
If the NCP1216 PWM controller has failed, get a replacement marked (1216AP06)—this version is rated for 65 kHz. Versions rated for 65–133 kHz are also available. Ours is 65 kHz.
For power supplies with a TL494, the common switching frequency is 50 kHz. Check the resistor on pin 6 of the TL494 and the capacitor value on pin 5 (enter them in the calculator). If your board uses the same component labels as the one in kirich's 36V-10A review, those are R17 and C15.
A common-mode choke; don't worry about the 10 A rating, it's small — CLICK.
Now, about ripple. I think it's a bit of magic and alchemy. Different "hands" and oscilloscopes give different results.
But you can clearly see the overall picture and the differences when testing different power supplies. Probes made by "kirich" — CLICK.
We'll be looking at a 48V-12.5A power supply (NCP1216) and a 48V-10A power supply (TL494) from Kirich's 36V-10A review, but labeled 48V-10A.
The popular 150W electronic loads are from Kirich. I first ran the power supply for 30 minutes with a load of about 200W. With the 48V power supply, at currents above 5A, I let the load cool down and then tested it at 6, 7, and 8A. The current during the test, no load — 8A. The capacitors have been replaced in all the power supplies and bypassed with ceramic capacitors.
On the 48V-12.5A power supply (NCP1216), I installed IGBT transistors G30N60A4 (instead of IRFP460A) as an experiment.
On the 48V-10A power supply (TL494), I replaced and rewound the transformer (I was experimenting and haven't finished yet).
I didn't take average readings; I got set up and simply ran the power supply, then saved the results. That's how it turned out.
For the 48V-12.5A power supply (NCP1216), it's 100mV per division. For the 48V-10A power supply (TL494), it's 50mV per division.


Now about the “war” between Kirich’s 150W DIY load and DPS modules. The “pornography” has nothing to do with the power supply’s circuit design.
It also has nothing to do with the transistors in the load (I checked, just in case). I suspect the potentiometer.
The “pornography” is always there, and it appears suddenly, all at once. In other words, it’s not there at 5A, but it is at 5.1A.
There is one pattern: the higher the voltage from the DPS, the later this effect appears (at a higher current). At 12V, you can use it up to about 4A.
For example: 12V-5A set on the DPS (no pornography). 48V-12.5A power supply. DPS5015.


Now, 12V-5.2 A set on the DPS (pornography present). 48V-12.5A power supply. DPS5015.


I add an output filter to the smaller DPS units (3005), between the DPS output and the probe terminals. I don’t add a filter to the DPS5015 or DPS5020.
For some reason, the smaller DPS units are noisier. Replacing the power supply doesn’t change things at all. I found a couple of inductors (toroidal cores with thick wire), put one in series with the positive lead, then added a ceramic capacitor and a regular 100uF capacitor (right across the terminals).
I use 2.5mm wire or 12-14AWG silicone wire (inside the enclosure). One of my smaller DPS units didn’t have thick wires, so I used 16AWG silicone wire and doubled up each lead. For wire, I use PuGV 2.5mm.
I’ve fitted all my DPS units with the same connectors — CLICK.
I use the original small DPS plugs only to connect to the power supply’s terminal block. For banana plugs, I use THESE screw terminals. Inside the enclosure, I use THESE (OT-10A). The current ratings are secondary; what matters is the hole diameter. I only use solder-type terminals. I file down the sharp edges, scrape the oxide off the copper with a sharp point, fill the terminal with solder, and solder in the tinned wires. Then I cover the connectors with heat-shrink tubing. For banana plugs (male), I use THESE. They have protective spring-loaded caps. They work with multimeters. For multimeter leads, I use 16AWG silicone wire and solder it to the connectors as well. Copper alligator clips: THESE.
The alligator clips are excellent; the build quality is much better than that of the usual ones from the store. I’d buy them even if they weren’t copper.

Now let’s get to modifying the transformer in a TL494 power supply. If you’re interested and feel like tinkering in your spare time (like I do), there’s nothing available for sale here, so I started looking on AliExpress. At first, no luck; I could only find ready-made (pre-wound) transformers.
You have to take them apart (boil them) and rewind them with copper wire. Something like THESE.
I was about to buy one, but finally found what we need. It took ages to find. Bookmark this store — CLICK.
The main thing is the transformer we need (with a bobbin) — CLICK. It fits Chinese power supplies (TL494) using the existing holes. No modifications needed. The same store also has blanks for output inductors (cylindrical cores in various diameters).
Inductor blanks (cores) — CLICK. There are plenty of useful parts; see what you need specifically.
Ready-made inductors (from another store) — CLICK.
The prices are great so far. Shipping, for example, is charged only once for the transformers. Buy a lot of 2 and make it 4 — order using the coupons available on AliExpress. The price comes out the same as for two transformers without a coupon (great).
And THIS transformer for a standard power supply is a maybe. You’ll have to check it in person; it might fit height-wise. As for the length, you can drill the holes where needed (there’s room on the board).
The Chinese manufacturers don’t use all the pins on one side of the transformer (the yellow dots in the picture). One pin is left unused for safety. Rather than damage the transformer, I drilled a couple of holes in the board (at the yellow dots).
We take the transformer we need. The holes line up. Compared with the original transformers, it’s slightly shorter but a little wider (2-3mm). Most importantly, its core is wider — 15mm instead of 13mm on the original. It’s designed for higher power (and weighs more). The bobbin’s build quality is much better than the original’s; the new one is clearly a higher-grade part.
What wire to use for winding the transformer. After experimenting, I settled on these sizes: 0.5mm for the primary and 0.6mm for the secondary (we’re not discussing the skin effect).
PETV-2 or PETD wire. This is for hobbyists without special equipment or winding experience. For example, 0.7mm wire is already awkward for me to work with. With 0.4mm wire, you need more strands.
Now, what the Chinese manufacturers wind and how they do it. The count may be off by 1 (or even 2) turns. I opened up a 48V-10A power supply. It’s from the manufacturer of the 36V-10A power supply reviewed by Kirich (wound with aluminum wire).
1) First, the primary — about 18 turns.
2) The secondary: 20+20 (with a center tap).
3) Then another primary winding, about 16 turns.
4) Finally, an auxiliary winding: 5+5 turns.
Surprisingly, there was varnish in places. The same wire diameter was used throughout; with the varnish, it was a little over 0.5mm. Let’s call it 0.6mm. The primary uses two strands. The secondary uses 4 strands — that is, two windings, each with 4 strands (20+20). The auxiliary winding (5+5 turns) uses 0.3mm-0.4mm wire.
Splitting the primary into two sections, as the Chinese manufacturers do, is the right approach.
Here’s how I did it. I'm no pro; I just needed to figure out what would be convenient and how much wire would fit.
You can download Lite Calcit 2000. Fill it in for your setup and see the exact number of turns.
I aimed for round (nice) numbers: 30 (primary) and 10+10 (secondary). The auxiliary winding is 5+5 (I also tried 4+4; it works).
In the end, I wound the primary in one go for now (0.5 mm), using four strands. The secondary (0.6 mm) was also wound with 4 strands. I coated each layer with varnish and
insulated it. For insulation, I used 15 mm thermal tape or 15 mm Mylar tape. I got the thermal tape from green-spark.
Mylar tape - CLICK. In this setup, the windings fit fine, with a little room to spare. I'll have to experiment with splitting the primary: wind it in two sections, with the second section between the secondary windings.
I also tried fewer than 30 primary turns (as suggested by the program), as well as 31 turns. For the secondary, I tried everything from 8+8 to 12+12.
In the end, the power supply works and makes no unusual noises. I always wound in the same direction. When winding the secondary as two windings (say, 10+10 turns), imagine that it's one wire (one winding), so you don't mix up the winding direction. You can also wind it as a single winding with a tap (center tap), but that makes for a long winding. Two separate windings are easier to wind.
I don't know much about this; it's a dark art to me. Choose the number of turns and winding order to suit yourself.
Here's a clear diagram to make it easy to understand.

Primary and auxiliary windings.
Auxiliary winding: 5+5 (with a center tap), using 0.4 mm–0.3 mm wire. I wound the primary with 27–31 turns, using four strands of 0.5 mm wire.
Pin 1 (and the yellow dot) - start of the auxiliary winding.
Pin 2 - end of the auxiliary winding.
Pin 3 - center tap of the auxiliary winding.
Pin 4 (and the yellow dot) - end of the primary winding.
Pin 6 (and 5) - start of the primary winding.
Secondary winding.
I make it from two windings (10+10, 11+11, 12+12).
Pin 1a - start of the secondary.
Pin 2a - end of the secondary.
Pin 3a - center tap of the secondary.
I also tried not soldering the center tap to the transformer pins, but simply bringing out the ends (twisted together) and soldering them directly to the board.
I drilled a couple of holes in the board, between the transformer and the inductor. That seemed to work fine too.
The center tap is where the end of one winding connects to the start of the other. So, start winding the first secondary winding at 1a (10 turns) and finish it at 3a. Start the second secondary winding (as if continuing the first) at 3a and finish at 2a (10 turns).
I always wound all the windings in the same direction.
In short (winding order): primary winding (about 30 turns), secondary winding (about 10+10), auxiliary winding (about 5+5).
If you happen to come across an inexpensive Chinese power supply with the right voltage and THIS design, buy it.
I'll show a few details using a 48 V-8.3 A test power supply. The transformer has been resoldered many times. The output inductor and common-mode choke have also been swapped. I used it to test 100 V capacitors (checking their size). I use it as a test bench. For some reason, nobody pays attention to this power supply. Yet in terms of ripple and temperature rise (at the same power), it beats power supplies with MOSFETs (if we're comparing off-the-shelf units). The components in this power supply are exactly the same as in the 48 V-10 A model (with a TL494), but it costs a lot less. With a $5 coupon on a $10 order, it's a no-brainer.
48 V-8.3 A power supply (for experiments).


Based on a couple of indirect clues, I guessed that this was originally a 36 V-10 A power supply. The Chinese keep a flexible stock of labels: when they need to, they make it a
36 V model; when they need to, a 48 V model. The first clue: an orange ceramic disc capacitor (104) is installed alongside the 1000 µF/63 V output capacitors. It's rated for 50 V. They don't install it in 48 V power supplies (only up to 36 V). The load resistor was supposed to be 1 kΩ (as in the 36 V power supply), but in fact they fitted a 10 kΩ resistor.
They mixed up the orange and red bands. Replace it with a 3.3 kΩ/2 W resistor. Almost by accident, I got an unexpected result with the output inductor in this power supply.
A while ago, I decided to wind copper onto this inductor instead of the original aluminum. But I didn't have the right copper wire. The store had either 1.2 mm or 1.4 mm wire. That diameter is awkward to wind around these two cores (1 mm works great). I thought I'd temporarily wind it with thick 1.2–1.4 mm copper wire as a stopgap, then rewind it later. I forgot about this inductor and used this power supply to test inductors from other power supplies (usually 1 mm wire in 3 strands).
But then I tested this inductor: I set the load to 8 A, then 9 A. I set the power supply to 43 V and replaced the trimmer with a fixed resistor. The protection didn't trip at 43 V-9 A (it trips above 10 A). The inductor barely gets warm. So I left it like that. Today I'd make this inductor differently, using the same wire. Back then, I tried to keep the amount of copper on the top and bottom of the inductor to a minimum, to keep its height down. But as it turned out, there's plenty of clearance under the cover. You can safely wind it with thick wire and simply mount it snugly against the board. The Chinese sometimes make this inductor touch the power supply cover.
Several sellers offer this power supply using the same picture. It costs about 1150 RUB; with a $5 coupon, it's a great deal - CLICK. But I don't know exactly what they're selling now.
One small thing to watch out for with the new transformers in these power supplies. At first I couldn't figure out why the cover wouldn't close properly on one unit after I installed the transformer. There seemed to be enough headroom. I figured it out: the new transformer is slightly taller, so the fan was resting on the edge of the core. By just 1 mm. Press the cover firmly against the end of the power supply (the side with the transistors), and it drops into place. It depends on the particular fan. If needed, just adjust the fan mounting holes a little. I marked the spot in the photo, along with the holes on the board where I brought out the midpoint using a twisted pair (I was experimenting).
The output diode assembly is a C25P40F; all the other components are exactly the same as in the 48V-10A power supply (based on the TL494). Use the price difference to buy decent capacitors.
For now, if you're looking for a high-power TL494 power supply on AliExpress, check the prices (and read the reviews) for DIANQI power supplies. Several sellers carry them.
Also check out Smun power supplies—several sellers carry those too. I can't find any photos of the insides of their 36-48V models.
There are some decent Sanpu power supplies—above-average quality, and priced accordingly. They're already harder to repair, though: they use SMD components and quality parts.
I'll explain how to choose/calculate the resistors for adjusting the output voltage of a TL494 power supply.
Open Kirich's review, “36 Volts, 10 Amps” — CLICK. Look at the color-coded schematic. We need the purple section, specifically resistors
R32, R39, and R40, plus the VR1 trimmer (usually 102, or 1 kΩ). Let's say we don't know the output voltage of this power supply.
Follow THIS link. Enter 2.5 V in the “Vout” field (use a decimal point). For R1, enter 30.68 kΩ (R32 + R39 = 30.68 kΩ; they're in series). For R2, enter 2.5 kΩ (R40 + VR1 = 2.5 kΩ; they're in series). Our VR1 trimmer can range from 0 to 1 kΩ. Start with its maximum value, 1 kΩ. Click “Calculate.” The minimum output voltage of the power supply is 33.18 V.
Now enter 1.5 kΩ in the R2 field (R40 + VR1 = 1.5 kΩ). Set the VR1 trimmer to 0 Ω.
Click “Calculate.” The maximum output voltage of the power supply is 53.63 V.
The same formula works for calculating resistors for the TL431. “Vout” is almost always set to 2.5 V. The rare exception is 1.25 V.
As for 24 V boards for soldering irons: first, here's THIS used board. People on Muska said that, judging by the markings, it was a board the Chinese had overclocked for 12 V. I bought it, and all the components and soldering were factory-original. The Chinese hadn't touched it. Check the soldering on the TL431 resistors and the diode assembly, as well as the current-sense resistor. The capacitors are factory-fitted, branded parts. The input capacitor is 68 µF/450 V. The output capacitors are rated for 35 V. No-load voltage: 24.1 V. The voltage barely sags under load. The protection trips at 3.4–3.5 A. The current-sense resistor has its original insulated lead. It has the full set of filters: two common-mode chokes, an X capacitor, a varistor, and a thermistor. There's an output inductor, too. The heatsink is additionally insulated in hazardous areas. A branded board and an excellent option for soldering irons. Cheap, popular 24 V power supplies: on Banggood (single-sided assembly) — CLICK. On AliExpress (single-sided assembly) — CLICK.
Keep an eye on KSGER's T12 soldering iron kits (STM32). They sometimes have pricing lapses. They get listings mixed up for a while, until they notice. You might find a kit with a board, connectors, wires, and a soldering iron that costs less than the bare board on its own.
For varistors on power supply boards, check the fuse ratings. If the board was designed to have a varistor, the fuse
is usually rated for a very high current. Even small used boards (around 60–80 W) come with a 6.3–8 A main fuse.
Usually the varistor is gone, and so is the second fuse (around 3.15 A), but the high-current fuse is still there. It's no use anymore. This sometimes applies to larger power supplies too (400–500 W): they use a 10 A fuse.
For Y capacitors, it's best to get ones marked X,Y 400 V right away. Some are marked 420 V and so on. They come in different sizes.
The 250 V ones are “small drops”; the 400 V ones are “big drops.” A common capacitance is 222. Y capacitors — CLICK.
As for common-mode chokes, I once wanted to wind one on a good core from the output of a computer power supply. But I looked at branded
power supplies and saw that these chokes are almost always wound on green cores. It looks like you can't just wind one on any old core.
There's a link to the green cores in the review.
A selection of branded used boards — CLICK. If your multimeter has a capacitance measurement function, check the 400 V high-voltage capacitor right on the board (before powering it up). It's best to power up the board for the first time through a safety lamp.
Try to avoid buying 9 V adapters on AliExpress. In the past, almost all of them were made from used phone chargers.
The Chinese replaced the TL431 resistors but left the original capacitors in place. You were lucky if they were rated for 10 V. Often they were only rated for 6.3 V.
How do we check whether a seller is “reliable” when it comes to shipping and tracking numbers?
Add a few small items to your cart (resistors, PWM controllers, for example). Check that the shipping cost
isn't added up (if there is a shipping charge). Sometimes shipping costs are added together even with a good seller. To double-check, click “Buy,”
but don't pay for the order (it'll stay in your orders). Message the seller: “Hey, your shipping costs are being added together—please fix it. Charge for shipping only once.” Good sellers will do that, no problem. Others start bringing up the rules and so on. You know what that means.
How to get a tracking number for free shipping (multiple items)
Again, add a few small items to your cart, but choose untracked shipping. Change the shipping method to “AliExpress Saver Shipping” for just one item. It'll add 50–60 RUB to the total order. Good sellers won't give you any trouble. It will ship the entire order with a proper tracking number. Other sellers will start bringing up the “rules” again.
If you happen to need a small 9V adapter, look in the usual stores. But don’t search for a 9V adapter;
look for switches that use 9V power. Buy one of those. Sometimes this is hard to explain to the girls—and sometimes to their husbands too.
You end up practically swearing at them: buy the switch, put the adapter in your bag, and you can throw the switch itself in the trash right there.
But you can also check for my favorite pricing mistakes right away. Here, they were selling some kind of Tenda for 310 RUB.
It turned out to be an 8-port, 1000 Mbps model (but damn, it’s white). They got one letter in the model number wrong. A switch like that is worth keeping.
As for 12V-2A adapters.
Let’s get a Smun adapter on AliExpress. For example - CLICK. The case is the right kind. It snaps together and is the same size regardless of what’s inside.
Good used boards fit in this case - CLICK. These used boards are indestructible, so if you have a soldering iron, buy one. I haven’t seen the inside of the Smun, but electricians said it can handle 2 A. The cable is the right kind, with a ferrite filter.
This might come in handy for someone. I was looking for ways to spend the small free coupons AliExpress was giving out. I ordered these (they’re on the way).
1) I liked the look of the BM91B multimeter (for the country house, the garage). With the coupon, it was under $10 - CLICK
I’m hoping it has a fast continuity test. I even found their website - CLICK.
2) I saw a “thermometer” on this site and found one on AliExpress at the right price (I adjusted it to fit the coupon) - CLICK.
3) Another coupon deal: a decent caliper - CLICK.
4) While I was at it, I got some proper-looking 16 A probes (the multimeter end is insulated) - CLICK.
5) And while I was at it again, I got a K-type thermocouple with banana plugs (for multimeters) - CLICK
The “lazy” way to install a DPS in a power supply. Standoffs: 18–20 mm. I bought the plastic ones locally and the metal ones on AliExpress - CLICK.
I found the case locally for under $3.


Open-frame load version.


Some old Soviet heatsinks (heavy), better suited to passive cooling (I got them for free). Used IRFP264 transistors - CLICK.
I was comfortable running it continuously at 200 W (48 V power supply). But when I went above 230 W, I was already expecting the load to burn out. Everything went fine, though. It feels reliable, with the limit somewhere around 200 W.
Enclosed load version. Tested up to 150 W; 130 W is reliable. . Standard 92 mm computer fan. Heatsink - CLICK.


That’s about it. Good luck, everyone.