Gospower G0966A-540-0278 power supply, 54 V 2.78 A (64 V 6 A)
¥96 ($15)
As usual, a reminder: to keep up with new posts on the blog, I recommend following me on Instagram, where I’ll post notifications about every new article and maybe write about something interesting too — link to my account
I bought this power supply from the shop of the fairly well-known Chinese modder 100MHz. As it happens, I think it was a little cheaper, though I could be mistaken.
I ordered it back in the summer. It took a long time to arrive, and I even considered ordering another one, but in the end the Chinese seller refused to sell it to me a second time, so I only reordered the EPP-500-54.
This power supply is also available on Alibaba, and it may turn up on AliExpress too. While looking for links for this review, I came across a very interesting power supply. Too bad it’s expensive—I’d be curious to get my hands on one :)
It was simply wrapped in soft foam.
The power supply is almost the same size as the EPP-500-54, which also means it could fit in a small RuiDeng enclosure.

The power supply is built around a U-shaped aluminum chassis that serves as a heatsink, with the electronics mounted on a circuit board.

It measures 127 × 82 × 38 mm. In the photo below, it’s shown next to a 180 W power supply and the better-known “people’s choice” model.

I’m sure you’ve already noticed that the chassis is a bit bent. Yes, the power supply arrived exactly like this. It looks as if someone played football with it—and apparently the seller shipped it that way, since the packaging was intact and the Mean Well wasn’t damaged either.

At first glance, it looked as if the side of the case had bent so far that one of the power components had come loose. But straightening it didn’t change much; most likely, the insulator itself is bending.
Even from underneath, you can see that none of the mounting holes in the chassis line up with the mounting holes in the circuit board...

Top view. The layout is fairly compact.

They clearly didn’t skimp on the input filter—you don’t see one like this very often these days.

1. In addition to the fuse, there’s a 560 V varistor with a 10 mm diameter.
2. There’s also a thermistor, tucked in among the other components.
3. The bridge rectifier is pressed against the heatsink through a thermally conductive rubber pad.
4. There are two input capacitors connected in parallel, rated at 56 µF and 450 V. The active PFC inductor is visible next to them, which is why the capacitors are rated for 450 V rather than the more usual 400 V. They’re Rubycon capacitors, and something tells me they’re genuine.

You’ve probably noticed something odd in the title. Yes, the seller lists this power supply as 64 V, 6 A—that is, 384 W. But judging by the label, it originally came as a 54 V, 2.78 A, or 150 W, unit. As they say, compare the difference!

But let’s move on.
1. Four power components are pressed against one side of the case: the PFC transistor and diode, plus a pair of inverter transistors. The controller board is next to them.
2. Interestingly, they fitted a pair of polymer capacitors on the hot side of the controller board.
3. The output diode assemblies are also pressed against the chassis.
4. There are four 63 V, 330 µF output capacitors: three before the inductor and one after it. These are Rubycon too, which is nice, but having 63 V capacitors in a power supply “overclocked” to 64 V doesn’t look great.

The output section. You can see a small transformer; the inductor is hidden under the board to its left.

Taking the power supply apart is very straightforward: remove the four screws holding the boards and the four screws pressing the components against the heatsink, then lift the board out of the case.
Under the board, I found a large piece of thermally conductive rubber, apparently intended to improve cooling. The only problem is that the heat has to pass through a plastic insulator...

There aren’t many components on the underside of the board, and it looks a little dirty too.

On the left is the current-sense shunt for the power factor correction circuit; to the right is a pair of transistors connected in a half-bridge configuration.

Once I’d removed the board, I could take a closer look at some of the components.
1. Bridge rectifier
2. Y capacitor
3, 4. You can also see that ferrite beads are fitted over the leads of quite a large number of components.

There is nothing unusual in the output section, except perhaps that the board is clearly designed to accommodate a synchronous rectifier controller, which means there are probably also low-voltage versions of this power supply.

While inspecting the unit, I got to the PWM controller that runs the whole power supply: TEA1716T.

This is an unusual—or rather, unfamiliar—controller, since there is nothing particularly unusual about it; it just isn't very common.
It's a controller for a resonant half-bridge power supply with an integrated active PFC controller.
Below is an example circuit from the datasheet. If you'd like to read more, there's a link above. Two things prompted me to look up what type of controller this was:
1. The half-bridge, although at this power level a single-ended flyback is more commonly used.
2. Given the half-bridge topology, I couldn't see the usual output choke, but there was one before the transformer—and it turned out to be a resonant inductor.

In simplified form, it looks something like this. Again, if you'd like to read more, search for “half-bridge resonant converter.”

1. When I inspected and disassembled it, I hadn't powered up the supply yet, but I already noticed that nothing had been touched around the feedback circuitry.
2. But it looks like the optocoupler had been replaced, and someone had soldered the board contacts for some reason.
3. The input terminal block had also been soldered for some reason... Why would anyone do that?
4. During reassembly, I noticed that the insulator was sticking up quite high and could get caught between a component and the heatsink. It turned out not to, although it was clearly very close. I probably should have trimmed it a little.

I power it up: 64 V at the output, although since nothing seemed to have changed externally, I expected 54 V.
I wouldn't call its no-load power consumption low: 4.4–5 W, compared with just 1.4–1.5 W for the more powerful EPP-500-54.

I wouldn't call the efficiency outstanding. Power supplies with over 95–96% efficiency have been around for a long time, and they're not even anything special. But the fact that this supply handled loads up to 6.5 A without any significant drop in efficiency deserves respect.
The horizontal divisions on the graph correspond to 0.5 A, so the maximum was 6.5 A, or over 400 W, despite the rated 150 W!

The efficiency measurement didn't take very long, but the power supply still had time to warm up a little.
The thermistor, transformer, and resonant inductor got the hottest.

The output current is a bit strange. During the efficiency test, the protection first kicked in at 6.1–6.3 A, but on subsequent attempts the supply shut down at 5.8–6 A. However, by increasing the current gradually, I was able to measure efficiency at 6.5 A, although the voltage had dropped to 62.5 V.
So far, so good—but when I wanted to capture the shutdown process dynamically, it turned out the supply could handle 7.5 A too. I stopped there because I felt sorry for it. And this was with the unit hot; it was about the same when cold. The only sign that the supply is overloaded is the noise: at 6.5 A and above, it starts buzzing.
Ignore the power reading; it's wrong because it's calculated using a voltage of 49.5 V, which is the maximum the electronic load can measure.

Ripple measurement.
At high frequencies, it was generally tolerable, if you ignore a slight oddity in how the waveform was displayed. Below are the results at 0, 1.5, 3, and 5 A. The vertical scale is 10 mV per division; the probe was connected without a C+C filter.

The reason the waveform was jumping around became clear when I increased the time base. Below is the ripple at 100 Hz; I had to increase the vertical scale from 10 to 100 mV.

While fiddling with the ripple measurements, I noticed the characteristic interference picked up by the probe, so I saved that too, just for fun :)
1. No load
2. When the current is raised to 0.2 A, the oscillation frequency changes abruptly.
3, 4. Ripple at 1.5 and 4.5 A.

Thermal test—of course.
The test was carried out under the usual conditions: each stage lasted 20 minutes, after which I measured the temperatures, took a thermal image, and increased the current by one step. The load was 1.5, 3 and 4.5 A here, which corresponds to 92, 192 and 288 W, respectively.
Measuring the component temperatures was awkward, since the transistors and diode assemblies are covered by caps.

Thermal images at the end of each stage. As you can see, the transformer got the hottest, but this was at almost twice the rated load, so the result is actually pretty good. I think that with active cooling, it should be possible to run at 6 A output current continuously.

As usual, I also measured how the output voltage drifted as the unit warmed up.
1. Cold power supply, no load.
2. The highest load I tested during the thermal soak.
3. Hot power supply at 4.5 A
4. Hot power supply, no load.
I won't call it “excellent,” but the result is definitely very good: the maximum drift was about 110 mV, which isn't much for a voltage of 64 V.

Conclusions.
On the one hand, you feel a bit cheated: instead of a 400 W power supply, you were sold a 150 W model, and on top of that its output voltage was raised from 54 to 64 V. On the other hand, this is a pretty good power supply, since it behaves well even under a sustained, twofold overload without forced cooling.
I have three real complaints:
1. The output capacitors are good quality, but 63 V capacitors on a 64 V output is a bit... much.
2. The 100 Hz ripple: it'll be interesting to see where it's getting onto the output so strongly. It's unlikely to be pickup on the probe cable.
3. And, of course, the external condition: the chassis is bent, and the input terminal block has been soldered :(
In short, I liked the power supply itself, and the price too, but those “little things” were a bit disappointing.
That's all for today. I hope this was useful, or at least interesting.
¥96 ($15)
This page was found when searching for:
бп 54 вольта, gospower g1342-0800wna-2шт, схема источника питания на 54,6 вольта, tea1716t блок питания, блок питания 63 вольта, https://www.kirich.blog/obzory/bloki-pitaniya/1066-blok-pitaniya-gospower-g0966a-540-0278-54v-278a-64v-6a, gospower g0684-50013240b, g1248-2300w gospower, g0966a 540 0278, плата блока питания 54в 3а
бп 54 вольта, gospower g1342-0800wna-2шт, схема источника питания на 54,6 вольта, tea1716t блок питания, блок питания 63 вольта, https://www.kirich.blog/obzory/bloki-pitaniya/1066-blok-pitaniya-gospower-g0966a-540-0278-54v-278a-64v-6a, gospower g0684-50013240b, g1248-2300w gospower, g0966a 540 0278, плата блока питания 54в 3а
Questions and answers
What voltage and current did the Gospower G0966A-540-0278 deliver?
The measured output was 64 V, although the label specifies 54 V and 2.78 A. The supply handled loads up to 6.5 A.
What load levels were tested on the Gospower G0966A-540-0278?
The author tested 1.5, 3, and 4.5 A, corresponding to 92, 192, and 288 W. The supply also handled up to 6.5 A.
What drawbacks did the author find in the Gospower power supply?
The chassis arrived bent, the output showed 100 Hz ripple, and the output capacitors are rated for 63 V while the output is 64 V.
The figures from this review are in the measurement database: Power supplies
Calculate manuallyAll calculators
Discounted products
You may be interested in
Discounted products
Support the project
All the content on the site and the advice are free, but we’d be grateful for your support of the project and the channel!
Comments2
Discuss on the forumHave a question or some experience to share? Discuss the article with the author and readers: the forum is a convenient place to share schematics and photos.
Go to the forumOnly registered readers can comment. It takes a minute to sign up, and you can log in with the same account you use on the forum.
Log inRegister



Comments are published in their original language.