Mean Well LRS-350-48 Low-Profile Power Supply

A year ago, I published a review of the LRS-350-24 power supply. Today, here's a brief look at a model from the same series, but with a 48 V output. Since these power supplies differ essentially only in output voltage and current, this review will be short, but I hope it will still be useful.


I got this unit along with an adjustable WZ5005 converter, and it was intended specifically for that converter. Since it's a completely separate device in its own right, I decided to take a look at it too, while I was at it, and run a few tests—but put everything in a separate review.

The power supply was bought on Taobao. It's also available, for example, locally for about $28, and on AliExpress for the frankly outrageous price of $57.

It came in the usual cardboard box, with the equally familiar power supply inside.


This model has the highest output voltage in the LRS-350 series and the highest stated efficiency, at 89%.


Connections are made via a screw terminal block. To its left are the output-voltage indicator LED and an adjustment potentiometer.
Since the power supply is designed for a narrow input-voltage range, there's a 230/115 V switch on the side. Before switching it on for the first time, I recommend checking its position—or better yet, removing it altogether.


The load capacity depends on the temperature and input voltage. The unit can continuously deliver the rated 350 W at an ambient temperature of no more than 50 degrees and an input voltage of at least 100 V with the switch set to 115 V, or at least 180 V with it set to 230 V.


The design is identical to the LRS-350-24, so I'll skip the details; they're covered in the previous review.

Cooling is active too, and depends on temperature and power. For example, with the unit cold and running at full power, the fan doesn't start right away. But when the power supply is warm and the load is completely removed, the fan may switch off.


Power supply board


Input filter, bridge rectifier, a pair of thermistors, a pair of varistors, and a low-profile transformer.


The input also has a pair of 560 µF Lelon capacitors connected in series, giving a total capacitance of 280 µF.


1, 2. As in the LRS-350-24, different diode assemblies are used on the output. Someone once asked me why thermal paste is applied when thermally conductive rubber pads are used. This photo is a good example of why it's needed there too—and the diode assemblies are in fully insulated packages.
3. The output has three 120 µF Nippon capacitors.
4. The remaining capacitors are made by Rubycon.


There's an output storage inductor, with the fan power supply inductor visible on the right and the Zener diode for the output overvoltage protection circuit on the left.


I didn't redraw the schematic, so I'll include the one from the 24 V version instead, since the differences between them are minimal. The only significant difference is in the low-voltage versions, which presumably use synchronous rectification.


The output was initially set to about 48 V. No-load power consumption is about 1 W, and the output-voltage adjustment range is 42–56 V, though the adjustment is rather coarse.


The power supply handles an output current of 7.5 A easily, with the output voltage rising by 5–8 mV compared with no-load operation. Heating has the opposite effect, reducing the voltage by roughly the same amount, which more or less compensates for the rise caused by the increased load.


Overload protection trips at around 9.4–9.5 A, well above the original maximum of 7.4 A. But what surprised me was that the protection is latching, i.e. After the protection trips, the input power must be disconnected.
I assume this behavior is related to the low output capacitance and the sudden voltage drop when it falls to the 10 V set on the load.


The power supply's efficiency approaches 91%, only a few tenths of a percentage point higher than the LRS-350-24. You could say the difference is within the measurement error.
Measurements were taken with load currents ranging from 0.5 A to 7.5 A; each horizontal division represents 0.5 A.


Ripple at no load and at currents of 2.4, 4.8, and 7.3 A was around 30–60 mV, excluding the “spikes.” That's very good, considering the low output capacitance and high output voltage.


I once came across a claim that adding a small amount of capacitance when measuring ripple can't affect the result. I repeated the measurement with 1 and 0.1 µF capacitors connected in parallel across the probe, as is usually recommended.


The result was essentially the same, but without the “spikes.” The vertical scale here is 20 mV per division, compared with 100 mV above.


Ripple at a slow time base with 50% and 100% load: direct connection on top, with the additional capacitors on the bottom.
The measurement with capacitors gives a result closer to reality, since there are always wires between the power supply and the load, and the load input often has a capacitor too. This makes the circuit act as an RC filter, which suppresses the “spikes” even more effectively.
The oscilloscope traces are shown at 100 mV and 20 mV per division as well.

In any case, the power supply gets top marks for ripple amplitude.


As I mentioned, unfortunately this power supply is somewhat “incompatible” with the load it was intended for. Its overvoltage protection causes it to shut down when the connected converter produces a voltage spike.
This problem can be solved by increasing the power supply's output capacitance, since the stock 360 µF is rather low. I could leave it at that, but there's a second issue I can't ignore: the possibility of a critical operating condition when connecting a load with high input capacitance—or, worse, a battery—to the converter's output....
One idea is to put a protection diode between the DC-DC output and the converter output, and take the feedback signal from after the diode. But that's for another time.


My conclusions will be much the same as in my LRS-350-24 review: this is a great power supply. The only real criticism I have is the low output capacitor value. At 360 µF and an output current of 7.3 A, that's far too little in my opinion—about 50 µF per 1 A of load current. I understand that this is partly due to the inverter topology, but still.

That's all for now. I hope you found this useful.