Repair Chinese psu faults with this easy fault finding guide


Tackling a dead power supply, followed by a low current fault

By Dorian Stonehouse.


A boxed Chinese model “Mean Well” S-400-12 (12V 33A)

Switch mode Power Supply


Layout is neat, but effective RF filtering is absent


What makes these power supplies tick?

The beating heart of this switching power supply consists of a pulse width modulation control chip: the Tl494CN (IC1), and a handful of discrete components.

IC1 will produce output square waves at pin 8 and 10 that vary in width (duty cycle), depending on the analogue control voltage level at pin 4

Timing is everything

The duration of these positive-going duty cycles generated by IC1, will determine the power across V+V-. Therefore, wider positive-going pulses will have less dead time, so will produce more power compared to narrower pulses, which have more dead time.


Duty cycle, dead time, pulse width, pulse gap 

These are all interchangeable terms (can mean the same thing) and are often used to explain the same circuit function.


Arbitrary  example only

Suppose a wide + pulse takes up the maximum time-slot of 1 second per second. Therefore, this pulse has a duty cycle close to 100% and little dead time.

As a result it delivers maximum power, a 100% (full-time) duty cycle to a load capacitor (let’s call it C1).


But what if the load suddenly needs 50% less power?

To accommodate this new power demand, the circuit, therefore, begins to shave 0.5 seconds per second (50%) off the width of the wide pulse.

This results in a dead time of 0.5 seconds, and a duty cycle of 50% compared to the wide pulses described above.


As pulse gap increases, output power decreases

And the resultant pulse gap now becomes 0.5 seconds per pulse, compared to the previous gap time of (almost) zero. Correspondingly, C1’s charge will fall by 50% of its previous value.


Back to the TL494CN chip (IC1)

IC1 works in a similar way to the above example.

Therefore, if the voltage at pin 4 of the chip is increased due to a partial short circuit at V+V-, IC1 output dead time will increase.

This will produce less power, so protecting components from current overload and irreparable damage.


Chinese “Mean Well” S-400-12 (12V 33A) switch mode power- supply

circuit diagram

This image has an empty alt attribute; its file name is SWITCH-MODE-S_400_12-scaled.jpg

Repair Chinese psu faults with this easy fault finding guide

Initial Faults encountered

With reference to the above circuit, V+ V- was zero, and fuse F1 had blown.

This was in part caused by a short circuit in one of the 4 bridge rectifiers.

Althought the above components were replaced, further checks revealed more faults.


A possible current surge had occurred

Switching transistors Q1 and Q4 were shorted out between collector and base, and were replaced.

As a result of these shorted transistors, all components in their immediate vicinity were checked.

As expected, base feed resistors R9 and R20 (1.5Ω ) were open circuit, so were replaced.

Luckily, no further short circuits were found on the DC side of the bridge rectifiers.


Always check those mains smoothing capacitors though

C2 and C3 (680 uf mains smoothing capacitors) were found to have lost 100 uf each in value and were replaced. Consequently, V+V- output was restored and adjustable up to 15 volts.


Load check reveals that only light current load was possible at V+V-

Two seperate faults will cause this low current fault, as the circuit “thinks” it is protecting circuit components from destruction:

  • A faulty component reduces pulse amplitude to transformer TR2, turning Q5 off, resulting in a higher voltage developing at IC1 pin 4. This causes longer pulse dead time at IC1 pins 8 and 11 and reduces power at V+ V-.

  • A short circuit at  V+ V-  will again switch Q5 off.  And the same mechanism as above reduces load current capacity.

Notice that in the above list, Q5 is common to both fault conditions.

Repair Chinese psu faults with this easy fault finding guide

The fault mechanism

Under low load conditions (transistor radio), the voltage across potential divider R31, R38 and hence V+ V- was 12 volts.

Voltage on the base of Q5 was 0.7 volts, Q5 was switched on with a collector voltage of about 0.38 volts (low).

0.38 volts appeared on IC1 pin 4, and the drop in voltage caused IC1 dead time to decrease.

The longer duty cycle pulses maintained V+ V- at 12 volts, with adequate current for the transistor radio load to work.


A real load shows up the fault

With a 6 amp bulb connected to this 30 amp power supply,  the voltage across potential divider R31 and R38 (V+V-) fell to almost zero.

As a result, Q5 switched off with its collector voltage rising to 2.43 V and D3 cathode settling at 1.92 V.

With D3 cathode connected to IC1 pin 4, the increased voltage caused the drive pulse dead time (pins 8 and 11) to increase. The result was a reduction in power at V+ V- falling to almost zero across the bulb load.


Checking out reference voltage source VCC

Rectified DC from tertiary winding of transformer TR2 (VCC) was constant at around 4.98 volts, eliminating fault/s on VCC source.

Repair Chinese psu faults with this easy fault finding guide


Making IC1 “think” that everything is working properly

This is a simple method of isolating a fault, involves switching Q5 on to simulate a correctly working load. This makes it possible to check IC1 voltages and waveforms, to see if the stage is working correctly.


Isolating IC1 from the rest of the circuit by using an external power supply

As explained, with a radio load at V+ V-, Q5 had 0.7 on the base (switched on) and 0.38V on IC1 pin 4.

This resulted in waveforms on IC1 pins 8 and 11 being about right, as shown below.


The 6 amp lamp is connected

However, with an external power supply supplying 0.7V to the base of Q5, the same results as above was achieved. But the lamp still remained off.


Repair Chinese psu faults with this easy fault finding guide

What did all this mean?

This meant that there were no faulty components from V+ V- to IC1, including IC1  itself.

Nevertheless a fault did exist after IC1 but before the switching transistors Q1, Q4.


Chinese “Mean Well” S-400-12 (12V 33A) switch-mode power supply voltages and waveforms on IC1


It’s always handy to have a working S-400-12 (12V 33A) switch-mode power supply on standby

Compared to a functioning “Mean Well” power supply, with Q5 artificially switched on, IC1 readings on both units were identical. Consequently, IC1 pins 8 and 11 revealed pulses >2.3 volts at 25.5 kHz.

Repair Chinese psu faults with this easy fault finding guide

Closing in on the fault

Q2 and Q3 (2 x 2SC2655) drive the transformer TR1, which drives switching transistors Q1 and Q4. 

These transistors then drive the main transformer TR2 at a frequency of 25 kHz. 

The resultant ac output is rectified by DB1 and DB2, smoothed by the electrolytic capacitors across V+V- 

As expected, waveforms on the base of Q2 and Q3 matched IC1 outputs, showing that the print circuit was intact.


The faulty components – revealed at last!

On the good  S-400-12 (12V 33A), waveforms on Q2, Q3 collectors were 15.8 and 17.8 volts respectively. 

On the faulty power supply, Q2, Q3 collector waveforms were < 10 volts.


Waveform values for a fully functioning Chinese “Mean Well” S-400-12 (12V 33A) switch-mode power supply


When it comes to testing semiconductors, don’t rely too much on your multimeter  

DC checks on Q2 ,Q3 did not reveal a fault. However, a transistor tester,  revealed one faulty 2SC2655 transistor.  

Two TIP31A (better) transistors took the place of the above transistors.

With V+ V- set to 12 volts, a 0.5 ohm nichrome wire got red hot, with no voltage drop across V+V-, thereby showing that the power supply functioned correctly.


Avoiding inrush current on the S-400-12 (12V 33A) switch-mode power supply

It became apparent that somebody had caused the above fault, by failing to take steps to reduce inrush current during switch on. This would have been easily addressed by connecting the load first – BEFORE switching on the mains supply.  

By failing to do this, residual heating current caused “Negative Temperature Coefficient” thermistor 5D15 (RT1) to drop <10 ohms.  As a result, with 230v ac already across the unit, once a load was connected the maximum pre-load DC voltage and current surge destroyed Q1, Q4, and other components nearby.


The soft start circuit (C24, R19, R29)

This aforementioned power surge will occur after the soft start circuit has completed its protective cycle. This is another reason to be careful, by connecting the load first.


When switching to another load, cool it!

To further add to this fault advisory, if the  S-400-12 (12V 33A) has been supplying a load for some time, and a swap-over to a fresh load is called for, always switch the power supply off at the mains first.

With the mains disconnect, leave the power supply cool down before reconnecting the new load and switching on.


Board modification with higher value NTC thermistor to protect circuit from destructive inrush current

Even before  RT1 has absorbed the inrush current, its resistance at room temperature is 10Ω.

With such a low room temperature value, under load conditions, RT1’s resistance can only go further down in value.

To obviate this problem, thereby increasing life expectancy of the circuit, a thermistor >200Ω at 25°C was substituted for RT1. The replacement component had a resistance of <10Ω under working conditions.

The thermistor was found amongst old television spare parts, but presumably a similar type would be available online.   


The slower the cook the better the taste

The thermistor was tested with the s-400-12 (12V 33A) connected to the lamp and switched on from cold.

As a result, it took 20 seconds for the lamp to reach 12 volts. However, the power supply was then stable, and the thermistor continued to cool down.


Fibre board from a scrap microwave oven

A replacement thermistor, however, must be heat insulated from the surrounding components and the circuit board. This was achieved by spacing the thermistor above several wafers of heat resistant fibre board.

Fire brick adhesive was then used to glue together the fibre board, allowing >1 cm spacing between thermistor and conductive cover!


Takes a bit of work to shift things around

Components which previously occupied space around the 5D15 thermistor were moved to the side of the component board (pictures below).  But the replacement smoothing capacitors from CPC electronics, were too long to fit upright, so they were installed horizontally.

Again, this meant moving components to make room for them. For example, the new bridge rectifier was made and fitted to the side of the switching transformer.  And the rear of the bridge was insulated using neoprene rubber backing and fibre board.

Below are a few pictures that show how the “mean well” s-400-12 (12V 33A)  modification was achieved. 


A word of caution

The modification was carried out by a qualified television engineer as an experiment, and should not be attempted by inexperienced people.


The Chinese “Mean Well” S-400-12 (12V 33A) Switch mode Power Supply prior to the modification


After modification was completed


Elongated view of the switch mode power supply, showing gaps between components and chassis wall



The thermistor modification is challenging. However, the replacement thermistor should markedly increase the life of the Chinese model “Mean Well” S-400-12 (12V 33A) switch mode power supply.

This has certainly been the case here, with the unit used since 2017 as a deep-cycle battery charger.

I hope the repair information proves useful to those wishing to further their knowledge on the ins and outs of this switch mode power supply.  And to those keen to tackle faulty switch mode power supplies generally, I hope to have been of service to you.

Best wishes to all readers and hope to see you all again soon



useful links: