12 Power supply repair

[blackbrayn]

Hi , i'm trying to repair(yet another  ) cheap chinese power supply with no success , i ran out of ideas and need some help.

Things already checked and replaced just to be sure:
-Main capacitor(hot side)
-Small capacitors(hot side around the IC)
-Diode in the hot side(1n4007) , had burn marks around the pcb , measured ok but put another one.
-Mosfet transistor(12N60)
-12V output filtering capacitor
-Output dual diode

Symptoms:
Power supply starts(somehow) but the green led pulses , aplying a load pulls the output voltage down , removing the load gets the power supply in "pulsing" mode again.

[PKTKS]

Those SMPS follow the pudim formula.. nothing too much
far from the basics.

But you lack vital information to the helpers here..

You must
= put the number of the supervisors and the details
= details are from the fb loop like opto coupler and error sampler

otherwise a pretty clear photo where those things can be reasonably guessed

in general they follow TL494 or UC38xx pretty much to the basic letter
which leads us to protection mode being entered..

like bad soft start or just crappy feedback loop where bad components
are totally out of spec and the fb loop brings the soft start back down

Paul

[blackbrayn]

The PSU uses UC3842D CI , i attached a photo of the area , just measured the Vcc for the CI , it oscilates from 10V to 13V , trying to figure more (looking for the start resistor).
Thx for the involvement.

[PKTKS]

the UC38xx layouts usually have a companion TL431 ...

check if you have one of this in general very close to the
sampler resistors on the secondary side or very close to
that opto coupler...

If not - you probably have a zener or 2 making references
those references are used in the fb loop to put the
UC38xx on the stable operation point.

It does have a soft start RC which slowly will put
the current "sense"  pin in proper operation mode

When that fails it goes wacky on and off trying to start

Paul

[Psi]

Find and check the value of the bootstrap resistor.
It will be like 300-800 kohms, it might be that 470k one i can see.

China loves to use low wattage resistors for the bootstraping circuit and they fail high-value from the heat after a while.
The PSU then fails to start up.

Used to see that a lot on computer psu and on metal cage psu module.

[blackbrayn]

Indeed it uses the 431 reference , i think i have some around , but i have better news , i managed to blow 2 diodes and the mosfet by tricking the feedback loop (put the dmm probes across the optocoupler (primary side), stoped the pulsing and i put a load , the mentioned parts blew) , now i have no voltage supply at the UC3842D(i replaced the diodes and mosfet)
Progress: -3 points. 

[PKTKS]

Unfortunately that shit happens with SMPS...

they are very tricky to debug and experience helps
but does not prevents to do some bad things...

testing SMPS requires some  of ahead thinking...

the dont's are
never cross the sides of the transformer as primary side is high voltage
never probe that crappy side which may have pulses of 1KV
a usable advice is to put a "winding" loop on the primary to try
to see the waveform... if required..

those parts are cheap.  you may replace them easily
but don't  blow the transformer you hardly will find equal easily

Paul

[blackbrayn]

Measured the resistors (startup resistors) that go from + side(hot positive) to the IC vcc pin , they seem ok , so probably  something went wrong with the IC  and it pulls the voltage down to 1 volt , will try to change it if i find one.

Sent from my mobile.

[blackbrayn]

Find and check the value of the bootstrap resistor.
It will be like 300-800 kohms, it might be that 470k one i can see.

China loves to use low wattage resistors for the bootstraping circuit and they fail high-value from the heat after a while.
The PSU then fails to start up.

Used to see that a lot on computer psu and on metal cage psu module.

Measured ok , thx for the sugestion.

Sent from my mobile.

[blackbrayn]

Started trobleshooting a second , identical supply that has the same symptoms , turns on and starts pulsing if i conect a load (seems more stable than the other one that pulsed the output even if no load was conected.)
Just noticed that the mosfet (12N60) gets hot fast with no load).

[strawberry]

Is RCD clamp working properly?
UC3842D need above 16V to operate
snuber capacitor across MOSFET DS can make it run hot

[blackbrayn]

Is RCD clamp working properly?
UC3842D need above 16V to operate
snuber capacitor across MOSFET DS can make it run hot

Didn't find the clasical topology of a rcd i just learned about (thanks)but the only 3 diodes on the hot side measured ok , there is no capacitor across drain-source , voltage on vcc on uc3842 is 13.9 V.


Sent from my mobile.

[tppc]

13.9V is low (maybe due to protection mode)
Did you check the Vcc capacitor? somewhere near the bootstrap resistor...

[blackbrayn]

13.9V is low (maybe due to protection mode)
Did you check the Vcc capacitor? somewhere near the bootstrap resistor...

Capacitor checked fine , it's a 220uF/35v in my psu , also the start resistors (it uses 3 in a series/paralel configuration , in the above  scehamtic it's the 56k one).


Sent from my mobile.

[shakalnokturn]

"RCD clamp" or snubber, will be minimalist on cheap designs.
You have one on secondary with 4.7 resistor and series blue capacitor. Check resistor is not open capacitor not leaky.

You may also have a snubber on primary side with C8+diode?
MOSFET runs hot but isn't it meant to be heatsinked?
Open bootstrap resistors will usually not cause hiccup, they just cause no starting or very random starts.

In case of hiccupping it is more the UC3842 steady supply that needs checking R19+series diode.
Seeing the quality of the PSU you may also want to check input fuse and thermistor for too high resistance.

[blackbrayn]

Got some free time and decided to verify more stuff on the PSU:

"RCD clamp" or snubber, will be minimalist on cheap designs.
You have one on secondary with 4.7 resistor and series blue capacitor. Check resistor is not open capacitor not leaky.  --->Checked resistor and replaced capacitor.

You may also have a snubber on primary side with C8+diode? --->Replaced capacitor , diode too.
MOSFET runs hot but isn't it meant to be heatsinked? --->Solved , put a test heatsink.
Open bootstrap resistors will usually not cause hiccup, they just cause no starting or very random starts.

In case of hiccupping it is more the UC3842 steady supply that needs checking R19+series diode. -->Checked resistor , replaced diode with another generic fast, low drop-out one.
Seeing the quality of the PSU you may also want to check input fuse and thermistor for too high resistance.  --> Shorted both of them for testing.

I will keep trying , thank a lot for the time and help.

Measuring again the UC3842D with the DMM , voltage oscillates between 10 and 16v , no load attached to the output of the power supply , i also made a video with the PSU start and load behaviour.

https://youtu.be/nTkPOHXzn5A

[PKTKS]

(..)
Measuring again the UC3842D with the DMM , voltage oscillates between 10 and 16v , no load attached to the output of the power supply , i also made a video with the PSU start and load behaviour.

UC  chips are tricky

They were designed to induce the transformer in CCM mode on the secondary
In other words  by sensing the primary current UC should be able to
drive an "average" CCM current on the secondary side

But for that to happen some steps should occur first - like soft starting
the primary inductor to leverage the secondary amperage enough
on the sampling FB loop by the "transcondutance" error amp done via TL431

that TL431 is very critical in the UC FB loop because it will induce
the proper CCM mode  - or not

So it may help to put some light load at start just in case you have
a wild deviated or even "forged" TL431 chip - which does happen

a LIGHT RESISTIVE LOAD will definitely helps to put stable loop
there are a plethora of forged chips out there which just go wacky
for no reason

the reason to load the regulator is fairly simple as to force a higher
current on the whole FB loop - mostly on the transcondutance error
amp which will induce a higher error current ...  if not.. it is crappy

the fb error current should vary with the load... a nice way to check
the loop is to measure the error amp current with the load variation

Some forged chips can be spotted some dont

Paul

[blackbrayn]

(..)
Measuring again the UC3842D with the DMM , voltage oscillates between 10 and 16v , no load attached to the output of the power supply , i also made a video with the PSU start and load behaviour.

UC  chips are tricky

They were designed to induce the transformer in CCM mode on the secondary
In other words  by sensing the primary current UC should be able to
drive an "average" CCM current on the secondary side

But for that to happen some steps should occur first - like soft starting
the primary inductor to leverage the secondary amperage enough
on the sampling FB loop by the "transcondutance" error amp done via TL431

that TL431 is very critical in the UC FB loop because it will induce
the proper CCM mode  - or not

So it may help to put some light load at start just in case you have
a wild deviated or even "forged" TL431 chip - which does happen

a LIGHT RESISTIVE LOAD will definitely helps to put stable loop
there are a plethora of forged chips out there which just go wacky
for no reason

the reason to load the regulator is fairly simple as to force a higher
current on the whole FB loop - mostly on the transcondutance error
amp which will induce a higher error current ...  if not.. it is crappy

the fb error current should vary with the load... a nice way to check
the loop is to measure the error amp current with the load variation

Some forged chips can be spotted some dont

Paul

Thanks for the advice , i will put a power resistor on the output and startup the PSU when i get home.

Sent from my mobile.