Two dead ATX SMPS's - TL494 and FSP3528 based

[imza]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

ok, thanks a lot selldoor.. i can't notice it's before.. but with ur explanation i was just realizes.. but how about the value of zinor diod 1-ZD1.. do you know it value.. cause in case i have missing that components.

[imza]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

and the value of zinor diod 1-ZD1.

[selldoor]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

I know the link is dead but do you understand what he said:
Found out that Q2 (please refer to generic schematic of HV side of SMPS here:
http://www.geocities.ws/luizissao/im...DO_ATX-250.gif ) d304x mosfet was dead/shortened.

Are yours burnt up are they labelled Q1 and Q2 on the board?

He says
Found out that Q2..... ........d304x mosfet was dead/shortened.
Therefore Q2= D304x
http://www.alldatasheet.com/view.jsp?Searchword=D304X

It is implied further down that Q1 and Q2 are the same - If you do not understand I
will try and explain it

[imza]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

i means, this two transistor on this pictures:

[selldoor]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Hello domas could u give me the spec of two transistor Q1,Q2 at the psu FSP ATX-250PA.

In post 1 he says:
Found out that Q2 (please refer to generic schematic of HV side of SMPS here:
http://www.geocities.ws/luizissao/im...DO_ATX-250.gif ) d304x mosfet was dead/shortened.

[imza]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Hello domas could u give me the spec of two transistor Q1,Q2 at the psu FSP ATX-250PA.

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

i am not an expert in different topologies and regulation methods

But it would be interesting to see how to improve that. It is just that FSP chips lack documentation. Is this p/s generally considered worthless POS?

[momaka]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Is this behavior normal with cheap p/s

Usually no. Most cheap PSUs will tolerate a lot more cross-loading than this. It's probably just a crappy design/platform. Older half-bridge Sirtec PSUs, for example, have/had the same problem; load them on the 5V and they are happy, load them on the 12V heavily and they oscillate.

Deer PSUs, on the other hand, no matter how crappy, always tend to regulate wonderfully. Also, low-end ISO PSUs from Channel Well have great regulation too, even though most of them barely have enough components to function. I have a totally gutless ISO PSU that I can load the 12V rail with 2 to 3 amps, and it is still in spec voltage-wise (not sure about ripple, though... do you think a single 680uF cap can do much there ).

or i should check caps of some 12V rail?

Well, if the caps are bad, that could cause the lower voltage and sloppy regulation too.
But from what I remember reading long time ago, FSP PSUs just aren't good with regulation. Someone please correct me if I am wrong here, though.

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

ok, bump an old thread..

Q-tec is fine and alive.. PS has received a voltmeter+ammeter combination from dealextreme quite some time ago, it has an ability to sense up to 50amps
http://dx.com/p/mini-digital-blue-re...e-shunt-219592
due to being a cheapo meter it does not show any current before it is more than 0.5a for some reason. But that was a cheapo project so it will be just fine.

The fsp bastard is annoying me for is inability to regulate.. I guess i need to put more load on 5V rail. I have done some testing, and it seems that either the crappy p/s is just too stupid and you have to load the two rails by the same load or it has its problems (full of teapo caps, non of them seem to have any external damage though).. This p/s is for testing things only, nothing vital, so no i will not recap it before it is actually dead or semi-dead. It seems that there is no load resistor on 12V rail.

There is one on 5V rail and i can't access it at the moment as it is pita to remove the board after my modification. Same with caps. That is why i will only remove the board when i really need to. And resistor is hearshrinked so i cant see the value. I do not care too much about 5V rails regulation as i use 12Vrail the most, and sometimes it goes to less than 11V!!!

I have few favorite load resistors around: 96ohm which if connected to 5V rail does increase 12V rail with 0.2V, which is not enough. And I have two power resistors of 1.8ohm for bad ass testing )

If i measure the resistance of GND TO 5V it is 42 ohms. Probably most of it is a load resistor.

12V measures to 322ohms and that is probably intereferring with the fan. I still do not care about this one as i will load it when i use it.

Specs without any addition loads (fan spinning at barely 4volts, and original 5v resistor intact):
5V: 5.08V
12v: 11.88V
---
Loading 5V with 96ohm resistor
12V: 11.94V
5V: 5.06V
--
Loading 12V with 1.8ohm resistor, should be about 6 amps:
5V: 5.38V
12V: 10.8V
---
Loading 12V with 1.8ohms / circa 6amps and 5V with 96ohm
12V: 10.92V
5V: 5.33V
---
now the bad ass testing
Loading each rail with 1.8ohm resistors, so ~6a and ~2.4a each
5V: 5.07V
12V: 11.81V
--
After the last experiment, you can see that if i want to use 12V rail with loads up to 6amps whth tolerances of 0.2V i should constantly load 5V rail with 2.4A not gonna happen.

Is this behavior normal with cheap p/s (i know it is low end device with cheap components) but they did not skip many of them or i should check caps of some 12V rail?

thank you

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

added an inductor, increased the cap to 220uF

Answers to my questions:
a) It was ripple created by regulator. After regulator mod my atx started making weird noise at 14,5V-15,5V. After improving my regulator mod, now it is silent there.. there is a bit of buzzing sound on lowest voltage setting, but it was always there as it is normal with adjustable atx supplies.
b) still not sure 100%, but i understand they wouldn't be needed with linear
c) it is easy, same color markings as for resistors, and they are marked as L
d) Now using 220uF, not 100uF anymore. Hopefully this will be enough

I haven't heatsinked the regulator, it runs just warm, maybe about 45 deg. celsius, which is much cooler that load resistors. My regulator is salvaged from some board so it is in DDPAK/TO-263 package, this would make it tricky to sink. Will see how this works.

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

I am amazed that regulator works partially, it is a switching regulator, it NEEDS the inductor to function as intended. I'm surprised that capacitor hasn't blown up due to the ripple current on that - especially without the inductor!

Nice answer I haven't actually stress-tested the regulator yet, just checked it for few minutes

Okay, it seems like I have to redo my voltage control )

Now I would like to ask the following:
a) Which ripple you are referring to? Is it the one caused by ATX power supply or the one that is actually made by the regulator?
b) If I was using a linear regulator, this ripple filters would not be needed?
c) Can you guys help me finding a correct inductor? Manual suggests 100uH. Never worked with the before. I can see some of them are very resistor-like http://www.taydaelectronics.com/100uh-inductor.html How are they typically marked on the PCB's? I mean like resistors are marked as R1, R5, caps are C2, C3 etc. Or should my inductor be a coil-based one?
d) Do I have to use 1000uF cap as manual suggests? As mentioned I am only using 5% of the regulators max current capacity.

[ben7]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Solved fan overvoltaging problem by adding a 12V regulator before the fan.

As mentioned before my voltage from Aux supply was 10.21V idle and it was 8.something when fan was running. So now I still have some undercooling in theory; when my atx is set to lower value than 8.6V it spinning at 8.6V anyway, but i crank my atx up to more than that value it is going together with output until 12V, and as I go past this value it still stays the same, 12V.

As I did not have 12V regulator with me I was forced to use an adjustable one - LM2576-ADJ. This meant that I had to use a small prototyping board for a regulator, cap, 1 resistor, and one trimmer (to get exact value of voltage needed). I have actually set it to run at 12.30V, just to make sure there is enough air flow. My regulator had one pin for turning it off - this is how I solved another problem of fan spinning on standby without turning PS_ON switch. As I had my switch for grounding PS_on with 3 pins, I was able to connect ps_on cable to switch off pin of regulator when the switch is in OFF position. There is enough voltage in mentioned cable to turn regulator off.

A simple question now: https://cdn.badcaps-static.com/pdfs/...c50180d683.pdf suggests that I should use a 1000uF cap on the output. It would be too big size wise there therefore i have put only a 100uF cap. And i haven't put an inductor on output as manual suggests. Again, I am running the regulator at max. 0.15 amp while it is rated at 3amps.. And it works fine. How bad is the idea of using a cap that has 10 times less capacitance and not using the inductor? It is just for running a fan, i do not care about ripple and shit for it.. Should I?

That is my funny little prototyping board, and it got wrapped in isolating tape to avoid shorts as you can see in last picture.

I am amazed that regulator works partially, it is a switching regulator, it NEEDS the inductor to function as intended. I'm surprised that capacitor hasn't blown up due to the ripple current on that - especially without the inductor!

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Solved fan overvoltaging problem by adding a 12V regulator before the fan.

As mentioned before my voltage from Aux supply was 10.21V idle and it was 8.something when fan was running. So now I still have some undercooling in theory; when my atx is set to lower value than 8.6V it spinning at 8.6V anyway, but i crank my atx up to more than that value it is going together with output until 12V, and as I go past this value it still stays the same, 12V.

As I did not have 12V regulator with me I was forced to use an adjustable one - LM2576-ADJ. This meant that I had to use a small prototyping board for a regulator, cap, 1 resistor, and one trimmer (to get exact value of voltage needed). I have actually set it to run at 12.30V, just to make sure there is enough air flow. My regulator had one pin for turning it off - this is how I solved another problem of fan spinning on standby without turning PS_ON switch. As I had my switch for grounding PS_on with 3 pins, I was able to connect ps_on cable to switch off pin of regulator when the switch is in OFF position. There is enough voltage in mentioned cable to turn regulator off.

A simple question now: https://cdn.badcaps-static.com/pdfs/...c50180d683.pdf suggests that I should use a 1000uF cap on the output. It would be too big size wise there therefore i have put only a 100uF cap. And i haven't put an inductor on output as manual suggests. Again, I am running the regulator at max. 0.15 amp while it is rated at 3amps.. And it works fine. How bad is the idea of using a cap that has 10 times less capacitance and not using the inductor? It is just for running a fan, i do not care about ripple and shit for it.. Should I?

That is my funny little prototyping board, and it got wrapped in isolating tape to avoid shorts as you can see in last picture.

[momaka]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Regarding 3 5V supplies: i know their uses, but my question was why have an separate output of 10.21V that is stabilised to 5V anyway? While there is main standy output when line voltage is present as well, at the same time as the strange 10.21V output is active. Couldn't the ICs be supplied directly by 5Vsb line??

I can think of one reason only: isolation. If there was static discharge induced on the 5VSB line, and if the PWM controller is hooked to that line, then this could possibly kill it. Having it's own line that is only available inside the PSU and nowhere else outside makes it isolated from external electrical noise. Also, if the load on the 5VSB produces any kind of ripple, it will not get to the PWM controller, and the PSU will be able to continue to work in a normal manner.

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Yes, the fan is running all the time when the mains cable is plugged and mains swich is turned on, as posted previously. As i said it is not a big deal, but i might want to put a transistor would be activated by PS ON switch.

I know that pins 13-14-15 have +5V stabilised by 494, but my fan is 12V I will have to go other way around.

Regarding 3 5V supplies: i know their uses, but my question was why have an separate output of 10.21V that is stabilised to 5V anyway? While there is main standy output when line voltage is present as well, at the same time as the strange 10.21V output is active. Couldn't the ICs be supplied directly by 5Vsb line??

[kaboom]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

I will try the alcohol thingy, thank you.
Now some more questions:

As my +12V rail is variable, the fan is varying according to voltage. in theory ATX should be undercooled if I set less than 12V on the supply (its hardly spinning at 5V). And it is getting ultra fast at 15.5V (thats the limit I've set)
Fan is connected to +12V directly and GND is connected over transistor and thermistor measuring temp of secondary heatsink.

I was ok with this speed control system (only i am planning to move thermistor to primary heatsink, as that is only one getting hot at the moment), but variable +12V is too bad for fan. Therefore I decided to connect positive lead of fan to 5vSB transformer. It has two outputs : one is going to to-220 style rectifier (main standby source) and another is going to a diode and is giving 10.21V. That is where I have connected my fan. It has at least 10.21V while main output is at 4.66V to 10.21V which is not too far from 12V. After I crank the supply up to more than 10.21V, voltage at that point is going even higher as there is a diode between the mentioned output and 12V rail. The problem is that fan is getting whole 15.5V at max voltage setting. I might want to put a 12V regulator to avoid overvoltaging the fan.

Another issue is that when I turn the atx on but do not turn PS_ON (i have 2 switches) the fan is spinning on standby as well. I might put a mosfet there to control it, as well as I might not do it Not a big deal

The question follows: why the f foes the 5vSB transformer have this stange output of 10.21V?? i have traced it to TL494 to 12th pin where it gets unregulated Vcc and it outputs a clean 5V output on 13-14-15 pins. Why the hell there is a need of 2 5V sources on standby (1 main over 3 pin rectifier and another over the chip)? Anybody has an idea? It means that every TL494 based atx has 3 5V outputs: main high current, secondary SB, and another one over the chip? Why the third??

As I have attached some pictures of final result it looks a little bit messy because of amount of cables. PS_ON, 2 small load resistors, 3 wires for feedback circuit and so on. There is a protoptyping board with 3 components attached directly to the var. resistor (for feedback adjustments), under all the cables.

The green heatsink is where I have put 12V load resistor (remember - fan is loading that rail only over 10.21V setting), before that fan was as a minimal load, and PS wouldn't work without 5V load that was on the pcb. Therefore i heatsinked bot of them (i would hate 5V load resistor overheating), zip tied to heatsink with normal thermal paste (didn't have any glue-based) and superglued (i know, ghetto) to zip ties, as they were not stiff enough. Green heatsink (salvaged from chipset in motherboard) would match the secondary heatsinks cuts - therefore they pressed into each other and i have fastened them with 1 screw in lowest part of green heatsink.

Pictures are attached. Hope this is useful for anybody

You've got the fan tied into the AUX supply for the '494. It will run anytime line voltage is present.

Pins 13, 14, and 15 have +5 on them from an internal regulator in the '494. This is also supplied by the AUX voltage.

Just tie PS-ON to GND permanently, and use the line switch, if you don't like multiple switches.

As for the three supplies: Any ATX PS will have them, not just '494 based ones. You've got +5SB for the motherboard, another one 5-20V for the PWM IC, and related components, opamp if present, and the main 5V output.

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

I will try the alcohol thingy, thank you.
Now some more questions:

As my +12V rail is variable, the fan is varying according to voltage. in theory ATX should be undercooled if I set less than 12V on the supply (its hardly spinning at 5V). And it is getting ultra fast at 15.5V (thats the limit I've set)
Fan is connected to +12V directly and GND is connected over transistor and thermistor measuring temp of secondary heatsink.

I was ok with this speed control system (only i am planning to move thermistor to primary heatsink, as that is only one getting hot at the moment), but variable +12V is too bad for fan. Therefore I decided to connect positive lead of fan to 5vSB transformer. It has two outputs : one is going to to-220 style rectifier (main standby source) and another is going to a diode and is giving 10.21V. That is where I have connected my fan. It has at least 10.21V while main output is at 4.66V to 10.21V which is not too far from 12V. After I crank the supply up to more than 10.21V, voltage at that point is going even higher as there is a diode between the mentioned output and 12V rail. The problem is that fan is getting whole 15.5V at max voltage setting. I might want to put a 12V regulator to avoid overvoltaging the fan. Another issue is that when I turn the atx on but do not turn PS_ON (i have 2 switches) the fan is spinning on standby as well. I might put a mosfet there to control it, as well as I might not do it Not a big deal

The question follows: why the f foes the 5vSB transformer have this stange output of 10.21V?? i have traced it to TL494 to 12th pin where it gets unregulated Vcc and it outputs a clean 5V output on 13-14-15 pins. Why the hell there is a need of 2 5V sources on standby (1 main over 3 pin rectifier and another over the chip)? Anybody has an idea? It means that every TL494 based atx has 3 5V outputs: main high current, secondary SB, and another one over the chip? Why the third??

As I have attached some pictures of final result it looks a little bit messy because of amount of cables. PS_ON, 2 small load resistors, 3 wires for feedback circuit and so on. There is a protoptyping board with 3 components attached directly to the var. resistor (for feedback adjustments), under all the cables.

The green heatsink is where I have put 12V load resistor (remember - fan is loading that rail only over 10.21V setting), before that fan was as a minimal load, and PS wouldn't work without 5V load that was on the pcb. Therefore i heatsinked bot of them (i would hate 5V load resistor overheating), zip tied to heatsink with normal thermal paste (didn't have any glue-based) and superglued (i know, ghetto) to zip ties, as they were not stiff enough. Green heatsink (salvaged from chipset in motherboard) would match the secondary heatsinks cuts - therefore they pressed into each other and i have fastened them with 1 screw in lowest part of green heatsink.

Pictures are attached. Hope this is useful for anybody

[momaka]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Nice work!

Yeah, the Q-Tek is more or less a standard (or "classic", if you will) half-bridge power supply design so as long as you have some schematics of one of those half-bridge power supplies, it's only a matter of checking components until you find the bad one.

By the way, here's an interesting voltage mod Th3_uN1Qu3 made to an AT power supply and got very stable results (read post #5):
https://www.badcaps.net/forum/showthread.php?t=22129

How do you guys clean up the black residuals after soldering? I guess it is some soldering paste that stays there.

Yes, that's just left-over flux. You can clean it with a small piece of rolled paper towel dipped in alcohol. I usually do it with fine tweezers. It's not the fastest method, but it cleans the board very well in my experience.

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

solved by replacing diode D12 as for the first schematic.

[domas]

Re: Two dead ATX SMPS's - TL494 and FSP3528 based

Nice one again, i managed to fix Q-Tec PSU by the way momaka suggested... again Thanks man. There were two 1ohm/1.2ohm (can't remember) resistors cooked to well-done though they did not show signs of heat until i removed them from PCB. They became something like 1.3Mohm and 0.3Mohm I have reinstalled old transformer for mosfet control (identified it as defective, while is was not) and BINGO. Works.

The problem was that i did not need to remove any zener diodes. It still had the same problem as before, while adjusting from 4.8V to higher value it wuold make and annoying sizzling noise (actually sounded quite extreme) and would be even louder at 12V, and extremely loud over the mentioned voltage. Would still shut down mid 13 volt range, different every time. Later I noticed that if I hold GND cable in my fingers and/or touch the case of ATX the sizzling sound would be way lower and would let me go up to 16V. I understood that ground was missing. I have reinstalled (as i was working with pcb on the table before) the PCB to the case and bingo - it has a little ticking sound at low voltages, but that is normal. I have desoldered output caps for -12, -5, +5 and +3.3 volt rails as they would not be used. Caps voltages had enough headroom but I decided not to put stress on them (all of the rails would adjust together with +12V line accordingly, except for 3.3V rail), i wouldn't use the rails anyway. The supply was still working and adjusting nicely, but i had to keep a little load resistor on 5V rail.

But I have managed to kill the supply again.

HOW I have killed it: as my fan is running on 12V rail, the fan would adjust from the output voltage. I wanted it to be controlled by the thermistor as before. There is a low 10-12V output on STBY transormer, i decided to use it. I was checking the voltages on TL494 (Pin 12) where unregulated voltage comes to chip and chip regulates it to 5V on pins 13-14-15. It was working as it should. While testing voltages with my DMM that has thick leads, i have shorted TL494 12pin with 11pin OR the junction of 13-14-15 pins, cant remember. The supply have shut down.

Current situation: every time i turn on the supply it gives a pulse of voltages to the the rails, ant then shuts down. if I turn it off and turn on again it does not give a pulse again, as I have to discharge 5Vsb trail before turning the PSU on. I have tested every diode on the pcb, and most transistors. I think i will desolder q3 and q4 for proper testing tonight - as now I was just looking for shorts there. Have not tested negative voltages for the turn on pulse, but it seems that every rail is giving a significant voltage on power on, except for 3.3V, as it was 0.03V, but again this might be because of slow DMM I have. Does anybody have an idea where should i look now?

I do not have a schematic for exact ATX but those two are very close to what I have:





Thank you in advance..

P.S. I know my posts are a little bit messy