PSU Repair: A Case Study

[FritzEiv]

I used to do this for a living

Don't tell my boss, I've managed to convince him that I'm only an expert at running benchmarks and writing about the results

Too bad he reads all the comments

 

[Daniel Sauvageau]

It is important that the replacement capacitors are 105C type, as the normal 85C type capacitors would not last very long.

Temperature is only something like the 6th most important specification unless you actually intend to run capacitors near their maximum rated ripple current while also near their maximum rated temperature. Tons of flyback SMPS use output capacitors that have nowhere near adequate ripple current rating. This is less of an issue with polymer caps which typically have 2-3X as much current rating for a given capacity and voltage.

Is low ESR caps wasted on normal 60hz power filtering(diode rectifier after a low voltage AC transformer)?

You would ideally want reasonably low ESR to mitigate losses and capacitor self-heating across the operating frequency range. While the input might be mostly 60Hz and its first few harmonics, its load, which would be the primary side of the transformer and its FETs, typically operate at 40-500 kHz, more if you consider the harmonics.

Oh... wow.

I didn't realise the US used black for phase and white for neutral - I'd seen the wire colouring in the odd power cable etc. that comes with US gear

Whatever the wiring color might be though, the IEC plug and connector pinouts remain the same and people would normally expect them to be wired properly.

 

[shrapnel_indie]

Sometimes caps do have other tolerance ratings, but they are usually marked when they do.

Technically, all components have tolerances on all parameters worth listing. Capacitors' nominal capacitance just happens to be notorious for having some of the widest tolerances in all of electrical engineering.

In the case of the cap that still read 600µF on the multimeter, it turned out to have about 4 ohms of ESR in my step test when I revisited it. The reason it let so much noise through despite having a "good" capacitance is the ESR: the capacitor is no good at suppressing ripples at 10+ kHz if there is a 4 ohm ESR in front of it. If there is 1A of ripple current from the flyback, you get 4V worth of ripple instead of the expected 0.1V.

I was purely referencing capacitance when I initially referred to tolerances. I do agree there are tolerances with the other measurements too, but things like ESR and leakage currents are usually much tighter than the "acceptable" capacitor value range for a given capacitance.

 

[Someone Somewhere]

Oh... wow.

I didn't realise the US used black for phase and white for neutral - I'd seen the wire colouring in the odd power cable etc. that comes with US gear

Whatever the wiring color might be though, the IEC plug and connector pinouts remain the same and people would normally expect them to be wired properly.

It was more the comment about black being connected to the chassis on one side and white the other... here, black is always going to be at or near ground, no matter whether you're dealing with AC or DC.

 

[DaveLT]

I want to read up on the repair of the antec aria's PSU so bad. lol.

An EE here.

 

[Per Hansson]

The fact that the neutral and live wires where reversed is very bad, but worse is the fact that a double pole switch (DPST) is not used but a single pole switch (SPST ).
Because with the very common Schuko power outlets in Europe you can insert the cable either way therefore swapping the live and neutral!

As for the 5VSB circuit it seems to be the dangerous two transistor design, seeing how interested you are in electronics theory you might want to follow it up with replacing the whole 5VSB circuit with a much safer DM311 chip for example
http://www.badcaps.net/forum/showthread.php?t=15276

 

[shrapnel_indie]

The fact that the neutral and live wires where reversed is very bad, but worse is the fact that a double pole switch (DPST) is not used but a single pole switch (SPST ).
Because with the very common Schuko power outlets in Europe you can insert the cable either way therefore swapping the live and neutral!

As for the 5VSB circuit it seems to be the dangerous two transistor design, seeing how interested you are in electronics theory you might want to follow it up with replacing the whole 5VSB circuit with a much safer DM311 chip for example
http://www.badcaps.net/forum/showthread.php?t=15276

Using a SPST (single pole, single throw) switch will always be cheaper than using a DPST (double pole, single throw) switch.

How much is the DM311 and its support parts vs two transistors and their support parts? Manufacturers will always go for cheaper (even if it is just a few cents) to meet a price-point (and a good profit-margin)... even if that few cents could save them quite a bit in warranties. (Apparently the difference in cost for the volume of units sold isn't cheaper than what they'd pay in warranties.)

 

[MarkW]

I want to hear about any times you thought you had fixed a power supply and then had explosive results later... LOL

 

[Daniel Sauvageau]

Using a SPST (single pole, single throw) switch will always be cheaper than using a DPST (double pole, single throw) switch.

How much is the DM311 and its support parts vs two transistors and their support parts?

An SPST switch might be cheaper but if local regulations require DPST due to the non-polarized nature of their standard plugs, then that's the end of that argument. For devices intended for markets where polarized plugs are the norm, SPST is good enough when wired properly.

As for the DM311 and similar monolithic flyback controllers with integrated switches, they require fewer support components than a discrete oscillator, they require less board space, provide many extra features like soft-start and under/over-voltage lock-out, require less engineering effort to put in and likely provide much higher efficiency. You want to meet newer EnergyStar and other standards that progressively reduce allowable standby current draw? You pretty much have to use standby PWMICs. In bulk, the DM311 costs less than $0.60 while the Fairchild 5N60C FET in the discrete circuit already costs about $0.50 on its own.

Costs were likely quite different 10 years ago when standby PWMICs were still relatively new and inefficient but today, they are the only logical way to do it unless you have unusual design requirements - several significant advantages and effectively no inconvenients.

I want to hear about any times you thought you had fixed a power supply and then had explosive results later... LOL

I haven't had that "chance" yet. As I said in my introduction, I had never bothered repairing a PSU beyond refreshing the output caps before. Most PSUs I have seen that had issues worse than that had either already gone through the smoke signal phase or blew up while I was doing basic load/turn-on tests before popping it open.

Once magic smoke has come out, the repair is almost certainly well beyond economically viable.

I do repairs mostly as a hobby for family, friends, friends of friends/family, etc. I was working on some test jigs, needed a power supply for them, got fixated on using the SL300 for it (I could have pulled a random adapter from my spare parts bin instead) but I could not get it to power up at the time, so I decided to give repairing it another shot now that I had an oscilloscope on my bench and most of the rest is in this story.

There were a few points I wanted to add after sending the story in to editors but I was too late for inclusion, so I plan to spin those off as separate stories later.

 

[jimofnyc]

You need to buy these and keep them on the shelf: http://www.amazon.com/125pcs-Values-Electrolytic-Capacitors-Assortment/dp/B00CWQVXAG/ref=sr_1_7?ie=UTF8&qid=1434325606&sr=8-7&keywords=capacitor+kit

 

[Aris_Mp]

Important: You should replace caps ONLY with caps that have the same ESR! Manufacturers exploit ESR of the output electrolytic capacitors to ensure a fast load regulation.

In addition high ESR of elec caps can be actually a good thing in some cases. They can for example get rid the dangerous oscillations of ceramic caps, so we usually use high ESR caps in parallel with ceramic ones.

Finally, ESR is frequency dependant so we have to take into account the frequency that a caps operates at, in order to measure its real ESR.

 

[Crashman]

I want to hear about any times you thought you had fixed a power supply and then had explosive results later... LOL

None of mine exploded. Power supply repair was part of my overall earning strategy back when I was buying broken Abit motherboards, re-capping them, and selling them for as much as $100 each (used!) on eBay.

The store that was selling my refurbished PCs (separate operation, buy a truckload sell them one at a time) always had dozens of blown PowMax power supplies and piles of circuit boards with hundreds of higher-quality capacitors, so the owner just gave me the junk and paid me $20 a unit when I returned them.

 

[Daniel Sauvageau]

Important: You should replace caps ONLY with caps that have the same ESR! Manufacturers exploit ESR of the output electrolytic capacitors to ensure a fast load regulation.

Not quite. And especially not on a flyback PSU's output where noise is directly proportional to flyback current and ESR, so maximum allowable ripple voltage effectively dictates the maximum permissible ESR. If you have 1A of ripple current on a 100mOhm ESR cap, you get 100mV of ripple because of ESR. Try doing a fast regulator when your feedback loop is contaminated with ESR-induced switching noise, you have to use a feedback filter with lower bandwidth and this increases regulator lag.

As for ESR to prevent oscillation or ringing, that is mostly for power decoupling on high speed digital and analog PCBs and point-of-load regulators. Because bulk PSUs typically operate at frequencies lower than the caps' self-resonant frequency, ringing is rarely a problem there.

 

[Aris_Mp]

The flyback topology is mostly used for low-power applications. 99% of desktop and server PSUs nowadays use double-forward, full-bridge etc.

[strike]Also in your example with 1A ripple current and 100mOhm ESR you have according to Ohms law (I = V*R) -> V = I/R = 1A/0.1Ohm = 10 Vrms voltage ripple or 14.1422 Vpp (peak to peak) for a sine wave signal.[/strike]

In addition ceramic caps aren't used only in the EMI filter or at the start of the APFC converter but in other places inside a PSU as well (e.g. on the modular board in some).

Besides the theoretical part I am aware of several cases where capacitors were changed with lower ESR ones and the result was worse than before (and in some cases the PSU refused even to start).

Now for PSU switching frequencies this depends on the topology they use and there can be huge variations from PSU to PSU.

There quite many good PSU related books that refer to all above.

 

[Daniel Sauvageau]

Also in your example with 1A ripple current and 100mOhm ESR you have according to Ohms law (I = V*R) -> V = I/R = 1A/0.1Ohm = 10 Vrms voltage ripple or 14.1422 Vpp (peak to peak) for a sine wave signal.

You got that backwards. Ohm's law is V = R * I. If the ripple current is 1Arms, then the voltage ripple is 1A * 100mOhm = 100mVrms.

Also, the 1.41 crest factor is only valid for monotonic sine waves while current waveforms around an SMPS transformer tend to be more triangular or square. In a flyback PSU, you get a somewhat triangular charge current waveform and a square discharge. Can't use the 1.41 AC/RMS crest factor here.

There quite many good PSU related books that refer to all above.

And I have read through a few of them. I even own "Switching Power Supply Design, 2nd Edition" from McGraw Hill.

 

[Aris_Mp]

yeap stupid of me sorry. Writing a review and responding on the same time in technical matters isn't the best way to do it. It is indeed 1 A. I still cannot understand how I messed with Ohm's law

You keep mentioning flyback PSU. This isn't the case here unless you mean flyback otherwise. According to Maniktala in a pure flyback topology all the energy transferred from the input to the output must have been previously stored in the inductor(s),

In the main transformer indeed the signal isn't pure sine but still isn't perfect square or triangular. It depends on the switching speed of the main switchers.

Good reads on this subject.

1) Gordon McComb and Earl Boysen: Electronics for Dummies. Wiley Publishing, 2005 (ISBN 0-7645-7660-7)
2) Keith H. Billings: Switchmode Power Supply Handbook. McGraw-Hill, 1989 (ISBN 0-07-005330-8)
3) Marty Brown: Power Supply Cookbook. Newnes, 2001. (ISBN 0-7506-7329-X)
4) Marty Brown: Practical Switching Supply Design. Academic Press, 1990. (ISBN 0-12-137030-5)
5) Switch-Mode Power Supply Reference Manual, On Semiconductor, Rev.4, 2014
6) Sanjaya Maniktala: Switching Power Supplies from A to Z. Newnes, 2006. (ISBN 0-7506-7970-0)
7) Sanjaya Maniktala: Switching Power Supplies from A to Z Second Edition. Newnes, 2012. (ISBN 978-0-12-386533-5)


I have most of them in hard copies and the last one is just perfect. But it needs a deep electronics background in order to follow it.

 

[Daniel Sauvageau]

You keep mentioning flyback PSU. This isn't the case here unless you mean flyback otherwise. According to Maniktala in a pure flyback topology all the energy transferred from the input to the output must have been previously stored in the inductor(s),

In the main transformer indeed the signal isn't pure sine but still isn't perfect square or triangular. It depends on the switching speed of the main switchers.

I think you got flyback confused with forward voltage converters.

In a pure flyback SMPS, the main transformer IS the inductor: the primary side pumps the transformer core with a magnetic field and when the switch opens, the collapsing magnetic field forward-biases the diode on the secondary winding and energy from the magnetic field gets dumped there since the secondary winding provides a lower impedance path than the opened primary-side FET.

The flyback transformer's secondary inductance dumping current in the load/output caps can be simplified as an LR circuit since the capacitor voltage is relatively constant. R is a relatively small value since it is the cap's ESR, which is often simplified as zero. What is the current wave shape of an inductor dumping energy in a capacitor with negligible resistance in the loop? A triangle.

Actual waveforms will not be perfectly triangular due to wiring, diode, ESR, etc. losses but they are usually fairly close.

 

[Aris_Mp]

I know how flyback works and I try to tell you that this topology isn't common in today's PSUs where mostly double forward is used (along with half/full bridge and phase shift ZVT). Probably is the language barrier, I don't know.

http://img.techpowerup.org/150615/Capture617.jpg

In flyback case the current shape is indeed triangular however in forward and double forward (and on all the rest popular topologies used in modern PSUs) is more to a straight line than to a triangle form.

http://img.techpowerup.org/150615/Capture619.jpg

 

[Aris_Mp]

This whole discussion encouraged me to find the time and test my new differential probes.

http://img.techpowerup.org/150615/Capture620.jpg

http://img.techpowerup.org/150615/Capture622.jpg


Voltage before bridge: http://img.techpowerup.org/150615/Capture624.jpg
Voltage after bridge: http://img.techpowerup.org/150615/Capture623.jpg

Voltage waveform right after the main switches (with them operating in PWM mode since the PSU was at idle. At higher loads they switch to FM mode into this particular PSU): http://img.techpowerup.org/150615/Capture625.jpg

Voltage waveform at the entry of the secondary side: http://img.techpowerup.org/150615/Capture626.jpg

I am afraid was too tired (and luck the time) to connect the PSU to a load and check on the Ampere waveforms as well.

 

[Daniel Sauvageau]

In flyback case the current shape is indeed triangular however in forward and double forward (and on all the rest popular topologies used in modern PSUs) is more to a straight line than to a triangle form.

Yes, flyback has fallen out of favor for driving high power outputs. But what powers the PWM powering the main outputs? Most likely a flyback converter of some sort. The SL300 uses a flyback converter for the 5VSB output along with the auxiliary output that powers the main (also flyback) converter. My LG repair from a few months ago almost certainly had a flyback regulator for the 5V output which I bet also powered the resonant converter for the 12V and 24V outputs although I did not trace the PCB to find out on that one, though I probably should since in light of the SL300 repair, its auxiliary cap is most likely about to croak and I should keep this in mind if I re-open my older LG display too. I bet the majority of electronic power bricks and USB adapters use [strike]forward[/strike] flyback converters too.

On forward converters, the current waveform is only square when the power supply is operating at relatively high output current where the DC current through the inductor becomes relatively constant, getting switched between the free-wheeling diode and secondary output rectifiers. It only looks square because the triangular ripple current from the transformer applying voltage to the inductor (di = (Vtrans - Vout)/Lfilter, Vtrans, Vout and L are all constant, so di/dt = k, which integrated over time yields, drum roll, a triangle) becomes an increasingly small fraction of DC current. At lower output power, the di/dt becomes an increasingly large part of total current, making the waveform more triangular. At very low load, you end up in discontinuous mode where you have a sawtooth pulse with dead time in-between, where output voltage is maintained entirely by the filter cap(s) alone.

When designing or characterizing a switching power supply, you need to consider its three operating modes: discontinuous at low power, continuous mode at some point up the power ramp, and the transition range in-between.

 

[RedJaron]

I used to do this for a living

Don't tell my boss, I've managed to convince him that I'm only an expert at running benchmarks and writing about the results

Too bad he reads all the comments

Could be worse, Thomas. You could be like me and just have a BA in BS. Once Fritz learns all I really do is talk a good game, I'm gone. Maybe some politician wants to hire me.

 

[Crashman]

I used to do this for a living

Don't tell my boss, I've managed to convince him that I'm only an expert at running benchmarks and writing about the results

Too bad he reads all the comments

Could be worse, Thomas. You could be like me and just have a BA in BS. Once Fritz learns all I really do is talk a good game, I'm gone. Maybe some politician wants to hire me.

Funny you should say that, I have a BS in being a B.A.

 

[Aris_Mp]

most small power bricks are just linear PSUs since they are least expensive to make compared to SMPS.

For 5VSB indeed many PSUs utilize a flyback converter since the power needs of this rail are restricted. The Power Integrations TOPSwitch controllers are among the most advanced ones nowadays for the implementation of a 5VSB flyback circuit.

Finally, when designing a PSU you have to take first into account the level of the input voltage that represent the worst-case scenario. Because according to this voltage you will decide on the design of the magnetics which are probably the most crucial choice into a PSU.

However I believe that we went the conversation to another level and although I enjoy such talks I usually do this with examples, something that unfortunately I cannot do at this period due to work overload.

ps. Another major difference between flyback and forward converter is that at the latter the secondary part of the main transformer conducts at the same time with the primary.

 

[Daniel Sauvageau]

most small power bricks are just linear PSUs since they are least expensive to make compared to SMPS.

Tons of small adapters no longer use linear supplies anymore because copper has become far more expensive over the past 15 years, people are expecting much smaller wall warts from all the plug-sized USB adapters shipping with nearly everything these days, and many countries have adopted energy efficiency regulations that have effectively banned linear bricks. You can get 10W AC-to-USB adapters for $4, try doing that with an iron lump transformer, rectifier bridge, huge capacitors, linear regulator with a sufficiently large heatsink to stop it from frying itself.

The only device I remember buying in the past 10 years that came with a linear power brick is my ADSL2+ modem.

 

[nukemaster]

I have to agree that more and more devices come with switching power supplies.

My 2 last capture cards, cable box(its a smaller unit without PVR). I think my router is still a normal non regulated brick, my switch is a switcher new cordless phones are switchers too. Soo much better because of the lack of heat.

I still have TONS of normal unregulated power supplies that run fine, but they are not the most efficient things to use. I even have transformers from old alarm clocks and things, but I do not even think they are good for more than a few watts.