SMPS Repair: Why does the smoke always escape hours later.

[tango17]

What am I doing wrong?  I've recently repaired some switch mode power supplies. The repairs have failed, sometimes spectacularly with loud bangs and smoke.  The most recent two were a car battery charger (failed under max rated load) and an LED TV CCFL driver (failed under normal operation 24 hours later).

Typically, I find the main mosfets shorted and sometimes the driver transistors and some diodes are also bad.  My repair procedure involves checking those components and verifying that all the nearby resistors are good.  I check and confirm that the transformer windings on the primary and secondary sides have reasonable continuity and resistance. 

I make sure I use heat sink compound.  Sometimes I replace the controller chip if it's not too expensive, but usually not.  Where possible I check any output transistors on the controller chip.  Usually I check optocouplers and other feedback circuit components.  Then I current limit the AC input (the old series light bulb trick) and use a variac to slowly ramp up voltage.  I check output voltages and current draw, then put it back into service.

Everything works great for a few hours or days, then BANG! The main switching mosfets blow again.  Often the damage from the second failure is worse than the original damage.

I try to replace all active components, from mosfets to power diodes to driver transistors, even when they test good, especially if its companion failed.

Why is this happening?  Is there some step I'm missing?  Why does it fail hours or days later?  Before closing up the repair I try to make sure the high power components aren't getting particularly hot.  Should I be putting the controller IC output signals on the scope to check them after the repair?

I've had my share of failed repairs during the initial testing with current limit and variac controlled voltage, but once it works OK on the bench and in service for a while, is it unreasonable to expect it to continue working?

If you see any problems with my procedures or can suggest something better, I'd appreciate your comments.  I buy my components from Digikey, so they should be good.  I replace the components with the same ones.

It's very frustrating and your comments would be much appreciated.  Thanks!

[bill_c]

Not sure if it is the problem, but you didn't mention capacitors at all.

[wraper]

Probably you did not replace some faulty part, as faulty MOSFET can be a consequence of initial problem, or may damage something else when fails.

[bdunham7]

Do you generally have schematics for the things you are repairing?  Sometimes a little shotgunning actually helps.

[Bud]

Not a power electronics guy here. But i would suspect the transistors get killed by high shoot-thru current, which may be caused by poor design not being able to maintain sufficient dead time for the output stage excitation signals. Say, close to the rated power the duty cycle may increase and cause an overlap in the gate control switching , which will have the opposite stage transistors open at the same time and as the result kill them.
You mentioned a ccfl TV inverter, but LED TVs may have dynamic background lighting, not constant one, which also may cause change of conditions on the controller chip and at some unique combination of the drive go wanky.
All in all i would think it may be difficult to measure dead time, unless you have equipment and experience, and given tbe issue is intermittent.

[tautech]

Not sure if it is the problem, but you didn't mention capacitors at all.

This ^. So so often it's the caps. 
You need be able to accurately measure caps and their ESR. Knowing what type of cap is used and where and why is critical to understanding the stresses on caps and why/when low ESR or standard caps are used.
Be prepared to pull caps and measure out of circuit.

Downloading and studying typical application circuit examples in the SMPS controller datasheets can bring about a new level of SMPS understanding of the subcircuits that make a SMPS work.
So so often actual SMPS designs are a nearly exact copy of the typical application circuit in datasheets. 

Be methodical and you will win.

[tango17]

Here's the schematic for the ccfl.  Q1815 and D1810 blew.  I replaced them and their opposites Q1816, D1811.  Everything else seemed fine.  I ordered identical P11NK50ZFPs but the YG963S6R diodes weren't available.  Since the diodes weren't available, I used some APT15DQ60KG diodes.  The specs seemed close enough other than the non isolated TO220, so I used an insulating mount for it.  It ran for about 36 hours before exploding.

In the explosion I lost the same components above as well as almost everything nearby - BAV70 diodes, ZD1804 zener 15v, the 2N7002 Q1813 and even the resistors.  Everthing on the bottom half near Q1816 tested good as did everything on the primary side - pins 2, 5 of T1803.

There aren't many caps other than the 1kv C1897 and C1839.  Would you have tried to test or replace IC1801 the LX1692 (obsolete) or caps near it?  Since the 5 transistors on the primary side were OK, I didn't go any farther, like testing caps or resistors on that side.

The datasheet for the LX1692 didn't show a circuit all that similar to my eyes, but I did look at it.

Feel free to comment on this specific repair (help much appreciated), but its really the procedure and approach to this type of repair I'm asking about.

I'm about to order from Digikey again and thought I'd buy the STTH15R06FP diode as a replacement for the YG963S6R s
(found in a xref substitute table) or the BYC15X-600PQ which looked OK.  The STTHs had only 2 left in stock at Digikey, which concerned me.

This time I'm replacing everything on the secondary (both of the switches Q1815-16 and all suport components) except for the high wattage resistor R1822 and the two caps C2897 C1839 and a few minor resistors that test good.  I measured capacitance and ESR on both those caps and they're both good.  Would you replace anyway?

Would you try to buy an ebay replacement for the controller LX1692 just to be cautious?  Would you look more closely at caps and resistors on the primary side?

Thanks for suggestions.

[tautech]

Oh, quite complex ! 
LV controller feeding drivers for T1803 with dual secondaries providing isolation from the HV real switching.
Zero current sensing ! 
From the initial blow up I might suspect some snubber not doing its job or insufficient/faulty mains filtering let a spike through.
500V rated devices right ? On a ~400V rail uprate them to maybe 800V.

Some might advise a full rebuild of the HV section and maybe that should be considered.

[tango17]

Thanks for the comments so far.

Remembering that I'm trying to get tips and tricks to improve my success rate in  switch mode repair.  I posted the ccfl circuit just to have a specific example to discuss, so let me comment and respond:

CAPACITORS:  My experience has been that electrolytic caps are the cause of general circuit failure more than half the time.  However, in the case of switch mode failure, that hasn't been true.  I always check any electrolytics for bulging and then in circuit esr or shorting, but other than a few tantalum caps, I seldom find other types of caps with any problems.  So in smps repair I seldom find caps that need replacement, unless I find a problematic electrolytic visually or with in circuit testing.  Do you agree?

UPRATE VOLTAGE RATINGS:  These are typically failures of devices that have worked for years, then my repair fails within days.  Possibly increasing the voltage rating would help prevent the once a year spike failure.  But since it fails almost immediately isn't it likely that there's another damaged component causing that second failure?  I'm anyways concerned about changing or substituting out an original component.  Do you often upgrade components when doing an smps repair or use original spec parts?

REPLACING TESTING CONTROL CHIP: The second failure always seems to involve a direct short through the switch FETs.  Like the dead time wasn't long enough and both switches were on at the same time to short the rail and let out smoke.  Do smps control chips fail that way? Do you usually test the control chip? If so, how?  Do you replace it as a matter of course?  Or do you think it's usually another component causing my problems?

Thanks for suggestions!

[bdunham7]

Thanks for the comments so far.

Remembering that I'm trying to get tips and tricks to improve my success rate in  switch mode repair.  I posted the ccfl circuit just to have a specific example to discuss, so let me comment and respond:

CAPACITORS:  My experience has been that electrolytic caps are the cause of general circuit failure more than half the time.  However, in the case of switch mode failure, that hasn't been true.  I always check any electrolytics for bulging and then in circuit esr or shorting, but other than a few tantalum caps, I seldom find other types of caps with any problems.  So in smps repair I seldom find caps that need replacement, unless I find a problematic electrolytic visually or with in circuit testing.  Do you agree?

UPRATE VOLTAGE RATINGS:  These are typically failures of devices that have worked for years, then my repair fails within days.  Possibly increasing the voltage rating would help prevent the once a year spike failure.  But since it fails almost immediately isn't it likely that there's another damaged component causing that second failure?  I'm anyways concerned about changing or substituting out an original component.  Do you often upgrade components when doing an smps repair or use original spec parts?

REPLACING TESTING CONTROL CHIP: The second failure always seems to involve a direct short through the switch FETs.  Like the dead time wasn't long enough and both switches were on at the same time to short the rail and let out smoke.  Do smps control chips fail that way? Do you usually test the control chip? If so, how?  Do you replace it as a matter of course?  Or do you think it's usually another component causing my problems?

Thanks for suggestions!

Just so you can weigh the importance of my experience, I don't repair that many SMPSs, but I do see some, and what I do see is quite a variety.  One a month or so, perhaps.  I recently repaired the SMPS that powers the monitor that I'm looking at right now.  So take that into consideration as I answer your questions.

CAPACITORS:  They do fail, and when they do it is usually a clear cut case.  They don't necessarily blow up, but they will have greatly reduced capacitance in most cases of failure.  Having an ESR that is 20% higher or 50uA of leakage is generally not the issue.  I find blown/defective caps in the majority of SMPS or on-board VRM problems.  I don't find that many bad caps in areas other than SMPS, the output caps seem to be overstressed by design in many cases.

UPRATE VOLTAGE RATINGS:  You're right, this probably won't help your short-term batting average.  Common reasons to uprate are 1) to get a new cap that fits properly in the old space in older products when caps of a particular rating were larger  2) to get lower ESR/higher ripple ratings and 3) where the voltage ratings are right on the edge.  In some cases, a higher voltage margin will increase the life of the caps as well, but this typically applies to caps rated for 100 volts and higher--apparently going from a 50V to a 63V rating in a 35V circuit doesn't help much.  Or so I've read.

REPLACING/TESTING CONTROL CHIP:  I have ordered these several times thinking they might have been blown up with the rest of the circuit, but I have yet to actually need one.  I personally doubt that control ICs are a failure point, but keep in mind my limited experience.  I don't know if you have a systemic problem, such as not using appropriate replacements for your power semiconductors, a diagnostic problem where you aren't getting to the root of the failure, or if you have just had a string of bad luck. 

Since I don't do enough of this to stock parts, I often order extras if the parts are cheap enough--that way if it blows up, I can try again.  On your CCFL inverter example, you're going to have to do some carefull investigation with a scope--and that might be tricky on that circuit.  I'm not even sure where I'd start.  Perhaps a transformer has a problem or your replacement transistors aren't fast enough or .   Is anything getting hot?

[tautech]

Thanks for the comments so far.

Remembering that I'm trying to get tips and tricks to improve my success rate in  switch mode repair.  I posted the ccfl circuit just to have a specific example to discuss, so let me comment and respond:

CAPACITORS:  My experience has been that electrolytic caps are the cause of general circuit failure more than half the time.  However, in the case of switch mode failure, that hasn't been true.  I always check any electrolytics for bulging and then in circuit esr or shorting, but other than a few tantalum caps, I seldom find other types of caps with any problems.  So in smps repair I seldom find caps that need replacement, unless I find a problematic electrolytic visually or with in circuit testing.  Do you agree?

No.

Yes commonly there are visual signs that caps are failing but this is often not the case. Degrading output caps not only result in high output ripple as if their leakage is severe they add additional current load to the supply and unless there is current sensing can overload the supply possibly resulting in catastrophic failure. With a SMPS like the topic of this thread the lack of current sense would concern me and I believe this unit and its later burn in failure is a good example of needing an infrared camera to assist finding problem areas before failure.
Back now to SMPS caps I have seen fail:
Primary side bulk cap: uncommon, normally no problem.
SMPS controller IC VDD bulk storage caps: common. These typically are ~20-50 uF and not low ESR. They primarily store and smooth the flyback supply to the IC. They drift low in value and go high ESR where they fail to do their job properly and prevent the SMPS IC from starting.
Output caps: Just nominal value checks generally disclose them failing however ESR checks are also required to confirm they are sound.

UPRATE VOLTAGE RATINGS:  These are typically failures of devices that have worked for years, then my repair fails within days.  Possibly increasing the voltage rating would help prevent the once a year spike failure.  But since it fails almost immediately isn't it likely that there's another damaged component causing that second failure?  I'm anyways concerned about changing or substituting out an original component.  Do you often upgrade components when doing an smps repair or use original spec parts?

Typically matched to OEM spec however with MOSFETs real care is required to do this well however modern FETS can offer lower gate capacitance and therefore reduce risk of Miller effect. Google it.
In general no substitution is done without careful comparison between OEM datasheet and substitution part datasheet. 

REPLACING TESTING CONTROL CHIP: The second failure always seems to involve a direct short through the switch FETs.  Like the dead time wasn't long enough and both switches were on at the same time to short the rail and let out smoke.  Do smps control chips fail that way? Do you usually test the control chip? If so, how?  Do you replace it as a matter of course?  Or do you think it's usually another component causing my problems?

Checking IC operation is not difficult but you do need the datasheet for UVLO and VDD specs.
All you need is a PSU with sufficient output to exceed UVLO and a scope to check the output drive to the MOSFET and this is done WITHOUT mains power to the SMPS.
After catastrophic failures there is no telling what damage spikes may have done to the IC so it is best to replace it.

As the SMPS in this thread is working at LV and its output is via a pulse transformer to the HV section it is safer to measure than a SMPS IC operating in the HV section where great care need be taken with all measurements.
Still, we must always identify if circuit Gnd is mains Gnd or an elevated value if we are to use a scope anywhere on a SMPS and it's safer to have a differential probe for SMPS work.

As I mentioned earlier in the thread IC datasheets and 'typical application' circuits are of great assistance to understanding SMPS.

[YetAnotherTechie]

ccfl drivers also fail because of bad lamps....

[tango17]

ccfl drivers also fail because of bad lamps

Hmmm!  This is something I had not sufficiently considered.  After years of operation perhaps the lamps are bad.  Can anyone suggest how to test this?  Can the lamps be tested directly?  Or can I test the drive current or other parameter once I've made the repair?

As to testing caps, I'll be more rigorous, but are non electrolytics a common failure point, or just the electrolytics?

As to testing the smps controller chip, I've got a cheap digital scope and an old analog 20MHz, but no differential probes.  I just use an isolation transformer. My bench PSs are limited and my signal generator is laughable.  I'm just not sure of the proper way to test a switching control chip before repair is complete.  After its running I can test it, but I'm usually so happy something works after a repair I just put the cover back on and put it into service.... then BANG!

Does anyone want to suggest how they would test the LX1692?

I'd really like to test the bulbs, too but, again, I'm not sure where to start.

I really appreciate the helpful comments here.  Thanks!

[tango17]

I thought I would update on my recent repairs and what I ascribe my problems to.

Starting with the ccfl circuit diagram posted above, I'm placing the blame for the first failed repair on myself for probably missing a damaged component.  It worked for an hour or so after thefirst repair, then failed.  On the second repair attempt I was more rigorous about checking resistances, capacitance, esr, etc.  I replaced with identical components everything that seemed suspicious, and it has been working fine for the last week.  Lesson learned.

On my failed repair of the battery charger, I place the blame on a bad component, the main switching mosfet.  I used a replacement I bought from Amazon. It worked for a short time, but failed under 100% load test.  It is now working correctly with reliable components from Digikey.  Lesson 2 learned.

[WattsThat]

With mosfets and igbt’s, gate drive is everything. Anything flaky in a gate driver circuit, like marginal caps, can show up later rather than sooner. If you don’t know what the gate drive signal should look like, you should always check the temperature rise of any replaced device(s) after running for say 15 minutes, then perhaps an hour later.

[aqibi2000]

What is an accurate way to test capacitors in circuit for leakage?

Electrolytic and ceramic ones