Diode on light load keeps failing

This question arises from a repair work I'm doing on a friend's plasma TV - a Panasonic TH-P42X30T. The rectifier for the 5V standby supply failed short.

The rectifier is a B3100, a schottky diode rated at 3A & 100V. The service manual I downloaded omits a detailed schematic of the power supply board but at least the load currents of the various voltages are printed on the board. The load on the 5V supply is given as 0.2A.

I did some measurements after replacing the diode. The DC output ia about 7.7V, further regulated down to 5V. My scope shows the negative swing at the diode anode to be slightly less than 40V. The rectifier is used as a half-wave, so the peak reverse voltage on the diode should be around 45V.

At first, I dismissed the defect as one of those repair jobs in which it's difficult/impossible to pinpoint the exact cause of failure. Random part failure, unusual mains spikes, ESD, etc. But then, after poking around on the internet, I came across other cases where the same diode failed.

Now why should a diode keep failing when subjected to less than a tenth of its maximum current rating and half of its PIV rating?

Reply to
Pimpom
Loading thread data ...

The simplest design of that sort of power supply used the diode to charge u p a capacitor once per mains cycle. The capacitor starts off at the peak su pply voltage, and sags a bit until the next cycle. Most of the current that flows into the capacitor through the diode flows in a very short time arou nd the peak of the waveform, and the peak current could well exceed 3A.

formatting link

gives the non-repetitive peak forward surge current limit as 100A, but figu re 4 in the data sheet chops that back to 20A after 100 repetitions.

A ferrite bead in the right place might be all that it needs to keep it out of trouble.

--
Bill Sloman, Sydney
Reply to
bill.sloman

Is that a flyback supply? Usually there's a small inverter transformer, and some sort of mains-side controller, which can be as crude as a single transistor (blocking oscillator), or more often these days, a TOPSwitch or the like (3-7 pin power device or DIP or SOIC with integrated HV switch).

Peak current is limited by the switching device's current limit * turns ratio, and peak voltage is determined by the transformer's characteristics when the switching device turns on.

Apparently, the peak voltage is fine, although there is the matter: what bandwidth is your scope?

Diodes die in three ways, I think:

  1. Thermal. Overheat it and it melts, duh! Can be excessive RMS current, or reverse voltage. (Keep in mind, schottky are leaky enough to be prone to meltdown in reverse! Triggered by heat, however it gets there.)
  2. Voltage. Excess voltage is supposed to be handled by avalanche (well, in the types that are rated for it -- many carry a rating though, which is nice). But relying on a non-TVS component for TVS service is a bad idea (that is to say, the avalanche rating isn't much energy, 10s of mJ usually). Blow that out, and the device might not be (externally) hot, but it's got a pinhole short inside.
  3. Current. Peak current can damage diodes, I think by electromigration. Stay under the repetitive peak rating. You might think it can handle a surge of, say, 200A for 10ns, but it will actually die under such service, without ever getting warm.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

Touching it with a scope probe can muffle flyback peaks a lot. Meaning you might see only 45V but when the probe is off it may be much higher. Also, as Tim said the scope BW needs to be quite high with a small flyback (if the standby supply it is a flyback).

The toughest diode failure case I had was at a client where diodes in a pulse-echo multiplexer board failed in the field and only after a while. Even with the fastest scope we had available (300MHz BW) I saw nothing. Only a SPICE simulation showed peaks. Fixed those in the drive circuitry, problem gone. At first they didn't want to believe this was really the case until they realized that field failures dropped away.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

How about soldering a 10pf to ground to reduce the peak voltage?

Reply to
Steve Wilson

Even better, add a snubber?

Reply to
Steve Wilson

If the design is lacking that, sure. Or zener plus diode. Though it would really surprise me if that fell through the cracks on a Panasonic design. What sometimes happens (only seen it with other brands) is that a snubber resistor is dimensioned marginally and eventually fails open. Then the unit often still runs but occasionally blows the transistor or something else. And the replacement, and then the next replacement, until the root cause is found.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I've closed the TV for the time being after doing a band-aid fix to see if there's anything else wrong. I don't have any Schottky diode with a PIV rating higher than 40V or an ultrafast diode rated less than 1kV (=high Vf drop). I did the test with the UF.

I didn't examine the topology closely but it seemed to be a flyback. I remember that it's fairly complex with a single main switching transistor on a sizeable heatsink. The rather hefty transformer also feeds a 15V rail with a total load of >5A if I remember the print on the PCB correctly.

I used a small 10MHz single-channel analog scope so I could well have missed the spikes. I didn't note the frequency either but it was lower than I expected. Judging from the scope settings I used, I think it must have been around 15-20 kHz, with a lot of ringing. Could that have been affected by the probe too? I used a Tek x10 probe.

Reply to
Pimpom

That could very well be the case. Please see my reply to Tim too.

I've faced some tough challenges too over the years. Gives you a nice feeling when it's solved, doesn't it?

Reply to
Pimpom

What is the cause of the failure? Power dissipation, or something else?

Reply to
Steve Wilson

Overvoltage. Some semiconductors can take that for a while (avalanching etc.) and one day they die. Other times the demise is instant.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I meant the snubber resistor. Why does it fail open?

Reply to
Steve Wilson

UF types with their 50-100nsec trr should not suffer reverse recovery heating in a 15-20kHz switcher.

Are you sure? They let that big transformer run for standby? That must be an older TV, I don't think that would fly with today's laws.

Yikes. I wouldn't want to be a dog or a cat in your home. They can hear up there.

Ringing would only be affected if in the MHz range but I'd be more concerned about an initial spike. Hard to see with a 10MHz scope.

A rectifier diode failing with a short usually means that it got dinged from excessive voltage.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

It really looks like overvoltage.

The best so far was for a client who was really in dire straits because of such a problem. They were so thankful that they sent me a very fancy

2lbs chocolate bar for Christmas. Absolutely delicious. Causes other problems though in that it isn't good for my waist line.
--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

It burns out from running too hot for a long time. With through-hole that can usually be seen because it develops an "uneven tan" but with SMT they sometimes fail while still looking ok from the usual distance.

It is similar with those start resistors that drop rectified mains into a zener and cap which then starts a primary switcher IC. Lots of gear is thrown out because "it doesn't work" and often only a small 470k resistor failed open which makes the main switcher not start.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Doesn't anyone put their finger on things to see if they are too hot? How about a small FLIR to measure hotspots? Or an ir thermometer available on ebay for under $10. Or temperature indicating paint? How about measuring the power dissipation in LTspice before going into production? Or just guess and derate the component by a factor of ten.

It seems there are so many ways to prevent this kind of failure that it should never occur. Why does it still happen?

Reply to
Steve Wilson

penny pinching. Most failures wouldn't happen if the consumer didn't mind paying twice the price.

Reply to
tabbypurr

By the time you get the unit the resistor has already failed so it won't become hot at all anymore.

The usual, the attempt to squeeze some more milli-Yuans out of the BOM because the boss wasn't happy with the profit margin.

[...]
--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Sure

Ok...

Makes no sense to me. Ferrite beads usually will only be doing something in the ns digits. Main power supply surges usually at the us to ms digits

-- Kevin Aylward

formatting link
- SuperSpice
formatting link

Reply to
Kevin Aylward

They still would happen except that now the manufacturer would pocket

3-5x the profit margin.

Often you don't what you get pay for. Other times you get way more than what you pay for. An example are my bicycle lights. Not that flimsy yet expensive plastic "name brand" stuff I had on there before but nice efficient LED front lights in cast aluminum. Yes, Chinese no-name versions. The light is almost on par with a motorcycle front light except they consume only 8W. They have gone through >10k miles of rough turf, rain, mountain bike crashes, scorching sun and whatnot. All working just fine.

Another less powerful light I now use as a flashlight or if I have to rent a bike somewhere. It had to tough out an upside down landing of my mountain bike. The whole thing smacked onto this lamp, crashing hard after an "involuntary high speed dismount". The cast aluminum got slightly dented (something I thought wasn't possible) and scraped but the light works just as always. Thick glass lens, not one chip came off.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.