High voltage diode blowing mysteriously

[Joerg]

Not even in the note that says "5.6k + 47pF"???

Whoops, didn't see that, was looking in your text. And shouldn't it say
snubbered, snubberated or snubberificated?

Best to put them on the schematics. 47pF is almost nothing. Are you sure
this quenches ringing enough?

Voltage on both sides, yes. Current, no. I soppose recovery current
could be interesting; of course, that will be snubbed quite
effectively by leakage, so it will certainly be finite...

Hmm, I know the peak overshoot. I should be able to estimate peak
current from that, if I know LL and C. Meh, it'll probably be within
50% of "it all comes from the primary", which isn't accurate enough
to estimate recovery.

UF4007 is rated for 20-30A peak. ...

That's for non-repetitive.

... They aren't heating up very much,
at least until failure. Do you think it could be peak current?

Possibly, because at 75nsec trr these are fairly slow. Since you
probably don't want to spring for four of those Cree super thingamagics
which would totally raid the beer kitty, why the 2x10mH common mode
choke? Tried separate chokes yet?

Of course, if you came into a wad of cash:

http://www.cree.com/products/pdf/C2D05120.pdf

BTW, for better schematics drawings you can use the free edition of
Cadsoft Eagle, as long as it's hobby and not for profit use. Later when
you want to buy it for business it's rather inexpensive. Not like Orcad.

 

[legg]

It doesn't seem to be. The waveforms on either end look identical and inverted. The currents aren't aall that different. 30mA seems to be enough to cause continuous conduction mode, so it's not trying to play funny on that side.

Tim

If the currents are unbalanced, you may need freewheeling rectifiers
to ground, before the choke, on both sides.

RL

 

[Tim Williams]

If the currents are unbalanced, you may need freewheeling rectifiers
to ground, before the choke, on both sides.

Ah, interesting.

The dead time voltages are near zero (give or take a small amount of squiggle), so it seems to be okay.

Tim

 

[Tim Williams]

Best to put them on the schematics. 47pF is almost nothing. Are you sure
this quenches ringing enough?

Yes, it gets hardly a cycle before the ringing is too weak to see.

Possibly, because at 75nsec trr these are fairly slow.

Slow!? Come on, I know you love dabbling with MHz stuff, but gimme a break, this is only 120k after all!

Since you
probably don't want to spring for four of those Cree super thingamagics
which would totally raid the beer kitty, why the 2x10mH common mode
choke? Tried separate chokes yet?

Hmm, "common mode" is a funny way to put it. The negative rail is inverted, so I guess it should be "differential mode".

I used that since everyone else uses multiwinding chokes in forward converters. Two chokes would take up more space and be more trouble to wind; this is only one part.

I'm not buying them and I'm not mass-producing them, so don't bother going all Joerg on me. ;-)

Of course, if you came into a wad of cash:

http://www.cree.com/products/pdf/C2D05120.pdf

With 45pF at 200V (probably with high Q), I'll need even bigger snubbers. The transistors might not appreciate trying to push around the 455pF "not-recovery" capacitance, either. Actually, 28nC charge is like 0.5A * 50ns reverse recovery, so they'd be just as bad! The only difference is, you get the capacitance back each time, whereas the reverse recovery is heat each time.

Anyway, even peak Irr were 0.5A, that's roughly 75ns * 0.5A * 4 diodes * 120kHz * 2/cycle = 36mA effective reverse leakage. At 250V, that's 9W, or 2.25W/diode, but recovery occurs at a few volts, not all 250, so the real power is something else. Evidently it's fairly small, because the diodes don't seem to heat up until the one dies...

No one makes 1A SiC rectifiers? What's up with that?

Tim

 

[Copacetic]

Ah, interesting.

The dead time voltages are near zero (give or take a small amount of squiggle), so it seems to be okay.

Tim

Experiment with the transformer core gapping to massage dead time. That
is likely one of the elements you need to fix this issue.

We have gapped cores between 0.5 mil and as much as 50 mils for one
weird configuration.

Zero gap is bad, and we usually got prime results with between 0.5 mil
and 10 mils. Truly, if your transformer is not gapped, it should be,
even if only slightly.

 

[StickThatInYourPipeAndSmokeIt]

Nunya wrote:



============================================
But I say to you that everyone who is angry with his brother will be
liable to judgment; whoever insults his brother will be liable to the
council; and whoever says, ‘You fool!’ will be liable to the hell of fire.

Matthew 5:22

============================================



father mike

And I say to you, asshole, the last time you posted that, you also used
one of your rib poke nym morph moves. That makes you guilty of that
which you point your little netkkkop jerk wanna be finger at.

You ARE what you bleat.

You are more a troll than I ever could be... and you are pathetic as
well. And no, you are not "a father", mike. What you are is
SCAT-in-the-hat. That is what is between your ears, and that is what
makes up your tears.

The only thing you father are the little dingleberries you spew around
the world. You are SCAT-IN-THE-HAT.

Bwuahahahaha!

 

[StickThatInYourPipeAndSmokeIt]

But I say

Go away, shithead. We don't need any of the scat that oozes from under
your hat.

 

[StickThatInYourPipeAndSmokeIt]

But Archie, you LOVED inhaling my little anal 'love puffs' just
yesterday. Have you found someone else?

You are so retarded that you quote scripture in the same post that you
jack off at the mouth with this stupid horseshit in.

You're a real prize, you are... yeah... sure...

Better hope you never find yourself in one of my confessionals.

 

[Joerg]

Yes, it gets hardly a cycle before the ringing is too weak to see.


Slow!? Come on, I know you love dabbling with MHz stuff, but gimme a
break, this is only 120k after all!

75nsec still is 0.45% of each phase. Quite a lot.

Hmm, "common mode" is a funny way to put it. The negative rail is
inverted, so I guess it should be "differential mode".

I used that since everyone else uses multiwinding chokes in forward
converters. Two chokes would take up more space and be more trouble
to wind; this is only one part.

I'm not buying them and I'm not mass-producing them, so don't bother
going all Joerg on me. ;-)

Well, then wind another one

With 45pF at 200V (probably with high Q), I'll need even bigger
snubbers. The transistors might not appreciate trying to push around
the 455pF "not-recovery" capacitance, either. Actually, 28nC charge
is like 0.5A * 50ns reverse recovery, so they'd be just as bad! The
only difference is, you get the capacitance back each time, whereas
the reverse recovery is heat each time.

.... and they can be bolted to a heat sink. But you said yours don't get
hot so that's weird.

Anyway, even peak Irr were 0.5A, that's roughly 75ns * 0.5A * 4
diodes * 120kHz * 2/cycle = 36mA effective reverse leakage. At 250V,
that's 9W, or 2.25W/diode, but recovery occurs at a few volts, not
all 250, so the real power is something else. Evidently it's fairly
small, because the diodes don't seem to heat up until the one dies...

At 2.25W per diode these would get friggin' hot ... phssss ... OUCH!
What is more dangerous is if you had current spikes into the 2-digit
range which is only rated for non-repetitive.

No one makes 1A SiC rectifiers? What's up with that?

Not for those kinds of voltages, and SiC is still expensive.

 

[legg]

Ah, interesting.

The dead time voltages are near zero (give or take a small amount of squiggle), so it seems to be okay.

Tim

Coupled output inductors are useful, in reducing minimum load for
quasi-regulated outputs - your -250V and 6V3 filament supply.

There are leakage effects that have some regulation benefits, but will
almost guarantee conversion frequency doubled reverse bias spikes at
turn-off - so a kilovolt diode has no margin here.

The negative rail, being quasi-regulated, will have the higher voltage
when loading swings to the regulated rail.

RL

 

[JosephKK]

Well that's obviously wrong. The transformer only makes 300V peak output (600p-p, not counting LL), and it's obviously full wave, not half wave. The driver is clearly half bridge and the transformer ratio is clearly 1:4+4.
http://myweb.msoe.edu/williamstm/Images/Tubescope_Calc.png
These are my actual design constraints, actually, and the waveforms are very familiar indeed. Add some laggy parasitic C, and some springy LL, and you've got the real thing. The only difference between this circuit and the actual circuit is the center tap and split choke, which do not affect the simulation results, and serve only to split the resulting DC output in half, hence +/-250V.

Tim

And your line of thought goes to heck in a handbasket due to the inductor
input filter! Put some capacitance right at the output of the bridge,
470 pf ought to do. I bet it cools off your snubbers too.

 

[JosephKK]

And BOTH are incorrect designations. They can be used in item
descriptions (xfmr), but reference designators have an industry
standard
and your remark that they do not is noncorrect in all circles excpet
those
where some stupid dope like you refuses to use the industry
standard(s).
So, essentially you hang out in some stupid clics. Designators have
had variances based on device function. Diodes are a perfect example.
We see "D1" or "CR1", where "CR" was derived from "Cathode
Rectifier".

You have a bit of hoof in mouth i see. CR was once used for Crystal
Rectifier, a reflection of its solid state nature and variety of chemical
compositions. Ever heard of copper oxide or selenium rectifiers?

 

[JosephKK]

Ran out of room to draw them. The note just above indicates the nature of said snubbers.


"Turn on"? I draw these in Paint. Refdes are a hassle

Call 'em "high voltage diodes", and not "low voltage schottky", "EHV diodes" or "FWB". I'll know exactly what you're referring to, just as well as "D17-D20" or whatever.

Tim

I know you have heard to heard of LTSpice. Not too bad for schematic
capture alone.

 

[Neanderthal]

You have a bit of hoof in mouth i see. CR was once used for Crystal
Rectifier, a reflection of its solid state nature and variety of chemical
compositions. Ever heard of copper oxide or selenium rectifiers?

I owned a '66 Honda Dream in the early 70s.

http://www.vf750fd.com/blurbs/brochures/1959-1969/ca77305.jpg

The old bikes had Selenium rectifiers in 'em.

That was just the earliest exposure I had to them.

 

[Neanderthal]

I know you have heard to heard of LTSpice. Not too bad for schematic
capture alone.

Hard to model multiple output winding transformers though.

Even with the 'real pro' sim packages, it is not a simple chore.

 

[JosephKK]

You're an idiot. Usenet has always been that way. You not knowing
that after all these years, shows everyone just how little attention
you
pay to the details.

Nope None-yah. BTW I have been on Usenet since the time before "Big 8".
Word wrap has always been a client issue, for good reason. If your
client is too incompetent to handle long lines get a better client (or
better user wetware).

The "that" which "you take", Jan, is the fact that you are an utter
idiot
where knowing about Usenet is concerned. You also lack skill in
electronics. 

Oh Great Whizzard, post some of your electronics "skill".

 

[Tim Williams]

^^^^
Typo. Oh, and plonk, of course.

Tim

 

[Capt. Cave Man]

Nope None-yah. BTW I have been on Usenet since the time before "Big 8".
Word wrap has always been a client issue, for good reason. If your
client is too incompetent to handle long lines get a better client (or
better user wetware).

No, idiot, it has always been an nntp server issue, and the clients
were supposed to be written to comply. Any client that does not is at
fault, and it is on the authoring side where the line length limit gets
imposed.

The fact that you are oblivious to it is a huge tell about you.

The fact that you think it is a mere reader application issue is an
even bigger tell about your aptitude... err... extreme lack thereof.

 

[Neanderthal]

^^^^
Typo. Oh, and plonk, of course.

Tim

Not a typo, idiot.

"Gotta find a woman..."

And "Right on..."

are lines from a famous song. For you to be absolutely oblivious to it
tells me that you were just as blind back in the seventies as you are
now. Here's another... "Hot Pants!"

Learn how to set your line length properly and you MIGHT get back a
point or two of your pathetic IQ back. The thing is that it will take
years. It is like fat. It will take you just as many years to extricate
the stupidity you have accumulated over those years.

You are an idiot, right NOW.

IFYEMFY

(I Fixed Your E-Mail For You)

Bwuahahahahahahahahahahahaahaha!

 

[Joerg]

Hard to model multiple output winding transformers though.

Why? Never had a problem with that.

Even with the 'real pro' sim packages, it is not a simple chore.

I found LTSpice to be rather "pro"