torturing caps

We tested too many silicon power diodes for use as Grehkov-type drift step-recovery diodes, DSRDs, and we generally didn't find any systematic difference between, say, 400 and 800 volt diodes in the same family.

We did conclude that, if you find one that works, buy the rest of the reel. Mouser helped us there, letting us buy a few but helping us get the rest of that same reel if they worked.

John

We think that there are usually two parts, low voltage and high voltage, with the cut at 200 or 400 volts maybe.

John

On a sunny day (Wed, 15 Jun 2011 05:32:50 -0700 (PDT)) it happened Tabby wrote in :

I dunno, they should not get hot because of the leakage, but si leakage is very low, so I would expect voltage breakdown first.

On a sunny day (Wed, 15 Jun 2011 10:08:11 -0600) it happened m II wrote in :

There is sequence thermal issues, like cooling and heating, that could perhaps shorten life of a diode. I am a bit rusty on the semi-conductor internal stuff, but if the reverse voltage is bigger than some barrier it should start conducting (like a zener). Of course if it does that and the current increases because of low circuit impedance, then it will be destroyed. Maybe there is an other mechanism too, but then I forgot about it.

On a sunny day (Wed, 15 Jun 2011 10:24:46 -0700) it happened John Larkin wrote in :

Yes possible.

On a sunny day (Tue, 14 Jun 2011 18:53:30 -0700) it happened Jamie wrote in :

I think for things like 1N?00X where X stands for voltage, they just make one diode at say the max voltage, and then print something for the X. Much cheaper than manufacturing different ones for each voltage. So then it is safe to use a 100 V part at 600 V :-)

----------------

You do realize that they cannot test every diode(much more expensive considering the rate they can pump these out)? When they test, say, 100 and

90 come back at 600V and 10 come back at 100V do they compromise and say it's about 500V? If they did they may get a lot of angry customers or a bad rep!

Remember that these are generally destructive tests so even if you could test them all you wouldn't want to. It's all fairly basic statistics...

hou=3D

Certainly, for silicon. Reverse leakage is harmless if its kept to an extremely low level, but I cant remember just how low it needs tobe. IIRC microamp level, but memory's not really there.

Taking a guess here, I'm wondering if the reverse bbreakdown occurs in one tiny spot only, leaving the thing able to only handle a tiny amount of power in this mode.

NT

Hi,

I think the ceramic capacitors capacitance drop with increasing voltage is due to the structure of the barium titanate dielectric:

It looks like that central titanium atom (orange) will get pushed by the electric field, and at a certain electric field intensity will be holding less potential energy as it cannot be pushed against the rest of the atoms anymore. That would explain the diminishing returns on capacitance vs voltage for ceramic caps. I guess there are higher performance ceramic dielectrics with atoms with more repulsive forces than titanium that can withstand higher electric fields and still store potential energy.

cheers, Jamie

I'm certain that every production diode is tested. AQL would be the pits if they weren't.

A good test machine could probably do 20 a second, as fast as a handler could move them. Testing forward drop and reverse breakdown isn't destructive.

John

On a sunny day (Wed, 15 Jun 2011 17:52:48 -0500) it happened "Stretto" wrote in :

I fail to see your point. I think they just make 600 V diodes, and label them like 100, 400, 600 or whatever. No testing anywhere. It is probably HARD to make a bad diode with their equipment.

whatever.

Every diode is tested. Do you realize they test (and _TRIM_) every resistor, and resistors cost roughly 1/10 of what a diode costs?

On a sunny day (Thu, 16 Jun 2011 07:01:10 -0400) it happened Spehro Pefhany wrote in :

whatever.

I once ordered 1000 or so BC109 from Philips (at that time). Once of those had an empty case.... Maybe it was tested, but for sure nothing was then done with the test.

I also had a resistors from Voti that had a different value from the color indication. A few in a strip. Forget about that testing. Maybe for mil-spec parts, and speed binning and such. Testing costs money.

At least you ended up well adjusted.

Every diode is tested. Do you realize they test (and _TRIM_) every resistor, and resistors cost roughly 1/10 of what a diode costs?

---------

riighht...

There is something called laser trimming for resistors and even then it is expensive.

Do you realize to run a test on a diode it takes power? Power costs money. If your pumping out 10's of millions of diodes you really think that management is not going to figure out how to cut by only testing a few? You know what six sigma is? Didn't think so...

The production line generally is ran with extremely tight tolerances so all the diodes will work in the first place and if there is some mechanical issue that might produce a run of bad components it will be picked up in the statistical tests.

Lets say it takes you 1 second to make 1 component. 1M components would take

11 years to complete. If you want to get any reasonable production time and costs saving you can't test every component. This is why they have such high production quality standards so they can do these sorts of things.

In any case this is a moot discussion because any destructive tests would be absolute useless to do. Of course this doesn't mean you can't extrapolate from other tests but this is just as useless if you are doing proper statistical analysis.

Now obviously they do have medical and military grades that may test every component but you do end up paying for that. These components are also not produced in the same quantities as commercial grades.

Sounds like he just does not have the tolerance for resistor work.

On a sunny day (Wed, 15 Jun 2011 17:52:48 -0500) it happened "Stretto" wrote in :

I fail to see your point. I think they just make 600 V diodes, and label them like 100, 400, 600 or whatever. No testing anywhere. It is probably HARD to make a bad diode with their equipment.

---------

Again, they do test some of the components. They have to. If the machinery failed at some point it could produce a string of bad components. With proper statistical analysis this could easily be checked and even fixed quickly(simply stop the production line to fix the issue then start it back up).

The main point here is that even though many of the components may work at much higher ratings doesn't mean they know that and even if they did it is only statistical knowledge which doesn't mean that every component will work like that.

For example, suppose you have 100 "identical diodes" from the same production line. The breakdown voltage is rated at 100V. You destructively test them all and all but 1 fails at a much higher voltage. OTOH, suppose you didn't test them all but just tested a few and they all had higher ratings. You assume you can use them at the higher rating.

Now suppose you use them in a product. Lets suppose 1 diode per product. Suppose the diodes costs 1c each. The product itself costs 100$. Because you are using them at a higher rating by your assumption you will have one failed product. Since that failed product costs 100$ and uses a 1c diode is it worth it? maybe, maybe not...

The point is that the manufacturer does the statistics on it and gives you the result. You can live or die by it and it's your choice. I just hope your not making medical or military equipment...

What do you do when Lady Gaga makes a pass at you?

Resistor, of course! ;-)

Cheers! Rich