RE: Testing High voltage puck diodes.

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "Leigh Copp" <Leigh.Copp@xxxxxxxxxxx>

Hi Jim,

In the last 20 years I've heard a lot of talk about diodes (and other 
devices) failing intermittently. I have yet however, to ever see a 
semiconductor that was anything in between good, and bad. They either 
fail from too much voltage, or too much current, or too much heat. 
Really large devices in cyclic applications, expecially with water 
cooling, can also fail from excessive thermal cycling. (Hence the 
research into silicon carbide and similar substrates on large power 
devices, becasue it's coefficient of thermal expansion is closer to 
that of silicon).

I am open to the benefit of  experience from others, but I have spent 
literally hundreds of hours in the field investigating claims of 
power semiconductors that would intermittently fail, or only fail at 
high current or voltage. I can definitively say that I have yet to 
ever fail to find other problems that were directly responsible.

High voltage passives (especially capacitors) can start to break down 
at higher voltages, and can appear to be fine at lower votlages. RF 
tank caps for example are not worth testing without a 15 kV hi-pot 
set; a 1000V meggohmeter hardly tickles them.

Once semiconductors fail however, it is for good, and I have yet to 
see a failed semi that couldn't be verified with a DMM, although an 
analog multimeter is usually a little more foolproof (IE the 
venerable Simpson 260) due to its increased current sourcing ability.

I've heard that in theory the difusion currents within the device 
will turn semiconductors back to beach sand if we wait long enough, 
but that happens at about the same speed as your window panes get 
thicker at the bottom (becasue glass is a super cooled liquid, not a 
solid), and almost as slowly as your wife's sparkiling rocks will 
revert to charcoal briquettes.

The only thing that can go strange on you, as I mentioned above, is 
that the die attach on single side devices (studs for example) 
becomes flimsy due to thermal cycling, and the attendant mechanical 
fatique that it creates. This will cause the temperature profile to 
be less linear across the junction, and they will do some neat things 
when close to their limits.

So for my 2.5 cents worth check them with a multimeter, and then install them.

Cheers,

Leigh


	-----Original Message-----
	From: Tesla list [mailto:tesla@xxxxxxxxxx]
	Sent: Sat 07/10/2006 10:18 AM
	To: tesla@xxxxxxxxxx
	Cc:
	Subject: Testing High voltage puck diodes.
	
	

	Original poster: "Jim Mora" <jmora@xxxxxxxxxxx>
	
	Hello All,
	
	I am preparing to install 6 sets of six series connected HDB7.5 puck
	and stud diodes in a full wave 3 phase transformer oil tank (given
	there is room once the filament transformer is removed). These diodes
	are rated at 7500 volts reverse each, with a forward current of 1.4
	amps. The recovery time is "standard". I called the engineering
	department and they said they should be fine without any balancing
	components at 60hz.
	
	Keeping in mind that these are take out surplus, I would like to test
	them under somewhat hardened conditions, not just the forward
	junction drop. I was thinking of using half of one of my 15KV neon
	transformers across a diode test setup with say a 30 ma load of some
	sort. I have a 15kv dc probe around here somewhere which I will also
	test thoroughly!
	
	What would be the wisdom of noting the waveforms of each diode and
	discarding the ones that are inconsistent? Naturally I will test the
	diodes with a meter first. Anyone know a better or safer way to do this?
	
	Thanks in Advance,
	
	Jim Mora
	
	There are a lot of bolts holding on the transformer/PS cover! And I
	would not like playing with these pucks once submerged!