[Prev][Next][Index][Thread]

Re: Diode Strings



Tesla List wrote:
> 
> Original Poster: David Trimmell <davidt-at-pond-dot-net>
> 
> Hello,
> 
> I am looking for opinions and suggestions. For my tube coil work I am in
> need of some robust, high voltage diodes. Recently in some tube coil
> experiments I have been sending back decent surges into my doubler diodes.
> Needless to say they have not fared to well. So, I am going to order some 6
> amp 1kV diodes from Digi-Key.

Having destroyed a lot of Microwave oven diodes at $15 a pop before
figuring out what was wrong, I suggest that you look at your circuit and
determine if there is anyway that you could be running excessive forward
current through the diodes when you have an underdamped load that rings.
I was charging big discharge caps with a bridge and a NST. After
discharging the cap in a bang, I noticed that the diodes were failing
(Just a few at a time in the encapsulated stack, so the PIV rating was
essentially dropping). I later figured out that when the cap was
discharging, it was underdamped, and the reverse voltage forward biased
the bridge, running many amps through my little 0.1A rectifiers. A
series resistor fixed this problem.

In any case, I have two rectifier solutions: 1- Call K2AW Silicon Alley.
He has inexpensive 1 Amp 14 KV blocks (about $15, as I recall). These
can be substantially overloaded if you run them in oil.
2) Get the 6 kV 200mA (50 cents each, cheaper in quantities of 10 or
100) diodes from All Electronics (www.allcorp-dot-com) and put them in
series with no snubbers, etc. Run them in oil (the package is small
enough that it will arc over the body in air. 

In general, the HV rectifiers are made from series strings of lower
voltage units (typically 3-6 kV each). If the diodes come from the same
lot, they will be sufficiently well matched that you don't need RC
snubbers. In fact, the RC snubber can actually cause problems as the
diodes turn off and the C supplies a low impedance source of energy very
close to the diode.  And, of course, what sort of resistor were you
planning on using? A standard 1/4 watt resistor is only rated at 250V. 
The resistors will also draw significant power.

Modern HV diodes are very consistent and very fast. The RC equalizing
chains illustrated in things like the ARRL handbook are really hold
overs from the days of poor consistency and junk box parts of unknown
provenance.

The idea behind the parallel C was that if one diode in the chain turned
off faster than the others, the inverse voltage across it would rise too
quickly and cause it to break down.

As a practical matter, the scatter in reverse recovery time will be
nanoseconds with modern HV diodes. Also, what happens if a diode does
turn off faster and  breaks down in reverse (something they are actually
designed to do, as long as the power dissipated isn't too high, it is
fine, that's what a zener diode is used for).  The entire current is
running through the reverse recovery of all the other diodes in the
string. Eventually, they will also turn off, so you essentially get a
little reverse current spike (and a little power dissipation spike) on
that one diode. Take some typical numbers.  Assume that the diode turns
off 1 uSec early (orders of magnitude larger than what is likely).
Assume further that you have 1 Amp to push through the string in reverse
(also unlikely, but....) You'll be dissipating a kilowatt or so for that
microsecond (3kV reverse voltage drop in avalanche * 1 Amp), but it will
only last a microsecond, so the actual joules is about 3 millijoules. If
this happens every cycle at 60 Hz, the total power dissipation is 180
mW, well within the dissipation capacity of the device potted or
immersed in oil.  And, of course, the actual reverse current flow will
be smaller (the reverse impedance is quite high) and much, much shorter
in duration.

Recall also, that in most power supply applications, the applied voltage
doesn't instantaneously go from one polarity to the other. It is a
sinusoid, so the reverse voltage gradually increases from zero.  And, if
you are feeding a capacitor input filter under light load (low ripple),
the forward to reverse transition is even slower.

You might be interested to know that they don't use snubbers inside the
microwave oven stacks, or in TV HV rectifiers.

I also recommend the Phillips Semiconductor Data books. Phillips makes a
whole raft of HV diodes just for such uses and has ap notes on making HV
rectifier stacks for things like X-ray power supplies.


References: