Re: Rectifier/NST Protection

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Jim Lux by way of Terry Fritz <twftesla-at-uswest-dot-net>" <jimlux-at-earthlink-dot-net>



Tesla list wrote:
> 
> Original poster: "Pete Komen by way of Terry Fritz <twftesla-at-uswest-dot-net>"
<pkomen-at-zianet-dot-com>
> 
> Josh,
> 
> According to my old ARRL manual, diodes in series need both a swamping
> resister and a bypass capacitor across each diode.  The resister balances
> voltages across the diode stack.  The capacitor protects against differing
> turn off times and passes HF spikes around the diodes.  If HF bypasses the
> diode, it must end up at the NST.


The operative word here is "OLD".... 
 
In fact, (although opinions vary) putting capacitors across the diodes may
actually make it worse.  Modern diodes, bought in a batch, will be very
closely matched (far more closely than any set of resistors or capacitors
you are likely to find...). If you were getting your diodes by unsoldering
them from scrap equipment, then the whole RC equalization network thing
would be worth considering. The 1995/6/7 Radio Amateur's Handbook explains
why rectifier 'equalization' is prone to cause premature rectifier
failure.[page 11-9, middle column, top]

Consider the capacitors, for a moment.  They form a voltage divider.
Typical ceramic caps have a tolerance of 20%.  I haven't actually measured
any, but the piece to piece variation might be a lot smaller.  However, if
you string them up in series, the voltages will divide inversely in
proportion to the capacitance, and hence, could vary up to 20%.  Not great
if you are running on the ragged edge for your diodes.  Same applies to the
resistors.

Another thing about capacitors is that they store energy, which, in general
with HV gear, is a "bad thing".  

A much better approach is just to overdesign the diode string.  Put a bunch
extra in series.  Say you need a 20 kV rectifier and you have 1000 PIV
diodes.  Use 30 diodes not 21...

Also, consider your application. if you are rectifying a sine wave, who
cares if they turn off at different speeds.  The voltage doesn't
instantaneously reverse on the string.  As the voltage crosses through
zero, the diodes will start to turn off.  If one turns off a bit faster
than all the rest, sure, it takes the entire voltage, but, the voltage
isn't coming up all that fast (in terms of the diode recovery time.. a few
nanoseconds), so by the time the voltage gets big enough to cause problems,
the rest of the diodes will also have recovered.  Commercial HV stacks are
made with soft recovery diodes by the way, just so that if one starts to
turn off early, it isn't as big a deal. (Note that if you have capacitors
in the stack, when that first one turns off, the capacitors across the
other caps actually slow their recovery down, aggravating the process). 

Another thing to consider is that modern diodes have fairly good avalanche
characteristics, so, if one DOES turn off fast, it just breaks down (like a
1000V zener), and conducts a bit of current while the other diodes recover.
 Sure, for that few nanoseconds, the power dissipation shoots up, but the
number of joules isn't all that big (unless you have all those bypass caps,
which is why stored energy is bad).  Likewise, if the stack is momentarily
overvoltaged, and one of the diodes goes into reverse breakdown and starts
to conduct, you don't want stored energy in the cap aggravating things.

A nice discussion is at: http://www.vcnet-dot-com/measures/D.amplifiers.3.html
> 
> I was pondering something similar using 1N4007s with a 3 megohm resister and
> a .01 microfarad ceramic capacitor in parallel with each diode.  3 megohms
> gives a current about 10x the leakage current. of the diode.
> 
> Pete Komen
>