[Amps] More Capacitors vs LC filters

[Jim Thomson]

Date: Tue, 17 Apr 2012 11:24:52 -0600
From: John Lyles <jtml at losalamos.com>
Subject: [Amps] More Capacitors vs LC filters

Even 2 kJ is dangerous. The large GE capacitors that we use at work,
run about 4 kJ each in operation. In 2001, we began to lose them, 
explosively. Each time, the metal can peeled open like a can opener did 
it - scary. Like others said, must prepare for the eventuality that a 
bug, fly, mouse decides to bridge across a capacitor terminal, to 
discharge that energy in one place, without a glitch resistor to protect 
it. All of our high power RF amplifiers use purely Capacitive power 
filters, on order of 100 - 250 uF, as they are pulsed systems and 
require stored energy to supply the pulse, i.e., pure economics. LC 
would have the ringing problem at each pulse, just like CW will do. It 
can be designed to move the ringing away from the excitation pulse rate, 
if careful.

Stored energy makes me wary every time I have to safe a cap bank and go 
inside. My present project is using 88 kJ of stored energy for 150 amps 
of plate current. I have only witnessed an accidental discharge once, 
through some resistance, thank goodness. I was seeing a white spot for 
an hour and everyone in the building came to see what the dynamite was 
about.

But on the other hand, in broadcast transmitter with tubes, class C FM, 
i did not waste time in deciding to use an LC filter (in 1980s) which is 
much safer. With constant load, it was a perfect match. And I did 
resonant choke design also. It worked very well. But not something you'd 
want with a pulsed load.

There are reasons for both approaches, so we don't need to beat each 
other up over it.
73
John
K5PRO

###  The one thing I did not get around too was to measure the ESR of my 32 uf and 47uf  OIL caps.  
I have since sold that 6700 vdc  supply.  It used  4 x 47uf oil caps in series–parallel..+ a 10 kva
pole pig.  In box #2, I installed  20 x 2000uf  @ 450 vdc caps.  Box #2 is completely sealed, nothing can get
into it.  Box #1, with the oil caps + xfmr  has a 234 cfm papst fan on the front of it, sucking air out.  I did have one
big bang event with the 4 x oil caps. They are 4.5 kv rated, and with 2 in series, plus a V divider, each cap gets 
3350 vdc.  The problem was, at the hot end, the 6700 vdc was too much  for the series 4.5 kv rated caps..and it blew out
below the  top insulator.  That threaded rod that comes up out of the oil and though the ceramic stand off had arced sideways
to the well grounded metal casing of the oil cap !  

## The fix was easy.  Just make damn sure the metal cases on the oil caps are NOT grounded.   I ended up placing 
all 4 x oil caps on a very large sheet of  .25 inch  thick micarta.   And then placed  .125 inch thick sheets of micarta
BETWEEN each oil cap.  The 1 x 1 angle aluminum that holds the caps front and rear was replaced with  1 x 1
micarta angle stock.   I treat oil caps like lytics, insulate from each other and from the chassis.  Once you have the
metal casings of the oil caps completely floating, you have now relieved the stress points.   Any bug or dust between B+  and
metal casing will not cause an explosion.  

##  On my buddy’s  24 x 3900uf cap setup,  one of the caps launched itself.  He had
used 2 x layers  of plywood, with holes in the 2nd sheet to hold the caps.  Huge crack in the wood below
one of the caps towards the hot end of the string.  The plywood was mounted to the steel frame of the rack.
I would have used a big sheet of micarta or uhmw myself. 
The fix in his case was to mount the large sheets of plywood on 4 x big glastic stand offs.   That was
the end of the big bang problem.    It was just a poor mounting design.  Sure, a big bang, but not like dynamite or anything.

##  I measured the ESR  of these various big lytics and using a B+K  875B.. they are typ .032 ohm  and as low as .017 ohm. 
Once you have 8-24 of em in series, the total ESR  goes up..like to  .4  to .7 ohm.  Add in the Z of the various HV wiring,and
at least you have a bit of Z  to limit fault current before the glitch R.   I highly suspect the ESR of a typ 50 uf oil cap
is virtually zero. 

##  In my latest creation, filters are in place on the intake side of the 4 x 120mm fans.  The 4 x fans blow air into the cabinet,
pressurizing it with clean air.  Screened louvers is the exhaust for the cab.  No bugs or dust can get in there.   
Each lytic is well spaced  from any adjacent lytics.  All interconnected wiring is teflon coated.  Large stand offs used
everywhere.   A 3rd HV fuse feeds the bank of series caps too.  The caps are never run at more than 71% of their V rating. 

##  The short of it is...as long as proper spacing is used, correct materials used, etc,  you wont have any problems with 
either oil caps or lytics.    My 100K @ 3 watt mof’s that are placed across each lytic are within a few ohms of each other.
Operating the caps at 70-75%  of their V rating ensures leakage is next to nil. End result is V drop across each cap is identical.
The ripple current rating is sky high on big caps, like 10-25 amps.  They just wont heat up with typ 1-3A of plate current.  
Temp inside the cab where the caps reside is the same as the ambient temp of the room.  
Insides of cabinets stay clean, dirt-dust-bugs cant get inside em when they are pressurized with clean,filtered air.  

###  with 7700 vdc no load and  24 x 3900uf caps in series, total energy is  4817 joules.  4817 joules divided by 24 caps =
200 joules PER cap.    There is one heck of a big difference between a lytic with 200 joules stored in it ..vs  the 4000 joules
stored in one of John’s big oil caps.   The ESR  on the big 4000J oil cap is  zilch  vs the individual lytic.   The massive 
explosions  just don’t happen with these high C series strings of lytics.  You still have to be careful though. 

later... Jim  VE7RF