[Amps] How to know when filter caps begin to fail

[Jim Thomson]

Date: Wed, 28 Mar 2012 22:17:07 -0600
From: "Jim Garland" <4cx250b at muohio.edu>
Subject: Re: [Amps] How to know when filter caps begin to fail
To: <amps at contesting.com>
Message-ID: <D5A8680215A44DC98F1462A3896EA0AC at Garland>
Content-Type: text/plain; charset="us-ascii"


My experience is that the only time electrolytic caps explode is either when
the polarity is backwards or when a rectifier diode fails and AC appears
across them. I've found other age-related failures to generally be more
benign. Generally they just lose their capacitance but don't start
conducting. Now tantalum caps are another story. They almost always fail by
short circuiting, sometimes catching fire.
73,
Jim W8ZR

##  I agree.   With really old caps, the C just slowly drops off.   On all new supplies, I measure the
ESR and also the D factor on em..b4 they get installed.   Then u have notes way down the road. 

##  if u look at both united chemi con and also  cornel dublier, EG eng notes, and formulae, etc
They will have specific examples of cap usage  for lytics.    Like caps for application XXX must be able
to handle 15 A  CCS of ripple current  20 hrs per day...and caps must last  200 K hrs. (or 300k etc). 

##  If u look at their leakage current  vs  max stamped voltage ratings, you will see that by operating the cap at less than
90%  of its ratings  has benefits....like leakage current drops.   At 75-80% of its ratings...leakage current is through the floor,
non issue.   You are just asking for trbl by operating  caps  at  95-100%  of their V ratings.    You don’t want to be operating
them at their max ripple current ratings either.  Ripple current is  typ 2.56  x dc plate current.   The air temp, surrounding the caps
should be the same as the room temp, or maybe slightly more.   A 100 k @ 3 watt  MOF is plenty for EQ.  The bleeder current
drawn will easily swamp out the leakage current by at least  a ratio of 15-20 or more. 

##  operate the caps at 75%  of their  V ratings,, and low temps inside the box where the caps reside,  100k eq resistors..of the 1% tol type ,
and  it will last a LONG time.   The  1% tol eq resistors I get from mouser are all exactly .2% on the low side.  They are all identical in value.
No need to match em up anymore.   The resulting vdc  across each cap is identical, within a few volts of each other. 
  High value caps  like 1000-10,000 uf caps  in 450 vdc  are the ones that have the   real high ripple current ratings.
My 2300uf and also 2500 uf caps both have 10A  CCS ripple current ratings.   The 3900 uf caps are even more.  My 10,000 uf ones are 30A CCS. 

##  If one cap in a string goes wide open, who cares.  It still has the  eq resistor across it.    Its not gonna blow up.   Now if any one cap  shorted, again,
nothing happens.   Since I run em at only 75%  of their rating... it’s a non issue.   Loads of headroom left.  The energy stored up in just one cap, operating
at 75% of its  stamped 450 vdc rating is not gonna be a grenade.   They blow up cuz of AC through em, or excess VDC across em..or  RVS polarity.  

##  here’s  how u get around the problem if  one leg of a rectifier fails shorted....say   one entire leg of a FWB.    Just install a  1N5408  or better yet,
the  bigger 6A10  ..and wire one across  each lytic in the string.   The single  safety diode  across each lytic is wired  RVS connected ! 
In normal operation, they sit there..and do nothing.     Now if  one leg of a FWB decides to short out, the normal deal is... you will be pumping raw AC
into the caps on half cycles.    When and if that ever happened,  the RVS connected safety diode across  each lytic will  conduct.. (all of em)..and  the 
diodes simply  short out  the  HV raw AC.   Fuse blows..end of event.   Nothing is damaged,  and you will never lose an electrolytic.   Diodes  like
the 6A10 are dirt cheap.  Rated for 1 kv piv  @ 6A CCS..and a  400A surge.  The actual  PIV rating on em is 1200-1400 V. 

##  for added safety, I install a  fast HV fuse between  on leg of sec  (either leg) and  input of FWB...like a Buss  HVU-3   Use the smallest size HV fuse
that will  just handle the normal average current.    The peak current through the FWB is sky high..and every 8.3 msecs.. but the average is way down. The
legs of the FWB  alternate, so you end up with a 50% duty cycle on the rectifier diodes anyway..even if using rtty-fm.  

##  I use a high power 50 ohm glitch R , typ a pair of 100-225 watt wire wounds, 100 ohm each, in parallel, to make the  50 ohm glitch R assy. In the
smaller amps, its usually just a single 50 ohm-50watt  WW.   Just b4 the  glitch R, another HV fuse is inserted.  ( 2nd hv fuse is the B+  fuse). 

##  any shorts, trbl in the RF deck, etc... and the  50 ohm glitch R will  LIMIT the  fault current to a safe value.   The HV  B+  fuse  will  INTERRUPT
the fault current..and do so in 2 msecs. 

##  OK, now if u had a short in the lytics..say B+ to  chassis  from dangling wires etc..... the B+  hv fuse will  do nothing.... its on the wrong side of the caps. 
in this case, the  1st hv fuse, located between  plate xfmr sec  and  FWB..will blow open asap, protecting your FWB.    I also install a  fast, magnetic hydraulic breaker
in the  240 vac primary, sized correctly of course.   I use the instant trip variety.  These types have no oil in the reservoir..and open off really fast.  
They are NOT heat activated  either. 

##  lastly,  I install    one B+ meter  directly between  the  B+  and the B-   of the string of caps.    The 2nd  B+ meter  gets wired in the RF deck, again directly between
the B+ and B-.     Now if the HV B+  fuse ever blows open.... the  HV meter in the  RF deck will read ZERO.    Meanwhile, the  HV meter in the  B+ supply will
STILL read 3kv   ( or whatever ur normal B+ is)       

##  Now if the AC  HV fuse ever opens up,  (located between xfmr and FWB), then  BOTH HV meters  will read ZERO.  

##  rauch claims   Ameritron uses....  “special lytics” .   Formed at  600 vdc... then tested at 525 vdc... and rated at  450 vdc.    Look up the procedure  used by CD
etc, and others..and it’s the EXACTLY the same.      BTW, that  525 vdc  “surge rating”   is only  for 2 secs or less.... like for  transients, spikes etc. 

##  The fault current is the same  with any  B+  to chassis  short..regardless  of the amount of C  used in the  supply.    Its just  B+  divided by glitch value. 
It does form a RC time constant though, but who cares.   The HV  fuse that precedes the  glitch R  will blow open extremely fast.   3 kv divided by say 25-50 ohm
glitch value  will result in  60-120A of fault current.    And 60-120 A  of fault current is gonna blow open a 1 A  rated HV fuse  lickety split....like in less than 2 msecs. 
Glitch stays intact every time.   Forgot one thing. I parallel a bunch of 6A10’s   to make one big diode.   then make a 2nd similar assy.  Then wire the 2 x assy’s   
between  B-  of  caps..and chassis.    Then the B-  can never wander more than  + or -   .7V    Stick some more across the plate and grid meters and its now  bomb  proof 

later...... Jim   VE7RF.