Creeping Sparks and Homemade Caps... (fwd)

---------- Forwarded message ----------
Date: Fri, 24 Apr 1998 15:43:13 -0400
From: Thomas McGahee <tom_mcgahee-at-sigmais-dot-com>
To: Tesla List <tesla-at-pupman-dot-com>
Cc: jimlux-at-earthlink-dot-net
Subject: Creeping Sparks and Homemade Caps...

> ---------- Forwarded message ----------
> Date: Fri, 24 Apr 1998 07:52:19 -0700
> From: Jim Lux <jimlux-at-earthlink-dot-net>
> To: Tesla List <tesla-at-pupman-dot-com>
> Subject: creeping sparks on surfaces
> Oddly enough, one can get really long sparks on a thin insulating film
> backed by a metal plate without much voltage.Bazelyan and Raizer refer to
> this as a "creeping discharge". They have a photograph and data on a 1
> meter long discharge on a 4mm thick piece of glass on a grounded surface at
> a voltage of only 100 kV. They attribute this to the dramatically lower
> energy requirements to propagate the spark compared to air.
> In view of this, when fabricating your capacitors, you should make sure
> that you don't have large areas of plate on one side of the dielectric
> without a matching plate on the other.

When building any kind of homemade capacitor we have to be *very* aware
of this tendency of sparks to creep along the surface. Many a 
homemade cap has died an early death because the extent of this creepage
was underestimated. 

A few months ago I reported to the Tesla List on experiments that I did
with flat plate capacitors. For those that have forgotten or never saw
the original posts, let me simply say that I found that upon 
disassembling GOOD working flat plate caps I found that the 
greatest corona damage was found at the ends where the aluminum
plates came out. This in spite of the fact that the distance
allowed at these ends was more than double the distance allowed
around the side edges. 

A few of my observations:

1) Greatest corona occured all along the ends where the plates exited.
   A) Large, LOW impedance end connections show more corona stress
      than thin, tab-type end connections
   B) Thinner end connections reduce the corona but adversly
      affect low impedance pulse characteristics of the cap.
   C) When tab-type connections are used, then it is imperative 
      that ALL sides with such tabs have a side clearance 
      *at least* DOUBLE the distance afforded the regular sides.

2) Sharp edges increased likliehood of corona and creepage.
   A) Curving the corners reduced corona stress.
   B) Burrs on aluminum flashing are particularly bad.
   C) Do NOT leave aluminum foil "ragged cut". Trim with scissors!
   D) It is preferable to use scissors to trim foil to size.
      Just folding the aluminum foil over is bad because the
      foil is now uneven. Uneven places and wrinkles give
      air bubbles a place to hide!!!

3) 1" free along long edge was barely adequate for 9KV NST useage.
   The spark creepage effect comes into play here.

4) 2" free along short edge was barely adequate for 9KV NST useage.
   The spark creepage effect is QUITE apparant here!!! My testing
   is leading me to believe that the sides where the plates or
   tabs exit probably needs about THREE times the length
   required for the regular sides, because the underlying plate
   literally attracts and pulls the corona along the dielectric
   surface. I believe THIS to be one of the major failure
   mechanisms for homemade caps. To make bullet-proof caps,
   increase the amount of free poly where the metal exits.

5) Pinholing through dielectric was greatly reduced when multiple
   dielectric sheets were used instead of a single sheet.

6) Corona damage was reduced when multiple layers were used, and
   this was reduced even further when "floating" sheets of
   aluminum foil intervened between dielectric sheets.

7) Caps run "dry" experienced very severe corona damage at all
   places where air is present.

8) Oil was excellent at reducing much of the corona formation. Air
   bubbles should be vigorously removed.

9) Adding Kraft paper greatly aided in getting bubbles out.
   There is a slight reduction in capacitance.

10) Kraft paper exhibited some discoloration with time, and helped
    me see clearly the pattern of corona stress. The color
    changes reflected the stress on the poly itself, but the 
   slight color changes made the corona stress easier to see.

11) Assembling a flat plate cap by first "painting" the components
    with oil greatly aided in producing a bubble free capacitor.
    This "painting" with oil also had a neat side benefit in that
    when building up a "set" of thin poly sheets to make one
    "thicker" sheet, the oiled sheets stick together and actually
    handle more easily. Besides painting with oil, you can also
    dip or soak component parts of the cap in oil prior to 
    assembly. I paint poly and plates and paper.

12) Subsequent compression of the assembly using a compression
    frame increased capacitance a bit less than 10%. I decided
    to try compression frames for several reasons.
    A) They keep the components together.
    B) They help exclude air.
    C) I had hoped to come up with some form of dry cap that
       would survive. To this end I oiled up all the parts
       and assembled, then tightened up the frame. This
       squished most of the oil out. After a few weeks of just
       sitting around, I noticed that air had begun to get
       between the edges of the poly. So I decided against
       using the thing dry, and immersed the assembly in oil.
       After de-compressing and re-compressing, this unit
       has been used only under oil, with excellent results

13) Polyethelene in oil will sometimes swell. This will usually
    make the capacitance drop slightly, as the plate separation
    increases. In my compression frame setup capacitance is held
    pretty steady... but I am not yet certain what kind of
    pressure increase is involved with the poly/oil, as I have
    not yet disassembled a compression frame cap after *extended*

Experiments continue, and if and when I have any other info I will
post it to the Tesla List.

Hope this helps.
Fr. Tom McGahee