Re: RCR filter mods

Hi Scott,

	I'll post this to the whole list too...

At 11:24 AM 09/17/1999 -0700, you wrote:
>Hi Terry -
>Fellow coiler Ross Overstreet and I have been using identical "Fritz" RCR
>filters for protecting our NST-powered 6" coils; Ross is using a bank of
>12/60 transformers, I'm using a bank of 15/60's.
>Our RCR filter consists of (2) big 5.6k Ohm wirewound resistors, a 500 pF
>cap, and safety gaps in each HV supply leg tron the NST's. We quickly found
>that the resistors were dissipating a lot of current, and obviously this
>loss was affecting coil performance. Ross recently posted the dramatic
>results of completely eliminating the filter: an additional 20 inches of
>arc length, to 66 inches.

The post by Ross was definitely a wake up call about filter design.  Let's
consider some things.

When this filter was originally posted by me, I was not considering that it
may be used with more than one NST.  Also, it was designed to give very
good filtering without too much thought given to how it would hurt
performance.  The fact that the resistors get hot (sometimes VERY hot) was
an indication that there may be excessive losses.  Also, if you double the
current through the filter (by using two NSTs), the losses quadruple
creating far greater streamer loss.


Since then, we have learned much more about all this and I think we can
re-address the subject and "update" the filter in light of the latest
information and modeling results.

The capacitors that the filters use, should be in the range of 1nF give or
take 0.5nF.  If they are higher, there is a chance of resonance
over-voltaging the transformers (you have seen this effect as you mention
later).  With the caps in the 1nF range, the two resistors between the NST
and the filter caps can be removed.  They only serve to stop resonance
between the NST and the filter caps.  There will be a little resonant rise
if the circuit goes open load but the safety gaps will stop it and it will
not be very high anyway.  So, removing those resistors cuts the power lost
about in half.  If the filter caps are much lower in value, they start to
loose effectiveness.  So the only question is what value of resistor do we

5K of resistance with 60ma dissipates 36 watts in two resistors (one in
each leg).  At 120mA, the loss jumps to 144 watts.  so the resistance needs
to be adjusted in proportion to the current from the NSTs.  We know too
large of value will definitely just waste energy as heat in the resistors.
We also know that there are some of us that don't use filters and the NSTs
do not instantly fail :-O.  One can run models all day long and calculate
turn to turn voltages in the NST windings given various transient pulses,
but such calculations would mostly be guesses at this point and would vary
considerably from system to system.  So what guide should we use??

I think you and Ross have found it already!  Spark length.  Too much
resistance and the spark length goes down too far.  To little and the risk
of "zero spark" (dead NST) gets higher...  So it becomes a game of how much
spark length are you willing to give up to protect your NSTs.  Obviously,
reducing the length from 60 to 40 inches is not acceptable no matter how
much safety it gives.  But a 2 inch reduction from 60 inches probably is
easily acceptable if it adds any protection at all.  So we have the numbers
and equations we need to play our game...

The power dissipated in the resistors is: 

P = 2 x I^2 x R

I is the current from the NSTs and R is the resistance.  Since there are
two legs we multiply be two.  Theoretically, we also must add the power
expended in draining the filter caps at every break which is:

P = 2 x 0.5 x C x V^2 x BPS

However, for 1nF at 21000 volts and 120BPS that is 53 watts.  Pretty darn
high!  So for a 60mA 15kV NST with 5k resistors we have a total loss of 36
+ 53 or 89 watts (125 in the full RCR filter).  Note that the power used to
discharge the filter caps is far higher than the simple current heating
from the NSTs.  This dissipation is independent of the resistance value
(for those that don't follow this argument... just trust me ;-)).  

If we charge a 27nF primary cap to 21kV and fire 120 times per second, the
input power of the coil is 714 watts.  So the loss in the filter is
significant.  Of course, the two NST case can be far worse.

There is something that two (or more ;-)) NSTs also have.  Lower output
impedance.  Thus, two NST are far less likely to be damaged from transients
than one NST because they are harder to "push around" (pigs don't push at
all!).  So lowering the resistance value is acceptable since two NSTs don't
need as much filtering as one NST (NST = Neon Sign Transformer ;-)).  

Ok lets minimize the losses....  53 watts is too darn high!  lets make the
filter caps 500pF and take a "little risk".  That will drop the discharge
loss to 26 watts which is about 4% of the total system power.  Since
streamer length is proportional to the square root of power, we loose about
2% of our streamer length or about 1 inch in our 60 inch streamer coil.
About right IMHO, but we don't want to loose much more (note that this 26
watts is constant regardless of how many NSTs or charging current is used).
 Now lets blow another... say 15 watts to the filter's dual RC network
resistors.  For 60mA, that is about 2k ohms.  For 120mA, 500 ohms....  So
we have 500pF and 2k ohms which gives a simple cutoff frequency of 160kHz.
The 1nF 5k filter was 32kHz.  That "may" sound like we have lost
substantial protection to get back 50% of our filter losses.  However, are
we blocking the fundamental frequency, or those nasty 4MHz, 36MHz, 120MHz,
+++MHz  signals?  If we are concerned about the high frequency stuff, a
160kHz filter is just as good as a 32kHz filter!  In fact, from that
perspective we could probably raise the cutoff (and lower the losses) much
further!  However, if the fundamental frequency is the killer, we may have
just shot our neon's dead!  The old filter lost 125 watts and delivered
714.  That may drop the streamer length 5 inches.  The "new" filter looses
41 watts which may be around 2.5 inches.  Note that this may affect
streamer length in many complex ways aside from my simple estimate here....

Soooooo....  You see how the game is played.  Quite honestly, I didn't want
anyone blaming "ME" for killing their neons.  So, the original filter is
pretty bullet proof and will stop anything.  Admittedly, that protection
comes at the high cost of streamer length.

So there is the question of "how much high frequency can a neon take?"
Lots of fundamental but no very high frequency?, Not even much
fundamental?, everything but very high frequencies....  We really don't
know.  The "data" is further complicated by the fact that my factory new
NST is much more capable of taking abuse than a 20 year old unit exposed to
the -30 to +150 temperatures of bone dry Colorado for 20 years....

I would guess (a big guess too) that a new factory fresh neon will take
practically anything but very powerful high frequencies.  An old beat to
death unit is lucky to hit 21000 kV at 60Hz at all....  Anything else will
easily push it over the edge.

To make a long story short.  If you want practically any NST to last
forever, use a high loss filter like to one I posted.  If your "feeling
lucky", use something less....  Please post what happens!  The only real
way to know for sure is just to see what people's experiences are.  I
notice that far fewer NSTs die now than a few years ago since all this
filter talk has come up.  The question is, "Did we go too far?"

Above all --->> ALWAYS USE SAFETY GAPS!!! <<---  I single high voltage
spike will instantly destroy any neon.  There is zero doubt about the
theory behind safety gaps and their circuit loss is zero.  There is no
reason not to use them!!!!

Of course, the RCR and RC filters are effective.  but, perhaps we need to
use our tricks to come up with a really well designed filter using Ls and
Cs only that would not have power dissipation.  the designs of the past
have has some problems but we are supposed to know what we are doing now ;-)

I am a bit surprised so many people use that filter...  After going through
all this, I now see that having a nice schematic is far simpler to
reproduce than thinking about all this ;-)

Sooooo.  What will "terry" do?  Well, I know the filter my LTR coil has now
wastes about 90 watts to heat.  If I could add that 90 watts to the ~800
watts of available power to my coil now (a not so simple redesign in an LTR
coil), I may get 900 watts in.  Assuming it all made in proportionally to
the streamer length, I could get 6% greater streamer length or about 3
inches.  However, the thought of packing my coil in the car, drive awhile,
setting it up, letting it run full blast, and then watching the neon go
bang after 5 seconds....  I will not change a thing...  Since I tend to run
full blast in high effort situations were "bad" failures are really "bad",
I will not take the risk for that three inches...  I can just crank the
variac up a bit past were I "should" and get that length back in a far
safer way.  I am a bit concerned that Ross found such a great difference
with the filter removed.  I suspect that is due to tuning or RF loss issues
in the resistors but that is still an unknown...

I guess these were the "hints" your were looking for.  I hope this gives an
idea what goes into filter design and the many things that can be
investigated still.  It is a deceptively complex subject that does not lend
itself well to the "trial and error" design approach.  Unfortunately, it is
not really easy to do "scientifically" either other than the bullet proof
overkill way I used originally.  However, do get rid of the two NST to
filter cap resistors.  They appear to be useless...



>I'm a little more conservative, and want to retain some vestige of
>protection for my precious transformers. I have obtained additional 100+
>watt wirewound resistors in the values of 2k, 1k, 500 and 300 Ohms. Would
>you consider substituting some of these values in your filter simulation,
>and commenting on the resulting filter performance?
>Also, I made an interesting observation  while trying to determine the
>cause of a slight performance loss in my coil. While checking the output
>voltage of each of my 15/60 NST's, I found that one was around 15 kV as
>expected, but that the other was around 18.5 kV, both at the same input
>voltage. I then realized that the higher-output transformer was still
>connected to the RCR filter board, and the additional 3.5 kV was apparently
>the result of resonance with the 500 pF ceramic bypass caps in each output
>leg (caps grounded to secondary coil earth ground). Disconnecting the
>filter board dropped the output back to 15 kV.
>Any hints on determining the optimum value for these bypass caps, balancing
>RF bypassing capability for transformer protection vs current losses
>through the cap vs possible voltage boost via resonance?
>Thanks for any comments,
>Scott Hanson
>(feel free to post any of this to the list if you think there would be any