Driver Circuit Description


designed by Rob, N3FT; with inspiration from the Duane Bylund article
in Radio Electronics, Sept. 1991, p. 33.

NOTE:  The schematic was made long after the circuit was built and 
perfected. Sorry.  That means I cannot guarantee that it is free of 
errors.  However, the schematic was compared to the original as best 
as I oould.


1.  The resistor network at pins 6 and 7 of the LM555 is incorrect.  
Instead of a fixed 200k and a 500k pot in parallel, it should be the 
200k fixed in parallel with a 500k pot and the resulting two in series 
with a 1k fixed.  Thus the 200k fixed established the maximum duty 
cycle, and the 1k pot establishes the minimum duty cycle.  You could 
try eliminating the 1K, but you might damage the LM555.


I advise doing what I did, which results in a pretty easy and 
professional looking driver:  Go buy some surplus half bridge or 
bridge switching supply, cut away the low voltage portions of the pc 
board, and then mount the new UC3825 controller on a small daughter 
board built out of a plated thru hole + ground plane vector board.  
Put the resulting circuit inside an aluminum case of some sort.

The surplus place "Wacky Willies" in Hillsboro, OR where I buy alot of 
my stuff still has lots of these NCR 500 watt switching supplies that 
I used to build mine.  I believe these even have a current mode 
controller, though I think its a UC3845 or similar.  The supply has 
the 4 IRF740's, huge ferrite transformer core,  and nice heatsinks- 
very, very pretty design.  However, I do not know if they will do mail 
order, you can try them at 503-642-5111.  These supplies were $20-30 
last time I was over there.  At that price it would be worth it for 
someone to make some arrangment with them-  only its not going to be 
me, sorry, I don't wan't to be in the mail-order business.

Jamesco also sells alot of half-bridge MOSFET 200 watt and up supplies 
very cheap.


1. The circuit from the AC line up to the MOSFETS is the standard 
power portion of any half-bridge 250W or greater switching supply. 
Setting jumpers either selects the bridge rectifier for 220v 
operation, or the so-called full wave voltage doubler for 120vac 

I added the GE V130LA10A MOV's to keep potentials between ground and 
the ac line to a minimum.

The construction info for the common mode filter inductor is my 
equivalent to what is actually on the driver, since I have no info on 
the actual core material of the on the unit.  All the caps are the VDE 
and UL rated polypropyene "across the line rated" types.

2. The original power supply was I believe a push-pull type so I had 
to cut some traces and add some wires.  However, the addition of the 
IR 80SQ035 schottkey diodes and the HER 305 ultra fast rectifiers are 
my addition to the normally seen type of circuit.  

It is absolutely mandatory to disable the internal source to drain 
diode of the MOSFETS.  Otherwise you may as well buy a bag of a 1000 
mosfets.  The reason they need to be disabled is that when connected 
to a circuit capable of storing huge amounts of energy, like a high Q 
coil, the circulating current tends to alternately drive the 
source-drain diode of the mosfet to extremely heavy conduction, 
storing up large amounts of charge in the juction.  Then a hundred 
nanoseconds or so later, when the opposite mosfet starts to conduct, 
the first mosfet looks like relatively large capacitance- or basically 
a short circuit for an instant. One or both mosfets blow.  Its 
possible to increase the dead time (thats the set off period where no 
mosfet can be conducting) by increasing the Ct of the UC3825, but for 
high speed operation, there is a practical limitation to this.  Also, 
the slow diode wastes power even when its not blowing up the fets.

The schottkey diode works by making it impossible to forward bias the 
source drain diode of the fets.  It only needs to be rated for the 
maximum pulse current present- the ones specified are probably 
overkill in the ratings.  You don't need to worry about the reverse 
voltage rating of the schottkey, since the HER305 conducts when the 
schottkey gets reverse biased.  And schottkey diodes are fast- much 
more so than the HER305- so there's another worry gone.  The HER305 is 
a 3amp 400 volt 50nS reverse recovery time diode.  Having it external 
to the fets also removes a source of dissipation in them.  IR makes 
some diodes calle HEXFREDS which are available in even higher voltage 
and power ratings but are much more expensive- would be better for a 
bigger setup.

3.  Output transformer-  more attention needs to be paid to its 
construction than any other part of the driver.  

First obain a core.  You don't have to pay a fortune for a new core.  
Find some 1kw computer supply and remove the transformer.  I have 
obtained several of these cores by simply putting the transformer in 
boiling water until the shellac softens, and then pulling apart the 
core.  Then you can take apart the transformer windings to get the 
bobbin and if you're lucky, some expensive teflon insulated wire.  
Forget any core or transformer that looks like its coated with epoxy 
or some other extremely hard enscapsulation. 

My core when both E sections are put together has the dimensions:
5.5 x 5.5 x 2 cm.  Anything that size or larger will work.  The bigger 
the core, the less chance with core saturation at low primary turns 
count.  Thus less secondary turns can be used.

Make certain any used core material has a low resistance.  The low 
resistivity ferrite core materials are suitable for transformer use.  
Any core with unmeasurable resistivity is probably actually from a 
choke, and won't work.  My core measured about 10k ohms when the VOM 
probes were about 1/4 inch apart when touching the core.   Type 77 
material is good if you can get a brand new core.

After obaining a core, get a hold of some some teflon coated wire- 
14-16 gauge.  Why cut corners?  Drill two holes on the bobbin just the 
size to let the primary leads thru and then wind the primary.  Secure 
the wires, then apply a coating of GE Silicone II sealant around the 
primary to there is about 1/8 inch thickness.  Use a spatula to make 
it perfectly smooth with no ridges. Let it cure overnight.

Then using #30 gauge Belden (again why cut corners, use real Belden 
magnet wire, if you can obain any special enamel types for severe use, 
but all means use them)  Wind 50 turns for the first layer- then coat 
with another 1/8 inch of silicone and let dry overnight.  Make certain 
that the magnet wire going to the spool sticks straight out thru the 
silicone, so you can easily start to wind the next layer.  Use 
gap-filling super glue and accellerator to secure windings as you are 
progressing.  Then after a night curing, wind the next layer, except 
now the windings will be adding so that when this layer is finished 
there will be a potential difference of 100 turns of voltage between 
the ends of the two windings.  This is done only to make winding 
easier.  If you want only 50 turns of voltage potential between the 
two windings, you can put two silicone layers, each half as thick, 
with the wire criss-crossing between the two windings.  Do this for 
the last two layers.  Then fit the core together, and add as much 
extra silicone as possible to fill the gap between the windings and 
the core going aroound the outside.  This provides extra safety 
against corona and flashovers.

I am sorry but I have not been able to find the  exact winding specs 
for my transformer.  I believe that I used 10 primary turns and 200 
secondary turns.  However, I can measure the resistance of my 
secondary and it measures 7.0 ohms.

I did quite a bit of experimenting.  A turns ratio of 10 to 1 puts 
much less stress on the mosfets (cw operation is possible), but 
streamers of 3 only inches or so.  A ratio of 30 to 1 is extreme, 
putting much more stress on driver, with really no increase in 
streamer length over the 20 to 1 ration  At the 20 to 1  ratio 
streamers of 8-12 inches are possible.

4.  Controller circuit: I used an UC3825 for several reasons.  First 
its optimized for super fast current limit and shutdown.  Secondly, it 
operates up to more than 1mhz.  Thirdly, it had higher mosfet drive 
capablity than any other device at the time I did the design (I 
believe 1.5A peak). 

Looking at the circuit, the input thru the diode is the on/off singal 
from the LM555 duty cycle circuit.  When this signal is low, the 
UC3825 error amp operates normally and the 10k panel mount control 
varies the PWM level- in reality this pot functions more like an on 
off switch- I never operate it at any other position than min or max.
When the LM555 drives it high the error amp goes low and the UC3825 is 

Pin 5 connects to several fixed resistors and a precision 10 turn 
panel mount put for tuning to the coil resonant freq.  Socket the 
fixed resistors so they can be switched out for coil experimentation.

Pin 6 connects to the  4700pf frequency determining cap and some 
additional components that function to stabilize the circuit.  I will 
refer the reader to the Unitrode Databook for a discussion of slope 
compensation for an analytical treatment, but basically, for stability 
reasons , one wants to feed some portion of the oscillator signal back 
into the ramp pin.  The 0.047uF cap is simply a DC blocking cap- so 
its really the 4.7k resistor that feeds some current from pin 6 into 
pin 7.  The 0.001uF cap makes certain there is a sharp rising edge to 
the signal.  You can empirically adjust the value of the resistor 
until the circuit does not squeal (i.e. stable) over the full 
revolution of the panel mount PWM control. Socket all these parts.

It is the ramp pin 7 that senses the ouput current to the coil and 
adjusts the PWM on a pulse-by-pulse basis to provide as much of a 
constant current drive to the coil as possible.   The current sense 
circuit between the MOSFETS and the tranformer uses a large 0.1ohm 
resitor and a small ferrite core to isolate from the HV, meanwhile 
feeding back a representation of primary transformer current.  This 
feedback is what makes the PWM panel mount control pretty much 
irrelevant. (except for checking for instabilities)

The sensed signal is also sent into the current limit pin 9 which 
shuts off the MOSFET drive on a pulse-by-pulse basis when voltage on 
the pin goes beyond about 1 volt.

There are two small pots that adjust the signal back into these pins.  
Start them at half of full scale to begin with.  Then once you are 
sure that the coil is resonant- back them off until the streamer 
length hits a maximum.

It is the dynamic current limiting that sets this circuit apart from 
all else, especially the Bylund circuit, since my driver can run with 
the coil out of resonance indefinitely, since the current mode control 
comes into play.  (thats as long as these two pots are adjusted 
correctly, and the slope compensation is correct)

5.  The LM555 Duty Cycle Circuit.  With more mosfets the circuit could 
function CW.  However, more bang for the buck can be obtained with an 
impulsive operation.  And it looks and sounds like a tesla coil. The 
streamers are much longer for a given amount of mosfets when the fets 
are pulsed.  With this circuit, the duty cycle can be adjusted from 
near nothing to a roaring flaming discharge.   Its a simple astable 
LM555 circuit running around 30hz from the data books.

6.  MISC:  Power and output voltage sources on the UC3825 have 
seperate pins, so I have utilized the feature and isolated both with 
ferrite bead choke.
   The schottkey diodes on the UC3825 clamp the outputs between 0 and 
12V to give crisper drive to mosfets.  It is also possible to use a 
bidirectional 15v transient suppressor (P6KE15) directly between the 
gates of the mosfets to source for additional protection to the 
mosfets against gate oxide pucture if any of the other fets fails, but 
I didn't include them on mine.


Thats all I can think of for now.  I will try to answer any additional 
questions, but now I'm kind of burnt-out on schematics and 
documentation and want to get my new fets so I can get this thing up 
and running again (don't try to drive any electric fences :) or 
anything other than a coil like I did).  Also, this driver with the 
Bylund coil seems to pack a wallup of low freq AC envelope.  I get 
quite a jolt just from the leakage on the ground connection.  The coil 
also seems to put out alot more current than an equivalent size tesla 
coil.  I have cut thru ceramic tiles with the arc.  (was trying to 
find an easy way to cut tiles when my roommate was installing new 
floor, ha ha)

Good luck. Rob.