Creating high voltage electrical sparks with a television flyback
transformer is a fun project for many electronics hobbyists. In fact,
the first few HV projects I ever worked on were flyback-drivers of different sorts (outlined
in another page here) -
their simplicity and ease of construction makes them great introductory
projects into the world of high voltage.
Some of the common flyback driver designs include the popular single transistor driver, as well as the
ubiquitous Mazili ZVS flyback driver. However, I wanted something more
robust, be capable of driving flyback transformers at
high power levels, yet still being reasonably straightforward to
This page outlines the design and operation of a half-bridge flyback
driver, and describes how you can build one yourself too!
If you are unfamiliar with flyback transformers and what they
are and how they work, please feel free to read my earlier page on this topic
here, where I describe their method of
operation in great detail. This page will focus primarily on the design
and construction of a half bridge flyback driver.
For result, please scroll down to the bottom of this page. Thanks for
ready and please be safe if attempting the project described here!
Design and Operation
Overall Project Goals
For this project, I create a few goals and overarching
ideas to guide myself when designing the circuit:
- Develop a robust and power driver
- Design a circuit which is simple, and easy to build
- Be flexible for driving a variety of flyback transformers
Based on the above guiding principles, I decided to use a half-bridge
driver topology, driven by a fixed frequency oscillator.
I could have gone for a more powerful full-bridge inverter, but using a
half bridge saves components and will be powerful enough. In addition,
I decided to go with all discrete components (i.e. not using any
programmable ICs) such that hobbyists can easily replicate this project.
Finally, I also wanted some sort of basic protection for overall
longevity and robustness of the system, so I implemented a
simple over-current detection which cuts off power when a set current is
Note that despite the fact that the half-bridge driver
is driving a flyback, it does not actually drive the flyback transformer
in typical 'flyback - mode', and instead more akin to a regular, albeit
high frequency, transformer. Specifically, flyback-mode means that the
primary and secondary conduction cycles are non-overlapping. In the
flyback mode, the primary coil acts essentially as an inductor, where
energy from a primary cycle is stored in the magnetizing inductance of
the inductor, and then released to the secondary coil.
One popular feature which people may be interested in is
to create a 'singing' arc, whereby the driver modulates power into the
flyback transformer based on an audio input. I decided to omit this
feature from this particular project, focusing instead on simplicity. However I do plan to create a much more advanced driver in
the future with this capability (and more!).
How the driver works
Here is a schematic for the completed Half Bridge
The Halfbridge Flyback driver system including design, schematics and
layout are available for use under the
Power is supplied via J1 which, in this
design, can range from 9 to 36VDC. This is the main power supply for the half-bridge
inverter, and also powers the voltage regulator generating the 15VDC rail which is used for the other
parts of the circuit. Note that the range of 9 to 36VDC is simply the
input voltage range to the small DC/DC converter (VR1) - you could
certainly run the driver at even higher voltages, limited by your main
switching MOSFETs, but then you will have to find another way to
generate the 15V rail.
Note: To generate the 15V rail, I'm using an
However you can also replace this with a standard
(TO-220) voltage regulator. However those typically have a maximum 35VDC
input and will get very warm (heatsink). Note that those also have a 2V
voltage drop at 1A I_fwd, so you will be limited to ~17 to 35VDC input.
For ultimate adjustability, the 15V rail can be generated with another
external power supply.
Next, a generic 555 timer (U2) is set up in the standard
astable oscillation mode with a duty cycle fixed at 50%. The operation
frequency is set by R6 and C6, as well as a external potentiometer
(5kR), connected to J5. In the schematice above, this
allows setting an operating frequency between about 23.4kHz and
101.3kHz. This square wave is fed into the input of the UCC37322 (U1),
which is a powerful 9A MOSFET driver.
Note: As shown in my schematic above, the 555
generated frequency can be calculated as freq = 1
/ (1.4 * (R6 + J5)*C6 ) where R6 = 1.5kR
and J5 = the value set by the potentiometer and C6 = 4700pF. Another
example is using the same 5kR potentiometer together C6 = 3.3nF, R6 = 2.7kR, giving us
80.2kHz (when the pot is 0 ohms) to 28.1kHz (5k ohms).
Instead of driving a FET directly,
the UCC37322 drives a hand-made Gate Drive Transformer, which then
drives the gates of two power MOSFETs. Using a 1:1.6:1.6 turns ratio
(for example 10 primary turns and 16 of each secondary turns), this
generates a +-12V gate drive signal across the gates of the power
MOSFETs (from the +15 0V across the primary winding of the GDT).
The maximum input to the half bridge is limited mostly
by the choice of MOSFET used. For example if the popular and cheap
IRFP260 MOSFET is used, something like ~150VDC can be used. In my case,
I've opted for a more expensive but better MOSFET (IXFH80N25X3 250V 80A)
which should be good for up to ~180VDC or so (though in practice 36V will
produce good results, and is already more than likely to be able to fry
most flybacks!). Likewise, the DC power supply used needs to be sized
accordingly. I've opted to use a 150W 36VDC DC power supply for this
project which should be more than enough for most applications.
Over Current Detection
A simple over-current-detection (OCD) scheme was also
implemented. This is described in the bottom-right block in the
schematic above, and can also be omitted if you do not want over-current
protection. J4 allows keeping the OCD circuit in system or out of
To achieve OCD, a home-made current transformer (CT) was
created which senses the half bridge output current. This was
constructed using a 20:1 turn ratio CT. The signal from this CT is
rectified through a bridge rectifier (made from diodes D1-4), and
burdened by load resistor R3. The voltage across this is sensed via a
Overvoltage Monitor IC (TPS3702CX33DDCR), which is essentially an
integrated comparator. In this case, the trip voltage for my particular
choice of IC is 3.60V. When the voltage across the burden resistor
exceeds 3.6V, the TPS3702 IC's OV (over voltage) pin is pulled low
As a result, with the current setup, the peak current
limit is set conservatively at 7.2A. To increase the current
limit, either increase the number of turns on the CT, decrease the
resistance of the burden resistor. For this particular IC, the
propagation delay for an over-volt event is 35us, and the time for
re-enabling is 19us when the voltage falls below threshold. Hysteresis
is 0.55% for the X part, with OV thresholds 3.60V and 3.43V.
With the design complete, I decided to do a quick lunch
time layout. One thing I thought I wanted to do was to fit everything
into a standard ATX computer power supply box. I decided to go with a
36VDC 150W DC/DC switching power supply and I managed to find one which
had a 5x3" footprint. In order to make mounting easier, I used the
same footprint as the basis for the board layout so I could mount
everything into the box in a tidy way (on top of each other).
After about an hour or two of layout, I got something I
was quite happy with. 5x3" is a lot of board real-estate so layout was
no problem at all. The main bus features a laminated bus structure.
Power comes in from the right into the 5mm screw terminals on the right.
This is how it looks like in 3D. Most computer ATX power
supply boxes come conveniently with a power switch, IEC input jack, and
a 12VDC fan. However we will not have any 12VDC supplies, so I replaced
it with a 80mm AC main directly to mains instead. This will provide
cooling for both the power supply and switching MOSFETs which have their
own direct-mount heatsink. Having separate heat-sinks allows for a
direct thermal contact to the transistor package casing and should
hopefully improve thermals. Not pictured here (or on board) is the
current transformer used for over current detection. The Gate Drive
Transformer is a self-wound GDT and cable-tied to the PCB via the 2
Putting the Half-bridge Driver Together
After waiting for a while for the PCB and components to arrive,
I was finally able to assemble the board.
There are not too many components so the driver was
completed fairly quickly. The two things that I had to do an extra
little bit of work was to tap and drill holes in the two heat-sinks for
the transistors - (these are very cheap 28 x 28 x 20mm square heatsinks
from Aliexpress / Ebay), as well as wind my own GDT.
For those not familiar with GDTs, GDTs are transformers
which allow generating isolated gate drive signals without requiring
isolated power supplies. They work great in this half bridge design
whereby the first of the secondary winding drives the lower FET and the
second secondary winding drives the gate of the upper FET. In order to
generate the +-12V desired, I opted for a 12:20:20 primary:sec:sec
ratio. The primary is driving at +15V, and should give us +-12.5V on the
secondaries. It is critical to make sure that the core does not
saturate, especially when switching at low frequencies. I opted for a
decent sized 1" diameter ferrite core (N30
material toroid made by Epcos / TDK and commonly found on Mouser or
Digikey), and I twisted 3 strands of wire together (to reduce leakage
inductance). 12 turns were wrapped around the core, then an additional 8
more for the secondary windings.
Care was also taken to ensure that the polarity of the
secondary coils were soldered in reverse (so we only have one transistor
on at any time otherwise a dangerous shoot-through condition occurs!). I
recommend verifying the polarity with an oscilloscope if you have the
Likewise, the current transformer (pictured above with
brown wire) was made using the same ferrite core. 20 turns were wound on
it. The number of turns can be changed in conjunction with the burden
resistor to change the desired trip current. With the board assembled,
it was then mounted on stand-offs on top of a 36V 150W AC to DC power
Finally, I placed all the components into a scrapped ATX
power supply case. The case is ideal for projects like these because of
their compact size, essentially 0 cost, and comes with a built in fan (I
replaced it with a AC fan driven directly from the mains), switch and AC
filtering, power receptacle, as well as a cable hole at the front!
A barrier block was mounted at the front, as well as the
enable/disable switch, a bright red LED, and the frequency control knob.
The driver is complete!
Choosing a flyback transformer
To test out the driver, I need a flyback transformer. But what
should I use? There are many kinds of flyback transformers. As
described in my flyback page, most
hobbyists prefer the 'wide' or 'fat' secondary winding kinds, as
opposed to the more cylindrical ones. I had a few on hand, but I
also picked up a custom made flyback back in 2014.
In 2014, a forum member on the 4HV forums (Fiddy) made a
custom order of a few transformers. Specifically and quoting Fiddy,
these flyback were constructed as follows:
• No diodes for true ac output.
• No primary coils at all.
• Insulating bobbin on-core for your own primary
• Large gap between secondary and primary for your own primary winding.
• 3000 turns of 28AWG secondary
• Potted and insulated secondary
• Thin plastic film between core joints
I bought myself one! For reference, above is a photo of
the Fiddy's Flyback (middle) with a F0241 flyback on the left and a
KFS230867 (replaces Koyod HR 42020) transformer on the right. I did a
quick weigh in and they weighed 352g, 668g (with the wire) and 140g from
left to right!
With the much nice construction (and lack of other
secondary windings), this large flyback will be my main flyback of
Preliminary Test - mini Jacob's Ladder
I plugged the driver in and it worked first time around,
both with regular flyback transformers as well as with the large custom
flyback! Here are some preliminary results.
first test, I created a very simple Jacob's Ladder out of two
wires and hooked it up to a flyback transformer connected to the
half bridge driver. The driver was fed with 34VDC, which drives
a 12 turn primary coil on the flyback transformer. The result
was some very beautiful arcs ranging from white hot (~30kHz), to
beautiful long purple arcs (~42kHz with some resonant action
Due to the high operating frequency, the arc is very
quiet and flame-like, dancing between the two wires in a mesmerizing
In this preliminary test, the results were very good.
The half bridge driver performed well showing really no signs of much
heating at all despite a continuous run of around five minutes. The
flyback transformer (windings and core) remained cool as well, with most
of the heat being from the ladder itself. Power level was measured
around 2A at 34V for a roughly 70W operation when arcs are drawn.
Time to turn up the power!
As usual, more to come soon!
Links and References
Here are a list of links to other great
hobbyists whose work I have learned from in creating this
project. Many thanks to them for pioneering the way.
Marko has a good write-up of a half bridge flyback driver
powered by a SG3525 PWM IC
Steven Ward built a simple and effective half bridge flyback
driver powered by a 555 timer as well
Uzzors2k also builds a 555 timer based flyback driver running in
flyback-mode, as well as other great push-pull drivers
Advanced Linear Designs has a
on 555 astable 50% duty operation
Back to main page
(c) Gao Guangyan 2017
Contact: loneoceans [at] gmail [dot] com
Loneoceans Laboratories. Copyright (c) 2003 - 2017 Gao Guangyan, All
Rights Reserved. Design 3.
Removal of any material from this site without permission is strictly
prohibited and will result in infringement of copyright laws.
Disclaimer: Projects and experiments listed here are dangerous and should
not be attempted.
... page generated in 0.00206 seconds