Wanting to take the step up from my Palmtop SSTC I decided to do a
proper, class E, 4MHz SSTC. For those who have never seen these before,
the basic idea is to use a switching topology commonly used in RF
amplifiers, instead of the usual half or full-bridge. The reason for
this is too greatly reduce switching losses in the power MOSFETs.
Richie Burnett was the first to use this topology in a Tesla coil, and
since then many have built similar coils. Despite it's small size, the
current draw of this coil is 270W. I'd hazard a guess that it's at least
90% efficient, so that's a lot of power in the plasma plume.
High frequency drivers
Original, discrete design
This is the original driver I developed, which has since been superseded.
It is tricky to tune, but on the other hand uses junk components one might
having lying around. For those buying new parts, see my newer design.
The driver used for the 4MHz class E SSTC is a string of several
amplifiers, which buff up the signal from a 4.096MHz crystal. It's
pretty straightforward until the first mosfet stage, where the high
frequency fun begins. The IRF630 gate is driven through a small
transformer, which is biased at around 3V. The bias reduces the amount
of voltage swing required to reach for gate potential, and the
transformer helps match the 2N390X stage to the gate impedance. Leakage
inductance in this transformer also play a part, by creating a resonant
circuit along with the gate capacitace. If tuned properly a nice sine
wave can be created on the IRF630 gate, allowing for proper gate drive.
Which is all well and good, but there are still two stages left to
tune! The next stage is actually a class E stage which is used just to
drive the IRFP450 gate. Again the gate transformer must be carefully
tuned for the best waveform, but in addition you want to tune for a
class E waveform on the IRF630 drain. (See below for how to tune class E)
To be honest here, I had already tuned the IRF630 when I discovered
moving the windings on the gate transformer had a profound effect on
the waveform amplitude. At this point I simply moved the windings until
the IRFP450 received a perfect sine-wave gate waveform, and simply let
the IRF630 drain waveform be. If overall system losses are a major
concern, you'll need to put more effort into tuning the IRF630 than I
did! One fun thing to note, is that at this stage we're already
generating roughly 5W of RF power just in gate drive. Without a
heatsink your IRFP450 could overheat just from gatedrive alone, given
time.
Once the IRFP450 gate waveform is perfected, you're only about halfway. The
final class E stage is the one that counts, and tuning it can be
tricky. At this frequency minor changes of the secondary or even
primary can destroy the tuning. I experienced this myself, as I had
originally tuned everything to working order, and then foolishly
decided to cut off any remaining PVC from the formers to get a more
deliberate look to the project. Little did I know that the 1cm of PVC
removed from the secondary and primary formers were adding vital
amounts of capacitance to the setup! I was forced to make a new
secondary, so make sure you don't repeat my mistake! In essence you'll
need to use a more or less set value of drain capacitance, and tune the
primary and secondary until you get both breakout, and a good class E
waveform. This process takes time, so be prepared to do some
experimenting.
IFRP450 gate, 5V div, 40ns
New and improved driver (25.04.2022)
After blowing up my driver board during a demonstration for a friend, this project sat
in a closet for years. The cause of the fault was something as silly as connecting the
driver power wires in reverse, due to unclear markings of the screw terminal I had used.
This has since caused me to put protection diodes in some of my projects, and mark input
terminals better. Anyway, rebuilding the old board proved difficult, and for my new board
I wanted something with less tuning, which would just drive a large MOSFET without much
hassle. An additional goal was making it easy to switch drive frequency, so I could use the
same design on any future class E coils I might make. This lead to an embarrassing amount
of hardware revisions and failures, but eventually I landed on something simple and reliable.
The driver uses a crystal oscillator module to provide the drive frequency, which
eliminates problems from external noise. This proved to be an issue in the previous
design, as it relied on a sensitive, high impedance oscillator circuit built around
a HEF4049 inverter. This also removed the need for a pre-amp circuit, already here
simplifying the design. It's not drawn in the old schematic above, but I had a fiber
optic input with a NAND gate to allow polarity switching of the fiber optic signal.
This allowed running in CW and "turning off" the coil when interrupt pulses came, rather
than the usual behaviour of turning the coil on when a pulse is received. This doubles
as a CW/interrupted mode switch if running with no interrupter too. This circuitry was
integrated into the new driver, using a D-latch with Q and !Q outputs. Since the latch
is synchronized with the clock source, we won't be interrupting the MOSFET in the middle
of a switching cycle. A minor improvement, but good to have. Finally the entire output
stage is replaced by a single NCP81074A low-side gate driver IC. All in all a simple,
small, and easy to configure driver. Changing frequency is a matter of finding the correct
oscillator module, with no other tuning components. That is left to the class E stage!
I tried various approaches at getting full push-pull drive with the NCP81074A chip,
but had to give up after several hardware revisions. The configuration where one chip is
run in inverse of the other simply failed, always causing one of the chips to run very
hot. I never found the cause of this, but assume it is related to the internal construction
of the driver chips. Perhaps a dead-time generator would have solved it. Either way, I
solved the problem by using a DC-restoration circuit on the secondary side. The MOSFET
gate waveform looks pretty good. I'm unsure if the duty cycle is still 50%, but this
doesn't seem to cause any ill effects in practice. I used a 1N5819 diode, with 100nF
capacitor, and 18R gate-source resistor.
Revision 4 PCBs, with mods to revision 5.
Since the DC-restoration circuit was now needed on the secondary side, I designed
a class E power PCB as well. This should work well enough for lower power designs.
A potential problem I've noticed with all revisions (even those without the DC-restoration)
is that when turning off the gate drive a long period of ringing is seen on the gate.
I assume this is caused by resonance in the gate drive transformer, and blocking capacitor,
but I'm not sure. The voltage can quickly become enough to turn the main power MOSFET
on, and at such a low frequency that it might cause problems. Something to keep in mind!
Left: Gate drive signal. Right: Trailing voltage on gate after turning off interrupter.
Magnetics
The main RF choke in the Class E output stage is something taken from an ATX SMPS
years ago. Despite this, I'm quite certain the core is some kind of powdered iron.
Given the color markings of yellow with white, then according to
this guide, it is:
MATERIAL #26 (ยต=75): A Hydrogen Reduced material. Has highest permeability of all of the iron powder materials.
Used for EMI filters and DC chokes. The #26 is very similar to the older #41 material but can provide and extended frequency range. See AC Line Filter and CD Choke sections for size, permeability and frequency range information.
Which makes sense. As for the GDT, I'm not too sure what type of core to go for.
I've used various EMI toriods taken from old equipment, and they seem to work
great. The usual toroids I use for lower frequency work do not perform as well.
Other than that I can't say. Send me an email if you figure it out!
Tuning for Class E
As mentioned earlier the premise for class E is to reduce switching
losses. This is achieved by turning on the switch with zero current and
zero voltage across it.
Despite having tuned my coil, I still don't have a good procedure for this. When you tune or adjust one
component, everything else that was previously optimized needs
readjusting. For this reason you'll need to tune and retune many times
before arriving at an optimal tuning. For starters wind a coil with a
resonant frequency somewhat below your target frequency. This will allow you to
remove turns later. Then estimate the values (or use the ones in the
schematic) of the class E components, and power it up while watching
the drain waveform. If it appears as a half-sine wave that is cut off
before reaching zero, you need to remove some turns from the secondary.
If it appears as a sharp spike, at less than 50% duty cycle, you need
to add some turns to the secondary. Once correct, you should have
breakout when run from 50V.
Now you'll need to tune the primary side. You'll need to experiment with
both coupling and inductance, and possibly the drain-source capacitance
and RF choke. Decent values of the RF choke and drain source cap can be
found even if your setup isn't tuned perfectly, so start here. Once
you've settled on values for these components, you need to experiment
with the primary winding. Some of the factors that come into play are:
Length of wire
Number of turns
Spacing between turns
Distance above base
The tuning guide below will help you see which direction to take things in.
Disclaimer:
I do not take responsibility for any injury, death, hurt ego, or other
forms of personal damage which may result from recreating these
experiments. Projects are merely presented as a source of inspiration,
and should only be conducted by responsible individuals, or under the
supervision of responsible individuals. It is your own life, so proceed
at your own risk! All projects are for noncommercial use only.
This work is licensed under a
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
