EM Gun - New PSU Ideas

[Bigglez]

Greetings,

I've been running a simulation for a new power supply
design, that would rapidly charge a capacitor bank from
a 10V battery.

This design exists only as a SPICE model, but its
performance is encouraging and not too far removed
from building a hardware prototype.

The concept is based on a flyback boost converter
with separate primary and secondary current-mode
feedback loops. The converter operates in DM
(discontinuous mode) to purge all stored charge
from the core before starting a new charging cycle.

The cycle time decreases as the capacitor charges,
with a starting frequency of about 1200Hz. Once fully
charged the PSU "burps" at about 60Hz to top off
the capacitor.

Additionally, the output voltage is set to 1kV and the
PSU shuts down when the output crosses the set point,
and starts up again when the output capacitor
discharges by 50V.

Components were picked form the current LTC library.
Nothing has been optimised.

Here's the simulation Macro model:


Here's the output voltage plot. For faster simulation
this plot is for a 200n cap shunted by 2.5meg (the feedback
resistors).


Finally, here's the current waveforms in the transformer.
Blue is the primary current.


If anyone is interested in the models and circuit files (only
in SwitcherCAD III format) I'll post or PM them.

This project was started out of curiosity, these are the
first milestone results offered for discussion. The circuit
works, the output voltage is regulated, and the smaller
capacitor charges very fast. Simulation of this design
took several hours to the one second mark, using a P4/2Ghz
PC. I'm not proficient in SwitcherCAD III to know if I
have optimised the simulation settings - I was just quite
happy to see it oscillate and make some high voltage!

Comments Welcome!

[MrAl]

Hi Peter,

This circuit looks interesting, although it also seems a bit complex
for what it does. After all, all we are trying to do is step up the
voltage to some level high enough to drive a coil, which means it
doesnt even have to be very well regulated.
Also, although the small cap may charge up quickly that doesnt
mean there will be enough energy stored in it to accomplish any
real propulsion. I think we need a larger output cap to get this
to happen. The original design had an equivalent 600uf on the
output i think. Maybe this would be a good starting point.

Perhaps come up with a simpler overall design and then we can
work from there?

Still interesting however.

Any waveforms from the original design anywhere on the web?

[Bigglez]

Greetings Mrai,

This circuit looks interesting, although it also seems a bit complex for what it does.

How would one simplify it?

Also, although the small cap may charge
up quickly that doesnt mean there will be enough energy
stored in it to accomplish any real propulsion.

Correct. My simulation runs at about one hour of PC
crunching for one second of circuit function.

For POC (proof of concept) I have scaled the load capacitor
accordingly.

After all, all we are trying to do is step up the
voltage to some level high enough to drive a coil, which means it
doesnt even have to be very well regulated.

But it does have to remain safe, so some form of over
voltage shutdown is required. Also, the capacitor
bank self-discharges and has to be periodically topped
off. The circuit runs from batteries and should be
shutdown when not required to save energy.

Perhaps come up with a simpler overall design and then we can work from there?

Does anyone have any suggestions?

Any waveforms from the original design anywhere on the web?

I don't know of any. The published design
draws about 450mA from the battery and requires three
minutes to charge the capacitor bank, according to the
author. Perhaps you can work from there?

Comments Welcome!

[MrAl]

Hi Peter,

There are some relatively inexpensive controller ic's out there made
for switching power supplies. These are not uC like PICs, but
ic's made for power supplies. They typically have voltage feedback
inputs and current feedback inputs too.
Does this sound like it might be interesting to you?

[Bob Scott]

Suggestions for a simple voltage booster for use in a capacitive discharge into an inductance? OK. How about the old Mark 10 CD ignition system. I owned up to three of these beauties at one time - an Archer Kit version from RadShack, a genuine Mark 10B and a Heathkit version. Delta, the manufacturer used a saturating transformer/oscillator section but it is not a flyback type. Since it was designed in the 60's they use PNP power transistors in TO3 cases. Just replace with 2N3773 and reverse the input polarity. Here's one schematic version:

[Bigglez]

Greetings Mrai,

There are some relatively inexpensive controller ic's out there made
for switching power supplies. These are not uC like PICs, but
ic's made for power supplies. They typically have voltage feedback
inputs and current feedback inputs too.
Does this sound like it might be interesting to you?

Sure! Beats designing from scratch any day.
Do you have any favourites? Any to avoid?

Comments Welcome!

[Bigglez]

Greetings Bob,

Suggestions for a simple voltage booster for use in a capacitive discharge into an inductance? OK. How about the old Mark 10 CD ignition system.

Funny thing I was thinking about these beasts over the
weekend. From what I remember there's an inverter
powered by the car battery to charge a capacitor to
around 400V DC. The ignition "points" (breaker) then
dump the capacitor into the standard ignition coil.

The 1.5uFd cap in your Mk 10, if charged to 400V,
would store 0.12j each cycle. So to get the magic
300j would take 2,500 cycles (give or take). I'm not
sure what the recovery time is for the inverter, it
has to keep up with a four cylinder engine at, say,
8,000 rpm (33Hz or 30ms). So a charge from cold
would take 75 seconds. Beats the 3min in the EM-15
design.

The transformer in a saturating core inverter is a
major design headache. Just getting bobbins and
cores would be too.

At 50% efficiency that 600j input equals 75 seconds
at 8watts (or 660mA at 12V).

The EM-15 design draws 450mA at 12V for 3min to
charge 300j, so that's 972j from the battery (31%
efficient).

I'm still interested in the flyback design, as it can
more readily adapt to the wide range of output
voltage that charges the cap. From cold the cap
is a dead short across the supply! As the energy is
stored in the inductor/transformer each cycle the
coil component will be larger than for a forward
converter design (where the transformer just steps
up the voltage, but doesn't store energy).

Another parallel technology is the photoflash using
Xenon tubes. Point and shoot camera flashes operate
from 200V to 350V, and have 50uF to 150uF
capacitors. (62.5mj to 9j). Most charge in
under ten seconds. Some studio strobes go to 20j,
but that's still only 7% of the EM-15 requirements.

I'm open to suggestions, (and a sanity check on
my reasoning and math) this is not a trivial project.

Comments Welcome!

[MrAl]

Hi Peter,

Well, the TL594 comes to mind, but im sure there is a more modern
ic that works almost the same as this one but is able to drive
MOSFETs directly. I'll look up the part number if you are interested.

This kind of chip has onboard oscillator, feedback amplifiers, and
output drivers, all in one package. This makes it attactive as a
power supply controller. Either 14 or 16 pins and comes in DIP package.

There's also an older chip SG35 something but it's not made for
MOSFETs.

[Bigglez]

Greetings Mrai,

Well, the TL594 comes to mind, but im sure there is a more modern
ic that works almost the same as this one but is able to drive
MOSFETs directly. I'll look up the part number if you are interested.

Wow, wasn't that one used on the Titanic?
My data sheet is dated January 1983.

This kind of chip has onboard oscillator, feedback amplifiers, and output drivers, all in one package. This makes it
attactive as a power supply controller. Either 14 or 16 pins and
comes in DIP package.

Well that's good if we wanted a fixed frequency converter.
Charging the cap is like driving a vehicle up top of a step hill,
starting from standstill. The first few cycles are into a dead
short across the supply, but as the capacitor fills up the
final cycles have little effect on the delta voltage (asymtopic).

The fall back position is to built a strong inverter and
charge the cap through a resistor, much like the CD
car ignition system Bob Scott suggested.

We know the fully charged cap is 1kV so a 1k resistor
would limit the initial current to 1Amp, and dissipate
1kW briefly as the capacitor charges. With a TC of
5CR the charging time would be 5*600*10e-6*1e3 or
3 seconds instead of three minutes.

An energy limited converter would deliver the same
size charge packet on each cycle until the capacitor
reaches 1kV. If the charge was 1j it would take
300 cycles to fill the 600uF capacitor. The flyback
topology offers this as the same amount of energy is
stored in the inductor on each cycle.

Control circuitry for this idea requires three control
loops (primary charge, secondary dump, capacitor
voltage).

There is a very attractive chip for doing something
like this, the MAX8622. Good luck getting one, its
not available to the general public.

Comments Welcome!

[MrAl]

Greetings Mrai,

Well, the TL594 comes to mind, but im sure there is a more modern
ic that works almost the same as this one but is able to drive
MOSFETs directly. I'll look up the part number if you are interested.

Wow, wasn't that one used on the Titanic?
My data sheet is dated January 1983.

This kind of chip has onboard oscillator, feedback amplifiers, and output drivers, all in one package. This makes it
attactive as a power supply controller. Either 14 or 16 pins and
comes in DIP package.

Well that's good if we wanted a fixed frequency converter.
Charging the cap is like driving a vehicle up top of a step hill,
starting from standstill. The first few cycles are into a dead
short across the supply, but as the capacitor fills up the
final cycles have little effect on the delta voltage (asymtopic).

The fall back position is to built a strong inverter and
charge the cap through a resistor, much like the CD
car ignition system Bob Scott suggested.

We know the fully charged cap is 1kV so a 1k resistor
would limit the initial current to 1Amp, and dissipate
1kW briefly as the capacitor charges. With a TC of
5CR the charging time would be 5*600*10e-6*1e3 or
3 seconds instead of three minutes.

An energy limited converter would deliver the same
size charge packet on each cycle until the capacitor
reaches 1kV. If the charge was 1j it would take
300 cycles to fill the 600uF capacitor. The flyback
topology offers this as the same amount of energy is
stored in the inductor on each cycle.

Control circuitry for this idea requires three control
loops (primary charge, secondary dump, capacitor
voltage).

There is a very attractive chip for doing something
like this, the MAX8622. Good luck getting one, its
not available to the general public.

Comments Welcome!

Hi again,

The chip has been around for a while yes, but as i said there are
more modern chips out now too that work directly with MOSFETs
and have many features that take a lot of discrete components to
duplicate.

Asymptotic charging isnt as big a deal as you seem to be making it
out to be, and you dont need to put a resistor in series with the cap
to diminish the charging current into a cap with no charge in it yet.
The idea here has also been around since the Titanic, and it's called
'slow start'. The idea is to start the converter up with a very low
duty cycle and increase based on either time or some other, perhaps
measured, criterion. The way this is usually accomplished is to twiddle
some other feedback (or perhaps the dead time setting) to force
a very low duty cycle output, and as a small cap charges the effect
is reduced until the regulator starts to regulate normally.
One idea that comes to mind is to measure output current and regulate
it to some acceptable level, then switch to standard voltage regulation.

There are quite a few chips out there now that do this kind of thing.
Perhaps check out some of the chips on TI or National or whatever.

If you would like to stick with discrete parts that's ok too, i was just
trying to simplify the design a little so the board layout is easier and
it's not too complex to build up. It's up to you what you prefer though.

[Bigglez]

Greetings Mrai,

If you would like to stick with discrete parts that's ok too, i was just
trying to simplify the design a little so the board layout is easier and
it's not too complex to build up. It's up to you what you prefer though.

To SPICE the design before building a prototype
requires either a descrete circuit or macro-models
of any complex functions. Linear Tech has models
for many (if not all) of their IC products. Third
party SPICE models are available for common
functions (555 timer, for example).

A simple model runs much faster, so adding a
complex sub-circuit or macro-model is better
avoided.

Modeling is much faster than, say, winding new
transformers, for each iteration, even if the
models are off slightly.

A soft-start converter would take care of the high
in-rush current of the cold capacitor. The flyback
converter topology does this automatically, as
described earlier.

The resistor-capacitor concept was presented to
demonstrate the challenges in an efficient design.
Start with zero volts at one amp,
end with one kilovolt at "no" amps.

Comments Welcome!

[Karl Williams]

Hey Peter

Nice design that you came up with.

I was thinking of using a PIC microcontroller to drive a transformer through something like the MC34152D. A couple of the PICs output pins could drive the device and the frequency could be programmed. One of the PICs analog to digital converters could monitor the charge on the Cap bank and another A-to-D to measure the current. The program could vary the frequency/duty cycle as needed to accomplish the 'slow start'. The Pic would also allow for a nice interface using a serial lcd display or even just more LED's for a nice bar display of the capacitor bank charge.

I haven't actually had the time to sketch out a design though...

The MC34152D data sheet says, "dual noninverting high speed drivers specifically designed for applications that require low current digital signals to drive large capacitive loads with high slew rates. These devices feature low input current making them CMOS/LSTTL logic compatible, input hysteresis for fast output switching that is independent of input transition time, and two high current totem pole outputs ideally suited for driving power MOSFETs. Also included is an undervoltage lockout with hysteresis to prevent system erratic operation at low supply voltages.
Typical applications include switching power supplies, dc–to–dc converters, capacitor charge pump voltage doublers/inverters, and motor controllers. This device is available in dual–in–line and surface mount packages."

Here is the data sheet: http://www.ortodoxism.ro/datasheets/mot ... 34152D.pdf

Take care, Karl

[Bigglez]

Greetings Karl et al.,

Nice design that you came up with.

It's at the concept stage. I (we?) need to understand
the physics of charging a big capacitor before designing
circuits.

I was thinking of using a PIC microcontroller to drive a transformer through something like the MC34152D.

This approach won't change mother nature, but having
a uC in the mix will allow other features (as you noted).

The MC34152 is a 20V buffer/driver which sells for $1.50 or so.
It's only role would be to replace the complimentary emitter
followers I use to drive Fets (about 26cents worth of parts).

Here's a model of a 10V input, 1000V output topology.
The 1uF capacitor on the output charges in about 300ms.





300j from a 12V battery is quite a tall order. A transformer
will be required, as an inductor only design will draw too
much current from the battery.

Here's a transformer coupled model, which reduces the
input current demand. The same load capacitor charges in
about 80ms. Also, a soft-start circuit reduces the
in-rush current.





Do you have a spice simulator? I can post the files for
anyone that would like to run their own simulations.

Comments Welcome!

[Bob Scott]

I did some research to get a handle on how powerful is 300 joules. I need some perspective. So I checked a Remington .22 cal long rifle bullet specifications:

33 grains
1500 f/s muzzle velocity
165 pound feet muzzle energy

Using the freeware "convert" program I used 2 ways to calculate joules. Once using the weight, velocity and 1/2MV^2, the other with simple energy conversion. A real 22 calibre bullet only has 223 joules! These are regular hollow point varmint bullets but not the mil spec ones that blow flames out the end of the rifle.

So, we are not exactly playing with a toy with a 300 or 600 joule EM rifle.

300 joules isn't very hard to suck from a 12V pack of D cells at 1 amp and it only takes 25 seconds at 100%efficiency, proportionately less time at higher current. I wonder what percentage of energy the EM rifle transfers to the projectile?

[Bigglez]

Greetings Bob,

A real 22 calibre bullet only has 223 joules!.....
So, we are not exactly playing with a toy with a 300 or 600 joule EM rifle.

Thank you so much for generating this data! I looked for
a comparison of gunpowder to EM weapon energy while
reading the article.

I wonder what percentage of energy the EM rifle transfers to the projectile?

That's the 64k dollar question.
I'm not sure how to work through the various issues, in
fact I'm not sure of their individual importance to the project.
Things I thought of include:
(1) Conversion of electrical energy to kinetic energy
(2) Loss of magnetic flux due to 'shorted-turns' in the launch coil frame
(3) Merits of a long or short launch coil
(4) Lack of rifling in the barrel

300 joules isn't very hard to suck from a 12V pack of D cells at 1 amp and it only takes 25 seconds at 100%efficiency

Probably should consider 50% efficiency as "best in class".
"D" Cells would provide the in-rush current for a more powerful
charger. The main advantage is the firing rate. I also thought
about a battery belt instead on on-board batteries. I've used
TV camera battery belts that are typically NiCad D cells in
either 14.4V(camera) or 28V(sun gun lights) configuration.

Comments Welcome!