The wisdom of the list
I've found it pays before I build anything to float the design on the TCBOUK list.
OTLC design proposal
>I'm aiming for a mega simple design
The simplest circuit I've ever seen is Derek's first OLTC design. If you
replaced the transistors and stuff with an IXYS driver chip you would have
something brick-worthy. That discrete driver is inverting but you can use a
non-inverting driver by swapping the Q, Q\, S, and R on the flip-flop
around. Using a flip-flop as a one-shot is a naughty trick but it seems to
work very well, I've started doing it too.
That is the record to beat. Maybe you can do it by using an inverting IXYS
driver and making it self-oscillate by feeding the output back to the input
<cringe> Then you would not need any chips apart from the gate driver and
the voltage regulator. But you would probably pay for it in reliability and
My big OLTC got complicated because of all the protection circuitry to make
sure the IGBTs could never be killed by transient spikes. The "dimmer
switch" for regulating power without a variac upped the complexity too, but
it was worth it to hear the gnarly buzzing noise it makes :)
The OLTC2 was a research project. When I developed it into a commercial
product for Finn it got even more complicated because it had to withstand
every possible power-up and power-down condition without blowing anything.
The hardest one was if the power was removed and reapplied quickly, causing
the gate driver board to shut down and restart with the DC link caps still
fully charged. But you don't have to worry about this in a hobbyist
situation where you always sequence the power by hand, turning the gate
driver on first and off last. (you do that right?)
>ill breadboard a circuit at say 23volts
23 should be all right. I tested the OLTC2 to see if the IGBTs were getting
enough drive and they seemed totally happy at 24V with 2500A current. It was
running on 24V gate drive at Derby IIRC.
The mini OLTC ran at 30V though because that was the voltage Terry used with
BUT.. I wouldn't be too keen on running a gate drive chip 1 volt below its
absolute max. The reason is that the L*di/dt (i.e. transient) spikes in your
layout could easily exceed 1 volt and cause the chip to fail.
I Don't bother to measure the primary frequency. Its not easy. I
believe you can put a sig gen across the (disconnected) transistor, and look
for a current peak at the resonant frequency.
My usual tuning method, is
1. loosen the coupling between pri & sec.
2. run the coil at low voltage and low BPS
3. lengthen the quench so you get 2 ring-up/down cycles
4. measure the field around the secondary with an open scope probe
5. change the primary cap until max output.
6. When you start to see a full ring up & down with a notch you're near.
6. add topload capacitance (by putting a hand near it) if you can increase
the output you have too much primary capacitance.
7. tighten the coupling again, and adjust quench for max output.
I need to really measure the frequency of the primary I have just
finished for the oltc and wondered how I should do it.
I知 thinking I should hook up a capacitor of known value and then sweep the
primary setup with my signal gen looking for a peak on the oscilloscope. is
this correct? also how do I look at the peak? is it by air or connected to
To make sure you are In the right ballpark use
I use this for my oltc's, It gives fairly accurate results for my coils, but
I find that the uncertainty痴 in the primary circuit can change the results by
quite a large amount, so don't take it as absolute gospel.
Please take the sections on voltage out and spark length with a pinch of
salt. They are applications of various formulas out there, and they are
If you want to see the spec of my OLTC's click the buttons on top. The
button for the OLTCII still gives the external primary not the new internal
OTLC charging choke
I've got some igbt bricks CM400HA-12H 400 Amp 600V and I've been playing
with numbers in the Roffesoft calculator - thanks Derek. Really liked the idea
of the huge secondary with the 122uF tank cap, that resonates at 5kHz and takes
out small countries with a single strike :)
More likely though is:
160mm diameter, 600mm winding of 0.315mm wire, modest toroid or IKEA
salad bowls. Single turn primary. One brick handling a peak current of 1000A,
and hoping for a two or three foot discharge.
A few questions though: Is the purpose of the charging choke purely to boost
the voltage, or is it also serving to "isolate" the tank circuit when it resonates.
If it was just about boosting the voltage then I would prefer to use say a 220:50v
500W transformer, primary across the mains secondary in series with the live to
boost to 270V i.e. 400V DC. now a reversed diode from the igbt to the + supply
could prevent any danger of over voltage (but it's never that easy). How do I
work out how large a choke is needed?
Can I really just increase the bang energy by having several copper pipe
primaries in //? How would I calculate the inductance if I used copper sheet?
Can the primary be wrapped around the bottom of the secondary form i.e.
just below the secondary winding or is the coupling likely to be too small.
How do you stick your salad bowls together?
Finally the easy one as the universe started in a big bang before which there was
no mass, energy, space or time, (or even a before) and as the rate of the
expansion of the early universe had to exact (to better than one part in ten to the
power one hundred) for galaxies to form then hasn't God made his existence far
All the best,
Hi Alan, haven't heard from you in ages!
Is the purpose of the charging choke purely to boost
>the voltage, or is it also serving to "isolate" the tank circuit when it
It does both. It also makes the system more efficient. The reason is that
you are charging a capacitor and if you did it through a resistive ballast
you would waste 50% of the power. With the inductor you get almost all the
power into the capacitor. (that's also why it ends up charging to twice the
>How do I
>work out how large a choke is needed?
A good rule of thumb is to choose it so that it resonates with the tank
capacitor at a frequency equal to half the break rate. But it's not
critical, a factor of two either way seems ok. If you make it too small
(like 10 times smaller than this) the capacitor will charge to more than
twice the supply voltage due to boost converter action. I lost a few IGBTs
If you make it too big the current will be continuous and this can cause
problems if you stop the IGBT drive suddenly.
>Can I really just increase the bang energy by having several copper pipe
>primaries in //? How would I calculate the inductance if I used copper sheet?
Putting several primaries in parallel lowers the inductance a bit so you
can use a bigger tank capacitor. But the reason I did it was because I had
such a high RMS primary current that a single 10mm pipe would have
overheated. I started out with a single 25mm pipe but gave up because I
couldn't bend it without kinking it.
>Can the primary be wrapped around the bottom of the secondary form i.e.
>just below the secondary winding or is the coupling likely to be too small.
Well it worked fine with me. Both my OLTCs had k= about 0.15. I used stubby
secondaries though, a tall thin secondary would not have so much coupling.
>Finally the easy one ...
>hasn't God made his existence far
Well if the rate of expansion of the universe hadn't been just right, we
wouldn't be here to wonder at how just-right it was, so it doesn't really
prove anything either way. Hooray for the anthropic principle. But I can't
help feeling that the design of the universe shows signs of... something
far too deep and meaningful for me to understand.
When the coil fires, the current will split between the primary
(pretty much a dead short) and everything else. Considering the
primary resistance is about zero, that is where 99% of the current
will travel no matter what. So just having "anything" between the
charging circuit and the primary will have enough resistance to force
the primary current into the primary coil. However, you may want to
worry about the primary oscillations damaging the charging circuit.
A coil is nice since you can resonate it with the primary capacitance
and about double the voltage (the DC resistance has to be low for good
Q which is a big factor!). But if you have enough voltage available
already, that is not a concern. The coil does provide a nice low loss
way to charge the primary caps and does provide good isolation when it
fires. If you do not do resonant charging, it's value could be just
> Can I really just increase the bang energy by having several copper
> primaries in //?
That lowers the DC resistance and the inductance which allows higher
currents in the primary which is our big goal.
>How would I calculate the inductance if I used copper sheet?
I don't know the methods right off anymore, but others do.
> Can the primary be wrapped around the bottom of the secondary form i.e.
> just below the secondary winding or is the coupling likely to be too
There are programs to calculate the coupling for just about any
situation like MandK.
Antonio has programs that can do it for multiple parallel primary tubes.
> How do you stick your salad bowls together?
Metal tape. They have aluminium tape here for heating pipes or copper
tape can be purchased too. Welding or soldering the thin stainless
steel needs to be done by someone very skilled!! I have destroyed a
many trying it myself. The metal tape is by far the best IMHO...
It looks like this choke is just too useful to take out, the problem that I was trying
to solve was how to avoid putting 660V across a 600V igbt. Maybe all I need is a
variac and self control :) And maybe a warning light set for 500V and a crowbar
set for 550V.
> >How do I
> >work out how large a choke is needed?
> A good rule of thumb is to choose it so that it resonates with the tank
> capacitor at a frequency equal to half the break rate. But it's not
> critical, a factor of two either way seems ok. If you make it too small
> (like 10 times smaller than this) the capacitor will charge to more than
> twice the supply voltage due to boost converter action. I lost a few IGBTs
> that way.
> If you make it too big the current will be continuous and this can cause
> problems if you stop the IGBT drive suddenly.
Good that's straight forward.
> >Can I really just increase the bang energy by having several copper pipe
> >primaries in //? How would I calculate the inductance if I used copper sheet?
> Putting several primaries in parallel lowers the inductance a bit so you
> can use a bigger tank capacitor. But the reason I did it was because I had
> such a high RMS primary current that a single 10mm pipe would have
> overheated. I started out with a single 25mm pipe but gave up because I
> couldn't bend it without kinking it.
It looks like suck it and see, I'll need to dig out my old physics texts I'm
sure there was stuff about calculating the inductance of wires and sheets
no it probably means doing calculus, I'll play and measure instead.
> >Can the primary be wrapped around the bottom of the secondary form i.e.
> >just below the secondary winding or is the coupling likely to be too small.
> Well it worked fine with me. Both my OLTCs had k= about 0.15. I used stubby
> secondaries though, a tall thin secondary would not have so much coupling
Yes it will give a starting point and then I can play about.
>When the coil fires, the current will split between the primary
>(pretty much a dead short) and everything else. Considering the
>primary resistance is about zero, that is where 99% of the current
>will travel no matter what. So just having "anything" between the
>charging circuit and the primary will have enough resistance to force
>the primary current into the primary coil. However, you may want to
>worry about the primary oscillations damaging the charging circuit.
>A coil is nice since you can resonate it with the primary capacitance
>and about double the voltage (the DC resistance has to be low for good
>Q which is a big factor!). But if you have enough voltage available
>already, that is not a concern. The coil does provide a nice low loss
>way to charge the primary caps and does provide good isolation when it
>fires. If you do not do resonance charging, it's value could be just
This stuff is good - Steve add it to your oltc faq.
I can't avoid the charging choke, nor have I been able to come up with a new
alternative topology. if its doubling the voltage then at least I know where
I am. Steve you were doubling the mains with the capacitor doubler,
then you had the charger doubling again were your bricks 1200V?
I think I'll give it a twirl.
All the best,
Rather than a wasteful crowbar,
you can just sense the voltage (voltage divider + comparator, SIDACs, zeners etc.)
and 'fire' the IGBT as soon as the voltage reaches 600V
This would produce maximum power throughput.
>Steve you were doubling the mains with the capacitor doubler,
>then you had the charger doubling again were your bricks 1200V?
Yes. The voltage doubler didn't quite double because it had a big choke in
the mains input to improve the power factor. The DC resonant charging
didn't quite double either because of kickback (see below) So in the end it
charged its tank capacitor up to about 950 to 1000v. The crowbar was set
for 1100 or thereabouts.
The IGBTs in my mini coil were 900v and it operated between 550 and 650v.
That was with a plain rectifier (no voltage doubler) but a resonant
charging inductor. I imagine if you used an inductor with a bit of DC
resistance, the voltage would stay below 600V all right.
About kickback: The DC resonant charging interacts with the tuning of the
Tesla coil itself. You only get twice the DC supply voltage if the coil is
perfectly in tune and has low coupling. Otherwise there is a kickback that
cancels some of the charging voltage.
I'm now thinking along these lines:
I should avoid combinations of choke value, cap value and breakrate which
give maximum voltage rise.
A fairly low resistor, say 0.5 ohm will drop the voltage sufficiently.
A passive system, even if it wastes some power, which ensures that the voltage
cannot get to 600V is preferable to an electronic system which may trigger either too
early or too late.
I think I'll build it and then make some measurements.
>Nice - I don't really like the idea of the crowbar
The system that Sulaiman suggested would probably work all right. The only
disadvantage I can think of is that you don't control the break rate
explicitly any more.
So what will the break rate be? Well remember one rule of power electronics,
the average voltage across an inductor is always zero. The voltage at one
end of the charging inductor is the DC link voltage and the voltage at the
other end is (on average) half what you set the firing circuit trigger
So in practice I reckon the break rate would increase until the DC link
voltage (minus whatever IR drop in the inductor) was dragged down to half of
the firing voltage. This implies you can adjust break rate by changing the
DC link voltage, provided there is some sort of resistance or volt drop in
This argument only holds for continuous current in the charging choke (i.e.
bps > 2/sqrt(charging inductor*tank capacitor)) but if you think about it,
it's not possible for this circuit to run at break rates below the
continuous current threshold. If the tank capacitor voltage hasn't reached
the trigger point by the end of a half cycle of resonant charging, it's not
going to get any higher after that.
Given that the DC resonant charging doesn't start working until the IGBT has
fired once, and the tank cap voltage may not go high enough to fire the IGBT
until the DC resonant charging has started, the question of how to get the
circuit going also needs some pondering. Turning on the DC supply suddenly
might be good enough.
>I should avoid combinations of choke value, cap value and breakrate which
>give maximum voltage rise.
There are no magic values because the de-Qing diode kills the resonance. The
voltage rise goes up monotonically as you make the choke smaller.
>A passive system, even if it wastes some power, which ensures that the
>voltage cannot get to 600V is preferable
The simpler the better. The mini OLTC was done this way by just choosing
devices with lots of headroom. I started out along these lines for my OLTC
II but I could not think up a passive system that would control the voltage
tightly enough, while still allowing me to operate over a wide range of
break rates. So I used thyristors to adjust the DC link voltage, a meter to
show the peak voltage across the IGBTs, and the crowbar as an emergency last
resort. The idea is that you set the gate driver running at the desired
break rate and then turn the voltage up as high as you dare.
The OLTC II seems complex but it is no more complicated than it needed to be
to do what it does safely and reliably (ish)
I'll try and get some more stuff added to the OLTC FAQ.
I've been playing with numbers in the Roffesoft calculator - thanks Derek.
Glad its been of some use, I use it quite a lot. If you get any strange
results compared to reality, please feed them back to me. It works provably
for my coils, but not for Steve's for some reason. This is most likely down
to latitude ??
RE CM400HA-12H 400 Amp 600V
Be careful how close to the limit you drive those bricks, Spikes become a
nightmare, limit the supply to 500V with loads of transient suppression,
mine and Steve's 1200V bricks are both clamped at around 1Kv for some
Re Salad bowls,
I use a 1/2 strip of conductive tape, I've looked at welding them,
but the rims of the bowls I use are not accurate enough in either shape or
diameter to make a good job, so tape is much easier..
I use copper bar as my primary, and I do have different results from similar
windings in wire, inductors is parallel are easy enough to calculate, but
wide bar or tape is not covered in many of the formula I have found, and the
ones I have tried give even worse results..
Re Inductance of a bar primary...
All of the calculations I have done with the "normal" formula for
primary inductance have given answers that are too high, when given the
"width" of a copper bar primary. I have just tried another method. If you
convert the dimensions of the stock you are using into a x-sectional area
and then divide by PI to give the diameter of a wire that would have the
same x-sectional area, I get much more realistic results. This seams to work
in Win Tesla and my own OLTC Calc.
This is only a theory, and I only have one primary made from bar to
test it with, Does anyone else have similar problems with non circular
x-section primaries, Does this help ?
"and then divide by PI"
Equivalent radius = Sqr(X * Y/PI) where x & Y are the x section of
I've now got a whole week off, sadly my wife has to work so I'll just
have to amuse myself coiling.
I've got the secondary wound, 605mm * 160mm, 0.3mm wire.
Primary is going to be a single turn probably 10" diameter, 3 to 5 loops
of 10mm copper pipe stacked vertically. (I'll start with 3 and add more
loops if required to change the primary coupling)
As always its a balance between coupling and flashover. I don't have a
suitable diameter of plastic pipe for the primary, I suppose I could use
a stack of plastic flower pots or similar, suitably trimmed.
6" secondary 10" primary coupling too loose, or too close for comfort?
>6" secondary 10" primary coupling too loose, or too close for comfort?
I think it would be about right (k= 0.15 about) if you had the bottom turn
of the secondary level with the primary. That rarely causes a problem with
flashover because the bit of secondary near the primary is more or less at
I don't think flashover is so much of a problem with OLTCs anyway because
of the relatively low bang energy. Although my mini OLTC does sometimes arc
from a point about 1" up the secondary to its tank capacitor, if I run it
without the breakout point. I think this is because the mini OLTC secondary
got pretty badly
The big OLTC did flash over a lot but that was due to a sharp point on the
strike rail. A streamer would burst out of the secondary about halfway up
and hit this point. (or vice versa?) I fixed that by putting a "corona
ring" on the offending bolt. But if I run with no breakout above 100bps,
the streamers do sometimes come looping down from the toroid and hit the
primary. That always seems to trigger the crowbar and shut everything down.
I've never lost an IGBT due to flashovers on either coil. The best thing
about the OLTC is that by the time the energy finds its way up to the top
there is no current left in the primary to do damage. Unlike a DRSSTC where
both circuits usually ring up together.
>I need to track down a source for 500V transorbs which I'll put across the
>igbt's. Not sure what I need.
I think when you get up to brick sized IGBTs the possible transients are too
fierce for any transzorb or MOV available off the shelf. If you're using my
driver circuit then the R-C-D snubber combined with the crowbar is supposed
to negate the need for a transzorb.
On my smaller coil I used two 400V transzorbs in series across each 900v
I've been thinking about the OLTC circuit and it seems to me that the main problem is
protecting the igbt's from over voltage, either from the charging choke or some
kickback or spike from the secondary coil.
An alternative to the charging choke might be, a 400V-450V supply for 600V devices.
Then replace the choke with another IGBT plus a 2 ohm resistor. This IGBT is normally
switched on, to charge the cap, and then switched off say 1uS before firing and
switched back on say 10uS after firing. The internal reverse diodes, perhaps
supplemented by a pair of faster diodes now limit the output to the rails.
The stress on the high side igbt's will be the high current at switch on, that switching
loss would be considerable at hi bps.
Have I missed something obvious?
>Have I missed something obvious?
Not really. The only drawback is that you lose 50% of your input power in
the 2 ohm resistor. (you always lose 50% when charging a cap through a
The really elegant way to do it is to use a switched mode cap charging power
supply. A simple flyback converter would do (a boost converter won't work
since it can't output less than the input voltage, and a buck converter
won't work unless you provide cycle-by-cycle current limiting) or a H-bridge
running along the same lines as Marco DeNicolai's Thor CCPS.
This gives the ~95% efficiency that you get from charging the cap through an
inductor, but without the voltage instability of resonant charging. You set
the peak voltage you want and the converter turns off as soon as the cap
reaches that voltage. The only problem is that it's a lot more complicated
than a diode, an inductor, and a pair of well crossed fingers.
There are a lot of unknowns in the method when you use it with a single turn
primary. The way you arrange the capacitors and the connecting wires
(including the wire going through the CT) can have a big effect on the
inductance. The leakage inductance of the CT can have an effect too, I'm not
sure how big.
In the case of a single 4" turn I imagine the connecting wires will add up
to a good deal more than the primary itself.
You need to take care to arrange all the components just the same as they
will be in the final coil (which means for minimum inductance) Or scale from
my results and Derek's to get a ballpark kind of idea.
I have a 10" primary with one turn that resonates at 66kHz with 14uF. The
inductance will change as the square of the diameter and the square of the
turns. but the frequency changes as the square root of the inductance (and
the square root of the capacitance)
So a 5" single turn would resonate at 132kHz with 14uF, about 190kHz with
7uF, and a 5" two turn would be back at 66kHz again.
Seriously- stabilising the charging voltage of an OLTC is a tricky problem
and I haven't seen any simple elegant solution yet.
The more I think about it the more I like Sulaiman 's idea of having the IGBTs
fired by the tank cap voltage. To deal with the starting and breakrate
control problems, we could make it an either/or thing, so it fires off a
bang if it gets a trigger pulse from the break generator, or if the tank cap
voltage gets too high. The effect would be the same as putting a safety gap
in parallel with a rotary gap, whereas triggering off the tank cap voltage
alone is like having a static gap.
It's a good deal better than my system where an overvoltage just crowbars
the IGBTs on permanently and blows all the fuses. Here the coil would just
keep going and burn up the excess voltage by letting off extra bangs. I
think you could easily modify the OLTC II board to do it.
Yes I like the idea of combining the two approaches, Its very easy
with my design to use the OV trip to trigger the monostable to issue a
pulse. The only issue I can see is If you have any interference issues, your
coil will run away, e.g. coil fires, makes HV pulse, HV pulse triggers coil
>The only issue I can see is If you have any interference issues, your
>coil will run away
Well what's the worst that could happen? I reckon it would be to blow the
power supply fuse (or breaker or whatever) which is what a crowbar would do
anyway. And you might get some well impressive sparks when it goes flying
off at 13,000 bps.
> Seriously- stabilising the charging voltage of an OLTC is a tricky problem
> and I haven't seen any simple elegant solution yet.
> The more I think about it the more I like Alan's idea of having the IGBTs
> fired by the tank cap voltage.
that was Sulaiman's idea.
To deal with the starting and breakrate
> control problems, we could make it an either/or thing, so it fires off a
> bang if it gets a trigger pulse from the break generator, or if the tank cap
> voltage gets too high. The effect would be the same as putting a safety gap
> in parallel with a rotary gap, whereas triggering off the tank cap voltage
> alone is like having a static gap.
Is there a third bit to this? after the caps have fired there must be a minimum
time required for the voltage to rise particularly if the choke hasn't had a lot of time to
energise. If the safety can't retrigger for say 100uS that might also avoid the gap firing
all the time syndrome that Derek spoke of. We would also be able to hear the change
in tone when this happened.
Another possibility would be to switch on an extra safety igbt across the igbt's with a big
non inductive resistor is series - an active dampener - this could be a faster smaller igbt.
Even 20 ohm should be enough - it can be reset by the next firing. Of course if it fires
too often it will fry the resistors. Maybe a counter - and a disconnect relay
It would have to be coupled to a suitable warning. How about a solenoid activated
broadside of party poppers? Lets really test those brown trousers!
I've finished the primary, I've gone with 6 // single turns of 10mm copper pipe, 26cm dia.
Hopefully that will sink the inductance to 4 or 5 uH and give max current at under 350V
instead of 450V. 1600A is a silly current to be aiming for! (2 of 400A bricks). Copper
bus bar came today and I'll really have to work at getting the induction out of my wiring.
> It's a good deal better than my system where an overvoltage just crowbars
> the IGBTs on permanently and blows all the fuses. Here the coil would just
> keep going and burn up the excess voltage by letting off extra bangs. I
> think you could easily modify the OLTC II board to do it.
I really don't want the crowbar!
>If the safety can't retrigger for say 100uS that might also avoid the gap
>all the time syndrome that Derek spoke of.
If you're modifying one of my boards, they have a feature that was
originally meant to help them deal with interference, but could be handy for
this. They ignore any trigger pulses that arrive while the bang is in
progress and for a short time afterwards. That has the effect of limiting
the breakrate to about 1k bps no matter how fast you pump in trigger pulses.
You can adjust the "ignore time" by changing a capacitor.
The schematic for this board is here. This is a slightly older version,
there is a mistake in the power supply section
>Is there any mileage in the active dampener idea?
I'm not too sure. I suppose it would work fine in theory but I think you
could just end up frying the resistor and then being without protection. I
looked at stuff like that like using arrays of MOSFETs and light bulbs to
burn off excess voltage but for some reason I decided against it (can't
>If I take out Q6 which applies the crow bar to the driver fets and take the
>to the trigger input (pin 4 ic5b) via a diode and a 100 ohm resistor, I
should be in
The only thing is that the crowbar comparator is designed to stick on and
needs the power cycled to reset it. You'd need to replace D6 with a largish
resistor. And the LM339 has open collector outputs so it couldn't drive into
100 ohms with a 1k pull-up resistor. You might need some ORing diodes
somewhere in there to OR the trigger transformer and the crowbar signal.
>Anything to do with putting 45V into ic1, the 317T?
No, the LM317's voltage rating is the differential between in and out pins
(it doesn't see ground) so that's ok. The mistake is that I forgot to put in
the resistor between the LM317 out and adj pins- so it doesn't work at all!
>just run the test inside your fridge
>with your fridge is immersed in liquid helium at +4 degrees K.
I'll have BOC fill my kitchen with liquid He and send you the bill :-D If I
dangle above it Mission Impossible style I should be OK. Woops it's after 12
Science quiz. What are the current names for the elements that were known in
the '20s as "Uranium X2" and "Actinium emanation"? First correct answer wins
10,000 used smoke detectors and a jumbo roll of lead flashing.
I'm making progress with the otlc.
I've got the cap built - pinched the double sided pcb idea. I discovered that I could
remove a neat circle of copper and leave most of the glass fibre by using the drill bits
intended for woodworking - which cut at the sides rather than the centre of the holes.
Drill the hole with the 1mm drill, then use 4mm and then 6mm wood cutting drills.
Managed to do all the holes without cutting through the pcb.
I'm now ready to wire up the primary power circuit, well copper bus bar up actually.
I reckon its worth doing a bit of metal work to get the primary feed diameter as small as
The primary 6 // loops of 10mm copper pipe, (26cm dia) spaced 2mm apart is also
The secondary is finished 60.5cm winding length, 16cm dia wound with 0.315mm
wire apart from the top 12cm which is 0.2mm, average is 0.3mm (based on number of
2 of cm400ha-12h bricks, 600V 400A, 800A peak
Derek's program reckons that this will go at 78kHz with a 20pF top load and a 6.4uF
primary cap, peak primary current is 1592A at 500V but I'll probably stick at 400V,
1300A and push up the break rate, or increase the top load.
Has anyone got these calculations in a spreadsheet - its nice to draw up a table and
see the effects of changes.
My initial plan - was to put the power across the cap and divide the primary turns
between the two igbt's - is this necessary or will they share naturally?
Powerex IGBTs should have a matching group letter usually hand written on
them in white ink. If your bricks both have the same group letter it's
probably ok to connect them directly in parallel.
Otherwise it's good to have a bit of resistance and stray inductance in the
collector circuit of each brick to "ballast" them. NOT the emitters which
should always be paralleled as symmetrically and closely as you possibly
can. I have the two emitter terminals bolted together with a great big piece
of copper sheet.
I used to think that my split primary arrangement acted as a current sharing
transformer. I was persuaded by Neil Thomas that it doesn't and any improved
current sharing is due to stray inductance and resistance in the individual
circuits. There are more strays than a single primary would have (i.e. k from
one primary to the other is not unity) so it shares a bit better.
One last thing.. since you're using one of my driver boards, don't forget to
leave some space in the primary circuit somewhere so you can get the
Rogowski coil round a conductor. It would be a shame to have to hammer the R
coil flat so it fits between your laminated bus work :)))
I've got your board finished and will give it a test on my mini coil
tonight. And then back to the DRSSTC development program :-9
On 7 Apr 2005 at 10:40, Steve Conner wrote:
> Powerex IGBTs should have a matching group letter usually hand written on
> them in white ink. If your bricks both have the same group letter it's
> probably ok to connect them directly in parallel.
These are Mitsibushi - no obvious group letter :( But there are two styles of labels so I'll
> Otherwise it's good to have a bit of resistance and stray inductance in the
> collector circuit of each brick to "ballast" them. NOT the emitters which
> should always be paralleled as symmetrically and closely as you possibly
> can. I have the two emitter terminals bolted together with a great big piece
> of copper sheet.
The IGBT's are side by side. 2 right angle brackets go from the emitters
and butt together - the doubled bus bar is then tied to one face of the capacitor PCB.
( ASCII art is the reason that the pencil was invented )
> I used to think that my split primary arrangement acted as a current sharing
> transformer. I was persuaded by Neil Thomas that it doesn't and any improved
> current sharing is due to stray inductance and resistance in the individual
> circuits. There are more strays than a single primary would have (i.e. k from
> one primary to the other is not unity) so it shares a bit better.
the problem that I see is that the coupling to the secondary gets greater for the higher
This will be offset a bit by the coupling between the primary turns.
I would interleave to reduce this affect. Maybe I should spilt them 145 236 where
1 is the highest turn and 6 the lowest. Or the way its laid out one igbt is 1" closer to
the primary so I give the higher turns an extra 1" of busbar to go thru...
I think I'll experiment and put a small non inductor resistor between the two collectors.
If it stays cool all is shared, if it smokes then all is not well.
I'm planning to do a bit of work with the primary running from a 30v psu and scope what
its up to.
> One last thing.. since you're using one of my driver boards, don't forget to
> leave some space in the primary circuit somewhere so you can get the
> Rogowski coil round a conductor. It would be a shame to have to hammer the R
> coil flat so it fits between your laminated bus work :)))
Yes I was going to bolt rather than solder the busbars. Not sure that I could make a
good job of soldering then any way - even with a solder gun. Will the coil fit round a
20mm wide bus bar - If not I'll get the file out - shouldn't be too hard to out the centre of
a Rogowski coil.
> I've got your board finished and will give it a test on my mini coil
> tonight. And then back to the DRSSTC development program :-9
Many thanks - I look forward to your megawatt DRSSTC.
All the best,
> o||o caps
> o||o caps
> o||o caps
It looks like a very surprised smiley.. but I think I understand what you
mean.. as long as the two emitters are connected together solidly. The
reason being that they end up connected in parallel through the gate drive
leads anyway- and you don't want any significant part of your primary
current flowing in the gate drive leads. Where does the other terminal of
your cap bank come out?
>the problem that I see is that the coupling to the secondary gets greater
for >the higher
Yes that's why I interleaved mine, turns 1 and 3 connect to one brick and
turns 2 and 4 to the other brick. But it's a hassle.
>Not sure that I could make a
>good job of soldering then any way - even with a solder gun.
It's pretty much impossible to solder large areas of copper sheet with any
normal soldering iron. I used an oxy-propane torch to do mine. An ordinary
blowtorch would probably have worked but you know boys and their toys!
>Will the coil fit round a
>20mm wide bus bar - If not I'll get the file out
Yes- it's just a piece of plastic tubing wrapped with copper wire and then
heavily heatshrinked- the tubing is 10mm diameter and when made into a ring
by joining the ends the "donut hole" is about 50mm. So you can bend and
squish it to various shapes.
I tested the board last night and all seemed to be well, it successfully
made sparks. I still have to check the crowbar- is there any particular
voltage you want it set to?
>Many thanks - I look forward to your megawatt DRSSTC.
Me too but with a little unease- maybe if I wear two pairs of safety goggles
at once? This one is only going to be a tabletop(ish) unit with sparks about
10 bean tins long. If all goes well I'll look at converting the OLTC II next
which should get me to the 25 bean tin mark ;)
Oops I should have posted this picture instead-
It gives a better view of the metalwork.
I have the oltc gate drive on breadboard and ran it tonight.
I scoped it with the one good input on my tek scope(waiting on paypal
buyer protection result) I have put a pic under colins gear at bottom
well I知 concerned that it looks too good! I was running 100Hz to
400Hz input from the signal gen at 8 volts pk/pk square wave.
the ouput is in the pic. if I dropped below 6.5V on the signal
input the wave waveform looked terrible. oh I should say I知 driving a
100nF cap to simulate a brick through a 10 ohm resistor.
I realise igbt waveform wont look like this due to slow turn off of
the device and such. this waveform looks promising though with long
lead lengths and tracks on the bread board.
let me know if I知 missing something really obvious please.
I知 using the ixd414 chip for this at 18 volts.
>well I知 concerned that it looks too good!
It's because you're looking at a whole cycle of the 400Hz wave. The
interesting stuff all happens a lot faster on a time scale of less than
microseconds and you can't see it.
Try zooming right in by using a faster timebase. Something like 500ns/div.
You'll only see the rising edge (to see the falling edge, change the
trigger mode to falling) Another thing you can do on that scope is to set
the sampling mode to peak detect and it will catch every single spike and
piece of grot.
Also try setting your generator to give short pulses instead of a 50/50
square wave, so you have a waveform that looks more like what you'll get in
service. If it's got a duty cycle control turn it down as far as it'll go.
I've just built another OLTC II driver and I took some shots of what the
output looks like driving a CM600 brick (taken at the gate and emitter
terminals with a 10x probe- the gate is getting 24v)
rising edge- about 800ns rise time
what the whole pulse looks like (note 50us/div)
This driver circuit is fairly slow particularly in the turn-off time. But
it still works fine.
>So my question is - are very fast rising and falling edges required?
You're right, they don't really have to be all that fast. At turn-on you
just need to ramp up the voltage fast enough to keep ahead of the current.
i.e. if the current starts when the gate reaches 3v, and the peak current
needs 20v of gate voltage to carry it, then you have to get the gate to 20v
before the current reaches its peak.
My coil resonated at 67kHz- one quarter cycle of that is 3.7us so the
turn-on is more than fast enough.
The turn-off can be as slow as you like and in fact there might be some
advantage to having it slow. If it turns off at the wrong time the spike
won't be as severe. However if it takes more than about a half cycle it'll
mess up the timing. By the time the device gets round to turning off the
current will have reversed and be flowing the wrong way.
>I've got a 6 amp circuit and I'll have 10A fuses is that enough if the
It's not going to hurt the driver board or IGBTs whatever happens. The
IGBTs probably wouldn't even notice a couple of hundred amps short circuit
current from your power supply. So you just use whatever fuse is needed to
protect the power supply and charging circuit components (diodes chokes
resistors etc) The inductance and resistance of the charging choke helps a
lot in taming the short circuit current.
I had a 10 amp fuse in the DC supply on my OLTC II and it never blew once!
Usually when the crowbar triggered, the protection IGBT would try to
interrupt the current, fail short circuit, and then the thyristor drive
would cut out on the next half cycle- if this failed to cut out then the 20
amp circuit breaker in the AC input would go. Bizarrely the 13 amp fuse in
the plug never blew either.
>Might get some low voltage testing done this week.
Have fun! I'm away from Wednesday till Sunday but I'll try to get some
documentation sorted before I go. If I don't here are the most important
things I can think of
All the trimmers work backwards- anticlockwise to increase- due to me not
knowing left from right :(
The "Rogowski Coil" LED lights up if the Rogowski coil got enough signal to
operate on the last bang. If there have not been any bangs since power up
the indicator is undefined. (it usually lights up in this case which can be
The trigger input needs a pulse of about 12v which shouldn't be longer than
a us or two or the pulse transformer could saturate. It is totally isolated
from everything else by the transformer.