Hello folks, I plan to use 2 semikron SKHI 23/12 boards to control 4 half bridge
1200V 200A Toshiba IGBTs. I want to use them as a high voltage- high curent switch (3000 - 5000 volts, 150 amps...low duty cycle ..50usec pulses at 50 hertz) for some experiments I want to do. I am theorising that I should be able to connect them in series from low side emitter to topside collector and get a switch that can cope with transients of up to 9600 (8 x 1200) volts. I want to get the pulses as square as possible (especially the front edge). I plan to static balance the voltages with resistors ( I have some 15KV 100M wire wound resistors...) connected across the load contacts of the half bridges. I know nothing about snubbers but I intend to learn...any good info out there? I was hoping that those with experience in these matters might be able to point out any flaws or give me any helpful suggestions they may have. What would be the best way to current limit this? The current will be passing through a non inductive load and some heavy diodes to ground. The diodes are to stop any reverse bounce of current as I want purely DC pulses if possible. I am collecting parts at present. I have the IGBTs and the SKHI 23s. I plan to build some heavy capacitors for the power supply. (I have a large roll of 23 micron mylar :D) thanks Dave
Yes I saw that. Bad luck! These parts are hard to get here in Australia. I have to get them on ebay and the shipping often costs more than the components! I will take pictures, but it will be a while before there is anything worth looking at. David
Thanks Terry, I have an article about using single iGBTs in series. I will be following that. A lot of these papers assume that you are an electical engineer. I have a physics background and I find some of the info impenetrable... I'll keep looking though David
"Conservation of misery" LOL I LIKE that! The same thing happens in programming! It means your design is screwed and you need to think about it all again... coils store energy as "current", thanks, that's a good insight gnight David
I am also messing around with Toshiba 200A IGBTs (bought from CTR surplus) and Semikron 23/12's, making a welding inverter. In fact I made a working one, that passed some basic tests at 200A, but then messed the snubber and let the magic smoke out.
I have no depth of experience, but want to suggest that you keep us posted and take pictures.
I got mine on ebay, I suspect from the same seller as you would.
They now sell a set of 4 IGBTs, without the heatsink, which could mean cheaper shipping.
They are honest, real guys (CTR surplus), they have 3 selling accounts.
It is better to start earlier... I know that I, with my minuscule knowledge, would not take on a high voltage project like that. I think that you can indeed cleverly switch 1200 voltages (perhaps made by two separate 1200V DC power supplies), so that you get +- 2400V on output.
A divider chain using high-value resistors will have little current flowing and be of limited use in balancing the voltages of series- connected switches because of the large dynamic currents involved.
Can you guide us to some of the better papers?
Yes, I have had a little to say, because I've been working with series-connected high-voltage FET circuits, and because I have had considerable experience playing with MOSFET avalanche, that being an item on my favorite-pastime list. MOSFETs happily have rather benign high-power avalanche characteristics, so that with a detailed understanding, one can safely incorporate intentional MOSFET avalanche into their designs. I suspect that this cannot be said for IGBTs. I suggest following industry practice and add a parallel stack of silicon transient voltage-suppressor diodes intended to conduct before the IGBT's voltage limit is reached.
TVS diodes have the happy property of conducting within a fast sub-ns timeframe, and they can safely handle high currents for a short time, transferring pulses of energy into massive chunks of copper that are part of their internal construction. These "zener" diodes have some self capacitance, so using a stack of them in your system can enhance any capacitive voltage-balance scheme, which is better suited to dynamic issues anyway.
You can choose small TVS parts to minimize dynamic power losses in a carefully-tuned system, or large ones to handle serious time misalignments in brute-force, wire-it-up-and-go pedal-to-the-metal setups. For example, 1.5ke200A are modest 200V, 1500-watt TVS parts, rated to handle 15A peak for 1ms (50% decayed) pulse, 100A for 10us, and 200A for about 3us. (If your crude IGBT gate-drive time mismatch exceeds these numbers, larger TVS parts can be used.)
Five small 1.5ke200A in series would make a nice 1kV limiter for use with large 1.2kV 200A IGBTs (subject to the conditions above), and since each of these 200V TVS exhibits 100pF at Vbr, the stack would have about 20pF of capacitance. This compares to 300pF to 1000pF for a Toshiba mg75q2ys50, rated at 1.2kV and 200A for 1ms.
You can choose a larger TVS, say 5kp90A, a 5kW part rated at ~105V and 47A for 1ms, or 200A for 55us. A stack of ten of these would have 40pF of capacitance at 1kV (and ~250pF max at low-voltages), and could handle almost anything you could throw at it. I've done extensive pulse testing and modeling of such parts to 300A, and used them in a 1800A FET switch I described here some years ago (that switch has been in continuous full-power use for 8 years).
In conclusion, using avalanche TVS parts like these, you can help make your 1kV modules robust, and tolerant of poor pulse timing. Yes, there are subtle issues of wiring inductance (dI/dt, oops!), thermal effects on dynamic resistance, etc., to worry about, but once one starts down a path of using TVS elements in such circuits, I believe life will become much easier. Recommended.
Conservation of Energy will go a long way - coils store energy as "current", capacitors as "voltage" resistors dissipate as do voltage drops. A very good guess of values can be worked out by hand.
There is another law in engineering called "Conservation of misery", i.e. the optimisation of one parameter set will cause another set to suffer, preventing any global improvement.
Like what - there are tried & tested approaches to a lot of things these days. There are many papers on solid state metal-vapour laser drivers, f.ex.
Maybe!
What topology do you have in mind? Does the pulse width ever need to change? Does the output voltage need to change? Does the frequency need to change? What kind of load is it? How "flat" must the pulse be?
F.ex. You can have a single big power supply, switch floating in series with load, Single big power supply, switch grounded, Maybe cascode-connected. Many transformers with primaries driven in parallel from low voltage and secondaries in series, Transmission lines.
My favourite topology for this kind of thing is the "pulse step modulator" - patented by ABB/Thomcast (The patent is on the switching algorithm used to control the voltage) - a string of switced 1 kV power supplies connected in series, each supply fed from a separate transformer winding (which must be isolated to the output voltage). But, this is not so hard with a staged winding layout.
Here is a link:
formatting link
to one. The advantage is that all components are low-voltage, common, industrial stuff and the driver, having power available, can be more intelligent.
The only real disadvantage is that the feeder transformer is a bit complex, but you need 5-6 stages with low average power of 50W which can relatively easily be fitted into a measly ETD 49 core with insulation and all. The trick is to wind all secondaries the same way, leave at least 1 mm/kv gap at each end for creapage and wire the secondaries so that the highest voltage is on the outside.
.
The squarer the pulse, the more trouble you will have!! If you want it really square then you end up with transmission lines, impedance matching and vacuum tubes. This is not fun. Then there will be the coupling to other things, proportional to dI/dt and dV/dt. Be absolutely certain about this because fast, high voltage, high current pulses will cause all kinds of misery.
I once worked with power supplies specified by physiscists (;-); They demanded a controlled risetime of 50us from 0-160kV. This requirement cost dearly in reliability and trouble with the feedback loop - all kinds of "fun" for years on dummy load. When real-life experience was gained, of course, it was found that *milliseconds* of rise time was just fine because the load only came on within the last 2% of the voltage rise anyway (of course the usec requirement stayed so that nobody would need to revise the papers "explaining" why it was so important).
I have been suspicious of specifications and scientific papers ever since.
RCD networks should work. You have to cosider the worst-case scenario where you trip the chain on overcurrent and due to spread in timing, one device will trip first. The load current goes into the snubber until the other stages trip causing the voltage to rise. That voltage needs to be below the max voltage of the IGBT. Maybe a Transorb could help, but at 150 A .... not for long.
You need a safety enclosure, Interlocks and fail-safe earthing/discharge switches or you will be dead one day. There is an IEC standard for "secondary high voltage supplies" that describes clearances and precautions in very useful terms. It is very good advice to follow the design rules in there - there is a lot of accumulated experience.
Link here:
formatting link
dbase down, so could not get number.
Measure the voltage drop across each switch and use it to trigger a two-transistor SCR that cuts the drive. Be aware that IGBT's have a "latch" mode: If the current rises too fast and too high, a parasistic thyristor will fire, latching the device on. It takes negative "gate" voltage to
*maybe* switch it off again. Hence, the drive needs more intelligence than a mere pulse transformer.
Don't bother. Buy/scrounce whatever capacitors you need. Make them as small as possible (this is where flatness of pulse comes in) because they will hold a lot of energy that really like to *kill* you.
Gentlemen, THANKYOU for your long and very helpful replies :D I am going to study your advice closely before going any further. Yes I am very aware that these devices are deadly and I will be enclosing everything live with interlocks etc. To be honest, this stuff scares me shitless, just like the big jointer and tablesaw I use in my workshop. CONSTANT VIGILANCE! to quote mad-eye moody. ( I still have all my fingers!) I am planning to have the switch floating in series with the load, which will essentially be a carbon block in series with a metal rod of various configurations, surrounded by plates and ariels. I understand what you mean about useless and trouble causing specifications- I'll just go as square as the SKHI 23 drivers will allow, until I know a lot more. What I am trying to do is duplicate (at a lower level) the high voltage DC spark-gap experiments of Tesla without the spark-gap. I want an adjustable pulse length from 100 usecs down to as low as I can go with the equipment I have. Frithiof Andreas Jensen, I will check out the pulse step modulator as an alternative topology, thanks. Win, the TVS diodes sound really like what I need to look into. Your experience will hopefully cut short a lot of the "Shrapnel" stage I would have gone through. You have given me a lot to think about and I think I will ask more when I am further along and have more intelligent questions to ask
there has been at least two papers published in IEEE trans. power electronics this century on series connection of IGBTs (for a 25kV switch). go to a decent technical library and have a good read. the problem is getting all the switches to switch simultaneously, so the voltage is shared evenly. without care, this will not happen, and one or more switches will go *bang*
Win has made a number of posts here on this topic wrt FETs this year. google....
A second method uses individual floating power supplies (e.g. charged capacitors), with half-bridge elements to switch each stage from 0 to Vx. Stages are stacked to get a high voltage.
One of my univ. buddies who works for the power generation company's associated research firm made it rule to keep *both* hands in his pockets whenever the big stuff was energized.
Metal watches can be troublesome too.
Best regards, Spehro Pefhany
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Better yet, put your left hand in a pocket. This may slow down that instinctive grab for a falling object long enough for your brain to override it.
Leather shoes often have nails holding the heal on. If the floor is conductive, check your shoes.
Also if you are working in short pants watch out that your right knee isn't resting against a metal leg when your hand touches the HV by mistake. Trust me on this.
When I was in the USAF (1968-1976), you were _required_ to take off your watch, any rings, and they seriously recommended that you take the dog tags from around your neck and put them in your pocket.
I worked for a large photographic manufacturer in the late '70s early '80s. We had an operation for making lenses that used 7.5 kW induction heaters (360 of them!) In that era they were of course vacuum tube units that ran on about 7.5 kV at about 1 amp. The tube was a Westinghouse water-cooled triode.
Anyway, there were all kinds of safety rules about troubleshooting these and of course a big shorting stick permanently attached to a cable to ground the innards when the covers were removed. You had to be certified and qualified to work on these, and needed a letter from the pope and a note from your mother. I wasn't qualified..........
On day I was watching a "qualified" yo-yo trying to fix one of these heaters. The thing was basically an Armstrong oscillator, and at 7.5 kV and
7.5 kW most troubles in the oscillator were immediately visible as black spots or missing pieces. The heater also had grid and plate current meters, so you could generally see at a glance whether it was running or not. There were a few extra chokes and capacitors in the thing so that the work coil would be at 0vdc as a safety feature.
This nut had the covers off and decided to power it up anyway. He jumps out all the interlocks (including the cooling water flow switch) and powers it up, leaving the shorting stick in place. I was standing a few feet away expecting a big bang, but nothing happened for about a minute. Then, there was a spray of water and sparks and excitement all over the place!
He had put the shorting stick on the wrong side of the output coupling capacitor. The thing was oscillating, but the output energy was being directed through the shorting cable which was carelessly wrapped around the quick-disconnects for the cooling water. After about a minute it melted the hose off the fitting, letting the water spray all over. Just what you need in close proximity to a 7.5 kV power supply!
Good thing the guy was qualified. I shudder to think what an UN-qualified individual might have done...
Win, Tesla got some very interesting results with HV DC pulsing across spark gaps with ALL AC components suppressed. He had an HV dynamo charging a large cap bank which he release across a magnetically blown out spark gap. He found that he got a stinging "slap" at a distance of meters that stung his face and hands in particular when the spark length was greater than 100 usec. Under this period the pain disappeared, but he got strong electron charging effects in copper conductors nearby. He called it radiant electricity. As far as he was concerned, he had disproved Hertz's transverse electromagnetic experiments ( he went to visit him and found errors in his methods), and he modelled the effect as a longitudinal pulse in the "ethers". I wan't to try to reproduce these effects without a spark gap if possible as they are a pain in general. Tesla found that the effect was much stronger when he used a capacitor upline of the spark gap, hence my use of the "BFS" (BF switch) topology. As to the safety aspects, I am using an old but functioning microwave oven transformer as the power source. It seems to me that the best way to use it is inside the oven casing with all the safety interlocks still functioning. I have removed the magnetron and intend to replace it with a large capacitor ~ 10uF at 3000 volts. (I am building it with offset conductors in a coil formation so that the whole edge of each conductor is exposed on opposite sides to allow large current flow and eliminate inductance). There is also a large squirrel cage impeller fan to cool the magnetron that I will use to cool the IGBTs instead! I also have the parts for a 1 metre long HV probe that I will be using to take the measurements from a safe distance when I have to have the cover removed. (10 x 100M resistors 3 " long in series with a 1 M resistor inside a 1 metre PVS conduit) Do you know of a good design for a capacitive voltage divider to use with the CRO? The tip about nails in the shoes is a good one,. I intend to build a discharger onto the large cap and follow a set procedure after every construction test run as part of the test. I have had several 240 volt shocks in the past due to accident or stupidity (very unpleasant!) and I am very aware that a 2400 volt shock at 1 amp will kill me outright. Thanks David
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.