Huge DRSSTC Project / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

Huge DRSSTC Project

1 2 3 4 next

Move Thread

LAN_403

Dzejwor

Fri Mar 18 2011, 10:37PM

Registered Member #2750 Joined: Sun Mar 21 2010, 08:47PM
Location: Poland
Posts: 46

Hi. I plain to build next DRSSTC. This time something like steve's drsstc2. At first I have question about secondary and frez. Steve says on his website that he used 8inch pipe and 26awg wire and wound 2300 turns also wtites that the dimensions of toploads are 26"x4.5" and 44"x9" He writes that the resonance is approximately 45kHz but I get around 38kHz with those dimensios I think this is too low frez for this coil. Next question is MMC. I have 75pcs of CDE942C20P15K. What is better configuration of those caps - lower capacity higher voltage (5 in series and 15 strings 450n/10kV) or higher capacity lower voltage (4 in series 15 strings or more 563n/8kV with 15 strings of 4). I plain to push around 1200A in primary at 700-800V DC on the bridge. I plain to use predikter and CM300DY24 IGBTs. I don't have an idea for gate driver yet but I probably use step down GDT - something like 2:1:1:1:1 or 4:1:1:1:1 povered by intermediate mosfet bridge with another gdt. I dont want to use swithing power supplies and I'm not entirely convinced of the solid state gate drivers like steve do. Next I want to know what is the difference betwen CM300 igbts produced by Powerex and Mitsubishi. From what I usually see people use Powerex - why ? I have bricks from Mitsubishi - datasheets of Powerex and Mitsubishi are identical.

Gregory

Fri Mar 18 2011, 11:45PM

Registered Member #2922 Joined: Sun Jun 13 2010, 12:08AM
Location:
Posts: 226

Hey man, good talk about this because I am starting my new coil that is equal to your with same configuration and same IGBTs!!!

1 - The frequency never is low, the lower the better for yours IGBTs!

2 - The better configuration is the higher voltage AND the higher capacitance (if you want big arcs, for this size of coil use 500nF or more!), later I show how to calculate the MINIMUM MMC voltage.

3 - In my system I plan to use step-down GDT too: 2:1:1:1:1 with 48v at primary, it will have ~24V at gates. For me GDT is better, more simples and more reliability!

4 - I will use the same IGBT.. I don't know exactly what are the differences but I think that the powerex have more internal substrate it means that it will handle better the "overdrive". (we will run 1200A in a 600A device..)

**2** - To calculate the minimum MMC voltage you need to know the frequency you will run this, you said 38kHz I wil usel it for example:

First you calculate the MMC resistance (this resistance * the current that will generate the high voltage across it), the AC resistance of a capacitor is called *reactance* and can be calculated by this formula: ** Xc = 1 / (2*PI*F*C) **

Now you have the MMC AC resistance, using the standard ohm law you can discovery the voltage across the tank.
Knowing you said that will run 1200A (the result is more accuracy with the RMS current) and imagine you use 500nF MMC:

Xc = 1 / (2*PI*F*C)

Xc = 1 / (6.28 * 38*10^3 * 5 * 10^-7)

Xc = 8.37 ohm

Now we know that a 500nF MMC at 38kHz have an AC resistance of 8,37ohm

using ohm law:

U = 8.37 * 1200A

U = 10044V

Your MMC will have 10kV across the terminals!

Dzejwor

Sat Mar 19 2011, 11:56AM

Registered Member #2750 Joined: Sun Mar 21 2010, 08:47PM
Location: Poland
Posts: 46

So I need more caps to build 10kV/540n MMC. I stil dont know what primary form is beter for this size of coil. I wonder between the cylindrical and conical. Conical look nice but is much more dificult to wind and I don't know how to define coupling. Cylindrical is easy to made. And next question - can I use wire with double (heavy) insulation or can it cause a significant deterioration in the secondary parameters ? (I have problem with buy wire with single insulation)

Gregory

Sat Mar 19 2011, 03:48PM

Registered Member #2922 Joined: Sun Jun 13 2010, 12:08AM
Location:
Posts: 226

1 - Yeah, the better for you is buy more caps

2 - I think that in my system I will use conical primary, I think is the best relation between k and flash-over problem. For calculate you can use the JavaTC, search in google.

3 - About the wire I don't know that is the differences of double insulation

Steve Ward

Fri Mar 25 2011, 05:42PM

Registered Member #146 Joined: Sun Feb 12 2006, 04:21AM
Location: Austin Tx
Posts: 1055

Though my website doesnt show it, all of my new coil designs use flat primaries. The coupling is generally high enough, and it has the best control of electric fields near the bottom of the coil, avoiding annoying flash-over between coils.

Also correct that the Fres stated at 45khz is off, its more like 41khz in real life (the actual # of turns is not counted, just calculated, so maybe its wrong). My DRSSTC-2 now uses a flat primary coil, ~10.5" ID and 23" OD with 9 turns of 3/8" copper tube. The smaller tube is actually about the same AC resistance as the 5/8" stuff i used originally, but with much less surface area to radiate heat i now must run water through the tubing to cool it. The original MMC made from the 942C series .15uF 2kV caps, has long since failed from overvoltage. I was running strings of 4 in series, but my the peak voltage across the caps was 9kV, so this was too much stress. The coil often is used for shows now, and its MMC has a 24kVDC rating, and operates at ~11kV peak, far higher reliability.

I made an experimental coil once with a 12" tube and 24awg wire, it resonated at ~35khz and with a flat primary achieved coupling of ~.18. This coil gave better performance with 12 foot sparks at 1500A primary current and ~5 - 8 kW (real power, not kVA) at 150-300bps. It also used a 1.25uF tank capacity and a short ON time of just 4 or 5 RF cycles. This mode of operation seemed efficient for long sparks with higher BPS, below ~150BPS the spark length would drop off dramatically, which might be considered a drawback.

Dzejwor

Fri Mar 25 2011, 08:34PM

Registered Member #2750 Joined: Sun Mar 21 2010, 08:47PM
Location: Poland
Posts: 46

Nice to see your reply here Steve. Where do you get those big epson B25856 caps? I already have 90pcs of 942C and I realy dont want to kill them because they are very hard to buy and expensive in my country â€“ not only 942C but generally capacitors suitable for drsstc. Largest expense in any of my coil, it is the capacitors. I build 540n/10kV cap and try to run my coil at 40-45kHz. Your flat primary looks very good. I try to make flat for my coil but I have some questions. First I dont like square bases but this square turn look cool

I plain to build round so I make round safety turn too. I dont want to pump water through the winding for cooling so I use 15mm tubing (This is the typical diameter in my country what is near to your 5/8". The next available is 18mm). Plastic sheet on primary is only for visual or for flash-over protection ? Is there a simple method of determining the coupling ? For example when I use cylindrical primary I have around .18 if secondary starts at the same point where is the top end of primary. Ofcourse diameter of primary is important too â€“ usualy around 2times bigger than secondary in my coils. I can calculate coupling using javatc but I donâ€™t know how much it agrees with the real world.

I made some calculations using javatc and I get coupling around 0.133 with flat primary like Steve's but using 15mm (~5/8") copper and 0.3" space betwen turns - outer diameter increased to 27" with 9 turns and slightly smaller inner diameter. I have a little more Inductance. I plain to use dual toploads because I lkie them - they are look better than single topload ;) So my coil look now as follows:

primary:
ID: 10.3"
OD: 27"
9 Turns of 15mm (~5/8") tubing

Seconndary:
200mm (~8") PVC tube
~2300 turns of 26AWG copper wire
120cm (~47") height

Bigger topload:
Minor diameter: 6"
Major diameter: 43.3"

Smaller topload:
Minor diameter: 5"
Major diameter: 27.5"

Tank cap:
90psc of942C - 540n/10kV

Bridge and electronics:
2xCM300DY24
4x 3300u/450V electrolytics configured to 3300u/900V
2x 5uF/1kV Wima MKP snubbers
8or9x 100V TVS in series at all four IGBT's in bridge
Three phase supply with full wave rectifier - it should give around 600DCV with low pulsation. In future I maybe change them to the inverter supplying ~800VDC
Hammertone predikter drive with intermediate mosfet bridge and 2:1:1:1:1 GDT

Dzejwor

Thu Jun 30 2011, 10:57AM

Registered Member #2750 Joined: Sun Mar 21 2010, 08:47PM
Location: Poland
Posts: 46

Hi. I colected some stuff and little changed parameters of my ciol. Now I have 100pcs of CDE and my MMC will be 600n/10kV. I get bigger electrolytic caps - 8200uF/400V and decided to make 4:1:1:1:1 GDT not 2:1:1:1:1 as I say before. Now I have finished secondary and collected all parts of electronic. I need to make toploads and primary and ofcourse put it all together

I add some photos of my stuff

Mads Barnkob

Thu Jun 30 2011, 01:32PM

Registered Member #1403 Joined: Tue Mar 18 2008, 06:05PM
Location: Denmark, Odense C
Posts: 1968

I noticed that internal current limiting isn't mentioned anywhere in the datasheet for the CM300DY24, does this mean that it is born without it?

Kizmo

Thu Jun 30 2011, 02:51PM

Registered Member #599 Joined: Thu Mar 22 2007, 07:40PM
Location: Northern Finland, Rovaniemi
Posts: 624

Mads Barnkob wrote ...

I noticed that internal current limiting isn't mentioned anywhere in the datasheet for the CM300DY24, does this mean that it is born without it?

It does.

Ward actually tested what this brick can survive in more or less typical drsstc conditions:

Finn Hammer

Thu Jun 30 2011, 02:52PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Dzejwor,

Great to see the Predikter on a pro. quality board.
The predikter was designed to be used with discrete gate driver boards, using optical isolation on the timing signals. Due to a faulty gate driver design, I never dared to use the original gate drivers with any real power, so instead I used 1:4 gate transformer and intermediate bridge.
This solution has drawbacks, the main one being that the predikter puts out a very short, IRRC 10nS, pulse, after the last cycle in a burst. It is due to a delay on the interrupter pin, which allows the oscillator input to turn the gate driver chip "on" just briefly before it is turned "off" by the interrupter pin. This short pulse was drowned in delay on the optical link, but with a gate transformer it transfers enough energy to the gate transformer to create a huge and long ringing oscillation which looks really bad.
I fear that it could turn on the gates unsequenced.
Both Daniel's Diablo Tonnere and Thumper use this drive scheme, and we damped the oscillation with a 5W (I think 50ohm) resistor across the outputs. Not elegant, but it works..
There are 2 possible solutions, the one I prefer is to use the new Gate driver boards that I have designed, and I will let you have the design if you want to use them. Discrete gate driver boards means that you don't have to fiddle with gate resistor/diode networks, and you can really dial in the perfect non-ringing switch times.

Cheers, Finn Hammer

EDIT, I just notice that you have experience with the Predikter already. Did we talk about these issues already? Did you notice the "trapped ripple " ringing in the gate transformers?

-Finn

1 2 3 4 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob