BigCoil / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

BigCoil

1 2 next

Move Thread

LAN_403

dude_500

Sat Oct 01 2011, 06:46PM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

I've finally run my big coil at nearly full power, so I'll finally post it in the projects area. This is the same coil that I've been using for a year, with a larger topload and a new larger bridge.

Bridge specs:
fd600r17kf6cb2 (1700V 1000A) Full-Bridge
Floating opto-coupled gate drivers driven by switching supplies
Prediktor driver

It was designed to give operation up to 1kV bridge input, leaving plenty of headroom for voltage transients. I ran it last night up to 850VDC (put out 8 foot ground arcs great), which it blew two IGBT's at after a few seconds of running. I'm not sure exactly what caused the failure, it is two transistors on one output leg. One of the TVS strings on this set appears to have been sheared off (or alternatively blew the lead off, but it looks like a mechanical failure to me). This fault doesn't quite make sense though, since I have practically no voltage transients due to the tuned prediktor driver, and certainly never experienced anything near 2x voltage spikes. It also appears in the video that it may have hit my current monitoring BNC cable leading from dedicated current transformer to scope the moment it blew up, however I'm not sure how this would kill the coil. Perhaps some of the coil's frequency offset by pi/2 was induced backwards from CT to primary, which then caused the driver to panic? Anyone had anything like this happen before? It would be re-assuring to hear that this is likely rather than just blindly replacing bricks and hoping.

Video of operation:

Here are some construction pictures:

Starting from scratch

Etched opto-coupled gate drivers

Milling bus-plate

More bridge bussing

Current transformers on one of the bridge output terminals

Adding snubbers and DC bus caps

Logic box with optical interrupter input

Completed big-coil driver with its little brother on top (a more recent project that is not yet complete)

Patrick

Sat Oct 01 2011, 09:45PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

dude_500 wrote ...

Here are some construction pictures:

Starting from scratch

Thats a nice extrusion youve got there !!!
Where the hell did you get that piece of aluminum?
Is that A Hammond aluminum box?

Marko

Sat Oct 01 2011, 11:42PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi sir,

Failure of such beefy expensive igbt's must come as a big frustration. I'm not sure how to help other than advising to look for all possible "stupid" reasons for failure first. Gate drivers latching up? TVS strings getting shorted out for some reason?

I think many people now don't believe isolated gate drivers are necessary for diving big igbt's anymore, as a properly designed gdt will do just as well.
GDT's are nice because they make physically impossible for 2 opposing igbt's to turn simultaneously on.

Marko

dude_500

Sun Oct 02 2011, 12:29AM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

Patrick wrote ...

Thats a nice extrusion youve got there !!!
Where the hell did you get that piece of aluminum?
Is that A Hammond aluminum box?

Found it in another shop upstairs, they didn't want it, took it. College is great.

It is a hammond box for the logic. It works great to shield all that and keep it nicely enclosed and safe.

Marko wrote ...

Hi sir,

Failure of such beefy expensive igbt's must come as a big frustration. I'm not sure how to help other than advising to look for all possible "stupid" reasons for failure first. Gate drivers latching up? TVS strings getting shorted out for some reason?

I think many people now don't believe isolated gate drivers are necessary for diving big igbt's anymore, as a properly designed gdt will do just as well.
GDT's are nice because they make physically impossible for 2 opposing igbt's to turn simultaneously on.

Marko

I have noticed a lot of people not using isolated drivers. I have always passionately hated gate drive transformers, and every time I try to use them, it ends up not working nearly as well as when I break down and use driver boards. It could be something to consider though. It would also probably eliminate whatever ringing I do have since switching would be slower.

Goodchild

Sun Oct 02 2011, 05:49AM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

God god man why such big IGBTs for a coil that size?!?! I looked up the datasheet for those bricks and man with rise and fall rimes in the uS range and a Vsat of near 4V, you could get away with much smaller faster IGBTs such as CM300s or CM600s that also have mush lower Vsat voltages in the <2V range.

A note about GDT's I use them very successfully driving a full bridge of CM600HA-24H in my large DRSSTC, easy putting out 12 to 13 feet of spark.

GDTs can be your friend if you treat them right

dude_500

Sun Oct 02 2011, 06:14AM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

Goodchild wrote ...

God god man why such big IGBTs for a coil that size?!?! I looked up the datasheet for those bricks and man with rise and fall rimes in the uS range and a Vsat of near 4V, you could get away with much smaller faster IGBTs such as CM300s or CM600s that also have mush lower Vsat voltages in the <2V range.

Because overkill is the name of my game in this project. My hobby is making big precision power electronics, and unfortunately this is the largest coil I can store for any length of time at college. The size of the coil is inconsequential, it's all about the bridge for me.

Also, I have both datasheets open for these bricks versus the CM600HA-24H's I used in my last bridge, and they have nearly equal or superior specs in nearly every regard... same C-E drop (2.6V vs. 2.5V), superior recovery diode drop by 1V, half the rise and fall time, 3x less input capacitance. Only place it's a little worse is turn-off delay which isn't that big of a deal. I'd take these bricks over a CM600 any day.

Goodchild

Sun Oct 02 2011, 08:33AM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

Datasheets don't tell the complete story a lot of times. Just look at the 60N60 vs the 40N60 SOT-227 IGBTs, the datasheet shows the 60N60s having a higher current rating than the 40N60s, yet the 40N60s can be run hundreds of amps higher than the 60N60s.

The CM600HA-24H has been used by many people and has been proven to run very reliably close to the 5KA mark in a DRSSTC. Finn's coil or example pushes close to 14+ feet I think with a full bridge of those IGBTs. My general conclusion is that 1700V silicon is SLOW and the 1000A rating means LOTS of paralleled IGBTs inside of the brick making it even slower and having a large internal L. I know what the datasheet shows, but your not comparing apples to apples, if you notice in the Cm600HA-24H datasheet the switching times are for "Resistive Loads" and the switching times in the fd600r17kf6cb2 datasheet are for "Inductive Loads" both datasheet assume hard switching also.

If you compared both IGBTs in a soft switched ZCS resonate inductive load scenario (like a DRSSTC) I would put my money on the CM600HA-24Hs being faster and a lot more efficient. Even a full bridge of CM600HA-24Hs would be way overkill for a coil that size. That's why I put them in my coil (which is of similar size to yours) so it would be overkill and I would have lots of head room. I don't have to use any TVS on my bus ether because the bricks switch rather fast and there are really no voltage transients more than a coupe volts.

I'm not trying to discredit your work in anyway here, I hope I didn't come across in that way. I think your craftsmanship and design is stunning!

dude_500

Sun Oct 02 2011, 04:29PM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

Don't worry, no offense was taken towards my work... I'm just really curious to get an intuition as to why these bricks are worse for tesla coil duty. It seems that apples are being compared with apples to me, two big bricks. Having smashed open both a CM600 and one of these, they both have 8 dies. Also, these bricks have what appear to be a much lower inductance design with a unique double-sided PCB wiring design that I haven't seen used in other bricks.

Also, isn't inductive switching slower than resistive, making these specs look even better than they already did?

Goodchild

Sun Oct 02 2011, 04:40PM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

Well inductive switching times are not always worse, but can be if your like hitting the sine at it's peak or something. So you say you demolished a bridge of CM600HA-24H as well? Dang I have to hand it to you that's a hard thing to pull off hehe

You must have something else in your design that is causing brick failure, CM600HA-24H should be good for way more than 8 feet. Even CM300s could push out 12 feet if you treat them right.

OK just a thought but is your driver logic powered by the same circuit as your bridge? If the breaker trips I have heard that the driver can latch ON all 4 bricks creating a scary cross conduction. This could short the bus cap right through the IGBTs killing them rather quick. I ask this because your bridge failed right as it got a ground strike, and a lot of times ground strikes are just enough to kick the breaker over.

dude_500

Sun Oct 02 2011, 04:50PM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

I did kill two CM600's, but that was my fault. I had a crappy snubber cap that overheated and blew half the bridge.

As for the failure of this one, I don't think it was any power failure. I have feedback loops from all power supplies including optical feedback from the four isolated supplies and if any one drops a few volts below normal operating, the entire system shuts down by locking out the interrupt input. The logic supply didn't trip the breaker, anyways. Still quite unsure what went wrong this time, still thinking perhaps a voltage transient issue since the half of the bridge that burned out is the half that had a sheared off 1500v TVS string. It just doesn't make sense that it rang that high though unless someone changed my timing pots or something (this possible ring is the motivation for my perfect phase leader).

1 2 next

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