Usage of deadtime in brick-driven DRSSTC's / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

Usage of deadtime in brick-driven DRSSTC's

Move Thread

LAN_403

Marko

Fri Apr 01 2011, 07:52AM

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

Hi guys,

Being completely out of sync with current status of DRSSTC technology, what are the most recent findings regarding the actual existence of shoot-through danger?
The last I remember was that it was "being researched" in finn hammer's "Predikter Driver" PDF, whether it is important at all. Like have any significant decrease in losses been observed?

Since isolated gate drivers have generally fallen out of favor to gdt's, applying dead-time through them is difficult. I thought of using a separate GDT for each device and two pairs of gate drivers, but can't stop thinking of it as an pointless overcomplication. Any thoguhts?

Marko

Goodchild

Fri Apr 01 2011, 06:07PM

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

I find a GDT not that hard to implement even on a large coil with big bricks. I use one large GDT usually about 2 inch OD and about 10 to 12 turns. I drive all 4 bricks with this one GDT. In my new big DRSSTC I drive 4 CM600HA-24H using this method and phase lead.

I find that shoot-through is not really much of a problem even with large bricks. I just use a little trial and error in finding the correct gate R. I usually start low at about 1.5 to 3 ohms and work may way up to get the best results. Every brick and coil has it's of personality so it's hard to know the correct gate R before you run the thing. My big DR with CM600s uses 5.1 ohms on the gate, same with my smaller CM300DY-12H twins. If I remember correctly the CM300DU-24F seemed to like 3.3ohms.

Generally speaking the slower the brick and the faster you try to switch it, the more gate R you are going to need. looking at the Toff characteristic of a brick will give you a good idea of how much dead time you will have to add to the Ton characteristics.

Just my point of view on the subject

James

Fri Apr 01 2011, 06:43PM

Registered Member #3610 Joined: Thu Jan 13 2011, 03:29AM
Location: Seattle, WA
Posts: 506

Having played with HV gate drive ICs, I'm recently really starting to prefer the transformer approach. It may be archaic and low tech, but done properly it's very robust and if you blow a mosfet, you won't destroy your entire circuit as can happen when the expensive and fiddly drive ICs blow.

There's a good spreadsheet out there, I know Uzzors has a link on his site. I've had very good luck with some ferrite toroids I got on ebay. Previously I had used salvaged interference suppressor toroids and found those to be far from ideal.

Marko

Sat Apr 02 2011, 12:01PM

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

Goodchild wrote ...

I find a GDT not that hard to implement even on a large coil with big bricks. I use one large GDT usually about 2 inch OD and about 10 to 12 turns. I drive all 4 bricks with this one GDT. In my new big DRSSTC I drive 4 CM600HA-24H using this method and phase lead.

I find that shoot-through is not really much of a problem even with large bricks. I just use a little trial and error in finding the correct gate R. I usually start low at about 1.5 to 3 ohms and work may way up to get the best results. Every brick and coil has it's of personality so it's hard to know the correct gate R before you run the thing. My big DR with CM600s uses 5.1 ohms on the gate, same with my smaller CM300DY-12H twins. If I remember correctly the CM300DU-24F seemed to like 3.3ohms.

Generally speaking the slower the brick and the faster you try to switch it, the more gate R you are going to need. looking at the Toff characteristic of a brick will give you a good idea of how much dead time you will have to add to the Ton characteristics.

Just my point of view on the subject

Hi,

How exactly do you determine when your gate resistance is sufficient (for preventing shoot through) and when not? Do you power your bridge with full DC supply and scope it for dimples or you just adjust the resistance to damp the ringing and don't really worry about shoot through at all?

When one looks at slowness of IGBT bricks it might appear that shoot through would be a major problem, but it seems suppresed for some reason. Am I right to suspect that it's actually the miller effect-added charge resposible for this apparent "auto-generated" deadtime that we apparently see in tesla-coil inverters?

I'm pretty sure this happens to an extent in SSTC's but DRSSTC's do not drive lagging current so the situation is a bit different. There might actually be a same effect though, especially if a "predikter" driver scheme is used.

Marko

Goodchild

Sat Apr 02 2011, 10:42PM

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

Shoot-through for me is already taken care of via the controller. I used Steve's Ward's controller that already has an RC dead time circuit built in so shoot-though is not really a problem.

As for choosing gate R, I basically keep going until I get the best looking waveform. If you add to much gate R you will get pulses at the beginning and end of each cycle because of the the diode starting to conduct.

Marko

Sun Apr 03 2011, 04:51AM

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

Goodchild wrote ...

Shoot-through for me is already taken care of via the controller. I used Steve's Ward's controller that already has an RC dead time circuit built in so shoot-though is not really a problem.

As for choosing gate R, I basically keep going until I get the best looking waveform. If you add to much gate R you will get pulses at the beginning and end of each cycle because of the the diode starting to conduct.

I'm sorry if I was unclear from start - I was specifically interested in whether such a circuit is required or not.

Marko

Goodchild

Sun Apr 03 2011, 07:36AM

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

Ok I see what you are asking, well I guess In my system I use a combination of both preset dead time in the controller via an RC dead time and gate R on the gates of the IGBTs (which is just another RC dead time circuit).

Steve put dead time into the controller to begin with, but because the controller is designed as a "universal" DR controller you have to "fine tune" your dead time with gate R on your IGBTs.

Overall you do need some sort of dead time in your system to prevent shoot-though. This is simply because IGBTs usually turn "on" faster than they turn "off" and if you don't compensate for this the signals will over lap resulting in shoot-though.

So the answer to your question is yes you do need dead time in your system one way or another. Until there comes a day that they make switches that are perfect and turn on instantaneously.

Marko

Sun Apr 03 2011, 01:35PM

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

Hi goodchild

Well, I'm asking that just because coils without deadtime generators seemed to work perfectly well before. Though one might expect slow bricks to explode the first second after turned on, for some reason it never happened, so I was wondering whether anyone ever made a proper analysis of this instead of just speculation?

Can shoot through actually be observed by supply volzage dimpling or some other way, or as decrease in losses/heating?

To elaborate my idea regarding GDT's: I didn't like the idea using a single GDT to deliver dead-timed drive, since this only turns off the igbt to 0V, is this adequate or should I look into driving the gates to full negative voltage when off?

This is possible with GDT's if two gdt's and two pairs of drivers are used, of course at cost of increase in complexity. The voltage would "walk" negligibly to equalize V/turn over few cycles, but that shouldn't be a problem. Nobody ever tested this so perhaps it's an overkill idea if just a normal single gdt approach works fine? Any input appreciated...

Cheers,

Marko

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