SSTC Controllers- PWM "dead time" allowances? / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

SSTC Controllers- PWM "dead time" allowances?

1 2 next

Move Thread

LAN_403

Bosko

Thu Mar 12 2009, 07:16PM

Registered Member #2021 Joined: Wed Mar 11 2009, 07:06PM
Location: St. Paul, Minnesota- USA
Posts: 4

I have recently been reviewing PWM motor drive designs, and in comparison to SSTCs, questioned the lack of Q1-Q2 switching dead time in PWM IGBT gate driver designs. Wouldnâ€™t you experience some â€œshoot-thruâ€?

To re-phrase the question, usually with PWM Motor controllers there is 2-10uS dead time designed into the switching scheme of the top & bottom half-bridge IGBTs. This is to prevent both top and bottom IGBTs from possibly conducting at the same time due to the IGBTâ€™s T-ON & T-OFF delays. I had noticed that in all SSTC designs, the opposing drive pulses to both bottom & top switch occur at the same time. Would there not be a chance of momentary conduction due to the inherent device switching delays?

Your comments/opinions would be greatly appreciated.

Thanks,

Bosko

Wolfram

Thu Mar 12 2009, 10:38PM

Registered Member #33 Joined: Sat Feb 04 2006, 01:31PM
Location: Norway
Posts: 971

Welcome to the forums.

In SSTCs, we normally use bipolar gate drive, and this is very effective in reducing shoot-through, more information on this can be found here:

.

Some times, diodes are also used in parallel with the gate resistors, to make the discharging of the gates faster than the charging.

Anders M.

Steve Conner

Fri Mar 13 2009, 10:10AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

The 2-10us deadtime is required because motor controller IGBTs are much slower than the devices we use in Tesla coils. This is partly due to the IGBTs themselves being big slow bricks, and partly because they are hard switching an inductive current, which leads to lots of stored charge and a long recovery time.

Any IGBT that needed 2uS of turn-off time would be pretty useless in a SSTC. The small coils that are most popular with hobbyists normally run at over 200kHz. Resonant operation ala DRSSTC helps them turn off faster than the datasheet would suggest.

MOSFETs don't have minority carrier effects, so they turn off as fast as you can discharge the gate, no matter what drain current is flowing.

Bosko

Fri Mar 13 2009, 11:41AM

Registered Member #2021 Joined: Wed Mar 11 2009, 07:06PM
Location: St. Paul, Minnesota- USA
Posts: 4

Thank you for the informative responses, it makes sense.

Bosko

Boomerang

Fri Mar 13 2009, 01:18PM

Registered Member #80 Joined: Thu Feb 09 2006, 11:36AM
Location: Melbourne, Australia
Posts: 4

This is quite an interesting subject, especially if IGBT characteristics are considered as distinct from MOSFETs (because their structures are not identical). As Anders.M points out, bipolar gate drive does assist significantly for MOSFETs and it does help with IGBTs, however that is not the whole story, especially when brick IGBTs are concerned. Most IGBTs turn off slower than they turn on, especially brick IGBTs and with these, shoot-through is still present.

The average total current due to shoot-through can be evaluated by driving an IGBT H-bridge under CW conditions (ie not Interrupted) at normal operating DC supply voltage but with NO LOAD connected, while measuring the total DC supply current into the bridge. Strictly speaking this measurement also includes the dv/dt current which charges the IGBT Collector-Gate Miller capacitance, however this is usually less than the shoot-through current and there is not much that can be done to reduce it in any case.

With no dead-time in the controller, it is not uncommon for the DC average of the Cross-Conduction current to be up around 5% or more of the rated DC current of the IGBT Bricks (Fuji Electric Devices IGBT Module Application Manual). For example under 50kHz CW driving conditions at 300V DC supply and with no load connected to a Brick IGBT H-Bridge, the author has measured 6.5A total current at the DC bridge supply. This represents Cross-Conduction and dv/dt Miller capacitance charging current for four 150A DC manufacturer rated IGBTs. With dead-time adjusted correctly in the controller, the DC supply current dropped to around 1.5A, representing the residual dv/dt current. With no dead-time and with CW driving conditions, this bridge would be dissipating 300V x 6.5A = 1950 Watts before a load is even connected !

The one saving grace of DRSSTCs is that they are normally operated in Interrupted mode with a typical duty cycle of around 0.03 (based on a PRF of 120Hz and T on of 250usec). In the example above this means the average DC supply current due to no dead time Cross-Conduction is around 6.5 x 0.03 = 0.2A and the Cross-Conduction power dissipation in the IGBTs is around 60 Watts total. Thus, on power dissipation grounds it is tempting to discount the need for dead-time.

However, remember that Cross-Conduction is actually the current pulses that flow during the small interval of time when two of the bridge IGBTs are conducting simultaneously (one is turning off and the other is turning on). Thus our 6.5A DC current is actually the average of many narrow but high peak current impulses. Assuming these are around 0.1 usec wide, at the 50kHz driving rate the 6.5A average translates up to current peaks of 6.5 x 10usec/0.1usec = 750A peak. Larger operating current IGBT Bricks and higher DC supply voltages will have proportionately larger peak currents!

Using deadtime to get rid of cross-conduction now introduces another problem, namely, the DRSSTC controller has a larger overall loop delay and the IGBTs now don't switch at zero load current. It is however, possible to have your cake and eat it too by introducing a delay compensation feature into the controller. The author's large DRSSTCs use this configuration and it works well, albeit with a bit more complexity in the controller.

speakerguy79

Tue Mar 17 2009, 12:46AM

Registered Member #1460 Joined: Thu May 01 2008, 12:45AM
Location:
Posts: 19

If you look around there are several 600VDC capable half bridge controllers with dead time set to 0.5us or below. L6384 and IR2104 come to mind.

Arcstarter

Tue Mar 17 2009, 02:01AM

Registered Member #1225 Joined: Sat Jan 12 2008, 01:24AM
Location: Beaumont, Texas, USA
Posts: 2253

Well, a gate resistor can add some dead time, while helping eliminate ringing on the gate. In fact, i had a 10 ohm gate resistor on one mosfet of a 300-350khz sstc, and 3 ohms which i though was 10ohms on the other. This blew the mosfet with the 10 ohm gate resistor very fast. I was Bad heating. This obviously means too little/much dead-time on one of the mosfets which causes shoot-through. I switched the resistor, and it does not blow anymore.

Bosko

Tue Mar 17 2009, 11:37AM

Registered Member #2021 Joined: Wed Mar 11 2009, 07:06PM
Location: St. Paul, Minnesota- USA
Posts: 4

Is the "Large DRSSTC Controller" schematic mentioned by "Boomerang" posted anywhere on the net?

Boomerang

Wed Mar 18 2009, 10:35AM

Registered Member #80 Joined: Thu Feb 09 2006, 11:36AM
Location: Melbourne, Australia
Posts: 4

At the moment I don't have any controller information posted, however I do have my own design notes. I'll try and make some time to tidy up the functions mentioned here if there is interest.

Bosko

Wed Mar 18 2009, 12:11PM

Registered Member #2021 Joined: Wed Mar 11 2009, 07:06PM
Location: St. Paul, Minnesota- USA
Posts: 4

Thank you, I would appreciate that very much.

1 2 next

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