RE: SSTC questions / body diode stuff

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Jan Wagner by way of Terry Fritz <teslalist-at-qwest-dot-net>" <jwagner-at-cc.hut.fi>


On Thu, 13 Mar 2003, Tesla list wrote:
 > Original poster: "Justin Hays by way of Terry Fritz 
<teslalist-at-qwest-dot-net>" <pyrotrons2000-at-yahoo-dot-com>
 >
 > Hi Everyone,
 >
 > Jan, I see your point, but will still have to kindly disagree with
 > you.
 >
 > Only after the freewheeling current spike has been dissipated,
 > it is safe to turn on the other MOSFET (in half-bridge) or pair of
 > MOSFET's (full bridge). If the MOSFET's are still turned on when the
 > diodes are still conducting...you get an "ON" DIODE on top, and an "ON"
 > MOSFET on bottom! This equals shoot-through current, similar to
 > having the wrong MOSFET's turned on at the same time.

Yup, precisely. That's what I tried to say too.

 > This is why I
 > think dead time could be added to allow slow diodes (body diodes) to
 > fully turn off before the next switching transistion happens. Indeed,
 > Trr goes up significantly with increased current, but the current
 > spike itself is so short, I think this effect can be dismissed.

Yeah, you maybe are correct!
My mistake there, with claiming otherwise. Probably. ;)
I constantly forget which way the voltage swings to support the current
flow - ok, in the opposite direction, so when the upper switch turns off,
the pri end which is now free swings negative and the lower freewheeling
diode conducts. Placing close to 0V accross the lower switch
(drain-source, collector-emitter, ...), btw. This lower switch is next to
turn on, and there are zero volts accross it, i.e. the long-sought ZVS.
Soft recovery of the diode into the switch. The requirement for all this
is that the driver is exactly at f_res.

If it is slightly out of tune. Hmm. Phase shift in the current, and the
freewheeling diode of the currently activated switch will conduct at some
stage (while the switch is still on). Switch turns off, current still
flows through the same diode, no pri voltage swing, 350V or-what-you-got
is accross the lower switch. Current keeps flowing through the upper diode
all the time during the drive dead time (until the pri current naturally
changes direction, but let's assume for now that the dead time is too
short for this to happen). Soon, the lower switch turns on. Transition
from 350V accross it to 0V accross it => forced recovery of the upper
freewheeling diode and potentially dramatic shoot-through if that diode
is "slow-recovery".

So the the only way to prevent this (and still use slow diodes) is to get
a good estimate for the maximum deviation of the drive frequency vs the
true resonant frequency. Then you'd know how much the dead time has to be
- i.e. enough time to allow the pri current to change polarity during this
dead time.

A (potentially unlucky) guess:

   t'=abs(1/f_res - 1/f_drive)
   dead time = 0.5*t'/T_drive_period   (*100 to get %)

  should work both below and above f_res, with dead time either after or
  before the ON time

Or maybe I'll have to think this over all again - arrrgh.... ;-)

I hope the above mental arithmetics / simulation wasn't too dense and
everyone lost track on it... Didn't run any software simulation on it,
and there could be some errors. Please feel free to correct me!

 > Also:
 >
 > Contrary to popular belief, oscillator-based SSTC's do NOT switch in
 > a ZVS fashion, even if the system is in perfect tune. The signal from
 > the control IC (TL494 in many designs) is not locked onto the
 > secondary coil output. The two signals drift in and out of phase in
 > respect with each other...the end result is that MOSFET's switch all
 > over the place in respect to the sinusoidal output of the resonator.
 > And since this current/voltage relationship is directly coupled to
 > the primary, you see what happens. Only feedback or PLL coils switch
 > in a ZVS fashion. And only PLL coils can switch at *perfect* zero
 > crossing. Why? Big MOSFET gates cause an RC time delay with feedback
 > coils. This has the tendency of pushing things out of phase, and more
 > towards non-ZVS. The bigger the MOSFET gate capacitance, the more
 > non-ZVS in a feedback (antenna) coil.

If you mean zero current crossing, ZCS, then definitely "yeah" to what you
wrote.
ZVS is switching ON with zero volts already accross the switch (lossless
turn-on). If one adds a small amount of dead time at, errm, the beginning
of each turn-on (delayed turn-on) you should get ZVS. Or was that, "before
turn off (earlier turn-off)"...
feel free to comment


happy coiling!
cheers,
  - Jan

--
*************************************************
  high voltage at http://www.hut.fi/~jwagner/tesla
  Jan OH2GHR