[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [TCML] Inverter dead times



Hi Herwig,

See my comments below:

On Tue, Aug 7, 2012 at 6:37 AM, Herwig Roscher <herwig.roscher@xxxxxx> wrote:
>
> Hello !
>
> In order to understand inverters for SSTCs better, I've studied AN1045 of International
> Rectifier and found in chapter "Full Bridge Inverter":
>
> "...All the IGBTs should not be OFF at once during switching, since the power source
> feeding the inverter continuously supplies current. It follows that a pair of IGBTs
> should be turned ON a time delta t before turning OFF the other pair; time delta t
> depends on turn-on and turn-off times associated to the devices (normally is some
> hundreds of nanoseconds long). As a result, for a finite short time, a cross-conduction
> of current takes place through the legs and the IGBTs must withstand it. During this
> interval of time, the only harmless path for the reverse current is through a “free-
> wheeling” diode and an IGBT..."

This sounds correct ONLY under one special condition, and that is that
the bridge is operating from a current source (inductor) rather than a
stiff voltage source (capacitors).  In this condition, the reason for
switch conduction overlap is so that the "flyback" voltage from the
inductive power source will not exceed the switch voltage rating and
cause failure (or excessive losses by avalanching MOSFETs).

This overlap in switch conduction is not common practice in SSTC
design.  Most SSTCs that ive ever seen have been voltage source
inverters, which means they have a DC capacitor that feeds them.  In
some cases it may only be a few uF of film capacitors, other cases it
may be kJ's of electrolytics.  In this case, you really want to be
sure that you *dont* have overlap in conduction between switches.
More on this later...

>
> In chapter "Half-Bridge Inverter":
>
> "...In this circuit, cross conduction must be accurately avoided. X1 and X2 must never be
> ON at the same time. This is accomplished by switching X2 ON a time delta t after X1 is
> switched OFF. Again, this blanking time depends on turn-on and turn-off times associated
> to the devices.

This is what id expect to hear.

>
> There is *no* freewheeling current. D1 and D2 are protection diodes which ensure that V
> EC is always well below the safety limit, in particular during switching transients. Very
> low current diodes can therefore be used, provided they are fast enough..."

Id have to read the AN to understand why there is no freewheeling
current, but in the case of a SSTC, there will absolutely be
freewheeling current!  Due to this, you want to take considerations
when choosing the anti-parallel diode for the switching device.

>
> My questions:
> 1. Is it really advisable to tune full bridge inverters having a short shoot-through
> time?

In general, NO.  If you know what you are doing, then maybe ;-). I
have seen some "current fed" inverter designs that work this way, but
ive never seen a SSTC designed this way (though its a pretty
reasonable thing to do, honestly... if you know what you are doing, if
not it can get ugly fast)

> 2. What is the correct design for the free-wheeling/protection diodes and could
> suppressor diodes be used (very fast) ?

Suppressor diodes are not a good choice to use as a wheeling diode,
they really arent made for much current carrying capacity, and if they
do have to absorb voltage transients, then they might already be
running too hot to also handle forward current.  You simply need an
ultra-fast recovery diode.  In the US, we have the MUR____ series
(like MUR3060, a 30A 600V diode).  The diode wont have to see as much
current as the switch usually, so its current rating can be
considerably smaller, like 1/2 the switch current at most, and maybe
as little as 1/10th the current rating of the switch.  It would be a
good idea to use both a TVS suppressor diode and an ultrafast wheeling
diode.

> 3. Need the IGBTs in a half-bridge inverter to have a V CES of 600 V, if the rail voltage
> is 400 V dc ?

Yes.  Each switch sees *at least* the 400VDC.  When you consider the
inductance of the circuit, the voltage stress across the switch can
easily be 2X the input voltage if designed poorly.

I should point out that IGBTs are often "co-packed" with a suitable
fast diode, and are great to use "as is" for a SSTC.  MOSFETs, always
have a diode as its part of the MOSFET itself.  Sadly this MOSFET body
diode, is not a very good diode, certainly not as good as an optimized
ultra-fast diode.  Its suggested that you add a schottky diode to
block the reverse current through the mosfet body diode, and then add
an ultrafast diode in parallel with the mosfet-schottky diode string.
The schottky diode only has to be rated for like 20V as it only gets
reverse biased by as much voltage as it takes to forward bias the
ultra-fast diode.  The schottky diode does however have to conduct all
the MOSFET current, which can be a lot, so the diode will need
heatsinking probably.

You can, however, control your switching so that the diodes never have
a forced recovery, in which case i think its safe to use the internal
mosfet diode without bypassing it.  Basically, you make your switching
lead the primary current.  If you think about how this works, you will
see that the diodes shut off when the switch in parallel with them
turns on, and this is considered a "soft-recovery" and i think its
very low loss.

Steve

>
> Looking forward to reading your comments,
>
> Herwig
> _______________________________________________
> Tesla mailing list
> Tesla@xxxxxxxxxx
> http://www.pupman.com/mailman/listinfo/tesla
_______________________________________________
Tesla mailing list
Tesla@xxxxxxxxxx
http://www.pupman.com/mailman/listinfo/tesla