Short answer, IGBTs are much better but not as good as MOSFETs.
In many cases they're good enough that the turn-off tail can be
ignored with very minor accommodation if any. In some cases an
IGBT may be superior than a MOSFET because you don't have to worry
about the switches RDSon drifting apart with temperature.

Though now days you can buy a 75V 120A 0.0004 Ohm MOSFET in a
T0-220 package... I guess RDSon isn't much of an issue.

The main problem with both IGBT and regular old bipolar transistors
in a push-pull circuit is the turn-off tail. The MOSFET does not
have a turn-off tail. There are two classes of IGBT, punch-though
and non-punch through. The punch-trough devices have better turn-off
times but are more fragile. Lately I've been using Trench Field Stop
IGBT's and they're very good.

There are several means of preventing imbalance of the transformer
in driven (not self oscillating) converters. The PWM will adjust
the on-time to compensate for the tail as it regulates the output
voltage. You can sense the differences with a circuit that converts
time to voltage (a capacitor and a current source) then make the
correction.

There are clever flux balancing windings that can be added. One
Unitrode app note describes how this can be done in the course of
presenting a half-bridge power converter. I can't recall the
document number.

In the half bridge and full bridge sometimes a capacitor in series
with the primary wdinding prevents saturation. I think you could
build a two capacitor divider across the input voltage and at the
center connect your transformer centertap lead. Then as the
imbalance increases the voltage at the center tap with shift
to compensate for it. I've seen half bridges built this way...
might work for a push pull... just guessing as I've never tried it.

A very small gap (0.001-0.003") will prevent saturation if the
imbalance is minor and not decrease the magnetizing inductance
too much. Sometimes any decrease is unwelcome though.
A distributed gap material like powdered-iron, koolmu, MPP
or sendust, might be useful if you expect to have flux imbalance
problems.

Pulse by pulse current limiting will mask the problem, so the
transformer is in saturation but not far into it and the current
limit keeps it from destroying the switches. Kinda risky to rely
on this alone but it's helpful combined with other measures.

In a current fed converter the transformer may saturate and then
switches are effectively connected directly to the current source.
No harm done! A current fed push-pull is a rugged topology. The
VAX8800 computer uses one for its control and start-up power supply.


I was reading up on push-pull topology of switching power supplies
and see that they have problems with flux imbalance. I used to work on
some power supplies that were push pull when I was in the USAF and the
driver transistors were always failing, now I know why. I see that
this isnt as much of a problem for FETs, How about IGBTs.


Jimmie

On Sep 18, 8:56*pm, Grumpy The Mule wrote:
Short answer, IGBTs are much better but not as good as MOSFETs.
In many cases they're good enough that the turn-off tail can be
ignored with very minor accommodation if any. *In some cases an
IGBT may be superior than a MOSFET because you don't have to worry
about the switches RDSon drifting apart with temperature. *

Though now days you can buy a 75V 120A 0.0004 Ohm MOSFET in a
T0-220 package... I guess RDSon isn't much of an issue.

The main problem with both IGBT and regular old bipolar transistors
in a push-pull circuit is the turn-off tail. *The MOSFET does not
have a turn-off tail. *There are two classes of IGBT, punch-though
and non-punch through. The punch-trough devices have better turn-off
times but are more fragile. *Lately I've been using Trench Field Stop
IGBT's and they're very good.

There are several means of preventing imbalance of the transformer
in driven (not self oscillating) converters. *The PWM will adjust
the on-time to compensate for the tail as it regulates the output
voltage. *You can sense the differences with a circuit that converts
time to voltage (a capacitor and a current source) then make the
correction.

There are clever flux balancing windings that can be added. *One
Unitrode app note describes how this can be done in the course of
presenting a half-bridge power converter. *I can't recall the
document number.

In the half bridge and full bridge sometimes a capacitor in series
with the primary wdinding *prevents saturation. *I think you could
build a two capacitor divider across the input voltage and at the
center connect your transformer centertap lead. *Then as the
imbalance increases the voltage at the center tap with shift
to compensate for it. I've seen half bridges built this way...
might work for a push pull... just guessing as I've never tried it.

A very small gap (0.001-0.003") will prevent saturation if the
imbalance is minor and not decrease the magnetizing inductance
too much. *Sometimes any decrease is unwelcome though. *
A distributed gap material like powdered-iron, koolmu, MPP
or sendust, might be useful if you expect to have flux imbalance
problems.

Pulse by pulse current limiting will mask the problem, so the
transformer is in saturation but not far into it and the current
limit keeps it from destroying the switches. *Kinda risky to rely
on this alone but it's helpful combined with other measures.

In a current fed converter the transformer may saturate and then
switches are effectively connected directly to the current source.
No harm done! *A current fed push-pull is a rugged topology. *The
VAX8800 computer uses one for its control and start-up power supply.



*I was reading up on push-pull topology of switching power supplies
and see that they have problems with flux imbalance. I used to work on
some power supplies that were push pull when I was in the USAF and the
driver transistors were always failing, now I know why. I see that
this isnt as much of a problem for FETs, How about IGBTs.

Jimmie- Hide quoted text -

- Show quoted text -

I work with a couple of pieces of equipment that synthesises a
repetitive waveform by playing back the waeform from a ROM. I thought
I could do something like this to control the on-off timing of the
IGBT. This would set a minimum time between turn on and turn off and
the rest would be controlled by PWM.

This is one advantage of the flyback and asymmetrical forward
converters. They will reset if there's enough dead time.
So in those cases your ROM circuit would do the trick.

Symmetrical topologies like push-pull, half-bridge and
full-bridge can saturate even if there is plenty of dead
time. The core is always being driven by the control
circuit, so it has no time to relax, If the drive isn't
equal and opposite for each half cycle there is an offset
which the core accumulates. Eventually it will saturate.
unless there is some means to compensate for the imbalance
of the drive (like a coupling capacitor.)
So here the ROM circuit will not help.



wrote in news:8e344554-a8d0-4796-ae20-
:

I work with a couple of pieces of equipment that synthesises a
repetitive waveform by playing back the waeform from a ROM. I thought
I could do something like this to control the on-off timing of the
IGBT. This would set a minimum time between turn on and turn off and
the rest would be controlled by PWM.

On Sep 19, 11:57*am, Grumpy The Mule wrote:
This is one advantage of the flyback and asymmetrical forward
converters. *They will reset if there's enough dead time.
So in those cases your ROM circuit would do the trick.

Symmetrical topologies like push-pull, half-bridge and
full-bridge can saturate even if there is plenty of dead
time. * The core is always being driven by the control
circuit, so it has no time to relax, *If the drive isn't
equal and opposite for each half cycle there is an offset
which the core accumulates. *Eventually it will saturate.
unless there is some means to compensate for the imbalance
of the drive (like a coupling capacitor.)
So here the ROM circuit will not help.

wrote in news:8e344554-a8d0-4796-ae20-
:



I work with a couple of pieces of equipment that synthesises a
repetitive waveform by playing back the waeform from a ROM. I thought
I could do something like this to control the on-off timing of the
IGBT. This would set a minimum time between turn on and turn off and
the rest would be controlled by PWM.- Hide quoted text -

- Show quoted text -

I think I get it. Would this explain why in a push-pull topology after
the transistors have been replaced a couple of times the power supply
just keeps failing for no apparent reason?

Are you saying that in other topologies it doesnt saturate or that it
doesnt matter if it does?

Ordered Abe's book.

Jimmie

Hard to say. Sometimes other parts are wounded and cause the
transistor on-times to be slightly different. Often a base
drive component, usually a resistor, changes value. Some
lousy designs just won't work without selected transistors
required to match the on-times. It only requires a small
imbalance for the push-pull transformer to accumulate enough
flux to eventually saturate. If the core is steel or powdered
iron which can be magnetized by the fault current of the first
failure sometimes (rarely) that causes problems too.

The other topologies we've discussed are more forgiving. BUT
if the switch is on for too long, or the voltage applied to the
winding too high, causing the flux density that the transformer
can sustain to be exceeded, it will saturate. The current will
then rise quite rapidly the sparks will fly.

It's just that they're not senstive to slight variations in the
on-time of the switch. They reset the transformer completely
during the dead time, so they don't accumulate any flux from
on-time imbalances.

An exception is current fed symmetrical topologies which are
just as senstive to imbalance. If the transformer saturates
the fault current is controlled by the inductor feeding the
converter and the current gradually increases. So the control
circuit can catch the fault before the transistors are turned
into lumps of glass.

Some topologies rely on saturation in order to function, like
the royer (and some forms of blocking oscillator supplies) where
saturation removes the positive feedback base drive and allows
the switch to turn off. Nasty things but sometimes useful for
low parts count, low-power, converters.

I'm sure you'll find Abe's book helpful. I still re-read it from
time to time.



I think I get it. Would this explain why in a push-pull topology after
the transistors have been replaced a couple of times the power supply
just keeps failing for no apparent reason?

Are you saying that in other topologies it doesnt saturate or that it
doesnt matter if it does?

Ordered Abe's book.

Jimmie

I think I get it. Would this explain why in a push-pull topology after
the transistors have been replaced a couple of times the power supply
just keeps failing for no apparent reason?

Are you saying that in other topologies it doesnt saturate or that it
doesnt matter if it does?

Ordered Abe's book.

Jimmie

Well I remember repairing Sony TV push pull SMPS I made a living from
replacing many a blown PP pair. Sony then went to PP pair in a single
package. That reduced the business for me but I had much experience
repairing them already. Trick with the Sony push pull was 2% timing
components. The PP had to be within 2% of 50% duty cycle. They used a
self starting multi vibrator design. After replacing the blown parts
I'd power the input up at 20vac and use a 12vdc supply for the start
up circuit, Then check the waveform on a scope to make sure it was
with 2% fo 50% duty cycle. There was no dead time in the Sony's. They
just varied the frequency to regulate the voltage.

73

n8zu

Topologies like push-pull, half-bridge and full bridge don't
require dead time to reset the transformer core. Though
that doesn't mean the switch's conduction times can overlap
which causes shoot-though current.

I think what kills the push-pull in this case is overlapping
conduction times not core saturation.


For amplifier power supplies it would simplify things to
do the regulation at a lower voltage and keep the HV parts
at a minimum. Phase controlled 60Hz switching is ok but
this might be a better way.

There's a use for push-pull or half-bridge or full-bridge
where the switches duty cycles are not modulated and the
frequency is fixed. The switches run as close to 50% duty
cycle as possible without overlap. It's called a "DC
Transformer." It's one of the building blocks of compound
converter topologies. Handy because it offers isolation
and a fixed ratio of step-up or step-down with a DC input
and output.

Since there's minimal dead-time and no output inductor is
required. The efficiency can be very high. The control
circuit is an oscillator running at 2F (Like a 555,) followed
by a flip-flop and a couple of gates to insure there is never
overlap. When using MOSFETs an RCD network on their gates
will work, though I favor using logic gates.

So if you built one of these with a 10:1 ratio you could
put 200VDC in and get 2000VDC out. Any regulation or
protection would be done to the 200VDC input. This might
not be a bad idea... your 2KV output stage now consists
of only rectifiers and a capacitor.

Doesn't have to be 200V, pick the voltage that makes it easy.

Just a thought.



raypsi wrote in news:f60045d7-f5a2-4dbe-a850-
:

Well I remember repairing Sony TV push pull SMPS I made a living from
replacing many a blown PP pair. Sony then went to PP pair in a single
package. That reduced the business for me but I had much experience
repairing them already. Trick with the Sony push pull was 2% timing
components. The PP had to be within 2% of 50% duty cycle. They used a
self starting multi vibrator design. After replacing the blown parts
I'd power the input up at 20vac and use a 12vdc supply for the start
up circuit, Then check the waveform on a scope to make sure it was
with 2% fo 50% duty cycle. There was no dead time in the Sony's. They
just varied the frequency to regulate the voltage.

73

n8zu