Driving an h bridge to only flip voltages back and forth

Thread Starter

aeftimia

Joined Sep 16, 2011
18
I am trying to learn how to use power mosfets (n channel) or igbts to invert a dc voltage, but I only want to flip the voltage back forth--no dead time. I would connect the gates together in a diagonal fashion:
1 2
X
3 4
gates 1 and 4 are connected and gates 2 and 3 are connected, but I was under the impression that would break the high side transistors. If tying the gates together like that is not an option (unless I am mistaken) what else can I do? I was looking to find a method that would scale to work with off 120V@15A after I feel comfortable at lower power.
 

crutschow

Joined Mar 14, 2008
34,285
You need a small amount of dead-time, otherwise there will likely be a small interval where both transistors on the same side of the bridge will be on, creating a direct short between the power and ground. This seriously stresses the transistors and can zap them.

You best bet is to use a driver specifically designed to drive an H-bridge that has built-in dead time. The dead-time can be short, on the order of a fraction of a microsecond, but it is necessary for reliable operation of an H-bridge.
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
What about the small amount of time lag that naturally occurs while in switching the transistors? I assume that is generally not sufficient?
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
For example, consider the transistors during their linear region of operation. For example, on could either switch two transistors on at the same time as switching the other two off, wasting ~E*integral from x=0 to 1 of x(1-x)dx
=1/2-1/3=E/6,
where E is the energy normally wasted while switching the transistor from off to on to off again. The alternative is to switch the other two transistors on, then switch the other two transistors off. This would waste ~2*E*integral from x=0 to 1 of x dx=E

So I am ~%83 more efficient per switch if I switch all the transistors at once rather than switching them two at a time. I also cut my switching time in half, but I figure that is probably not that big a deal as long as the time I spend not switching them is much larger than the time I spend switching them.
 
Last edited:

praondevou

Joined Jul 9, 2011
2,942
On/Off switching times of commonly used IGBT/MOSFET are in the few tens to hundreds of nanosecond range.
What's the problem with introducing a few us long deadtime? This can be done digitally or via RCs.
Btw, you cannot just tie upper and lower FETs gates together, unless you work with P and N FETs and the power supply voltage is not higher than the max allowable gate voltage.

If you tell us your application we can help you better.
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
I am working on a full bridge version of the zvs driver as seen here: http://www.teravolt.org/zvs-driver/
I would like to either build an induction heater or a flyback driver (I think I can get away with using a very similar driver for either.) However, I like to learn from the bottom up, and I am more interested in optimizing a circuit that drives power across a transformer as efficiently as possible than any one application in particular. I would like to test the design by trying it as an induction heater driver or flyback driver.

So, if I burn out a couple of transistors in the process, that is ok. I am looking for the optimum design so I can modify it into something more practical later.
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
Could I drive a high side and low side transistor at the same time if I connect a potential difference across gates alone? Example:
1 2

3 4

Connect a potential difference V across gates 1 and 3 and -V across gates 2 and 4. That should (I think) cause either 1 and 4 to turn on or 2 and 3 to turn on--assuming all are either n or p channel mosfets.
 

praondevou

Joined Jul 9, 2011
2,942
Could I drive a high side and low side transistor at the same time if I connect a potential difference across gates alone? Example:
1 2

3 4

Connect a potential difference V across gates 1 and 3 and -V across gates 2 and 4. That should (I think) cause either 1 and 4 to turn on or 2 and 3 to turn on--assuming all are either n or p channel mosfets.
Can you please post a H-bridge picture, define if you want to use p/n channel or only n-channel FETs, give them numbers and tell us what will be the DC-BUS voltage and switching frequency? It gets much easier with a schematic.

There are quite some things to consider but one of the main things is to make sure that you never turn on the lower and upper FETs of one leg at the same time. This is what deadtime is for.
 

crutschow

Joined Mar 14, 2008
34,285
For example, consider the transistors during their linear region of operation. For example, on could either switch two transistors on at the same time as switching the other two off, wasting ~E*integral from x=0 to 1 of x(1-x)dx
=1/2-1/3=E/6,
where E is the energy normally wasted while switching the transistor from off to on to off again. The alternative is to switch the other two transistors on, then switch the other two transistors off. This would waste ~2*E*integral from x=0 to 1 of x dx=E

So I am ~%83 more efficient per switch if I switch all the transistors at once rather than switching them two at a time. I also cut my switching time in half, but I figure that is probably not that big a deal as long as the time I spend not switching them is much larger than the time I spend switching them.
It isn't the normal switching energy that's the issue. It's the large energy dissipated when two transistors on the same side of the leg are on at the same time. That can generate a very large short-circuit current pulse and high dissipation in the transistors.

If you want to blow transistors by not putting in any dead time, go for it. But that's certainly not an optimum way to drive the bridge. :rolleyes:
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
I was referring to the energy wasted when electricity goes down one leg as the transistors on that leg are switching.
Also, I will post a schematic shortly. Let me draw one first.
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
Also, lets say we are switching a maximum of 300volts at a maximum of 15 amps. That should cover anything that comes either directly out of the wall or out of the wall and through a step up transformer and/or a voltage multiplier.

OK, I might be able to get 600 using both a transformer AND a CW assuming I need enough current to drive another transformer.

Say I am switching at a maximum of 100Khz. Again, I might be able to get better, but so far, that is the best the circuits I plan on using for zero voltage switching can do.

As for the transistors. I have four of these:
http://www.fairchildsemi.com/ds/FG/FGP15N60UNDF.pdf
, and I am getting four of these:
http://www.fairchildsemi.com/ds/FG/FGA25N120ANTD.pdf
 

praondevou

Joined Jul 9, 2011
2,942
I am working on a full bridge version of the zvs driver as seen here: http://www.teravolt.org/zvs-driver/
I would like to either build an induction heater or a flyback driver (I think I can get away with using a very similar driver for either.) However, I like to learn from the bottom up, and I am more interested in optimizing a circuit that drives power across a transformer as efficiently as possible than any one application in particular. I would like to test the design by trying it as an induction heater driver or flyback driver.

So, if I burn out a couple of transistors in the process, that is ok. I am looking for the optimum design so I can modify it into something more practical later.
It's kind of hard to propose something useful if the exact goal is not known....like input, output voltages, effieciency, maximum part cost,

As someone else proposed, for the 300V 15A I would use a dedicated gate driver for each leg (or each FET). There are quite some options, one of the more commonly known the IR2110. Properly layed out it will do the job.

There is quite some information about induction heating here: http://www.richieburnett.co.uk/indheat.html
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
I attached a picture of my gate driving suggestion. This is an independent ac voltage source driving the two gates of one of the legs at once. Would this work? It did not seem to on Falstad's circuit simulator, but I am a bit skeptical that this is right.

Also, thanks for that links. I am reading that paper on switching losses of full bridge zvs topologies. I have only started reading it but it looks extremely relevant.
 

Attachments

Last edited:

Thread Starter

aeftimia

Joined Sep 16, 2011
18
Yes. I suppose so. I want to start with ~120V DC (maybe more if driving a high voltage flyback) and convert it to either high current or high voltage (depending on nature of the transformer I use) AC current at a frequency such that I maximize the power transfer across the transformer.
 

R!f@@

Joined Apr 2, 2009
9,918
I asked cause as I recall, I repaired a H bridge Inverter some time ago.

That never had a transformer, as the output is taken from the H Bridge.
 

Thread Starter

aeftimia

Joined Sep 16, 2011
18
There is not anything wrong with putting a transformer on the output is there? Like I said, I am trying to maximize the power transfer across the transformer by adjusting the frequency of the inverter.
 

praondevou

Joined Jul 9, 2011
2,942
I attached a picture of my gate driving suggestion. This is an independent ac voltage source driving the two gates of one of the legs at once. Would this work? It did not seem to on Falstad's circuit simulator, but I am a bit skeptical that this is right.
An N-FET is driven with a voltage from gate to source. Yours are not. This will not work.

I also recommend reading this for the basics: http://www.irf.com/technical-info/appnotes/an-978.pdf
 
Top