Transistors for 3.6 kW inverter circuit

[cas111]

Hello all I have a fairly straight forward question about transistors. I am an engineer focusing in power generation systems, and currently working on designing a test inverter for a new type of DC generator.

I have extensive knowledge of DC circuits and power systems, however this is my first foray into the AC realm. I have already designed all required circuitry for the inverter, sans the switching mechanism. I currently have relays in place to handle the switching however they are not reliable enough for me to use. So I have been pointed in the direction of using transistors in their place, and would like to know if anyone has any pointers of where to start looking.

On to specs of the design, the inverter will be a 1 to 1 inverter meaning there will be no transformer in use. The output Voltage of the power source is 170Volts and the maximum output current is 42 Amps. The current design is using relays to flip the DC power source inputs from positive to negative to create a square wave voltage. This square wave voltage is then passed through two different 3 pole filters the first being a shunt band-pass filter and the second being a low-pass LC filter. These filters turn the square wave into a suitable sine wave output.

I have constructed a simple 60Hz oscillator circuit out of a NE555 Ti timer chip. The output of the chip is boosted and then used to trigger the relays to switch. I would like help on using transistors to directly switch the output of the generator instead of the relays.

The maximum DC power output of the generator is 7,200 Watts. Maximum Power output of the inverter needs to be at the very least 3,500 Watts. So you guys know I am very aware of the safety issues when dealing with power levels and output levels of this nature. Also the entire project is being performed in a laboratory environment, with the latest in safety equipment and regulations.

 

[yungman]

I did not design at that high power. I did one linear supply of 50A but lower voltage. I used HEXFET by International Rectifier (IR). I don't remember the exact part number, but they are big. Because it was a linear supply ( super high accuracy of 10ppm), I use linear supply and power dissipation is horrendous. We had water cool copper plate to mount the big transistors. Start worry about the cooling, that is going to be a bigger job for you...much much bigger. You'll find a transistor, I don't think what you ask for is that much and you just pair more transistors together. I think I used 6 or 8 of the big guys.

Is there any reason you use only 60Hz? Go for true switcher of 40KHz or so. There are plenty of reference circuit and MOS driver for that.

 

[Studiot]

Get hold of a copy of

Power Electronics Semiconductor Switches

by

Ramshaw.

To work at higher power levels you need practical additions to circuitry appropriate to lower powers.

Apart from the theory and practical stuff you need he has representative device data for power Transistors, GTO, IGBIT, HEXFET, Triac, diodes, SCR etc

 

[Bobbywhy]

It seems to me difficult to imagine an inverter switching 3,500 watts at 60 Hz using mechanical relays! As for the selection of solid-state switches you will find many suggestions and options at these two sites:

http://www.truepowerresearch.com/2011/03/tutorial-power-semiconductor-switches-classification/

http://en.wikipedia.org/wiki/Power_inverter

 

[Studiot]

It seems to me difficult to imagine an inverter switching 3,500 watts at 60 Hz using mechanical relays!

Before solid state, relays and tubes were all that were available at this power.

Rotary converters were and still are used for really high powers.

http://en.wikipedia.org/wiki/Rotary_converter

 

[jim hardy]

here's a guideline on using such transistors.

http://www.onsemi.com/pub_link/Collateral/AN1541-D.PDF

 

[cas111]

Thank you guys very much for the replies and reference material. The Relays are switching the power just fine however they are loud and if left on for too long will eventually fall out of sync. which is why I am looking into the transistors. As to which ones I will be using I think I have found my replacements and it is all thanks to you guys.

Have a good one guys.

 

[jim hardy]

just for curiosity's sake, let us know what you use and what hurdles you had to figure out.

I'm all for learning from other people's experience.

I would have used SCR's because that's what i am accustomed to in that size machine.
IGBT's seem to offer advantage over SCR's wrt commutation
and over bipolars wrt current gain
and maybe over mosfets wrt dissipation i was unsure,
but i never used any - looking forward to a lesson here.

thanks, old jim

 

[uart]

Thank you guys very much for the replies and reference material. The Relays are switching the power just fine however they are loud and if left on for too long will eventually fall out of sync. which is why I am looking into the transistors. As to which ones I will be using I think I have found my replacements and it is all thanks to you guys.

Have a good one guys.

Hi Cas. At the voltage a power levels you are dealing with IGBT's would likely be the best choice. Mosfets can be more efficient at lower voltage levels and/or much higher switching frequencies, and Thyristors can be better suited at very much higher power levels, but you're right in the zone there for something that can very easily be handled with IGBT's. These are reliable and readily available.

If you're not familiar with switching transistor circuits then you may not be too familiar with "high side drivers". Fortunately there are several very good IC high side drivers available these days that make the job (of interfacing the low level switching signal to the high side power devices) very easy. Google "high side drivers" to find.

Your filter banks must be rather large. In the limited dealing I've had in building high power filters, they've always turned out to be a bulky and awkward (not to mention expensive) part of the system. Once you get your basic power circuit sorted out you should look into "sinusoidal pulse width modulation" as a far more efficient way of generating your sine wave. It will drastically simplify your filter requirements.

 

[cas111]

As of right now I am looking at MOSFET's and IGBT's for the switch mechanisms. The biggest hurdle that I seem to be facing is the voltage and current requirements, in regards to cost and reliability. Temperature also is a concern, however the addition of water-cooling would not be difficult seeing as the system already has water-cooling for other components. I have also been toying with the idea of using SSR's seeing as they are simple and easy to implement, however their biggest draw back is again cost. So I will most likely be using MOSFET's as my switches they are low cost and simple to implement they also have a fairly good reliability. As to using SCR's I am worried that due to the fact that they have a latching current which needs to fall below a certain amount before turning off would not be suitable for this project. Please correct me if I am wrong.

As to the pulse width modulation idea. It has been an option floating around for a bit, however we want a pure sine wave output to reduce any harmonics to an absolute minimum, and the simplest implementation of switching and filtering possible. The cost of the filters is actually not as bad as you would think, with careful selection of inductance and capacitance values. The biggest hurdle we faced in terms of the filter design was making the filter not need a bandwidth regulation resistor for the high power side of operation. The supply device happens to be a current source so this added some simplicity while adding complexity if that makes any sense. What I mean by that is due to the relatively constant nature of the current meant that a bandwidth resistor placed in series with the output was necessary to keep the filters functioning correctly and producing the sine wave output with low resistance or high power devices.

We have also been thinking about buying a pre-fab inverter circuit that has no transformer and no signal generator, which seems to be extremely low cost in terms of prototyping and debugging. However many of these are not designed correctly to do 1 to 1 conversion of DC to AC so it seems we are stuck making our own.

 

[Mike_In_Plano]

I've worked at these power levels and greater.

Though a lot of engineers achieve these designs on printed circuit boards, there are a lot of "bad things" that happen while learning. Since you're working at low frequencies, I'd having the power devices on a PCB. Instead, use chassis mounted parts, such as the int-a-pak, or common SOT-227(ISOTOP) modules. These, with ring terminals, and good, direct wiring will better serve your cause.

Removing heat is best accomplished when simple. Heat sinks with moderated (60 to 100CFM fans) are a goal. To achieve this, you need your power dissipation to be below 60 watts and prefferably 45 watts. While you can achieve this design with SCRs and IGBTs, they will no achieve such low heat dissipation. So for this application, I think MOSFETs are best suited. Four IXYS IXFN180N25T would likely do a great job.

They will need isolated gate drivers. These can be constructed with transformer circuits, and boot-strap drivers, but I highly recommend against it - unless your ready to burn time on some catastrophic debugging.

I suggest a simple, isolated gate driver, such as the Toshiba TLP351F. Place around 47 ohms in series between the driver and MOSFET gate and drive the gate with a twisted pair of leads from the driver to the gate to minimize stray inductance.

Power each isolator seperately with a 3kv rated, 12 or 15v isolated DC-DC supply. The CUI VFSD1-S12-S15-SIP may do the trick, though I haven't tested it.

A simple signal source is the TL494 power supply controller. You can set the frequency and dead time between turn on / off of the FETs.

Capacitors are important! You can easily destroy your FETs if you fail to provide enough close-proximity capacitance to meet their current switching needs.

What typically works is to have a 1uF metallized polypropylene between power and ground in close proximity of the MOSFETs and a larger value electrolytic connected in series and mounted nearby. 2200uF to 4700uF at 250V is about right for this bad boy, depending on the impedance of your source. This big cap will require inrush limiting which is typically a big resistor in series with a time delay relay. Some people simply power the relay from the cap and it pulls in automatically when the voltage is sufficeint.

This is enough to make you dangerious. Enjoy!

- Mike

 

[yungman]

My experience with those big MOSFETs were they don't burn easily due to power dissipation. They die instantly when the circuit is unstable. They die quietly. IF you have any closed loop feedback, be careful to do the stability analysis.

In the beginning, we contracted the supply out, they came back, it was a mess, those big fets burn left and right. Nothing explode or anything, they just quietly die. If you blow a fet due to power, it can be quite dramatic!

For power, parallel fets together is not a bad thing just be careful on balancing the power between the fets. That's one thing good about using very big panel mount big fets as you can get by with fewer. Also the cooling is more efficient with water cooling. I believe we used indium film under the fet to the cooling surface to get the best heat conduction. I don't think the normal heat gasket or those grease is good enough.