Keeping DC Bus Voltage within Range 3kW Power

I need to create a circuit to keep a DC Bus (from a generator) within defined voltage limits.
It is for a test bench to measure a new Power Point Tracking algorithm for a generator, so we are trying to produce as much power as possible using the algorithm and FOC control and then burning this off through a resistor.
Present requirements are Maximum Continuous power dissipation of 3kW and DC bus voltages up to around 150V and we will control it to within +-2V. Later we will need to go to 600-750V DC bus voltages (same power dissipation).
We will be using a 2Ohm load.
I was considering a Mosfet or IGBT and a freewheelDiode and then to simplify things a fet/Igbt driver. Initial tests will be using it in a hysteretic brake system, namely on at eg 110V and off when voltage is below 108V. If this produces too much ripple we will need to PWM the chopper.
Any recommendations appreciated as I have little experience in designing and specing components for this power level.
Thanks
Peter

ps an Active Load would be lovely , but prohibitive cost wise for this power and voltage. :frowning:

750V with 20 ohm = pulses of 37.5 A
Some mosfets in TO-220 case are rated 50A or 100A, but that is not practical. Since the pins can be too thin for that. You should also have a large safety marging.

See for example the datasheet of a NGTB50N120FL2WG. That is a 1200V, 50A IGBT. the advantage is that it can have large peak currents, which is a lot safer that just mosfets. There are others like that of course.

I don't know if IGBT's can be used parallel. It is possible with mosfets.
You could for example set 10 mosfets parallel, each with 100 ohm resistor to gate, and a mosfet-gate-driver-chip to amplify the Arduino output. For example ten mosfets like the STW12NK95Z (950V, 10A) parallel.

Or use a complete module. However, I searched for a module, but they are older and have less good specificiations than a single IGBT.

With such voltages and currents, you must use a optocoupler between the Arduino and mosfet/igbt. That way the usb bus via the Arduino (with ground to computer) is not connected to the high power part.
I see it like this: Arduino -> optocoupler -> gate-driver-chip -> igbt.

Hi Peter_n.

Actually it's a 2 Ohm Load not 20 Ohm.
I have a feeling I will change the Load resistor when I go to higher Voltages to reduce the current flow significantly.
The 2 Ohm was calculated on our system to allow a 60Amp peak but more constant current of around 20-25Amps at 110V DC. (Which is why we chose a 3KW resistor).

I agree with teh opto-isolating. Fortuantely it will be connected only to a screen and NOT my PC. :slight_smile:

Was reading up about parallelling IGBT's and Mosfets and got lost to be honest. Was going to use the IR2110 Mosfet/IGBT driver as it kicks out a few AMPS which should suffice and seems fast. Also TTL level control. Still confused though on the parallelling and how I am meant to work out teh resistor values etc.

P

A IR2110 as a driver ? That seems okay. It pumps 2A into the gate of a mosfet, that is not a lot, but I think you don't need high frequency PWM anyway.

I found a schematic (the second one on that page) : Transistors–MOSFET | skaled
Here is another one at the bottom of the page : http://www.mschrod.de/Elektronik/Bauteile_Beschreibungen/Mosfet/Mosfet.html
All the mosfets parallel, but each with its own gate resistor.

Can you specify what the maximum current and voltage will be.
When you need 60A max, I would use a mosfet or igbt that can do in theory 100A or 120A or even 200A. You might have to add a heatsink.

The flyback diode has the same rating, that should be the same as the maximum current. And not a normal diode for a rectifier, but a fast switching diode, because the mosfet or igbt is fast switching.
I was looking at the site of farnell, and there are diodes like this one : http://uk.farnell.com/ixys-semiconductor/dsei120-12a/diode-fast-109a-to-247/dp/1427253
So the diode is not a problem.

Thanks for those links.

Yes on one side it seems easy and then I read the following application note http://www.irf.com/technical-info/appnotes/para.pdf and started to wonder.....

I think we will be slow PWM , up to a mx of maybe 16KHz, if it really proved necessary for smoothing the bus to satisfactory levels and making sure we don't create interference or noise. So the idea is to design for worst case if it works so I have the flexibility. Thsi is very much a one-off test rig.
60AMp peak would be the maximum we would ever need, this would be using a DC bus of around 120V. If the motor supplier confirms we can increase the bus voltage safely we will and therefore reduce the current.
IF I can do it with one design, which could take me up to 750V DC bus, but then from the resistor we'll get inductive spikes so it might be very difficult and costly to cover this extreme. (With 60AMps for the lower voltage). This was one of the reasons of thinking of parallelling many together and using cheaper and more available units.
Still cannot work out how to calculate the reistor used when parallelling up the Fets?

I assumed that 100 ohm gate resistor would do (for high current low voltage).

16kHz is a high frequency. That driver chip of 2A might not be enough for just one mosfet. Parallel mosfets with gate resistors can not be used, the resistors will make it too slow. I must say that 16kHz with such voltages and current is beyond my knowledge. Perhaps your best option is to find a IGBT that can do that. I expect it will be expensive.

At that power level a gate driver v important. 60A means paralleling devices, good heat sinking. Gate resistors perhaps 5 or 10 ohms?

When mosfets are used parallel, I would also use the driver chip parallel. So every mosfet has its own driver chip. Otherwise I see troubles with the 16kHz.
After reading about it, I feel a lot more comfortable with IGBTs.
I just learned that Infineon has an Application Note about "Connecting IGBTs in Parallel (Fundamentals)". So IGBTs can be used in parallel, nice. And they can have gate resistors as well.

So I did a search at farnell again, and sorted for a cheap IGBT. However, they have a massive capacitance at the input. Sometimes 4.7nF.
This one has 2.6nF at input : http://uk.farnell.com/fairchild-semiconductor/fga20s120m/igbt-n-ch-1200v-40a-to3pn/dp/1885744
You could use 4 parallel, to have a good margin also at higher temperatures.
This one is only 1.5nF at input. That's already a lot better: http://uk.farnell.com/infineon/ihw20n120r3/igbt-diode-1200v-20a-to247/dp/1832349
I still would use 4 of them, all on the same heatsink so they have the same temperature. Either with a very strong gate driver, or seperate drivers for each IGBT. And you need 12V for the gate drivers.

This one is similar, with 1.4nF at input : http://uk.farnell.com/infineon/ihw30n110r3/igbt-single-n-ch-1-1kv-60a-to-247/dp/2443500

Well, after all, this project is straightforward. You have to take special care about the wiring. Better extra thick wires and sturdy connections than just within the specs. I just wonder what RF energy the 16kHz will cause in the wires. Maybe someone with a pacemaker should not be around.

Thanks Peter_N and MarkT.

As far as 16KHz, what would you think was a more reasonable level, as maybe we can happily go down that route if I study it more.
Parallelling gate drivers, I was concerned that they may not switch at exactly the same point and therefore the full load might go through a single IGBT or Mosfet if it switched on fractionally before the others.
Will study the "Connecting IGBTs in Parallel (Fundamentals)".
I already have an isolated 24V DC , so will drop it down if necessary, or if teh gate driver allows 24V, then will use it directly.
I am still new with these IGBT's in a faster switching environment etc. How would I calculate the gate driver power allowing for the various capacitances, extra resistors and PWM frequency?
At those prices, maybe could design an Active Load instead............ Another project I think... :slight_smile:
Yep, extra thick wiring will certainly be the order of the day. We also have some LARGE Capacitors on the DC-bus too.
I'll see if I can measure the RF energy once it is all running. In the lab we'll only be t lower voltages and power, for live testing we'll go up to the higher levels.

I don't know how to calculate the power needed to blast current into the gate. Sorry. I know others on this forum can, with the Coulomb/Joule or things like that of the gate. You may have to start a new topic for that.

For the frequency, I was thinking about the Arduino default of 500Hz. The 16kHz is something totally different. I have seen designs for such frequencies that use a transformer or coil and a sine current for the gates, to reduce the amount of current that is wasted by filling and emptying the gate capacitance. In your case, a sine current would be wrong of course, you have no choice but to waste energy for the gate capacitance.
On the other hand, some class-D audio amplifiers use mosfets with a (very) square wave at a frequency of 500kHz.

Using the gate drivers parallel would require to know the differences in delay and the allowed peak current by the IGBT. You can soften the peak current with a gate resistor to slow down the rise time and fall time, that makes it a little more tolerant for timing differences. Luckely those IGBTs can have a large peak current.

Roughly speaking the switching time = total gate charge / gate driver output current.

So for instance 22nC charge and 0.2A gate current gives 110ns switching time. Treat this
as a rough indication though.

Power dissipated in gate driver + gate, like all clocked FET circuitry is given by f Q V,
so for instance: 16kHz, 22nC, 12V = 4.2mW.

If you have the gate capacitance C rather than charge Q, use this identity:
f Q V = f C V^2

BTW if you parallel power MOSFETs or IGBTs you have to use gate resistors or risk a differential
oscillation mode that is highly destructive (as well as to make switching shared more evenly). Well
that's going by comments in an early MOSFET databook.

@MarkT Thanks for those rule of thumbs, very useful.

Finally looking at the datasheets , trying to decide if the IGBT is Positive Temp or Negative Temp as obviously I want PT if I'm parallelling so as not to have to worry about runaway.

I am looking at the datasheet for theFGA20S120M FGA20S120M datasheet pdf and looking at page 3, I assume I will go from Negative Temp to Positive Temp at 5Amps if I look at Figure 3. Have I read this correctly? I will be operating it in Saturated mode. Is there another way to check for the temperature coeffcient?

I want to make sure that if I parallel a number of IGBT's , that I will always be in the PTC region.

Thanks again.