Homemade BLDC Drive Circuit

[jnnewton]

I have seen so many designs on the web for sensored / sensorless, H bridge / half bridge, mosfet / igbt setups for driving a motor. I need some direction as to which design would best apply to what I want to build. I want to use a dsPIC to control the speed/position/torque of a bldc that is rated at 325VDC. I want to know what the circuit will / should look like between the pwm ouputs from the pic and the motor leads. Does anyone have any direction or experience with this?

The second part of this will be rectifying 110-240VAC to 160-320VDC for the motor power. Any suggestions here would also be greatly appreciated.

 

[dknguyen]

Decide what type of half-bridge you want, the snubbers you want, and the gate drivers you want. THat's pretty much what sits in between.

 

[jnnewton]

I believe that the decision for half bridge is between mosfet and igbt correct? I found this
and it seems that the mosfet is the choice for lower voltage (under 200V) and igbt is for over 1000V. The gray area is where i am at around 300 V max. It seems that the Igbt is better at low frequency and the mosfet at high frequency. If my motor moves at anywhere between 0 and 2000 rpm, that gives a frequency of 2000*1min / 60sec = 33 hz so I would choose igbt or mosfet. How do I decide?

 

[jnnewton]

I also have found that the mosfets have a source to gate voltage of 3-30Volts when I search on digikey. To me this says that if the source is 300V, the gate signal has to be 270-297 correct?

 

[dknguyen]

I also have found that the mosfets have a source to gate voltage of 3-30Volts when I search on digikey. To me this says that if the source is 300V, the gate signal has to be 270-297 correct?

Remember that there are PINS on a IGBT and MOSFET called source,drain,gate. THe source it refers to is the source pin and not the voltage source. So no. It means that the voltage across the source-drain (the load voltage if you will) can be up to 300V. THe gate-source itself can only handle a maximum of 30V. It probably needs about 10V to turn on fully (it will say in the datasheet).

Your comment is too vague, especially at these voltages where a floating high-side driver is almost definately needed. You need to say whether this gate voltage is being referred to ground or source because there tends to be floating voltages involved to drive the high-side transistors. From your confusion about the (very important and fundamental) source and gate voltages, perhaps it's best to review a regular half-bridge (or h-bridge or whatever else contains a half-bridge circuit) first before proceeding any farther.

YOu just have to sit through your power calculations and the V-I levels required to figure out IGBT or MOSFET. MOSFETs switch faster than IGBTs BTW.

 

[jnnewton]

Ok, two transistors, connected to the motor, one connected to voltage source, the other to ground. What power calculations will i need to do? the max voltage is 325, the max current is 25A, so the max power is 8125W. How do you decide which data sheet to look at. I search on digikey, for drain to source voltage of > 325, drain current of > 26 and am left with 160 possibilities in stock. will any of these do? how can i further narrow the selection?

 

[dknguyen]

no. When they say source they mean the SOURCE PIN on the MOSFET. As in the 3 pins on a MOSFET: source, drain and gate. NOT the power source, voltage source, current source, battery or any other power supply.

So if the gate voltage must be higher than the source by 15V, this means the voltage measured at the gate, relative to the source *PIN* of the transistor is 15V.

 

[jnnewton]

The max available power is around 1000W for these mosfets. how can the current be 30A, the Voltage 500V and the power only 1000W?

 

[dknguyen]

THe power being passed through the FET Is different than the power being dissipated in the FET.

I will use a diode as an example just because the voltage drop is constant so it's easy to deal with. A diode that passes 500V at 500A might have a voltage drop of 1V.

The total power is 500Vx500A = 250kW
The power being dissipated in the diode is 500Ax1V = 500W.
THe power passing through the diode is 499Vx500A = 249.5kW.

THat number might also be taking into account the switching losses (and definately conduction losses) and other things robbing the input energy of 30A@500V from getting to the output. Learn not to trust transistor datasheets and read them critically, especially things given as a single number. YOu have to look at the test conditions and circuits they using to get that number for it to be meaninful. Most of the time I just go to the curves now to find out what I want rather than reading from the table of numbers unless I'm just doing a quick comparison between different devices trying to narrow things down.

But you should be looking at the datasheets critically for your scenario, not just those numbers from some arbitrary pulsed tests. Actually calculate the power dissipation. Because a lot of transistor datasheets say "I can handle 300V max and 50A!". But when you calculate the losses in the transistor (using current and voltage drop/resistance) and multiply it by the thermal resistance and consider the max temperature, you end up with a current far far lower than what is written.

THe chain is only as strong as it's weakest link. THe silicon might be able to handle 150A. But if the package can only handle 30A, guess what? THe device can only handle 30A. The 150A silicon is just more efficient for conduction (not necessarily switching!) and lets you handle 30A in a smaller package than normal.

 

[jnnewton]

Ok, i understand that. Thank you.

calculate the losses in the transistor (using current and voltage drop/resistance) and multiply it by the thermal resistance and consider the max temperature

This is a little vague for me. I have a data sheet that says:

25A, 330V, RDS(on) = 0.23Ω @VGS = 10V

How can I use these to get what I need. you said using current (25A) and voltage drop / resistance (10 / 0.23 = 43.48) now I have two currents, 25A and 43.48A, but these don't really tell me much and they are also based at 10 or 25 deg C which I don't know what my operating temperature is until i put this stuff on a board, in a box and try it.

the thermal side, I have:

RθJC Thermal Resistance, Junction to Case 0.5 oC/W
RθJA Thermal Resistance, Junction to Ambient * 40 oC/W
RθJA Thermal Resistance, Junction to Ambient 62.5 oC/W

how do i take these numbers into consideration?

 

[dknguyen]

THis is for steady-state only (the act of switching can also causes a LOT of losses which should be accounted for by adding them into the conduction/steady-state losses).

THe voltage/resistance you use for your conduction losses is the voltage drop or resistance ACROSS the source-drain (the primary power terminals of the device). So you do not use the at the gate 10V since the other conduction and switching losses (rather than the gate capacitance charging losses) dominate, especially for power devices.

For example,
Ploss = RI^2 = (.23R)*(25A)^2 = 144W (for a 0.23R MOSFET which behaves like a small resistor when used as a switch)
Ploss = RI^2 = (0.7V)*(25A) = 17.W (for a 0.7V IGBT which behaves like a diode when used as a switch).

THe thermal resistance is how many degrees the device will rise above ambient for every watt of power that is dissipated in it (hence the units).

So without a heatsink, and if the heat path is from the silicon to the air (junction-ambient):

Trise = Ploss * ThermalResistance = 144W * 62.5C/W = 9000C

If the ambient was 25C, then the device temperatuer is 9025C. Which puts it well above the temperature limit which is probably 125-150C.

If you have a heatsink then the heat must flow from silicon (junction) -> case-> heatsink. Then the thermal resistance you use is the thermal resistance of junction-to-case + heatsink-ambient thermal resistance.

Obviously 9000C is a bit ridiculous so you'd need to find a much better MOSFET and/or heatsink to bring it within the device limits. ANd remember, you aren't even accounting for switching losses yet which, when the device is sized properly should be about equal to the conduction losses.

If the max device temperature is 150C and your max ambient temperature is 40C, then you can only allow a 110C temperature rise at most. So you must use this 110C in your temp rise calculation and find the appropriate heatsink (and if need be, find another device). So if you can estimate your maximum ambient temperature, device losses, and thermal resistance you can estimate your operating temperature.

Now here is one difference between IGBT and MOSFET. MOSFETs act like a small resistors when used as a switch and IGBTs act like diodes. For a fixed, lower current, the losses in the MOSFET will be less (since the voltage drop across the resistor is less that that across the diode). But as you increase the current, so too does the voltage drop across the resistor and so do the losses. But the IGBT's diode voltage drop stays the same. Since the power loss in a resistor increases by I^2 and the power loss in a diode only increases by I, this means that at a certain point (usually a very high current since you can get MOSFET with very very low resistance) the losses in the MOSFET become greater than those in the IGBT and continue to skyrocket from there.

In reality, the reason IGBTs are used over MOSFETs is because IGBTs are able to handle the higher voltages used in industrial drives. Higher voltages also mean lower current for the same power and lower current = less losses.

 

[dknguyen]

What are you building anyways with a really high voltage DC motor that is 8000W? THings get complicated and expensive at that voltage and current level and the little non-idealities you could ignore at lower power levels all come into play. THings that blow up motors/components and produce lots of noise. THe result is more switches, snubbers, isolated drivers, and multi-level inverters (really complicated).

It doesn't sound like you know enough to practically (and safely) proceed with this, because the things I just explained to you in this thread are don't even have to do with motor drivers. THey have to do with the basics of transistors (not even the physics of the transistors). And I know nowhere near enough to implement a drive of this power level.

I'm designing a BLDC here that has components capable of handling a max of about 2kW, but I only plan to use it at about 200W because I don't know if it will work properly up to 2kW (heck, I don't know if it will work at all). It's already costs me more than $600.

 

[jnnewton]

Yeah, doesn't sound like it to me either, can you recommend some sources where I can learn what I will need to know. I don't want to do anything that is unsafe, but I don't really have any interest in building a circuit to drive a small motor at low voltage. There are many people who have already done that and I could just copy a design and build it, but where's the fun in that?

 

[dknguyen]

Yeah, doesn't sound like it to me either, can you recommend some sources where I can learn what I will need to know. I don't want to do anything that is unsafe, but I don't really have any interest in building a circuit to drive a small motor at low voltage. There are many people who have already done that and I could just copy a design and build it, but where's the fun in that?

THen don't copy a design make your own or refine someone else's, add in feedback, modify or use another BEMF sensorless circuit. Most of what I learned was running around reading articles on H-bridges, half-bridges, 3-phase motor drivers, inverters, gate drivers, and sensorless BEMF detectors. Because a BLDC driver is realy just a 3-phase inverte rwhich is really just 1.5 H-bridges or 3 half-bridges.

Or study all these designs you seem to be able to find until you understand the basics. THen make your own. No one says you have to copy anything (and even if you did, a lot of people don't document their work completely enough for you to blindly copy their design without knowing what's going on).

 

[jnnewton]

Could you please elaborate as to what basics I will need to know or where to find them. Here is what I have so far (answers to steps in ()).

1. Basic Circuit Layout (3 half bridges made of mosfets or igbts driven by a gate driver, driven by motor control pwm from a dspic.)
2. How to choose a mosfet or IGBT(Idrain > Imotor for mosfet, Ice > Imotor for IGBT, Ploss = RI^2, Trise = Ploss * ThermalResistance Tambient+Trise < Tlimit.)
3. How to choose a gate driver(???)
4. switches, snubbers, isolated drivers, and multi-level inverters (really complicated) (???)
5. Feedback (Use QEI from microchip. Fairly straight forward)
6. Software (use pid from microchip, plenty of examples.)

 

[dknguyen]

Yeah, feedback and software aren't really problems.

Snubbers:
https://www.electro-tech-online.com/custompdfs/2008/03/design.pdf

Gate Drivers:
**broken link removed**
**broken link removed**
http://www.irf.com/technical-info/appnotes/an-978.pdf
http://www.irf.com/technical-info/appnotes/an-950.pdf

Multi-Level Inverters (it's probably best right now to just look at the half-bridge diagram and understand how it works).
**broken link removed**
**broken link removed**

BTW a multi-level bridge let's the bridge output more than just +V and -V(GND). It allows you go also go V/2 and other things in between so you can make more "blocky" approximations of a sinusoid to reduce harmonics which is important for higher power levels.

Flyback (there are 3 pages in this link on "motor secretes")
http://www.modularcircuits.com/h-bridge_secrets1.htm

 

[jnnewton]

Perfect, reading time for me. Thanks for the info and the discussion.