[Teaa-l] Heatsink for a Curtis controller

[capowski at email.unc.edu]

						July 8, 2008
Hi Everyone,

Re:  Heat Sinks for a Curtis Controller

Pete wrote us a few days ago about an unsettled issue of what type of 
heat sink to use for a Curtis controller.  He continued that one of the 
heat sink options was a flat, one-eighth-inch-thick aluminum plate.  I 
don't know where this idea came from, but I do know that it violates 
many of the rules of electrical engineering and thermodynamics.  Here 
in North Carolina, where the summers are hot, this issue is very 
important to EV owners, so I would like to offer the next few 
paragraphs.

I'm an electrical engineer with some experience mounting, cooling, and 
(unfortunately) destroying power transistors.  Back in 2000 I converted 
a Chevy S-10 to electric drive, using a Curtis model 1231C controller, 
whose maximum continuous current is rated at 500 amps.  Based on these 
experiences, here are some suggestions buttressed by undergirding 
reasons.

The Curtis controller contains 36 power transistors wired in parallel.  
These switch the motor current on and off according to voltages 
generated by the controller's logic board, which in turn is controlled 
by the driver's foot.  The logic board generates almost zero heat and 
can be ignored for our purposes here.

A power transistor used in a binary (on-off) mode is like an imperfect 
light switch.  When the transistor is off, no current flows through it 
and it generates zero heat.  When the transistor is on, current flows 
through it, but there is a small voltage drop across the transistor due 
to its internal resistance, and that voltage drop times the current 
equals the power lost in the transistor.  All this lost power is 
converted into heat which must be removed, or the transistor will burn 
out.

In an electric car, the traction battery pack provides a fixed voltage, 
e.g., 120 volts.  When you drive the car, however, you must vary the 
voltage that is applied to the motor, so that you can control the speed 
of the car.  You do this by depressing the gas pedal (gotta love that 
now-expensive name!) some amount.  The controller senses your command 
and applies a commensurate voltage to the motor.  To continue to build 
our example, let's assume that you wish to accelerate at a rate that 
requires the controller to supply 80 volts to the motor.  This is done 
by applying 120 volts two-thirds of the time.  The Curtis controller is 
called a "chopper", for it divides time into 15,000 periods per second, 
where each period lasts 66 microseconds.  During each period, the power 
transistors are turned on for some fraction of the period, called the 
"duty cycle".  For our example, the duty cycle is two-thirds (80 volts 
divided by 120 volts), so the transistors are turned on for 44 
microseconds, then turned off for 22 microseconds.  The result is that 
120 volts is applied to the motor for 44 microseconds, then 0 volts is 
applied for 22, microseconds.  This process repeats continuously.

Depending on the voltage applied, the rotational speed of the motor, 
and the amount of torque that it is generating at that instant, the 
motor draws current from the battery pack through the controller.  The 
motor current does not vary 15,000 times per second; it changes only 
gradually as the car varies its speed and as the controller varies its 
duty cycle according to the driver's foot.

For our example, let's guess that the motor is drawing an average 
current of 200 amps.   Thus, the controller must provide 300 amps 
during the time that the transistors are turned on and zero amps when 
the transistors are turned off. The data sheets of the transistors in 
the controller show that the voltage drop across the transistors is 
about 3.0 volts when 300 amps pass through them, so their power loss is 
900 watts.  Of course during the off state of the duty cycle, the power 
loss is zero.  In comparison to the speed of the flow of electricity, 
the speed of heat transfer is very slow, so we may average the results 
to a power loss of 600 watts.

Since the Curtis 1231C controller is rated at 500 amps, its power loss 
at 100 percent duty cycle is 1500 watts.  This is a lot of heat, more 
than a laundry iron or hair dryer provides.  The controller mounting 
system must be devised to get rid of this heat, even in North Carolina 
summers and even at near-standstill when there is no air generated by 
the movement of the car.

Several principles apply to this heat-transfer system.  First, heat 
transfer from metal to air is proportional to surface area of the metal 
and proportional to the difference in temperature between them.  
Secondly, heat flows far more readily in metal than it does in air.  
Finally, metal is a excellent heat reservoir, and the amount of heat 
that can be stored in a piece of metal is proportional to its mass.  We 
need to take advantages of all these principles.

To maximize surface area, mount the controller on a heat sink with 
fins.  Its surface area is five to ten times that of a piece of sheet 
aluminum.  Curtis sells a heat sink designed for its controllers at a 
cost of about $100, an expense that may save a $1500 controller.

To keep the temperature of the air as low as possible, blow new ambient 
air through the fins whenever the controller is energized through its 
key switch.  I mounted my heat sink on the electronic shelf so that the 
bottom of the fins is about one-quarter of an inch above the top 
surface of the shelf.  I drilled two five-inch diameter holes in the 
shelf and mounted two five-inch electronics fans to blow air upwards 
into the bottom of the fins.  The air passes through the fins and 
escapes out their ends.  Air to the fans comes from the bulk of the 
engine compartment, a large quantity of ambient air.  Never enclose the 
controller in a container, and don't locate it in the corner of the 
engine compartment where air flow to it is restricted.

Use a massive heat sink, such as the Curtis-provided one.  Not only do 
its fins provide a large surface area for heat to escape, but it also 
provides a large heat reservoir.  Since heat flow occurs over seconds, 
and since the most heat-critical portion of EV operation is the burst 
of energy to accelerate the mass of the car from standstill when 
ventilation for the controller is poorest, the heat sink provides 
temporary storage of the initial heat generated by the controller at 
startup, heat that is later dissipated when the car is moving.

Before mounting the controller onto its heat sink, apply a thin film of 
transistor mounting paste onto the bottom of the controller or the top 
of the heatsink.  Its function is to maximize the heat transfer between 
the two devices, and it does so by filling any voids between the two 
approximately flat metal surfaces that are theoretically in continuous 
contact.  I'm not an expert on transistor mounting paste, but I have 
successfully used the white transistor mounting paste that is sold in 
electronic supply houses.  It is silicone-based goo, laced with metal 
oxide particles that provide a path of low thermal resistance.  I 
believe it is called type "Z9" paste.  It's available from electronic 
supply houses, and has been used for decades for exactly the purpose we 
are discussing here.

Finally, drive to save the controller.  Don't accelerate with 500 amps 
when 300 will suffice.  Never use the electric motor as a brake when 
standing still, as when stopped on a hill waiting for a traffic light.  
This generates controller heat while its air flow is minimal, plus it 
wastes energy from the traction batteries.

Good Luck,
Joe Capowski