Turquoise 3-Phase Brushless DC Motor Controller

Preliminary
Craig Carmichael, January 2012

with:
IR2133 triple high voltage MOS gate driver chip,
4070 quad XOR gate chip &
12 IRFP3206 MOSFETS (60 volts, 120 amps)


1. Description
2. Layout
3. Current Ramp Modulation
4. Controller Operation
5. Assembly: Layout and Power Circuit Considerations
6. Cable Convention; Tec Tips
7. Chassis
8. Bits and Pieces

DISCLAIMER: Anything you do with this information is done on your own responsibility. I make no waranty of correctness, suitablility or applicability of any of it.

1. DESCRIPTION

   This device is a solid state 3-phase brushless DC motor controller incorporated into a small wiring chassis, created for the Electric Hubcap 5 KW pancake motor. As implemented it should work for motors up to about 6 KW, 18 to 42 volts battery supply (nominal), under about 130 amps, but it's only been tested with the electric hubcap at up to 42 volts and at under 100 or so amps.
   Other motor ratings or voltages will require different choices of power components, different resistor values, etc, though the controller can be used as-is down to about 14 volts (12 V sometimes works if the battery is well charged) if the amps are also within limits. (It can be used at 12 volts nominal if the 12 volt regulator is bypassed. Below 10 or 11 volts supply it will crap out.)
   A zenor diode/power transistor voltage regulator circuit drops the motor battery voltage to 12-15 volts to power the controller circuits - the two integrated circuit chips, passive components, and the power MOSFET gates. A 7805 five volt regulator supplies voltage for the 5V max. logic inputs of the IR2133.

   Certain aspects of the construction are considered "unfinished".

2. LAYOUT

   The integrated circuit 'chips' and associated components are mounted on a small printed circuit board. The power MOSFETs (along with a dozen filter capacitors) are mounted spread out on a heat sink in two rows and are wired to the circuit board with short lengths of hook-up wire. (Same as previous controllers.) The board is mounted as close as possible to the MOSFETs, consistent with spreading the mosfets out for good heat dissipation. (Note: The shorter the wires are the better. Especially the gate drive lines must be as short as possible and must be twisted pairs with their respective Vs or Com. The layout is designed to minimize the length of the power wires.)
   Two cables of small wires connect to two .1" header pin strips on the circuit board. A 6-pin one goes to the motor for magnet position and temperature sensors, a 10-pin one (may become 11) goes to the operator controls and readouts.
   The controller is mounted as one side wall of an aluminum chassis. Cooling is passive. Heatsink fins made of aluminum roofing flashing are clamped to the outside of the chassis opposite the power transistors inside, eliminating the need for custom heatsink extrusions. The entire chassis also acts as a heatsink.
   The side wall with the motor controller can be removed from the chassis for servicing or replacement without dismounting the rest of the chassis, allowing the chassis to be used as the wiring junction box for the system. Box dimensions are kept minimal to accommodate the often limited available mounting space.


An open controller (V1 PCB) showing the controller base with heatsink bars and aluminum roofing flashing fins
(This one has a specially cut-down chassis to fit a motorcycle - normally the back side here mounts on a wall.)

3. CURRENT RAMP MODULATION
(CRM)

   CRM is my own term for variable frequency modulation based on varying the current supplied to the motor rather than the voltage as in pulse width modulation (PWM). CRM provides maximum torque at stopped and slow RPMs, while the switching speed and losses - and hence the heat in the controller - are greatly reduced at higher speeds with high loads. CRM is analogous to direct torque control, since torque is directly proportional to current.
   With the motor stopped or greatly underspeed, when a coil is turned on, the current ramps up very quickly as there's little or no back EMF and the full DC supply voltage is effectively applied to the coil. With PWM, either the switching speed must be set very high (eg, 10-50 KHz) to prevent excessive motor currents, or the current limiting will trip the motor off early in each PWM cycle. Since PWM cycles are fixed periods, this means the motor is "off" most of the cycle, which greatly limits low-speed torque. Low speed is usually the worst time for low torque in a transportation motor.
   On the other hand, once the motor is turning at speed, the high speed, high power switching makes considerable heat in the controller, and there's no need for high speed switching at higher motor speeds.
   With CRM, the current ramps up to the maximum value (set by the control potentiometer) whenever it does, and the cycle is immediately terminated with a short fixed off period, eg 10 to 50 microseconds (as in, below 100 kilohertz to below 20 kilohertz). The next cycle starts immediately after that. At low RPM with fast current ramp-ups, the modulation is thus at a high frequency. As the motor speed increases, the motor's back EMF lengthens the time it takes for the current to rise to the assigned level, so the "on" time lengthens and the modulation frequency drops. The frequency is whatever it needs to be to provide the torque the control is set at - at any speed.

In broader terminology, CRM is an inverse form of pulse frequency modulation (PFM), where a pulse of a fixed period is repeated at a variable frequency. In CRM, the "off" pulse is the fixed period, and the variable "on" period modulates the frequency.

   A drawback to CRM in some applications is that it's hard to set a constant speed without feedback, except with a constant load where speed varies with torque. For vehicle transportation, the "electron pedal" operates quite like the familiar gas pedal, and drivers should have no trouble using it.

4. CONTROLLER OPERATION

   International Rectifier probably doesn't realize that their IR2133 triple MOS gate driver chip can be used as a single chip three-phase brushless motor controller with advanced 'current ramp modulation' control. At least, they are too modest to mention the fact, and they haven't taken a couple of simple steps needed to provide for both forward and reverse motor rotation, hence necessitating the second XOR chip.

   In explaining the unusual aspects of how the controller works, first, let's look at the chip and its pins:


IR2133

(ITRIP) I-Trip  .1

28. Fault (FAULT)
(FLTCLR) FaultClear  .2

27. LowIn3 (LIN3)
(CAO) CurrentAmpOut  .3

26. LowIn2 (LIN2)
(CA-) CurrentAmpMinus  .4

25. LowIn1 (LIN1)
(CA+) CurrentAmpPlus  .5

24. HiIn3 (HIN3)
(SD) ShutDown  .6

23. HiIn2 (HIN2)
(VSS) Ground  .7

22. HiIn1 (HIN1)
(COM) Common  .8

21. Vcc (VCC)
(LO3) LowGate3  .9

20. VBias1 (VB1)
(LO2) LowGate2 .10

19. HiGate1 (HO1)
 (LO1) LowGate1 .11

18. MotorPhase1 (VS1)
(VS3) MotorPhase3 .12

17. VBias2 (VB2)
(HO3) HiGate3 .13

16. HiGate2 (HO2)
(VB3) VBias3 .14

15. MotorPhase2 (VS2)

Note: for convenience and to avoid varying system interpretations, I'm going to use underline instead of overline for active low signals.

   Pins 9 to 20 are the three half-bridge, 12 volt, high power N-channel MOSFET drivers, with the three floating high gate drive outputs referenced to the output voltage of the motor coils. This critical area is where International Recifier's chips singularly excel, and since blowing up MOSFETs and whole controllers from power handling glitches is the usual problem with motor controllers, it's the reason for choosing IR chips in spite of any logic peculiarities.

   Pins 22 to 27 are the six inputs corresponding to each of the six driver gate outputs. (They are inverted from the outputs, which is of little consequence except for reversing the already arbitrary sense of the forward-reverse switch.) The fact that there are six inputs instead of three seems to show IR's weakness in conceiving their logic designs: there are six outputs, but the HIGH and LOW side of each pair can't both be ON at once without creating a short circuit from B+ (motor power) to common through the MOSFETs. Internal logic already prevents such shorts for bad combinations of inputs, so the extra three inputs appear to have little real value - at least for our application. The three pins could provide more useful functions.
   As it is, the three signals from the magnet sensors in the motor go to both the high and low inputs, in a skewed arrangement that tricks the IR2133's internal logic to do the common six-step commutation that it could have done just as well with three pins:

MagnetSensor1: HiIn1 & LowIn2
MagnetSensor2: HiIn2 & LowIn3
MagnetSensor3: HiIn3 & LowIn1

   Ground (Vss) and Vcc need little explanation. "Common" (COM) is the 'ground' connection through the current sensing resistor to the low side of the power MOSFET bridges feeding the motor coils. Since, in a high powered motor, the current sense resistor is just a straight length of thinner copper wire or other shunt, Common is "almost ground". (straight wire - NOT coiled into an inductor!) Actually I'm presently using nickel-brass ("nickel-silver") wire - it doesn't conduct as well, so it's shorter for the same resistance.  It's also heating up. I'm presently trying out 2mΩ, making 2 mV/A or 260mV for the max 130 A. (BTW I had the solder at the bus end of this 'resistor' break loose on one controller, and (like most faults), most everything fried. 100% good connections are vital in a solid state, high power circuit. Hence the wire now has a clamp at both ends.)

   The "current amp" on pins 5, 4, and 3 is simply an op-amp, intended to drive an external current indication signal. We have a better use for it as a voltage comparitor, eliminating an external LM339 chip.

   Pin 6, "SD" or "ShutDown" is intended to be an input for externally detected "fault" conditions, that shuts the power drivers off to the motor. Our "external fault" is that the motor run switch is in the "Off" position.
   All the "fault" conditions internal and external are latched and prevent the output from turning the mosfets on again until pin 2, "FaultClear" is asserted. Pin 1, "I-Trip", is an input that causes a "fault" if it goes above 0.5 volts. It is intended that this be connected as an analog signal (directly or indirectly) to "Common" to sense "too much" current going through the sense resistor. Pin 28, Fault, is an open drain output that indicates the outputs have been turned off and the motor is unpowered - for any reason.

   In this controller, the current amp "+" input is connected to "common" to sense the motor current. Nominally, the sense resistor is .002 Ω (), so the drop across it is 2 mV / Amp. The "-" input is connected to the control potentiometer ("gas pedal" or ???) that indicates how much current the motor is supposed to receive, from 0 to (nominally) 254 mV for 0 to 127 amps. The op-amp output goes to I-Trip - as an "on-off" signal rather than an analog level - to turn the motor coils off when the current has ramped up to the desired point. (We could as easily swap "I-Trip" and "SD": either one going high turns the motor off. But we won't.)

   The Fault output is connected via a 15 microsecond R-C delay circuit to FaultClear, providing the "short fixed off period" that ends a current ramp modulation cycle. After the 15 uS period, the motor is turned back on - unless the potentiometer is entirely off for zero amps, in which case I-Trip will stay high and hold Fault on even as it asserts FaultClear. The op-amp's reading of "common" is slightly biased up (15 millivolts, via a 100 ohms, 100 K ohms resistance divider to the +15 volt supply) so the potentiometer can definitely hold its input slightly lower than the current sense at the zero end.

   Let's "rename" these pins in accordance with these usages:

   "ShutDown" becomes the "OFF/ON" switch. A pullup resistor ensures that the motor stays OFF if the control cable is disconnected.
   "CA-" becomes "ControlVoltage".
   "CA+ becomes "CurrentSenseVoltage"
   "CAOut" becomes "CurrentSwitch"
   "Fault" becomes "OffTime": the output has turned off for the rest of the modulation cycle: this is asserted during the "short fixed off period".
   "FaultClear" becomes "StartCycle" to initiate the next modulation cycle.

   Since OffTime connects to StartCycle, the 'short fixed off period' ending the current ramp modulation cycle is set by an R-C time delay between them.

   There's a reason "Fault" can be triggered besides "I-Trip" and "SD": if the voltage driving the mosfet gates drops below about 9 volts (which would result in most mosfets partly conducting and dissipating much more power than they can handle), the "undervoltage detect" will shut it off. For transportation propulsion, shutting it down while (eg) a car or airplane is in motion is usually the most dangerous option, so the best way to handle it is to end the cycle and try again -- exactly what we're doing anyway. Also, undervoltage is unlikely to happen since the 36 to 42 volt battery supply has to drop to about 14-16 volts before the regulated 12-15 volt supply voltage would drop. The 2uF ceramic capacitor on the high side charge pumps should keep them going for quite a while - there should certainly be no problem in normal operation.
   Note: Smaller capacitors and low motor voltages, eg 12-18 volts, may cause hickups as the motor starts. Once it's spinning much smaller values would be fine. This will vary by motor and voltage. IR recommends nothing less than .47uF.

   Forward/reverse is the one vital function that can't be implimented on the IR2133 chip. This is especially where it would have been nice to have just three input pins - to have a pin left over for this function. Reverse is implimented simply by inverting the three magnet sensor signals. That way, "north" and "south" magnets appear to be reversed, and the coils turn on the other way. The simplest way to invert a signal or not depending on switch position is with an exclusive or gate. The three magnet sense signals are optionally inverted by one fwd/rev switch (per the circuit schematic), so a quad XOR gate is the choice.
   Since we're operating at 12-15 volts, a 4070 15 volt CMOS logic XOR gate would seem to be the simple choice, but frustratingly the logic inputs on the IR2133 work only up to 5 volts. So we have to add a 5 volt regulator and use a 74xx86. (74LS86, 74HC86, etc.) That's the second chip. (I used a SOIC to fit it on easier.)
   NOTE: The first version uses a 4070 at 15 volts, modified with diodes to make the outputs "open drain". It works fine, seems reliable. There was a hickup in getting PCBs for the second version and it hasn't actually been made and tested yet. (2012/01/20).

Schematic (version 2):


PC Board (V2):


Eagle circuit board layout program files:
IR2133.lbr
IR2133-Ctrlr-V2.sch
IR2133-Ctrlr-V2.brd

The solder pads on the PCB for the phase A mosfets now go:

A hi gate ---\ twisted pair
A out      ----/

A low gate --\ twisted pair
common ----/

and the same for phases B and C. B+ and Ground are separate solder pads.



MOSFET/Heatsink component wiring
Notes: 1. I now only use one [set of] AGC/AGO fuses on the power line (now that things are more reliable)
2. A, B and C out now have a separate balancing wire from each pair of MOSFETs to the terminal block.


5. Assembly: Layout and Power Circuit Considerations

   There are some considerations peculiar to high power circuits. I learned many of them the hard way, finally finding there are explanations for 'inexplicable' failures, and I'm explaining them here so you'll see there was logic behind the way things were done. That way you won't arbitrarily, eg, use thicker wire in a heavy current circuit where the thin wire specified was used for a reason (as would be my own inclination). In particular, in order to minimize stray inductances, most of the signals need to be kept short and straight. Minimizing stray inductances minimizes spikes and glitches in the signals, which become extreme at high powers driving an inductive load such as a motor.

* The DC power should be well filtered right at the mosfet half-bridges. This controller uses 12 mosfets instead of 6 in three legs: each mosfet is paralleled, doubled up, to handle the high powers and currents, which switch on and off suddenly, creating considerable spikes. Across each bridge is a 270uF, 100 volt capacitor with as low a series resistance as I could conveniently find: six capacitors, 1620 uF total. I have no close idea of what's optimum, but when I used three 100uF/64V capacitors early on (a value given in an app notes example circuit), they actually blew up quite soon -- across a 36 volt DC line coming straight from batteries. 1200uF might be considered minimal.

* The 3-wire external power cable from the controller to the motor should be no more than 3 feet long. This means the controller should be mounted quite close to the motor. If the motor is on a car wheel, the controller should be just inside the car next to the same wheel.

* Varistors between the motor phases are used to help dampen the worst spikes coming back from the motor. Present choice is 36V80Z. These were chosen for 36 volts and might gradually start popping at 42 volts, which would necessitate going to the next voltage up. It would be best to have several in parallel as the mosfets are only rated for 60 volts. (For 42 volts consider going up from IRFP3206s to 70 volt rated IRFP3207s.)

* If the gate wires (between the driver chip and the mosfet gates) "are over about 2 inches long they should be run with twisted pair wires." (International Rectifier: AN-947. They aren't kidding - I had failures from shoot-through currents spuriously activating MOSFETs until I started twisting wires together.) It is suggested elsewhere that the gate resistors be as close to the mosfet gates as possible. (Allegro: STP 02-3)
   However, the power mosfets need to be well spread out on the heatsinks for cooling. The most compact wiring arrangement found (shortest power wires) consitent with good cooling was to place them an inch and a half apart in two lines: 6 low side drivers and 6 high side drivers, on two 9" long heat sink bars. Thus it's 1.5" between gates of the doubled-up mosfets of each position. And the IR2133 chip itself is over an inch long, so the runs are up to almost two inches before they even leave the small circuit board. The gate resistors are best mounted on the PC board too, to keep the hard-wiring from getting messy. It can be seen the two inches will be considerably exceeded, and beyond the resistors. This is particularly true of phase "C". The circuit board is bolted onto three mosfets in common with their holding bolts. Since there are six pairs of mosfets and two pairs per phase, one phase especially - I made it "C" - will be more distant from the board. The board is bolted to low-side mosfets 2 ("A"), 3 ("B"), and 4 ("B"), leaving 5 and 6 ("C") more distant. (The "V2" board layout shortens "C" gate wires over V1.)
   The gate and source sense voltage wires run as a twisted pair from the board to the nearest mosfets on both low and high side.

* When using two paralleled power transistors to handle heavy loads, it is necessary to ensure they are balanced, that is that each one takes about 1/2 the load and dissipates about 1/2 the power. This is a bit tricky, because once one has heated up a bit more than the other, its internal resistance drops more, and it starts taking more of the load and heats up even more, which continues in a positive feedback "thermal runaway" loop. If the load is too heavy for one transistor, the one that started heating up first will blow. (Someone says I have this backwards, that for MOSFETs it's a negative temperature co-efficient. I should check this out, but it's still best to balance them in case of non-identical component specs.)
   Balancing is done by using small, equal value resistances in the power side of each paralleled circuit. If the mosfets (cold) are .002 ohms when ON, and balancing 'resistors' of .002 ohms are used, then even if the internal resistance of one mosfet should drop to .0005 ohms by heating, the total resistance in each leg is .004 and .0025 ohms, instead of .002 and .0005. Without the balance, the one mosfet would take 80% of the load; with it, it takes just 61% even in this extreme case. And because of the balancing, it probably won't get this extreme. ("Typical" value for IRFP3206 mosfet is .0023 ohms cold.)
   But what is used to obtain such tiny resistance? The answer is: equal lengths of copper wire on each side, of a thinner gauge than would normally be selected for the given current. These are the wires used to connect the mosfets bridge outputs to the motor cable connector block. Again we want to keep the wires to the motor as short and straight as is practical, so we pick the gauge based on the distance of the mosfets from the motor connector block. For four inch wires, we use #18 AWG wire to obtain roughly this resistance, or #16 for 6" wires.

6a. Cable Convention

   The control cable probably won't work unless it's wired right, and the motor sensor cable is wired straight through, so there's no need to dwell on them.
   The 3-phase power cable to the motor could be wired arbitrarily, and the motor side depends on the relation between the coils and the magnet board in the motor, which is likely to be pretty much random by the time three wires are wired to nine coils and are sticking out of the motor. There are six ways to connect the two 3-wire plugs, and only one of them is right. Essentially, pins are swapped until the motor runs properly in both directions. (There are two ways that it'll seem to run okay in one direction but poorly in the other, drawing a lot of current with no load.)

   If both the motor and the controller were wired arbitrarily, the motor would only run right on one controller out of six. However, if the controller end is standardized, then once the motor is running on one controller, it will run on any one. So I have defined a convention for wire colors and the controller side of the APP connector.

   Inside the controller, the three phases I arbitrarily call "A", "B" and "C", go from left to right, looking from the front side of the circuit board where the header sockets are.
   The wire colors, whatever they are, (resistor color code values) are used in numeric order, so that A is the lowest numbered wire and C is the highest. Thus, with a cab tire cable having white, black and green wires, Black (0) is A, green (5) is B, and white (9) is used for C.
   At the other end, looking into the open end (the end that mates with the other plug) of the APP plug with the plug pins on the bottom side, A, B, C go left to right - in this case: 0, 5, 9, black, green, white.

6b. Tec Tips

   At the heatsink, the B+ supply and the common return busses are #10 solid wire above the MOSFETs. These have a bunch of capacitors to filter the DC supply to the MOSFETs. The heavy wiring is above the MOSFETs; the light control wiring to the PC board is underneath.
   The easiest way to get at the control wiring is often to unbolt the 12 MOSFETs from the heatsink. Then (with chassis connections undone) the entire circuit assembly will pull free of the chassis.
   However, when installing more than one or two MOSFETs, it's best to have the heatsink bars on to line up the new ones with the mounting holes. In this case a good way to get better access is to unbolt the heatsink bars from the chassis, which again allows removal of the circuit.

   It is important to remember that almost any flaw, any short or missed or bad connection anywhere, will most often result in destruction of the circuit - usually the MOSFETs and-or the IR2133 driver chip. Be sure of every connection. Check everything, including for shorts from each MOSFET drain to the chassis, ie, for a bad heatsink insulator. Also be sure the bolts are tight, or the MOSFET is likely to overheat and blow under load.



7. Making The Chassis/Wiring Box
(Note: last updated Oct 2010)
[January 2012 comments]


A larger chassis - motor test unit for the shop

   My current visualization (after considerable experience) of the most practical layout is an aluminum box about 6" wide, 10" tall, and 4.5" deep [shrinking to 4" x 9.5" x 4.25" - Jan 2012], with the motor controller occupying one removable side. More generally, the box has to have the height and depth of the motor controller side piece, currently 4.25" x 10". The depth depends on the circuit board size and on how closely the two rows of heat sink are spaced. It could all be done in less space, but the power transistors should be well spread out from each other for good cooling, to avoid destructive thermal run-away. The width of the box is "whatever" as long as there's room for the components and wiring. The test unit for the shop is 7" wide, the one in the car is 6", and the one for the outboard motor, if it is to fit under the hood of the outboard, must be only about 2" or so - I'll be looking for a very small circuit breaker!
   It is intended that the box be mounted on a vertically oriented surface, eg, the rear face screwed to a wall of the car. The box can be oriented vertically (preferred), sideways with the heat sink fins up, or to any diagonal position that fits well and lets rising air flow past the heat sink fins -- or to any position if the cooling seems adequate.

   In the prototypes the motor controller was mounted in the front cover of a shallower box. This was a poor arrangement. With the heat sink fins attached, it was effectively quite deep anyway, the power wires had to be longer to attach them to the open cover, and the wires were being moved around as the cover was being closed, without being able to ascertain they were in good positions (eg, not pushing on circuit board components) once in place.
   If, on the other hand, the motor controller had been mounted in any non-removable part of the box, the whole box would have to be dismounted from the car to service it.
   For that reason, the removable side idea was chosen. Even if an off-the-shelf aluminum box is found, one side will have to be replaced with a removable aluminum panel, about #12 gauge.

   The front and back of the box can be flat aluminum covers that screw on to the "frame" side pieces. The back should be about #12 gauge; the front cover can be thinner, #14 or #16 if desired. Alternatively the back, top and bottom may be one piece (#12), bent around, with 1/2" lips to hold the top and bottom edge of the front cover.

   On the left (or right) side panel (10" x 4.5") the motor controller is mounted. This whole side with the controller can be unscrewed and removed from the rest of the box for replacement or servicing. The aluminum "roof flashing" heat sink fins are on the outside of this face. (Note that the screws for the back cover, and probably for the top or bottom or both if used, must be screwed on from the inside to be accessable once the box is mounted.)
   The breaker/switch is located on the top or the other side from the controller, as convenient, or an external battery post mounted switch may be used.
   The solenoid, terminal blocks and grounding post are attached to the back of the box.
   The filter capacitors sit in the bottom of the box, clamped in place.
Wire and cable entry holes are made where practical and convenient.

   Of course, Turquoise Energy Ltd. plans to offer ready-made wiring boxes complete with controller and all parts as soon as it’s possible to do so. (That just isn’t going to happen this month, April 2009!)

A convenient construction of the box is with the following aluminum sheet metal pieces:

1. a flat 6" x 10" back cover piece, #12 gauge.
2. a flat 6" x 10" front cover piece, #12 or #14 gauge.

   There are a number of options for the edges. If there is no access to sheet metal bending equipment, the edge pieces are cut to just the 4.5" depth of the box, and six feet of ‘L’ angle of about 1/2" x 1/2" (or 3/4" x 3/4"), by about 1/8" thickness, are purchased to attach the sides to the front and back covers. Another 12" - 16" or so of ‘L’ can be bought to connect the edges of the frame together at the corners.
   Otherwise, the pieces are 6" (or 6.5") across and 1/2" (or 3/4") lips are bent along the edges forming a shallow bracket ( ‘[’ ), leaving the 4.5" remaining width for the box depth. (Note: see drawings for the latest dimensions)

   I got excited when I found extruded aluminum brackets (‘[’) were made for windows, 2.5" wide with 3/4" lips. Surely four short lengths of that would frame the controller box, with flat covers on both sides and no aluminum bending required! Then I decided the motor controller should go in the side, requiring a deeper box, and I couldn’t find similar but deeper 4.5-6" wide strips. (There are such shapes more or less, but they are "structural", eg with aluminum 1/4" thick.)

3. a 4.5" or 5" ('depth', per above) x 10" side piece of about #12 gauge to mount the motor controller on. This may be extended to 11" and 1/2" lips bent into the ends to meet (and maybe screw into) the other sides.

4. The three remaining 'frame' edges may be formed from one piece, two or three, and may be a thinner gauge than the piece the motor controller mounts on, eg #14, if desired:

(a) a single 'depth' x 25" piece is formed into a shallow bracket, '[', the folds being .75" each and the main side area 4.5" across. The edges are then cut at 45º angles to make 'V's at 7.5" and 17.5". The piece is bent 90º at those locations to make three sides of a rectangular frame. For the last .5" at each end the overhangs are cut off and the ends are bent over 90º to form .5" lips onto the fourth side.

(b) Three separate pieces, 6", 6" & 10". Again 1/2" can be added as required for tabs to attach the frame pieces together.



The aluminum pieces for the box: Flat (a couple of pre-bent lips); bent; fitted; motor controller side removed.

   The size of the controller was at first 8" x 10". When I layed out the parts, it didn't seem to need that much room, and I cut it down to 6" x 10". It's looking like it could easily be smaller yet.


The motor controller (top side) in the chassis box


Box with controller side removed.


    The next three diagrams show the pieces of sheet aluminum. (CAD? What's that?) All the pieces should be about #12 gauge (~.080") except the cover, which may be #12 or #14. If thinner pieces are used, the screws that hold them together and hold the equipment in place may strip and things will come loose. (I discovered this using a thinner piece (#14 or #16) for the long side opposite the controller.) The front cover may take bashing by luggage in the car, so it shouldn't be too thin either.

   The cover in the photos has seven holes, but four is plenty as per the diagrams. The side holes are really extras, and the middle hole on the controller side puts the screw right against the circuit board and should be omitted. Likewise, the center screw through the back cover has the screwdriver rubbing up on the PCB and should be omitted.
   Another good idea would be to make the side flaps of the motor controller piece 1.5" deep like the bottom flap, with the screws at 1.25" also like the bottom. This would get the screwdriver a bit farther from the circuit board. Note that there are no holes near the bottom of the side flaps - it would just be too hard to get the screws in and out.







   I finally fabricated this motor controller enclosure conception in July 2009, after having done circuit boards in June.
   Final dimensions of this enclosure are 10" x 6" x 4.5", excluding the aluminum "roofing flashing" heat sink fins.
   The 10" height allows speading the power components for physical space and for good heat dissipation.
   The 4.5" depth is to the inside of the two covers, ie, the wall heights. The motor controller won't fit if it's less. (A narrower PCB, eg with surface mount components, would allow a slightly shallower box.)
   The 6" width provided enough interior space. (Originally I cut it 8" wide, but it seemed to have waste space and I cut it down. Worse than cutting the aluminum was having to re-do 2 of the 3 layout diagram sheets.)




8. Bits and Pieces
(Last updated Oct 2010)

[a few Jan. 2012 comments]



The Wiring Box with all components & heavy cables
(Also see the wiring diagram of this at the end of the chapter)

[2012 January]
Added later: Wire Clamps.
[I still don't think these are finished products, just better than what I had!]

Buss Bar Clamps: fat nickel-silver wire on right is the current sense resistor to ground.
The bottom of the wire was flattened with a hammer, looped around and drilled to form a clamp.


Powerline/Fuses Clamp & bits.
Holders for AGC/AGO automotive fuses are just sockets for spade lugs.
]

This box [refer to top image of the three above] goes behind the right rear wheel of the station wagon.
* The battery leeds + (+36V, red), and - (zero V, black) enter at the bottom left. The red goes to the bolt-in side of the breaker (black, left), and the two breakers (it's a double 100 amps house breaker) are tied together in parallel. A green grounding wire comes through a separate hole on the left. The left header connector goes to the motor sensors and the right one to the operator controls at the front of the car.
* The ground, battery "-" and motor controller ground wires all tie onto the one grounding lug bolt on the back cover of the box, seen towards the left. The three-phase power from the controller, via filter capacitors, and to the motor, ties into the copper triple terminal block, center-right.
* The power from the breaker comes out the plug-in connectors (normally these are the input when it's mounted in a breaker box) to two pieces of thick sheet metal that fit securely, and goes to one contact of the "solenoid" relay. (I later put in a longer #10 wire so the relay could be bypassed for testing.)
* The motor cables (three phase power and commutator optics) enter at the lower right through a hole behind the capacitors.
* Lower right, at the front: three 7uF spike filters (motor "run" capacitors from a motor shop) [DITCH THESE] across the three phases (delta), also with metal oxide varistors ("MOV"s - V36ZA80 [KEEP, solder at terminal blocks] soldered across them for additional protection.
* Inside the "top" side (left or right side when mounted as convenient) is the motor controller. "Above" the box are the aluminum "roofing flashing" heat sink fins clamped on the outside of the cover.


 
The Motor End
5-pin "trailer lights" socket & plug for the motor optics, and the Anderson "PowerPole" 75 amp connector (red) to the motor coils. The "A.P.P." plug & socket are identical - one is simply flipped upside down and they fit together. So far, these are the only suitable connectors I've found. Prior to finding them, I had resorted to making my own 'thin line' three-phase plugs and sockets! Note the holes that allow bolting the socket onto something. I plan to fill the one on the car with grease since they aren't intended for exposure to weather.


Main Power Switch or Circuit Beaker
   There are a number of possibilities for this switch. There are battery switches that clip to the terminal post of an automotive battery. One of those could be used and would be external to the wiring box. There are also switches made for heavy battery loads that could be mounted on or in the box.

   But a circuit breaker offers protection if things go wrong and is highly preferable.
   I have used salvaged dual and triple 100 amps house main breakers as switches. (And they have never blown even when things went wrong in the controller.) A breaker rated for 240 VAC should be able to handle up to about 40 VDC. They are, however, physically large and really require a taller chassis box than shown.

   The best breaker-switches I've found are Blue Sea Systems panel mount breakers, rated for 48 VDC and made for marine use. I got the 150 amps model, BSS7004. (BSS7048 has a nicer toggle handle but costs more.) 150 amps is probably a bit larger than needed, but you don't want it to trip when you floor the car! To mount this requires cutting a large rectangular hole in the box. They sell a bezel to cover inaccurate cutting. I used a file.

   For vehicle use, I think at least an accessable manual power switch is indispensable so the car can drive on gas if there's an electric motor or controller problem, without powering up the electric drive circuit. A circuit breaker/switch seems to me to be much the best solution.


Mounting an ordinary "house" circuit breaker: two long bolts and a piece of sheet metal for a clamp, plus a sized hole that fixes it in position, are the key.


Internal Wiring

   I used some #8 and even #6 jumper wires between components inside the chassis. I think I would now just use #10 wires unless they seem to cause problems. They're short, so loss will be slight, and they're considerably easier to work with. Very flexible wire is an asset, especially to avoid stress to the wires to the motor controller, which are attached to mechanically vulnerable components. Wire with many fine strands is the most flexible.

Power Relay: "Solenoid" or "Contactor"

   An inexpensive "solenoid" (the misleading name for 12 volt coil contactor relays) for switching the power via the car key is Tekonsha P/N 53536, available at RV stores. It's a little under spec on nominal current rating (70 amps), but the motor (we trust) won't be running when it's switched on or off.

   Cole-Hersee make "solenoids" with higher ratings, and I recently found these, at a marine store. They seem rather costly.

   Be sure to get one rated "continuous duty". The coil of one that's only made to be on for a moment (eg for a car starter motor) is likely to burn out.

   From the power disconnect breaker/switch (recommended) and then the power "solenoid" relay (if used), the power line proceeds to the bus bar of the power mosfets. From here, each phase goes through fuses to the power mosfet transistors, and their associated filter capacitors. In the case of the Electric HubcapTM, four common 30 amp AGO (= "AGC") automotive fuses: 60 amps per transistor, 120 amps per phase. There is also a thin wire from the bus bar to the circuit board to power the control logic.
   Here the main battery power bus ends.

   I confess to having bypassed the solenoid for testing so far, as the "ACC" circuit in my car doesn't seem to work, and the car won't be a hybrid until the mechanical torque converter is working properly.


Connector Blocks

   I originally used marette connectors ("wire nuts") to tie together the heavy wires between the motor controller, the motor drive cable, and the spike filter components. It was a lot of work each time trying to fit the three heavy wires into each marette.
   I couldn't find suitable terminal blocks. I finally made these "copper cube" connectors from a .5" x .75" copper bar. The holes for the wires are first drilled right through the bar's 3/4" side. (IIRC, I used a 3/16" drill for #8 wires.) Then holes were drilled and tapped for #8-24 machine screws into the wire holes. Another #8-24 hole goes right through the center to bolt down the cube. (Although only three wires connect, I drilled four screw holes anyway except on the first block.) After drilling, the piece was cut off the end of the bar, 3/4" long.




 

   Finally, the cubes were bolted onto a piece of dogwood, shaped and drilled as shown, via another threaded hole at the center. (High temperature plastic would be better - or maybe a sheet of mylar between the copper and the wood.) The holes in the wood under the cubes are recessed so the hold-in screws don't short to the wiring box. The triangle pattern isn't for the benefit of the wires: it simply makes for a firmer mounting to the box when it's screwed down.



In the first controllers, the wiring terminal blocks were in the wiring box. They are better placed on the motor controller as shown here. The base of the block, however, is angled so screwdriver handles don't hit the other side of the box, avoiding considerable aggravation.


An alternative "brass pipe" connector block.

   I also had an idea for an easier to make a unit using one inch long pieces of "1/8 inch" threaded brass pipe. (Actually, the inner diameter is big enough for a #6 wire or two #8 wires.) Having thought of it, I decided to try it and made one connector and a mounting block from a piece of lilac wood.

   I found the pipe tended to strip when the screws were tightened, though it seemed to work okay when I used a smaller than usual drill size to ensure maximum metal in the threads. (It was hard threading the holes, but I didn't break the threading tap.) The insides of the pipes are bound to be corroded, whereas the copper cube holes, being freshly cut, are clean. Cleaning inside the pipes - which I would want to do to ensure good connections - might be more work than drilling the holes in the copper cube.

   Personally, I think the copper cubes is much the superior connector system, though the pipe system is very compact and the piece of wood is easier to make. There's lots of thread length in the cubes so the screws won't strip with repeated insertions and removals, and being screwed down, they can't rattle with car vibration.
   I later found some commercial "bus connectors" in marine supply stores, but they aren't as good or as compact as the cubes I made. They generally tend to be long strips of screwdowns.


Varistors

   The varistors, which are pretty small, may be simply screwed into the copper terminal blocks with the wires.
   For those not familiar with metal oxide varistors ("MOV"s, "varistors"), these interesting devices conduct exponentially more current as the voltage rises, their resistance varying inversely with the voltage. That helps to "short out" damaging higher voltage spikes. The varistors I'm currently trying are "P36Z80". I tried the next voltage rating down, "P33Z70", but a couple of them burned out during regular motor tests. I want to use the lowest rating that lasts, to help best protect the mosfets.

   The reader should be warned that they have been known to catch fire. Like the other electronic components, they should be mounted only inside the metal wiring box.

   Seeing the sparks that result from connecting and disconnecting batteries directly to the motor coils gives one a lot of respect for the power of the transient energies of switching the coils on and off, of the capacity of the filters and of the electronic components to withstand such energies repeatedly and reliably, and why power mosfets keep going up in sparks and flame until everything is relatively perfected.



Power Wiring General Connections.
Note: DO NOT USE the 7uF "motor run" capacitors in the drawing. It works better without them and the spikes coming back from the motor coils seem about the same. (And the chassis has more room without them!)


Notes on the diagram:

1. The diagram of the motor shows the 3 sets of 3 coils in parallel. Although it shows the phases together, the parallelled coils are the ones physically 120 degrees apart from each other.

2. Doubled varistors V36ZA80 are shown. They are currently only single and nothing has blown up - yet. It also remains to be seen whether they will withstand long use or whether the next voltage up will need to be selected.

3. I've bought an inexpensive "solenoid" relay for the power via the car key as shown, Tekonsha, P/N 53536, but I haven't got it hooked up yet. [May 18th 2009.] It's a little under spec on current rating (70 amps), but the motor (we trust) won't be running when it's switched on or off.

4. Battery switches are available from marine supply stores.
I much prefer a switch that's also a circuit breaker as a disconnect in case of any short circuit.

5. Ditch the 7uF capacitors - no value at all.