48V Hybrid E-Torque Automotive Motor Generator Reverse Engineering

Just like the 48V Audi and Jeep Altermotor project, we have a new contender, a 12V Hybrid motor? Lets take a look and see if we can wake up this gadget.

Is this the new small 12V go kart propulsion system in a box?

Some Project Background:

The e-torque system is a variant of 'belt alternator engine start-stop' hybird, where the electric assist motor lives in parallel with the ICE engine. This is one of the first common 48v hybrid, which is incredibly exciting. There were two vehicles initially supporting this route, the 2019-2021 Jeep Wrangler and the 2019-2022 Dodge Ram 1500.

This beautiful device was born in late 2018. Its a liquid cooled induction motor with a built in inverter, its automotive grade and its presently fairly low cost on the used / surplus market. Just look at that thing. Its gorgeous. Have you purchased a hobby BLDC outrunner, and found that it exploded under load? Whats the half-life of hobbyking? While this is no where near as lightweight as a 7kw outrunner, it is comically ruggedized and has an actual coolant jacket. Behold the savior of electric go-karts, I bring you the e-torque.

Here it is, quite adorable. The motor consists of three face mounted mounting points, a built in spring based belt tensioner. To translate ~15kw into a serpentine belt you do need to play some tricks. The wrap-around on the belt is nominally the contact area for force transfer. Having a high contact area and adequate tension is very important to get this working. The key 'specs' we know about this motor [link] are: 60Nm of torque, 14-16kw of mechanical power peak and a maximum of 20k RPM. the continuous rating of 4-6kw is curious, but i think this is due to the 'coolant' being engine coolant, or basically 100C glycol-water. I imagine if this was running 25C water the continuous and peak powerwould be much closer. The coolant lines are simple hose-ring-clamp style connections, instead of an obscure quick-disconnect fitting which is nice, but seriously electric motors shouldn't be cooled with 100C water.

The motor itself is 'rare', its both relatively high power and low voltage. Its fundamentally difficult to provide continuous duty high power from a low voltage system due to the high current required. 48V hybrids operate at the edge of SELV or Safety Extra Low Voltage [wikipedia link]. This is the safety barrier to having more exotic connectors, insulation gloves and PPE. If you notice most items you interact with, aside from mains AC powered, are 48V (60v) or below, this includes POE (power over Ethernet), USBC-PD and portable power tools. When you push past this barrier from SELV to LV you get into a more scrutinous spot for UL approval when it comes to DC hazards.

There has been a huge push for over a decade to move away from 12v ICE automotive power for auxiliary systems and engine start, however it perpetually looses traction due to a number of barriers. 48V vehicles need to be able to 'jump start' a 12v vehicle, so they need to support a 12v system. That jumpstart process needs to be clear, such that the 48V bus is not used when jumping a 12v bus. The real goal of 48v mostly motivated by copper. High current wires are fundamentally expensive, whereas insulation for 48v systems is cheap. A 4x increase in system voltage allows for a 4x reduction in system current, so wires can be smaller and cheaper. Starter motors are the main high current draw for ICE vehicles, pulling upwards of 600A on engine start, and supporting that high current draw for such a short duration is somewhat wasteful. Reducing that cost is visible everywhere, have you noticed a return wire running to the starter motor? no the vehicle chassis is the return path, to save cost. Literally hundreds of amps wrap themselves around the bellhousing to find your starter motor every start, it generates a bunch of odd behaviors on sensitive instruments (analog fuel gauges generally display incorrectly during vehicle cranking).

So what controller does this monster use? 15KW at ~50VDC is 300A, what mosfets are they using? Oh baby that is a good question:

Behind that very very sealed in rear metal case is this masterpiece. I did not have any particular expectation but this is some remarkable engineering. This is a motor controller directly bonded to the moor casing for heat-sinking. You're seeing everything on the power electronics side, a three phase bridge, a central dc-link capacitor and external interface hardware. I will refer to this capacitor below as a 'bus capacitor' as that seems to be the colloquial term for local capacitors for motor controllers. The controller is remarkably compact for something that can pump 15000 watts. lets zoom in and take a look at everything.

This is a single motor phase H-Bridge, where the induction motor phase is ultrasonically bonded to the motor controller bus bar. At the very bottom we have a custom capacitor module that provides 48V bus distribution for each half bridge leg. This is pretty excellent and is probably very necessary for regenerative current related voltage spikes protecting the onboard fets. We do see some other interesting parts, namely that the mosfets used for the half bridges are directly wire bonded to the substrate that's being used for the motor control power board. The case of the motor is the major thermal path out, so having the fets adequately thermally connected is key. The motor has three stages of H bridges, one per phase and the remaining side is for DC Link input and other hardware. We also see a few bond wires that provide voltage sensing for the center of the half bridge and gate drive for the high and low side. My favorite part of all of this is seeing the normal ~0603 / 0805 gate drive resistors sitting plainly inline with the gates of each fet. Its like seeing a burger king wrapper on the international space station, its so normal in this very fantastically engineered assembly.

Lets take a look below:

How the DC-Link makes it to the central bus capacitor is curious. The 48V battery link comes in through a stud with a waterproof housing and is then bonded to a bus-bar. Thats where I loose the trace as the remainder of the system is 'potted' in plastic. I've seen a lot of epoxy potted electronics, but this appears to be an ABS-PC injection moulded assembly, which makes it incredibly difficult to teardown. Lets dig into see what else is going on behind the scenes on the remaining side of the motor controller. To disassemble this i needed to heat up the plastic to get a chance at chiseling it away. I used my handy 300W sparkfun hot air gun at various points to try and get *some semblance* of the plastic to soften up.

After significant chiseling, and what amounts to an archeological dig, we get a better picture of what is going on here. We know that this motor has to see 300A peak through the 48V feed, so copper bus-bars are used to deliver these amps to the central motor controller bus capacitor. There is a 'mystery box' in-line with the 48V input connection. The main capacitor has two copper bus-bars that appear to be ultrasonically welded to the DC link input. The motor controller (-) connection has to connect to the aluminum case, and due to dissimilar materials, it looks like a stud weldment is used, shown as the 'Case Ground Connection'. We also learn about that central capacitor, which is conveniently labeled. We have a 63V 470 uF capacitor from Vishay. The part number is clearly custom, but we get the important details. 63V is somewhat telling, as that is a standard capacitor voltage. For film and electrolytic the DC 'voltage steps' for capacitors are generally: 35, 50,63,75,100 V and I'm fairly certain these are defined by the thickness of the commonly available separator material. Vishay makes both electrolytics and film so we do not know the internal composition. Given the environment and potential temperature its likely a film capacitor. The voltage rating for a DC bus capacitor is an interesting problem, you maximize the stored energy at a fixed bus voltage when that bus voltage approaches the capacitor maximum operating voltage, however, the lack of overhead results in likely premature failure on over voltage. When you picture DC bus capacitor over voltage its important to think of it as a transient event, and as such the lower the capacitor impedance, the more tolerable it is to that over voltage event. Furthermore the proximity of 'the battery' to the dc-link bus capacitor and the impedance of that link is also very important. If you imagine a scenario where your low impedance battery is physically in the motor controller, the need for a local bus capacitor disappears, as the path for transients is short and readily soaked up by the battery.

Initially, a 63V capacitor seemed low for a '48V system'. Conventional '48v' lead acid systems have a fairly wide operating voltage. If you picture four lead acid '12v' modules in series, you end up with 14.4*4 or 57.6V, awfully close to our 63V limit. This is not a lead-acid system, so lets do some quick math to verify that statement. The main battery is a 13S (13 series cell) lithium ion pack, so 4.2*13 = 54.6V is the maximum battery voltage. Its unlikely in an automotive application the cells are charged to 100%, as that generally reduces cycle life, as such its more likely this is a 4.0*13 = 52V battery feed. We have a 11v overhead, which is less than ideal but not as terrible as initially observed.

Can we sort out who our mystery sensor is though? I chiseled out more of the three dimensional bus-bar routes to try and get a better hint. The only decipherable part of this was the VAC logo, which is from Vacuumschmelze, which sounds like a Justin Roiland auto generated planet, filled with vacuum cleaner people who use humans to clean the floors.

Vacuumschmelze Appears to be a magnetics company, so this could be a very bizarre common mode choke, or a current sensor. I was mostly concerned that this was a relay that would need 'enabling'. The closest part I could find from their catalog is this [link] 100A current sensor. Its interesting but a keen eye would notice that the initial picture showed a different part number, it turns out that this is an over-mold, the vac part number shown below. If this is a current sensor we do need to patch some output signaling to the control board,

• What hardware programmer do I need
• What is the development environment
  
• How long will it take to make a hello world 'toggle a gpio pin'

Once these are sorted the next step is to determine some pin-mapping, ie what port / pin on the controller maps to gate drive, current sensor, rotor position sensor, etc. I have a trick for this, and it toggle the pin at 1khz, and wave your scope probe over the board to find out the path of your EMI generator. Ideally whatever is on the other end is not going to get borked by having 1khz pumped at it.

The datasheet gets deeper, this thing is clocked at ~260 mhz. That was initially eye opening, but then i remembered the lowly ESP32 has a 240/320mhz core clock.  The datasheet references the MPC5777C Reference Manual, so lets see what that is. Oh boy the reference manual is... 3393 pages [link]. We learn some more interesting tidbits:

• 2+1 lockstep core, or basically there are two identical cores onboard for fault tolerance, that run in lock-step [pp 108]
• 500+ interrupt sources
• 70 12-bit ADC input pins, 4 actual sigma-delta ADC's
  
• Also four 16 bit sigma delta adc's ? [pp113]
• Four Can modules ?
• Two M_CAN modules ?
• Nexus development interface (NDI)