Making a Third-Party Traction Controller Work With Your Motor and Your Display Over CAN

Two versions of this call come in most weeks. One: someone has specified a good off-the-shelf traction controller and wants their own instrument or service display on it. Two: someone has a controller and a motor that have never met, and cannot get the pair to produce usable torque. Both are expected to take an afternoon. Both usually take a week, and rarely for reasons to do with how hard the software is.

Part one: getting it onto your display

Establish the protocol and bitrate from the unit, not the manual

Controller Protocol Bitrate
Sevcon Gen4 CANopen — DS301, DS401 and DSP402 profiles, ISO 11898-2 physical layer Configurable: 1 Mbit/s, 500, 250, 125, 100, 50 and 20 kbit/s. Current units typically arrive at 500 kbit/s
Curtis 1232E / 34E / 36E / 38E “CANopen compatible”; other 11-bit or 29-bit protocols configurable through VCL Up to 1 Mbit/s, set in the CAN Interface menu

Do not trust the manual for the default. Gen4 units typically ship at 500 kbit/s today, but older revisions of the Applications Reference Manual list 1 Mbit/s. Depending on which PDF you found and how old the unit is, the figure you plan around may be wrong.

Read the bitrate off the unit with the vendor’s tool, then set the rate for the whole machine deliberately before anything goes in the vehicle. A controller at the wrong rate does not fail quietly — you get error frames, and a bus that was healthy is not any more.

On Gen4 that tool is the DVT USB programmer. On Curtis it is the 1313K handset or 1314K PC kit.

Three architectures — start with the humble one

  1. Your display listens. The controller transmits what it already transmits; your screen decodes and draws. Nothing your display can do to upset the drive. Start here — a real screen in a day, and it shows you what the controller actually publishes versus what the manual implies.
  2. Your display participates. A CANopen node proper: SDO configuration at boot, PDO mapping, possibly NMT responsibility. More capable, and now your screen can break the drive.
  3. A gateway in the middle. A board speaking CANopen to the controller and your own protocol to everything else. Extra hardware, but it isolates the drive, lets you change display vendors without touching drive integration, and gives you somewhere to log.

Option 3 is right more often than people expect once a BMS, charger and accessories also want to talk. See CANopen, J1939 or your own protocol.

PDO for data, SDO for configuration

PDO is cyclic, unconfirmed, up to 8 bytes, pre-agreed content — speed, current, temperature, status and fault words belong here. SDO is confirmed addressed access to any object dictionary entry, for configuration and service.

Do not poll SDOs to drive a live display. Twelve values at 20 Hz over SDO will saturate the link and still look sluggish. Map them into TPDOs once and let them arrive.

The Curtis persistence trap

On the Curtis 1232E family, parameters written over CANopen by SDO are not retained through a key cycle unless you first set CAN_EE_Writes_Enabled (0x332F, sub-index 0x00) non-zero, make your writes, then set it back to zero. Curtis warns against leaving it non-zero in normal operation, so treat it as unlock–write–relock. The symptom is maddening: everything configures, the machine runs, you power-cycle in front of someone who matters, and it is back to defaults.

Termination

Exactly two 120 Ω terminators, one at each physical end; a correct bus measures about 60 Ω powered down. Curtis controllers offer a local 120 Ω, 0.5 W termination by wiring CAN Term H to CAN Term L — convenient until the harness already has one at that end. 40 Ω means three terminators. 120 Ω means one, and it will pass on the bench and fail in the field.

Part two: making it actually run your motor

This is the harder half and the one almost nobody writes about. A Gen4 is not a motor-agnostic box you bolt on. Until it holds a correct electrical model of your specific motor, you get cogging, audible noise, torque that collapses under load, thermal cutback long before the ratings suggest, or a controller that will not commutate at all.

The honest answer about autotune: there is one, and it is not for your motor

Sevcon’s self-characterisation routine — the SCWiz software — is for induction motors. If you are running a PMAC motor, which most modern traction builds are, there is no wizard that measures your motor and fills in the blanks for you.

That is the single most useful thing to know before you start, and it explains why you cannot find a procedure written down anywhere. There is no procedure. On PMAC it is manufacturer data where it exists, measurement where it does not, and iterative map building and loop tuning on a bench with instruments on it. It is know-how, not a routine.

So the first question is which kind of motor you have, because it decides whether there is a tool at all:

Motor type Gen4 software Characterisation route
PMAC PMAC No autotune. Manufacturer data if available, otherwise measured; maps built and validated by hand; loops tuned on a bench
Induction, closed loop Flux Vector Self-characterisation via SCWiz, from nameplate data
Induction, open loop Open Loop Slip Control Slip and flux factor maps; self-characterisation assists

Note also that the controller has to match the feedback device fitted to the motor — a Sin/Cos unit will not commutate a motor with an AB encoder however good the rest of your data is. The Sevcon guide covers feedback matching, and getting this wrong is a hardware purchase you cannot configure your way out of.

If you are on the induction path, two things from Sevcon’s own instructions: self-characterisation spins the motor, so the vehicle must be jacked up with the driving wheels free to turn; and SCWiz is not compatible with the 110 V Gen4, so induction characterisation has to be done with a 48 V or 80 V controller. If your production machine is 110 V you need a different unit in the loop purely to characterise.

What a PMAC setup actually involves

Where the data lives, so you can judge the size of the job:

  • Electrical parameters at 4640h and 4641h. Ask the motor manufacturer first — some will provide a full set, many will send a datasheet with three numbers on it, and then the values have to be established by measurement.
  • Motor maps at 4610h to 4615h. On PMAC: 4610h flux map (torque vs magnetising current), 4611h power limit map (max torque vs speed), 4612h voltage cutback map, and 4615h a second power limit map that overrides 4611h on the trigger in 4652h. The saturation map at 4614h is Flux Vector only, so it is not part of a PMAC build.
  • Feedback setup — encoder data at 4630h and pulses per revolution at 6090h. Sevcon warns that an incorrect pulses-per-rev may stop the controller braking the motor effectively, which is a fault that does not announce itself until you need the brakes. Commutation alignment relative to the rotor has to be established and verified, not assumed.
  • Protection and feel — stall detection at 4626h, and the 4650h feature bits including low-speed speed-loop gain interpolation between 25 and 50 RPM instead of the standard 50 to 200 RPM, plus current control gain scheduling. That last pair is the difference between a machine that whines in an aisle and one that does not.

Getting the power limit and voltage cutback maps wrong is what produces a machine that feels strong on a test stand and runs out of torque on a grade, or that pulls its own pack voltage into a fault. Those maps are judgement against measured behaviour, which is precisely the part that cannot be looked up.

Have it done instead

Motor characterisation and control development for Gen4

This is our core service on these units, and it is genuinely bench work: we characterise the motor, build and validate the maps, tune the current and speed loops for your duty cycle, set the feedback and protection parameters, and hand back a validated parameter set with the measured data behind it rather than a guessed configuration.

Where it makes sense we supply the result as a working package — controller and motor, characterised together, shipped configured and bench-tested as a pair, so the box you open runs.

The practical part: we need the motor in front of us. Characterisation is measurement, so there is no remote version of this.

If it is a motor we stock — see electric motors — we already have it and you need send nothing. If it is a motor we do not sell, you ship us the motor and we return it with the controller and the validated parameter set. Tell us what it is before you ship anything and we will tell you which case you are in.

Controller application software and accessory integration

Being precise per vendor, because the two are not the same thing:

  • Curtis — VCL is genuine application code running in the controller, and we write it: custom throttle and brake behaviour, switch and sensor logic, interlocks, drive modes, your own CAN messaging, accessory sequencing. If the stock parameter set cannot express your machine’s behaviour, VCL can.
  • Sevcon Gen4 — Sevcon’s firmware is not user-modifiable, and anyone offering to rewrite it is misleading you. What is available is the full object dictionary, the motor maps, and supervisory logic in an external node of our design commanding the drive over CANopen. In practice that does everything a custom application needs while keeping the drive on supported firmware.

Accessories — pumps, lifts, steering, lighting, thermal management, telematics — usually belong on that supervisory node or a separate bus rather than hung off the traction controller.

Displays, gateways and the rest of the system

Operator and service screens, CAN gateways between a CANopen drive and your vehicle bus, BMS integration and derate strategy, and NMEA 2000 for marine work — through schematic, PCB and firmware to production.

Bring-up order that saves time

  1. Controller on the bench with the vendor tool. Read the bitrate it is actually set to, set it to your system rate, confirm it is alive in the vendor’s own software.
  2. Motor model established — data entered, or measured and mapped. Validate torque and thermal behaviour on a stand before the machine moves under its own power.
  3. Passive CAN sniff: what does it transmit unprompted, at what rate, with what IDs.
  4. One value decoded end to end onto your screen, verified against reality rather than expectation.
  5. The rest of the receive set, checking scaling and sign against the vendor’s own display.
  6. Transmit, configuration writes and NMT, one capability at a time.
  7. Fault injection last: harness pulled mid-drive, power cycle, display killed, controller killed. Each must behave as written down beforehand.

For step 3 you want a real interface and a breakout — the IXXAT adapter and a 23-pin AMP Seal breakout box, alongside the programming tools in diagnostic and service tools.

Tell us what you are trying to run

Send the controller part number, the motor make, model and whatever data you have, whether it is PMAC or induction, the feedback device fitted, the pack voltage, and what the machine has to do. That is enough for us to say whether you are facing a parameter problem, a characterisation job, or a hardware mismatch no amount of configuration will fix — which is worth knowing before you spend another week on it.

Custom electronics and control system development · matt@electricdrivesolutions.com · (717) 421-9988. You reach the engineer who does the work.

Object dictionary indices, map definitions, the self-characterisation routine and the SCWiz 110 V limitation are from the Sevcon Gen4 Applications Reference Manual; CANopen and VCL behaviour from the Curtis 1232E/34E/36E/38E manual. Commutation alignment and encoder details vary by feedback device — confirm against the encoder and commutation sections of the manual revision for your exact part number and firmware. Shipped defaults change between production runs and manual revisions, so read settings off the unit rather than relying on any published default, including ours.

Back to blog