FAQ: 96V FOC inverter for IPM PMSM (ABZ encoder): engineering criteria & CAN BMS integration

Id/Iq, MTPA, and field weakening

With an IPM PMSM (interior magnets), the inverter controls two current components: Iq (torque-producing current) and Id (flux current). Iq produces traction torque, while Id is used to maximize torque per amp (MTPA) and to enable field weakening above base speed (negative Id).

Practically, an “IPM-ready” controller must allow a clean base speed → field weakening transition, without torque plateaus or discontinuities, and with explicit limits so that field weakening is not used “at any cost” (efficiency drop, heating, magnet stress).

SIA155-64 case: a versatile motor (urban / enduro / track), typical usable torque ~70 Nm, max speed 7–8 krpm. At 96V, field weakening around ~4,000 rpm makes the transition a key tuning area.

Why setup quality dominates outcomes

On industrial platforms (ZAPI, Curtis), the difference between “it runs” and “it runs well” is mostly in parameterization. A poor setup typically shows up at the base speed ↔ field weakening transition, more than at very low speed.

• Plateau then sudden “release” during acceleration: field-weakening transition not correctly tuned.
• Step-like regen: interaction between FW/regen tuning and battery/BMS limits.
• Underwhelming performance: sensor phasing, motor identification, or current limits too conservative.

The engineering goal is to learn and reproduce a stable calibration, then refine it in the real context (inertia, drivetrain, aerodynamic load, mass distribution).

Single-ended + Z index: compatibility checks

The SIA155-64 uses a single-ended ABZ incremental encoder. Channels A/B (quadrature) provide incremental position and direction; the Z index (one pulse per revolution) provides a reference and is commonly used in traction setups.

ABZ is valued in traction environments for its EMC robustness: edge-based digital decoding tolerates noisy power environments better than analog Sin/Cos signals. The trade-off is that ABZ single-ended support may depend on specific inverter references.

• Confirm native support for single-ended ABZ and Z index handling.
• Validate logic levels and encoder supply per encoder spec (typically TTL/CMOS), plus harness requirements (shielding, routing).
• Prefer platforms with sensor diagnostics: signal coherency, loss detection, offsets/phasing.

Repeatability and per-motor recalibration

In practice, manufacturing tolerances (motor + encoder mounting) justify a simple rule: new motor = recalibration. The goal is not to “redo all FOC”, but to secure sensor learning and intrinsic motor parameters.

1. Commissioning with free shaft in a safe configuration.
2. Keep UVW wiring consistent; reverse direction logically in the inverter (not by swapping phases physically).
3. ABZ encoder learning/characterization at multiple speeds: coherency, offsets, measurement range, Z reference.
4. Auto-tune when available: R_s, L_d, L_q, flux, etc.
5. Torque optimization: adjust electrical angle/advance to maximize torque at a given current (ideally with locked shaft for clean measurement).
6. Loop tuning with special attention on field weakening (the most sensitive transition).
7. Bench validation then vehicle refinement (inertia, drivetrain lash, mass, aero): “perfect at no-load” is not a field truth.

Rideability and battery limits

Industrial inverters originate from material-handling applications: rideability is achieved through mapping and ramps. Choices differ between speed command (industrial, marine) and torque command (vehicle).

• Speed control: typically linear (command = speed), common in marine / pumps.
• Motorcycle: “soft then aggressive” throttle map: first half progressive (fine control), second half steeper (strong response).
• Throttle-off: EVEA’s typical approach is freewheeling (not one-pedal).
• Regen: triggered by the brake (switch), with a dedicated ramp; optional brake-pressure sensor depending on the project.
• Regen is limited by battery acceptance: do not command more charge current than the pack can absorb.

VCU usually optional

A robust 96V traction architecture is often CAN BMS ↔ inverter. The BMS provides max discharge current and max charge current, and the inverter limits traction torque and regen accordingly.

Two “feel” effects are expected and should be designed-in:

• Full battery: regen reduced/disabled → it may feel like “the vehicle no longer brakes electrically”.
• Low battery: traction limited → it may feel like “the vehicle no longer pulls”.

These are not defects: they are direct consequences of battery protection using dynamic limits.

Continuous vs peak, duty cycle

Current sizing must match the use case: continuous (thermal) vs peak (dynamic), and the field-weakening strategy. For the SIA155-64, typical operating orders of magnitude are:

In field weakening, the goal is not to “hold speed at all costs”. If the application forces negative Id to the point where efficiency and temperature collapse, it is usually a sign that motor/application sizing should be revisited (or higher cooling capability).

ZAPI and Curtis reference set

For a 96V traction integration with SIA155-64 and a single-ended ABZ encoder, the following references are solid baselines:

• ZAPI BLE2 96V: particularly suitable platform (rating depends on version).
• Curtis 1236SE 96V: a coherent alternative for IPM traction.
• Curtis F2 / F4: possible options depending on generation and project constraints.

Important: single-ended ABZ is supported natively through specific references (not an add-on option). This impacts selection and sometimes availability.

BLE2 vs 1236SE vs F2/F4

The matrix does not replace commissioning: it frames platform selection. Final performance depends on calibration and vehicle integration.

5 symptoms → likely causes

• Plateau then “release” on throttle: base speed ↔ FW transition not tuned (FW, sensor phasing, motor parameters).
• Step-like regen: FW/regen tuning + CAN BMS charge limit; brake map to review.
• Weak pull / “soft” behavior: incomplete auto-tune, encoder offset/phasing, current limits too conservative, or BMS limiting.
• No regen on full battery: BMS max charge limit (normal behavior).
• Limited traction on low battery: BMS max discharge limit (normal behavior).

• Motor: 96V IPM PMSM, target torque/speed envelope, base speed ~4,000 rpm, FW up to 7–8 krpm depending on the project.
• Inverter: single-ended ABZ + Z index support, commissioning tools, sensor diagnostics, IPM parameter handling.
• Currents: 125 A continuous; typical dynamic 250 A / 10 s with a back-to-safe strategy.
• Commissioning: multi-speed encoder learning, auto-tune Rs/Ld/Lq/flux, torque optimization, FW tuning.
• CAN BMS: charge/discharge limits applied to traction/regen torque; anticipate “feel” at SOC extremes.
• Command: progressive motorcycle throttle map, coherent ramps and interlocks.
• Regen: brake-triggered (switch / pressure sensor), throttle-off freewheeling when relevant, bounded by battery acceptance.
• Validation: bench + vehicle tests (inertia, drivetrain lash, aero), iterate on base speed ↔ FW transition.

Why does ABZ narrow inverter choices?
Because single-ended ABZ is less standard than other sensors on some product lines; compatibility often depends on specific references.

What is the Z index used for in traction?
It provides a once-per-revolution reference that helps with indexing and certain learning/phasing sequences.

Why does regen disappear on a full battery?
The BMS limits allowable charge current. When the pack is full, acceptance drops and regen is reduced or disabled.

Why recalibrate per motor?
Motor and encoder mounting tolerances shift phasing and identified parameters. Recalibration ensures stability and performance.

Why is the base speed → field weakening transition critical?
This is where control switches to negative Id strategy. If tuning is off, it is felt immediately (plateau, steps, non-linearity).