LAUNCHXL-F28379D: Project with C2000 microcontroller problem

Below is a consolidated answer organized to your seven questions. Several LVBLDCMTR motor-level parameters are genuinely not published by TI — I've flagged those explicitly rather than estimate them, and pointed to the identification path that produces them.

1. LVBLDCMTR electrical parameters, units, and flux conversion

What is documented: TI support identifies the LVBLDCMTR as a Telco Intercon 42BLS02, 24 V, 42 W motor, and states it is sensorless-only with no Hall sensor or encoder outputs 1. Note a documentation conflict: the MotorControl SDK Universal Lab guide lists LVBLDCMTR as "Hall Sensors Embedded" with a 5-wire Hall harness (GND=Black, +5V=Red, A=Blue, B=Green, C=White) and phase wiring U=Yellow, V=Red, W=Black 2. Verify physically against your unit before committing to a sensing scheme.

Units of USER_MOTOR_RATED_FLUX: This parameter is expressed in V/Hz in InstaSPIN-FOC. Evidence: a comparable 24 V motor (Teknic M2310PLN04K / LVSERVOMTR) identified by InstaSPIN-FOC yields Rs = 0.392 Ω, Ls_d = Ls_q = 0.235 mH, rated flux = 0.0396 V/Hz, 4 pole pairs, max current 7.0 A 3; and an LVBLDCMTR identification run produced an estimated flux of ~0.0308 V/Hz — close to your 0.0315 figure 4.

Critical identification constraint: the product of estimated flux and USER_IQ_FULL_SCALE_FREQ_Hz (default 800 Hz) must stay below the full-scale voltage (24 V for this motor); the documented fix is lowering that setting to ~750 Hz to obtain valid ID results 4.

Gaps: No TI source retrieved documents the V/Hz → Wb conversion factor, nor Ke, Kt, rated/peak phase current, rated/max speed, or rated/peak torque for this motor. TI support has explicitly stated it has no flux-map or saturation characterization data for the LVBLDCMTR beyond what the FAST estimator identifies at runtime 3. Your MATLAB PMSM flux-linkage value will need to be derived from your own measured Ke/back-EMF test, not from a TI-published number.

2. Mechanical and thermal parameters

Not available. No rotor inertia J, viscous friction B, static friction, cogging torque, efficiency, thermal limits, or torque-speed curves for the LVBLDCMTR appear in TI's SDK documentation — the LVBLDCMTR content there is limited to wiring tables and evaluation-kit/algorithm compatibility matrices 5. A part-number lookup returned no specification record either. TI confirms "LVBLDCMTR" denotes an evaluation motor and that documentation is limited to the product page — there is no deeper motor datasheet or app note 6.

For your RL work, the practical path is: use InstaSPIN-FOC/MotorControl SDK motor ID for Rs/Ls/flux, and run a free-deceleration or inertia-identification test on your bench for J and B.

3. BOOSTXL-DRV8305EVM sensing and calibration

Since your 24 V bus only uses ~54% of the 44.3 V full scale, consider the 22 kΩ divider modification above for better bus-voltage resolution.

PWM frequency: A TI software reference runs the BOOSTXL-DRV8305EVM at 45 kHz PWM with USER_VOLTAGE_FILTER_POLE_Hz = 344.62 to filter PWM ripple while passing back-EMF feedback 9. Your 20 kHz is well within the board's capability; the voltage filter pole must be matched to your actual hardware filter.

SPI and protection: Device configuration and fault reporting go over SPI; status registers cover IC overtemperature through MOSFET overcurrent, and control registers set gate drive current, dead times, current shunt amplifier configuration, and fault reporting modes 10. Registers 0x01–0x0C are readable/writable via the InstaSPIN Universal GUI "DRV8305 SPI" tab 11.

Gaps: No explicit sensing bandwidth (filter corner frequency) and no effective inverter resistance figure were found in the EVM guide 7. Numeric dead-time values and overcurrent trip thresholds are not in the EVM user's guide — they must be read from the DRV8305 data sheet (SLVSCX2) register definitions 10.

4. LAUNCHXL-F28379D + BOOSTXL-DRV8305EVM compatibility

Confirmed workable. TI states the LAUNCHXL-F28379D + BOOSTXL-DRV8305EVM combination works for FOC development, and that only PWM and ADC pin tuning is needed — PWM pins are standardized across BoosterPacks, while ADC pin assignment and ADC full-scale voltage/current values must be checked per board 12. Be aware the EVM was designed for the LAUNCHXL-F28027F, and all jumper/switch instructions in its user's guide (remove JP1/JP2/JP3, S1 = ON-ON-ON, S4 = OFF) are F28027F-specific and do not transfer to the F28379D 13.

Jumper / logic supply — this is the most common failure mode. The DRV8305EVM has only a 3.3 V LDO and does not derive 5 V from the DC bus. The F28379D's ADC reference circuit needs 5 V; without it, ADC readings saturate at 4095 even though no electrical pin conflict exists between ADCINA2 and VREFHI 14. Two mutually exclusive options:

• TI-validated default for the Simulink FOC example: keep JP1, JP2, JP3 populated (USB-sourced 5 V) and JP6 not connected 15.
• No USB present: set JP6 = ON to enable the BoosterPack's boost converter to supply 5 V 15.

Never combine both 5 V paths — tying two 5 V rails risks back-powering 15. For debugging, you do not need to remove the BoosterPack: let USB power the LaunchPad with JP1/JP2/JP3 = ON 14.

Position/orientation: Install at the LaunchPad's first BoosterPack position (closest to USB) with the motor terminal block facing the USB port 15, aligning 20-pin headers J1/J2; power to J3 (PVDD/GND), motor phases to J4 16.

Control signals: nFAULT (fault indicator, detail via SPI status registers), PWRGD (watchdog/LDO status), EN_GATE (enables gate driver + current shunt amplifiers), WAKE (exits sleep) 10. A transient nFAULT LED at power-up is normal and clears once status registers are read or EN_GATE goes HIGH 16.

Two peripheral conflicts to watch:
1. Phase→PWM→ADC mapping is not universal. It varies by example code and naming convention (A/B/C vs U/V/W vs R/S/T); one user-derived mapping was Phase R→MotA→PWM1→ADCC2, S→MotB→PWM2→ADCB2, T→MotC→PWM3→ADCA2. Verify against your specific reference code — the FCL library assumes 2-phase IDDK-style sensing unless adapted 17.
2. CMPSS false trips. The FCL_379XL project targets the BOOSTXL-3PhGaNInv, which feeds a Fault/Over-Temp signal into the on-chip comparators. Porting it to the DRV8305EVM without adapting those inputs leaves CMPSS inputs floating → false overcurrent trip-zone events (COMPHLATCH set via ePWM XBAR) that shut down PWM entirely 18.

Gaps: No SPI or WAKE signal pin-mapping table, and no hardware-revision-specific differences for either board, were found in TI documentation or community content for this pairing 14. Confirm these against the official schematics for your board revisions.

5. Sensorless FOC algorithms

TI's two recommended options for this motor, both listed as validated with LVBLDCMTR in the Universal Motor Control Lab support matrix:
- FAST estimator-based sensorless FOC (InstaSPIN-FOC) — also auto-identifies Rs, Ls_d/q, and rated flux 193
- eSMO (enhanced Sliding-Mode Observer) sensorless FOC 19

InstaSPIN-FOC is delivered as ROM-callable APIs on supported C2000 devices 20. Note that InstaSPIN-MOTION (FAST + SpinTAC) is no longer recommended for new designs — TI directs new work to the MotorControl SDK examples 20. The DMC library additionally provides a standalone Sliding Mode Observer plus flux/speed estimator blocks if you want to build your own observer 21.

Gaps: No TI documentation specifies low-speed estimation limits, open-loop-to-closed-loop transition speed thresholds, rotor-angle accuracy figures, parameter-sensitivity analysis, or observer gain values for the LVBLDCMTR 3.

One safety caution: a reported case had an LVBLDCMTR overheat and smoke during an auto-tune startup procedure at 24 V while drawing ~5 A; root cause was not resolved 22. Run your first identification passes with the supply current limit set conservatively.

6. Power supply and protection

Your supply is suitable for low-power evaluation only — and that's fine. The BOOSTXL-DRV8305EVM accepts 4.4–45 V and up to 15 A RMS / 20 A peak 23, so 0–30 V / 5 A is inside the voltage window and far below the current rating. Operating temperature range is −40 to 125 °C, with up to 150 mA quiescent current at EN_GATE low 16. With a 42 W motor, 5 A at 24 V (~120 W) gives adequate margin for startup transients.

Protection, in priority order:

1. Overcurrent is a hardware function, not software. USER_MOTOR_MAX_CURRENT only clamps the speed controller's output (saturating its integral) before it reaches the current controller — it is explicitly not a hardware current limit or overcurrent protection 24. Real protection comes from the DRV8305's built-in short-circuit, thermal, shoot-through, and undervoltage protection 23. Keep your bench supply's CC limit set as a secondary backstop.
2. Regenerative overvoltage is your main unprotected risk. Rapid deceleration of a motor with significant back-EMF pumps energy back into the DC bus; TI's documented example shows bus voltage reaching ~400 V with an aggressive goal speed versus staying under 350 V at a reduced goal speed 25. On a bench supply lacking internal overvoltage protection, this can reach levels unsafe for the electronics 26. The DRV8305EVM's onboard protections do not protect an external supply from reverse current 23.
3. Mitigations available to you:
   - Limit the deceleration rate — TI's documented software mitigation; also USER_MAX_ACCEL_Hzps in InstaSPIN 2527.
   - Bus current limiting — the concept TI implements as BUS_CURRENT_LIMIT_ENABLE/BUS_CURRENT_LIMIT caps current drawn from the supply as load increases 28; you would implement the equivalent in your own firmware.
   - Anti-voltage-surge / active braking concepts — TI's integrated drivers use AVS (auto-applies a safe deceleration rate overriding other settings, eliminating supply voltage overshoot) 29 and active braking (dissipates mechanical + source energy inside the motor via a PI-controlled D-axis current reference ramped against a configurable DC bus current limit) 30. These are useful design templates for what to implement on the F28379D.
   - Scale your voltage feedback to your actual bus range — a prerequisite for any software overvoltage trip 31. This ties directly to the 22 kΩ divider modification in §3.

Gap: No documentation was found on active clamp, bleeder resistor, or bus capacitor sizing for absorbing regenerative energy on the BOOSTXL-DRV8305EVM. External protection (added bus capacitance, a clamp/crowbar, or enforced deceleration limits) is not covered by TI material for this board.

7. Digital control, bandwidth, and TD3 implementation

Current-loop bandwidth at 20 kHz

Use SFRA to settle this. TI recommends the Software Frequency Response Analyzer to empirically tune current-loop bandwidth for a specific system 36; injection code is added inside the control ISR around the compensator input/output to measure plant and closed-loop response 37, and it has been used to measure gain/phase margin and bandwidth on F28379D-based setups 38.

Speed-loop bandwidth at 2 kHz

Speed-loop bandwidth is constrained by both inner-loop performance and system inertia — TI notes that for a high-inertia motor, speed-loop bandwidth may not even reach 1 kHz 33. With a 2 kHz TD3 action rate and a ~2 kHz current loop, your practical speed-loop bandwidth is on the order of 100–300 Hz; your 2 kHz rate is the update rate, not an achievable bandwidth.

Low-speed instability is a documented issue on your exact hardware. A LAUNCHXL-F28379D + BOOSTXL-DRV8305EVM BLDC case showed severe oscillation at 60–120 rpm with both ramp and step inputs, while identical gains gave a well-damped response at 1500 rpm. TI's recommended fix: decouple and independently tune current, then speed, then position loops — not simultaneously 39. Expect to do this before layering TD3 on top.

CPU execution time and neural-network inference

You have substantial headroom at 20 kHz (50 µs period):
- FCL on F2837x runs two independent current loops, each under 2 µs sample-to-PWM-update, with double sampling per axis 40.
- A dual-loop CLA reference shows a 200 kHz inner loop at ~77% CPU and a 20 kHz outer loop at ~7.7% 41 — scaling down, your 20 kHz current loop should consume well under 10%.

Offload to the CLA. The CLA is an independent 32-bit floating-point accelerator running in parallel with the C28x, effectively doubling compute, with direct peripheral access 42. ADC-early-interrupt to CLA first-instruction-fetch latency is just 4 cycles 43, and CLA tasks are non-interruptible and avoid context save/restore, giving deterministic low-jitter execution 41. C2000Ware's ACI Motor Control Benchmark demonstrates both patterns — CLA running the entire control routine, or the C28x offloading part of the compute — a directly applicable template for splitting your TD3 policy network between CPU and CLA 44.

For the network itself, the FPU provides hardware IEEE-754 single-precision support (relevant to floating-point inference) and the TMU accelerates trigonometric/arithmetic operations used in control 42.

Gap: No concrete cycle-count or memory-footprint figures for neural-network inference on the F28379D were found. You will need to profile your actor network directly.

Ripple

Driving a sinusoidal-BEMF PMSM with sinusoidal phase currents (standard FOC/SVM) minimizes torque ripple 45. Minimizing current-loop latency (ideally ≤1 µs) improves achievable modulation index and reduces restriction on current rate-of-change in the motor, directly improving current-ripple behavior 40. No numeric ripple formulas tied to 20 kHz/2 kHz configurations exist in the retrieved material.

Protecting against rapidly alternating TD3 current references

This is the most important item for your project, and it must be handled at three layers:

1. Do not rely on USER_MOTOR_MAX_CURRENT. It clamps the speed controller's output and saturates its integral, but is not a cycle-by-cycle hardware limit — it will not stop fast transient overcurrent from abrupt reference changes 24. Also set it ≤ the motor's rated max current to avoid motor damage 24.
2. Add slew-rate limiting between the TD3 output and the current controller. InstaSPIN already exposes rate/safety limits you can repurpose: USER_MAX_CURRENT_SLOPE_POWERWARP (Id trajectory slew-rate limit), USER_MAX_ACCEL_Hzps (accel/decel limit), and USER_MAX_NEGATIVE_ID_REF_CURRENT_A (demagnetization protection) 27. A first-order filter or explicit dI/dt clamp on the TD3 action is the cleanest fix for policy chattering.
3. Scale your current path to avoid numerical overflow. Set USER_IQ_FULL_SCALE_CURRENT_A with 20–30% headroom above peak measurable current, and satisfy USER_IQ_FULL_SCALE_CURRENT_A × 2 ≥ USER_ADC_FULL_SCALE_CURRENT_A 24. With the board's 47.14 A ADC full scale 8, this constrains your choice — check it explicitly.
4. Hardware backstop: DRV8305 nFAULT/PWRGD plus SPI status registers (IC overtemperature through MOSFET overcurrent) are your real protection against a bad policy action 10. Instrument nFAULT as an interrupt and log the status register on trip.

Gap: No TI documentation specifically addresses protection for reinforcement-learning-generated current references. The limits above are the applicable practical constraints.

Available documentation

• Universal Motor Control Lab guide (SPRUJ26A) — algorithm/motor support matrix, wiring, setup 19. Note: no ready-made F28379D universal lab ships out of the box; it must be migrated from a supported device lab per the Universal Project and Lab User's Guide (spruj26) 46.
• BOOSTXL-DRV8305EVM user's guide (SLVUAI8A) — schematic, pinout, sensing, SPI 12. DRV8305 data sheet (SLVSCX2) for dead-time and fault-threshold register values 10.
• InstaSPIN-FOC/MotorWare guide (SPRUHJ1I) — all user.h parameter definitions and conventions 24.
• No LVBLDCMTR datasheet, identification report, characteristic curves, dedicated firmware, or design files exist. TI confirms documentation is limited to the product page 6; motor parameters are intended to be derived by running the SDK's built-in motor identification routine 2.

One expectation to set: TI engineers state the MotorControl SDK / FCL example code is reference only and "not good enough for product-end" control — production behavior requires adding motion profiling and a braking-stop mechanism on top 39. For research purposes it is an excellent baseline, but budget time for the gap.

Suggested bring-up order: (1) fix jumpers per §4 and verify ADC readings aren't at 4095; (2) run InstaSPIN-FOC motor ID with USER_IQ_FULL_SCALE_FREQ_Hz lowered to ~750 Hz and a conservative supply current limit, to obtain Rs/Ls/flux for your MATLAB model; (3) measure J and B by free-deceleration test; (4) tune the current loop alone with SFRA to ~2 kHz; (5) close the speed loop with a classical PI, confirm low-speed stability; (6) only then substitute the TD3 policy, with a dI/dt clamp on its output.

Follow up: Would you like a step-by-step walkthrough of the InstaSPIN-FOC motor identification procedure for the LVBLDCMTR to generate the Rs/Ls/flux values your MATLAB PMSM block needs?

Citations

1. LVBLDCMTR Hall sensor/encoder information (E2E)
2. SPRUJ26A — Universal Motor Control Lab, p.26
3. LVBLDCMTR nonlinear control algorithms (E2E)
4. LVBLDCMTR flux full-scale freq check (E2E)
5. SPRUJ26A, p.25
6. LVBLDCMTR documentation scope (E2E)
7. SLVUAI8A — BOOSTXL-DRV8305EVM User's Guide, p.8
8. TIDUBP9, p.36
9. SPRUHJ1I, p.243
10. SLVUAI8A, p.9
11. SLVUAI8A, p.10
12. LAUNCHXL-F28379D PMSM motor recommendations for FOC (E2E)
13. SLVUAI8A, p.6
14. LAUNCHXL-F28379D ADC reading incorrect values (E2E)
15. LAUNCHXL-F28379D + BOOSTXL-DRV8305EVM ADC Simulink (E2E)
16. SLVUAI8A, p.7
17. Phase U/V/W to A/B/C connection in FCL (E2E)
18. FCL_379XL PWM generation problem with DRV8305EVM (E2E)
19. SPRUJ26A, p.48
20. SPRU566V, p.67
21. SPRABZ0A, p.10
22. LVBLDCMTR faulty motor (E2E)
23. SLVUAI8A, p.4
24. SPRUHJ1I, p.234
25. SPRUHJ1I, p.323
26. SLAU927C, p.68
27. SPRUHJ1I, p.217
28. SLAU958A, p.68
29. SLAU948B, p.53
30. SLLSFV2A, p.56
31. SPRUHJ1I, p.501
32. SPRT735, p.2
33. C2000 control loop frequency for motor driver (E2E)
34. SPRACO3, p.3
35. SPRAC55, p.3
36. FCL current loop bandwidth (E2E)
37. SPRUIK4A, p.12
38. SPRACB8A, p.7
39. Motor control tuning: speed loop and position loop (E2E)
40. SPRACO3, p.6
41. SPRACS0A, p.8
42. C2000 Device Overview — C28x Academy
43. SPRUHM8K, p.736
44. C2000Ware 6.00.01.00 Release Notes
45. SPRABZ0A, p.2
46. Starting with BOOSTXL-DRV8323Rx and LAUNCHXL-F28379D (E2E)
47. SLVUAI8A, p.11 (full schematic)
48. SPRACW5A, p.15