topic Re: Board recommendation for MBDT BMS development + edge AI inference on NXP hardware in Model-Based Design Toolbox (MBDT)

Board recommendation for MBDT BMS development + edge AI inference on NXP hardware

miraj — Thu, 30 Apr 2026 10:55:59 GMT

Hello NXP Team,

I am working on combining two NXP-based development streams:

**Stream 1 — Model-based BMS development:**
Using MATLAB/Simulink with Embedded Coder and the Model-Based Design Toolbox for S32K3, I need to deploy a Battery Management System model as generated C code on an NXP S32K3-series microcontroller. The key requirements are: full MBDT for S32K3 support, SIL/PIL/External mode validation, FreeMASTER data logging, CAN FD interface, and ASIL D functional safety rating. I have been briefly looking at the S32K344 family as the target MCU.

**Stream 2 — Edge AI inference:**
Alongside the BMS controller, I need a second NXP board running quantized INT8 neural network models (TCN architecture, TF Lite format) on an NPU for real-time SOC estimation and short-horizon temperature prediction. I have been looking at the MCX N947-based boards for this purpose given the eIQ Neutron NPU.

**My questions:**

1. Given these two requirements, which specific NXP development boards would you recommend for each stream? I have done some research and have two or three boards in mind for the S32K344 side, but would highly prefer and value your recommendation.

2. For the edge AI board, is the MCX-N9XX-EVK or the FRDM-MCXN947 more appropriate for a context where I also want to measure power consumption during NPU inference?

3. Can the recommended boards communicate directly over CAN FD with a simple 3-wire bench connection, or is additional hardware required? I hope my details are sufficient. If you need more information, please let me know. Thank you very much in advance for your guidance and time.

Re: Board recommendation for MBDT BMS development + edge AI inference on NXP hardware

SorinIBancila — Wed, 06 May 2026 08:15:44 GMT

Hello,

1. The recommended hardware for the Model Based Design Toolbox for BMS are the 400V HVBMS S32K344 and the 800V HVBMS S32K358. I would pick the hardware based on architecture required (400V vs 800V). Both of them are well supported in the toolbox, with various examples.

2. Unfortunately, these boards are not supported by any toolbox we provide. I don't have any experience with these boards, so I can't really recommend one.

3. The hardware needed for CAN communication usually consist of two components:

CAN Controller: For S32K344, S32K358, it is included in the chip itself.
CAN Transceiver: For evaluation boards for S32K344, S32K358, it is an external IC (TJA1043, TJA1145 etc) which is mounted on the evaluation board.

Most of the evaluation boards include all the required hardware for CAN, but I recommend you to check these before buying.

Best regards,

Sorin Bancila

Re: Board recommendation for MBDT BMS development + edge AI inference on NXP hardware

miraj — Wed, 06 May 2026 13:56:10 GMT

Hello Sorin,

Thank you very much for your detailed reply.

I would like to clarify my use case further. My work does not require live battery cell measurement through an AFE chip — I am using dataset replay over CAN FD. Therefore I believe I only need the MBDT for S32K3 (general codegen toolbox), not the MBDT for BMS add-on. The official MBDT for S32K3 documentation lists the S32K3X4EVB-T172 as a validated board. Could you confirm whether the S32K3X4EVB-T172 is appropriate for: Simulink model deployment via Embedded Coder, SIL/PIL validation, FreeMASTER data logging, and CAN FD communication — without requiring any AFE cell controller integration? If so, this board seems to meet all my work requirements without the complexity of the HVBMS reference kits.

Additionally, I plan to connect the S32K3X4EVB-T172 to a second NXP evaluation board — preferably something among MCXN947 series (eIQ Neutron NPU) — via CAN FD. The S32K3X4EVB-T172 would act as the BMS controller sending data and the MCXN947 board would receive them and run neural network inference on its NPU. Both boards appear to have on-board CAN FD transceivers. Could you confirm whether a direct 3-wire CAN FD connection (CANH, CANL, GND) between these two boards is feasible on a short bench cable, and whether any termination resistors or jumper configuration is required on either board for this setup?

For the NPU based inference side of my work, I have looked at the MCX-N9XX-EVK. My work requires: INT8 quantized TF Lite model deployment via the eIQ toolkit, real-time CAN FD data reception, NPU inference latency measurement and power consumption measurement during NPU inference to report energy per inference. Is this evaluation board best suitable for my work?

Please also recommend other boards apart from two I mentioned when they are more suitable for my work. As mentioned by you regarding the edge AI boards, if possible, could you please point my query to the concerned person from your team?

I hope my detailed explanation helps you and please let me know for more information when needed. Thank you very much again for your time and guidance.

Best regards,

Miraj

Re: Board recommendation for MBDT BMS development + edge AI inference on NXP hardware

SorinIBancila — Fri, 22 May 2026 09:24:52 GMT

Hello,

1. Yes, S32K3X4EVB-T172 is suitable for your application and I don't see any reasons why it can't fulfill your requirements. We provide example for each subject you said (sil/pil, freemaster logging, CAN) and I recommend you to check them first. Make sure you install the latest S32K3 toolbox in this case (1.8.0).

2. I am pretty sure that you can connect these 2 boards together via CAN FD without any extra hardware setup (such as resistors). About the jumpers, you have to check out the Quick Start Guide/Schematics for each board, to make sure the headers are connected onboard to the CAN pins.

3. I will try to ask around about the MCX-N9XX-EVK but I can't guarantee that I can help you on this.

Best regards,

Sorin Bancila