Automotive
Automotive Embedded Systems
Raonebytes prototypes automotive and vehicle-adjacent embedded systems: telematics and fleet devices, control electronics and ECU prototypes, EV subsystem boards, aftermarket modules, AGV platforms and bench test rigs. We work at the prototype and low-volume end of the market, where the goal is a unit that behaves correctly in a real vehicle and has a defensible path toward production, rather than a Tier 1 series programme.
The short answer
Automotive embedded systems development covers the control electronics, vehicle networking, firmware and diagnostics needed to put a working module into a vehicle, a fleet or a test rig. A mid-complexity automotive device such as a telematics or logging unit typically costs $8,000 to $20,000 and takes 8 to 14 weeks as a prototype, with most of the engineering going into the 12V electrical environment, temperature range and bus integration rather than the application logic. Prototype, aftermarket and fleet work is a different regime from series production ECU development for an OEM, and the two should be scoped separately.
Common builds
Telematics and fleet devices
ECU and control prototypes
EV subsystem electronics
Diagnostics and CAN tooling
Bench and HIL test rigs
What Raonebytes delivers
Vehicle electronics
Schematic and PCB design for CAN FD, LIN and Automotive Ethernet nodes, with automotive-grade part selection, transient protection at the connector and EMC-aware layout rather than a consumer BOM adapted after the fact.
Embedded firmware
Bootloaders, CAN and UDS diagnostic stacks, control loops, event logging and secure over-the-air update, written so diagnostics and update are architectural from the first build instead of bolted on before shipping.
Telematics and backend
Cellular connectivity, GNSS, trip and event capture, store-and-forward buffering, plus ingestion, fleet dashboards and APIs validated against the actual device rather than against a simulator.
Test rigs and validation
Bench harnesses, CAN simulation, HIL-style rigs and environmental test plans mapped to the relevant ISO 16750 loads, with DFM review and manufacturing handoff documentation.
What makes this hard
The constraints that decide whether a design survives contact with the real world.
The 12V rail is not a clean supply
Cold crank can pull the nominal 12V rail down toward 6V while an unsuppressed load dump drives it far above nominal until it is clamped, and ISO 16750-2 treats both as defined test profiles rather than edge cases. Wide-input conversion, reverse-polarity handling and transient clamping have to be designed at schematic stage, because retrofitting them changes the board and usually the enclosure with it.
Temperature range rules out most consumer parts
Cabin electronics generally need -40C to +85C, under-bonnet placement pushes to +125C, and engine or exhaust-adjacent positions reach +150C, which is exactly what the AEC-Q100 grades describe. A part picked from a consumer catalogue may be rated only to +85C and can be changed or discontinued without notice, so component selection is a schedule risk as well as a BOM decision.
Safety classification changes architecture, not just code
ISO 26262 assigns QM or ASIL A to D from a hazard analysis based on severity, exposure and controllability, and anything above QM brings in a defined safe state, diagnostic coverage, hardware metrics and independence requirements. Discovering late that a function is safety relevant usually means redesigning it, so that classification conversation belongs at concept stage even when the deliverable is a prototype.
Bus access is a commercial problem as much as a technical one
OBD-II standardises only the emissions-related parameter set; most genuinely useful vehicle signals sit on proprietary manufacturer CAN messages or behind UDS security access that the OEM controls. Aftermarket and fleet projects need that access route settled before design starts, and read-only designs avoid a class of risk that transmitting onto a live vehicle bus does not.
Standards and protocols that shape the design
Identifying these early is what keeps them from forcing a redesign later.
- ISO 26262 and ASIL A to D
- A hazard analysis and risk assessment assigns QM or ASIL A to D from severity, exposure and controllability, and anything above QM requires a safety case: a documented argument supported by safety requirements, architectural measures, hardware failure metrics, verification evidence and confirmation measures, not a single test report. Full compliance is a programme-level activity run jointly with the OEM or Tier 1 supplier, so what a prototype can realistically do is stay compatible with that path by defining the safe state early, preferring parts that ship safety documentation, and keeping requirements traceable from the outset.
- AEC-Q100 and AEC-Q200
- AEC-Q100 is the stress qualification suite for automotive integrated circuits and AEC-Q200 covers passives, with temperature grades running from Grade 3 at +85C ambient through Grade 1 at +125C to Grade 0 at +150C. Consumer-grade parts fail this regime on two counts: the rated range does not reach under-bonnet conditions, and there is no qualification or change-notification discipline behind them, so a silent process change can invalidate a design that already passed test.
- CAN, CAN FD, LIN and Automotive Ethernet
- Classical CAN (ISO 11898) carries 8-byte frames at up to 1 Mbit/s and CAN FD extends the payload to 64 bytes with a faster data phase, while LIN provides a low-cost single-wire sub-bus for slow body functions such as seats, mirrors and simple actuators. Automotive Ethernet 100BASE-T1 and 1000BASE-T1 run full duplex over a single unshielded twisted pair and appear where camera, radar or zonal traffic exceeds what CAN can carry.
- OBD-II and UDS (ISO 14229)
- OBD-II standardises the diagnostic connector and the emissions-related parameters every light vehicle must expose, which is what makes generic aftermarket readers work at all. UDS is the far richer service layer used for reading data identifiers, running routines, security access and reflashing, carried over ISO-TP on CAN or DoIP on Ethernet, and most of its useful identifiers are manufacturer-defined rather than public.
- CISPR 25, ISO 11452 and ISO 16750
- CISPR 25 sets component-level conducted and radiated emission limits so a module does not desensitise on-board receivers, and ISO 11452 defines the radiated immunity methods a module must survive, including anechoic chamber, bulk current injection, TEM cell and reverberation methods. ISO 16750 covers the operating envelope: electrical loads such as supply variation, reverse polarity, cold crank and load dump in part 2, mechanical vibration and shock in part 3, and thermal cycling and humidity in part 4.
Typical technology stack
Compute
Automotive microcontrollers such as the NXP S32K, Infineon AURIX and automotive-grade STM32 lines for control and gateway roles, selected on AEC-Q100 grade, integrated CAN FD and available safety documentation. Linux-class SoCs appear only where the function genuinely needs them, for example a telematics unit running a cellular stack, local storage and OTA update.
Vehicle networking
CAN FD transceivers with bus fault protection and wake support, LIN for low-rate sub-buses, and 100BASE-T1 PHYs where bandwidth demands it. On the tooling side, CAN interfaces with DBC or ARXML signal databases and an ISO-TP stack underneath anything that speaks UDS.
Software architecture
AUTOSAR Classic layers software into MCAL, basic software, RTE and application components so integration happens against defined interfaces, and Adaptive AUTOSAR targets POSIX-class ECUs with service-oriented communication over SOME/IP. That overhead pays off when a module has to slot into an OEM programme; fleet, aftermarket and rig hardware is usually better served by a lean RTOS such as FreeRTOS or Zephyr.
Power and protection
Wide-input automotive DC-DC conversion covering the 9V to 16V band with headroom for cold crank dips and clamped load-dump transients, plus reverse-polarity protection and filtering at the connector. 24V commercial vehicle and 48V mild-hybrid rails change part selection outright, and high-voltage EV domains add isolation, creepage and clearance and interlock requirements that constrain the mechanical design as much as the schematic.
Sensing and positioning
GNSS with dead reckoning for tunnels and urban canyons, IMUs for motion and event detection, wheel-speed or encoder inputs, and isolated current and temperature sensing around battery and motor subsystems. Camera and LiDAR enter the stack when the application is perception rather than telemetry.
Connectivity and data
LTE Cat-1 and Cat-M or 5G modules for backhaul, BLE for local configuration and technician access, and store-and-forward logging so nothing is lost in a tunnel, a ferry or a depot dead zone. Signed images and rollback-safe OTA update are baseline, because a fleet device that bricks in the field costs far more than the hardware it sits in.
Why this matters
Most automotive prototypes that fail do so on the physical and electrical side rather than in application logic. The vehicle supply is noisy and transient-prone, the temperature envelope removes most of the consumer component catalogue, EMC is tested at component level before anyone sees a vehicle, and the data you actually want is often behind a proprietary CAN message or a security-access gate. Being explicit about the regime matters too: a fleet, aftermarket or rig device can be designed and validated quickly, whereas a series production ECU for an OEM is a programme with functional safety, qualification and supplier processes that sit with the OEM and Tier 1. Raonebytes designs the electronics, firmware, enclosure and backend as one system so those interactions surface on the bench, and our autonomous robot platform, underwater ROV and quadcopter builds involved the same combination of motor control, sensor fusion, ruggedised packaging and battery-powered embedded control that vehicle work demands.
Frequently asked questions
What does an automotive embedded prototype cost?
A simple enclosed device starts from $5,000, and a mid-complexity unit such as a telematics, logging or control module typically runs $8,000 to $20,000 for a working prototype covering PCB design, firmware, enclosure and bench validation. Autonomous vehicle and AGV platforms are a different scale at $40,000 to $75,000 and up. Cost is driven mainly by how many buses and signal types the device handles and how harsh its mounting location is.
How long does an automotive build take?
Most prototypes take 8 to 14 weeks from kickoff to a tested unit. Schematic and layout usually take 3 to 4 weeks, fabrication and assembly 2 to 3 weeks, and firmware plus in-vehicle integration the remainder. Bus integration against a real vehicle is the most common source of extra time, particularly when signal definitions have to be confirmed empirically.
Can you deliver an ISO 26262 certified ECU?
No, and any supplier claiming otherwise for a prototype is overstating what certification means. ISO 26262 compliance is a programme-level activity: the ASIL comes from a hazard analysis of the item in its vehicle context, and the safety case is assembled with the OEM or Tier 1 across the full lifecycle. What we can do is design a prototype to stay compatible with that path, by defining the safe state, selecting components with safety documentation, and keeping requirements and design decisions traceable.
Can you read data from a vehicle we did not build?
Emissions-related OBD-II parameters are standardised and readable on any compliant light vehicle. Beyond that, most signals live on proprietary manufacturer CAN messages or behind UDS security access controlled by the OEM, so the practical options are a licensed or documented signal database, empirical characterisation of the bus, or an added sensor that measures the quantity directly. That decision belongs at concept stage because it determines what the product can actually see.
Do you use automotive-qualified components?
Where the mounting location and the intended path to production justify it, yes: AEC-Q100 for ICs and AEC-Q200 for passives, with the grade chosen to match the ambient temperature at the mounting point. For an early bench prototype that will never leave a lab, consumer parts can be acceptable and cheaper, but that choice has to be a deliberate one, because swapping to qualified parts later can change footprints, tolerances and layout.
Do you take on EV and high-voltage subsystems?
We work on the low-voltage control, instrumentation and telemetry side of EV subsystems: battery and motor monitoring, isolated sensing, 48V mild-hybrid rails and CAN interfaces to existing controllers. Designs that sit across a 400V or 800V traction domain bring isolation, creepage and clearance and interlock requirements that constrain the mechanical build, and high-voltage pack and traction certification is a specialist activity we scope with the client rather than assume.
Related capabilities
Build with Raonebytes
Send your requirements and get a fixed quote, typically within 24 hours. NDA on every project.