solectrix GmbH

With our years of experience as an innovative development service provider to the automobile industry, we can manage the complete value creation process and assume responsibility for the complete development of new car electronics systems for OEMs and suppliers in all phases of the V model, from the concept and the hardware and software development to testing. Here you’ll find practical examples and innovative approaches for your central issues. Access more information by simply clicking on the respective topic. Interested in indirect vision? The topic is the complete replacement of common rear-view mirrors by an equivalent and suitable camera monitor system. Besides the fact that such a system allows for entirely new design concepts, improved sight and the integration of additional ADAS functions, the main challenge here is that the rear-view mirror itself is a safety-relevant component for safeguarding indirect vision for the driver. To replace this part, various criteria in the fields of functional safety, security, road traffic licensing regulations and image quality in general need to be met. Solectrix conceives and develops all of the image processing – starting with the setup of the first functional prototypes up to the actual series development taking functional safety into account. We support our customers in bringing together various aspects of high demands regarding image quality with those of functional safety and certification-relevant norms. To realize this, on the one hand we rely on model-based development as the basis for dedicated and customized image processing. On the other hand, we design and integrate the safety measures of this defined standard functionality exactly at the neuralgic points within the image processing chain. Unique features of the image processing of a mirror replacement system based on Solectrix IP:

• HQ HDR image signal processing
  • Recording and processing of HDR image data for maximum contrast and true-color display, including anti-flicker mechanisms
  • Customized and optimized noise filters
  • Adaptive tone mapping operators for dynamic compression
  • Dynamic adaptation of the parametrization and weighting for day/night modes
  • At appropriately high resolutions and frame rates
• Versatile scaling and display options
• Low latency and defined dynamic behavior regarding brightness control and automatic white balance
• Advanced & secured overlay engines
• Optional video compression via H.264 or MJPEG
• Video analytics functionality based on
  • Contour feature extraction
  • Sensor fusion and Kalman filtering
  • Flexible connection of AI cores for more complex analyses
• Functional safety up to ASIL C for typical safety goals like the avoidance of frozen images, delayed images or safety-critical image interference
• Integrated safety measures like, e.g., autonomously running consistency checks

Thinking about security for a new product? The increasing flexibility of system architecture results in car networks and interfaces having more potential weak points for cyber attacks. They can also entail considerable risks regarding functional safety. The main challenge is the protection of standard functionality against unauthorized access and manipulation. The integration of security safeguards ensures, among other things, that information is transmitted completely and unchanged and that only certain recipients have access to this information. Solectrix provides solutions in the field of camera-based systems. In this, however, we don’t simply add security to an existing device. Security requires a holistic concept, thus we incorporate this topic into the complete development process from the start („security by design“). Ultimately, we consider this a cross-cutting function of a system. Besides the classical in-car communication, this also affects any form of video data transmission, especially when driver assistance systems make decisions based on the video data downstream. We use pre-validated system components where hardware and software complement each other and don’t provide weak points themselves. Unique features of a security solution for video applications based on Solectrix IP:

• Use of „white box“ hardware-based image processing chains
  • Fully defined and attuned image processing
  • Limited options for manipulation of the processing chain
  • Hardware-based safeguarding of the complete video transport channel
  • Autonomously running consistency checks
• Use of a suitable Hardware Security Engine (HSE) with respective features like AES-128/192/256, RSA, ECC and the necessary hardware accelerators, real random number generators etc., with the following goals:
  • Isolation of safety-relevant information from the application and autonomous processing of the cryptographic operations
  • Acceleration of cryptographic processes with dedicated coprocessors
  • Forcing of various safety measures at the application level
• Supported security use cases:
  • Secure boot
  • Secure communication
  • Component protection
  • Secure storage
  • Secure update
• Security by design according to ISO 21434 as a general rule

Need your system architecture to be more flexible? In the driver assistance system field in particular, the challenge of the future will be to provide a high flexibility on the level of an adaptive system architecture. FPGA-based solutions with expandable, configurable data path functions will play an important role here. They clearly show their advantages through flexible hardware accelerators and an optimal mix of determinism, security, latency and isolationaspects. The basic option of using the same technology basis from simple to complex topics constitutes an immense advantage of FPGA-based solutions. Solectrix handles complete FPGA development, from FPGA-centric systems to the implementation of hardware accelerators within the FPGA fabric. Series development projects are based not only on VHDL development compliant with Automotive SPICE and thus on the same lifecycle as classic software. Safety-relevant systems furthermore developed according to the ASIL classification (Automotive Safety Integrity Level) or other norms that are taken into account as early as during conceptualization and the first development steps. Unique features of an FPGA solution based on Solectrix IP:

• „Enabler“ and thus key technology in all groundbreaking applications regarding:
  • Integration of customized image processing and algorithms
  • Fast digital signal processing and high data throughput at low latency
  • Safety and security for protection from cyber attacks and for ensuring functional safety according to the ASIL level
  • Determinism and integration of hard real-time requirements
  • Flexible connection of AI cores or any hardware accelerators
  • SoC designs + rapid application development & prototyping
  • Flexible high-speed I/O interfaces
• Technology matrix
  • Use of the most suitable manufacturer depending on the requirements: AMD, Intel, Lattice, Efinix or Microchip
  • Use of the most suitable CPU depending on a system architecture analysis: Aurix, Cortex-M, Softcore-CPUs or integrated FPGA SoC solutions
• IP portfolio and expertise:
  • Sensor interfacing and control of relevant manufacturers and technologies
  • Video and image processing (HQ HDR image signal processing))
  • Video compression via H.264 or MJPEG
  • Video analytics functionality
  • Stereo vision & sensor fusion
• Established and certified development process according to Automotive SPICE, which is used for FPGA development in the same way and with similar tools as in classic software development. I.e., independent from the implementation, topics like guidelines, coding styles, static code analysis, unit testing etc. are clearly defined.

Thinking about developing a smart Ethernet camera or a video streaming application? The greatest challenge lies in uniting high-quality image processing, functional safety, security, video compression and video analytics on the basis of a flexible system architecture to meet the increasing demands of the market. Solectrix develops the complete software of Ethernet-based smart cameras which require the integration of image processing in the tiniest installation space. Furthermore, we implement the additional output of compressed imaga data streams in a central control unit for downstream functionality. In both cases, established standards in the automotive Ethernet field can be supported, like SOME/IP or DoIP. Of course we support our customers in all areas, starting from optimized smart cameras up to flexible system architectures. Unique features of a smart Ethernet camera based on Solectrix IP:

• HQ HDR image signal processing
  • with variable scaling and de-warping,
  • separate luminance and chrominance noise filtering
  • and adaptive tone mapping operators for dynamic compression
• Full HD video transmission at up to 60 fps
• Support for sensor resolutions up to 4 MPixels
• Advanced overlay engines with GPU support
• Video compression via H.264 or MJPEG
• Flexible integration of application specific command, diagnosis and update mechanisms
• Video analytics functionality based on
  • Contour feature extraction
  • Optical flow estimation
  • Further camera-side integrated analyses of uncompressed video data
• Safety: Functional safety up to ASIL B
• Security: Integrated security measures like, e.g., secure boot

Thinking about custom-tailored imaging? For special use cases there are often ready-made standard ISP solutions that are a good fit and cover all in-depth requirements for this particular, narrow field. But if your development requires flexibility to combine cross-cutting aspects like, e.g., high-quality image data processing, functional safety, security and video analytics matters, in an individual and customized manner based on a flexible system architecture, an original, dedicated image signal processor solution is needed. High reliability regarding object detection in particular requires a closer look at the special aspects of an ISP, as it is important to know every processing step in detail and to avoid having a „black box“ with unknown behavior within the processing chain. Solectrix transfers the customer’s wishes, taking into account all requirements and parameters, into a customized image signal processor with a number of original, fundamental IP cores. For this, the model-based development approach is used. First, we create a first model on the system level that can be adapted iteratively in very quick and short intervals. Thanks to technologies like Halide, such models run in real-time on prototypical car PCs. The clearly defined and attuned processing chain is then refined piece by piece and derived for the final target technology and hardware architecture (like multi-core CPUs, GPUs, DPSs or FPGAs). Unique features of a customized image signal processor based on Solectrix IP:

• Optimized and clearly defined high-quality image data processing, e.g.:
  • Recording and processing of HDR image data for maximum contrast and true-color display incl. anti-flicker mechanisms
  • Customized and optimized noise filters
  • Adaptive tone mapping operators for dynamic compression
  • Dynamic adaptation of parametrization and weighting for day/night modes
• Flexible high-speed I/O interfaces, e.g., MIPI CSI-2
• Flexible image display with variable scaling and de-warping
• Support of sensor resolutions of over 8 MPixels
• Latency-optimized application
• Fast digital signal processing and high data throughput
• Flexible connection of AI cores or any hardware accelerators for video analytics
• SoC designs + rapid (application) development & prototyping
• Optional video compression via H.264 or MJPEG
• Functional safety up to ASIL C for typical safety goals like the avoidance of frozen images, delayed images or safety-critical image interference
• Integrated safety measures like, e.g., autonomously running consistency checks

Working in research & predevelopment? These days, the topic of convolutional neural networks (CNNs) figures prominently, as the implementation of image analysis tasks like object detection or image segmentation through neural networks already often considerably surpasses the common classic image processing algorithms regarding accuracy. The disadvantage of CNNs is that they often require an extraordinary amount of processing power. This is one more reason why it should always be weighed if or where exactly within a system to integrate such networks. But even if data is „only“ provided to another device, the necessity of pre-processing or data aggregation needs to be assessed. Solectrix offers solutions in both directions, with a framework for optimizing a potential AI toward achieving performance targets. This begins with aspects of model selection and model refinement and ends with the goal of minimizing the inference time on the target platform. At the same time, we continue to deal with topics of classic image recognition within the field of 3D surround view surveillance using stereo vision or 2D video analytics. Unique features of an AI solution based on Solectrix IP:

• Support at every step of the AI development process
  • Setting up a prototypical AI for, i. a., smart data acquisition on NVIDIA basis, incl. prototypical implementation of the complete camera system
  • Image classification and labeling
  • Adaptation of the model architecture and setting up of the training sessions
  • Pruning, optimizing & quantization of the model to reduce the number of calculation steps
  • Preparing the model for inference
  • Compilation for the Xilinx target platform
  • Creation of the complete embedded vision application
• „Ready to use“ development platform and tool kit „AI Ecosystem“ for the areas:
  • Model design: Iterative definition of the neural network
  • Training: For execution of the actual teaching of the network parameters
  • Supervision: The training process is observed, creating statistical information that map the development process of the training and its accuracy
  • Pruning: The trained network is simplified/trimmed so that a reduction of calculation steps is achieved without a loss of accuracy
  • Deployment: Execution of object detection in the final camera application

Our services portfolio ranges from research and predevelopment topics in the algorithm development field to the development and setup of prototypes and demonstrators and the actual series development of digital ECU cores. and suppliers, we are involved in numerous technology and customized series projects. These include, for example, the pre- and series development in the indirect vision field, as with the first series-production digital mirror replacement system for trucks alongside , the pre- and series development of Ethernet-based camera systems and of FPGA-based or microcontroller-based embedded solutions, as well as the technology development in the field of image processing. As an innovator and independent full-service provider, we develop and manufacture demanding embedded electronic solutions on customers’ behalf and offer off-the-shelf electronic components and systems developed in-house.