Methodology
In order to answer major challenges, the SOPHOS study will address algorithm, software, and hardware design. The algorithms designs will be optimized SAR data compression and SAR image formation to advance the state of the art in onboard SAR data processing. The SOPHOS software development efforts will be divided into: Management software, High- performance scientific and signal processing libraries, and the actual application software. Together with the application related algorithms the management software and the high-performance libraries will integrate into the application software.
Due to most satellite missions’ processing tasks of mixed criticality, SOPHOS’s design will combine industrial grade System on Chip (SoC) processors in combination with industrial grade FPGAs, with the design taking into account the need to allow a simple path to use space qualifiable components as needed in future development by selecting EEE parts that have qualifiable configurations. This will allow the SOPHOS system to deliver fault tolerance and data integrity while providing scalable performance and cost reduction for the system overall.
Hardware modules will be designed, manufactured, assembled and integrated. ECSS design rules will be considered during the development, however will not be required to allow the hardware design to maximize performance in the small required volume.
After manufacturing, the actual hardware modules/PCBs will be functionally validated in a laboratory environment reaching (TRL 4) for each of the hardware modules/PCBs. SW (management software and libraries) developed within SOPHOS will then be integrated with the hardware and validated in laboratory environment in terms of functionality and performance, bringing each integrated HW-SW sub-system to TRL4 (embedded SW pre-product prototype).
As a last integration step, an entire prototype system (for potential use with a nano-satellite platform) including any application software developed for the purposes of testing the SOPHOS system will be fully integrated and validated in terms of functionality and performance. After integration into the prototype system and successful laboratory validation of the management software and software libraries (respecting where applicable CCSDS and ECSS standards), the software including application software will be “frozen” (no further changes will be applied).
In a last validation step results of the project will be tested and demonstrated in relevant environment. For SW this means that the released product is tested in simulated environment (e.g. by means of SAR data emulation). All hardware module prototypes (entire PCBs) developed in the project will be subject to thermal tests that are representative for possible thermal conditions in operational environment. Radiation analyses for critical components demonstrate the usability of the HW for the reference mission profiles.
Overall SOPHOS will deliver sub-systems with TRL6 and also a prototype demonstrator for the full on-board processing and storage chain.
SAR-based Earth Observation

Today, SAR sensors support a huge variety of applications in both reconnaissance and security as well as in Earth observation of land, ocean and cryosphere processes. Depending on application, different operating modes, frequency bands and/or polarisations are used. Copernicus missions like Sentinel-1 and ROSE-L aim at supporting a multitude of long-term monitoring processes on a global scale (e.g. subsidence, deforestation, glacier retreat, ocean currents, etc.). In addition, ESA is launching SAR sensors within the Earth Explorer missions (e.g. BIOMASS and Harmony) to specifically address very particular scientific data needs (e.g. biomass distribution, ocean currents and glacier flow). Here, timely access to the data is less important and the focus is on the highest possible quality of data products.
On the other hand, national space program satellites like Cosmo/Skymed and TerraSAR-X support in addition also reconnaissance and security applications, requiring near-real time availability of the data. However, there is the bottleneck, since up to now, (compressed) raw data have to be downlinked to and processed in the ground segment before reaching the final user. This is also the case for New Space SAR sensors developed by private companies (e.g. IceEye, Capella- Space) targeting mostly time critical security applications. Thus, there is a twofold need for improvement: optimized data compression onboard, alternatively, onboard SAR image formation. Future SAR sensors should therefore support both user categories, i.e. they must implement reconfigurable onboard processing depending on sensor mode and multi-user needs. In this context, SOPHOS intends to foster all these application classes and thus will implement, validate and benchmark algorithms for both, optimized SAR data compression and SAR image formation (see Figure 1-1).
SAR data compression excellence: Nowadays, one of the most widely used methods for SAR raw data digitization is the block-adaptive quantization (BAQ). BAQ is a lossy data compression method which employs a space-varying estimation of the raw data blocks to set up the quantizer dynamic. This information is then exploited to determine the quantization decision levels that best match with the observed backscatter statistics. Being a simple scalar quantizer, BAQ offers a good trade- off between scheme complexity, achievable compression ratio, and resulting image quality. For these reasons, it represents an attractive solution for data quantization in spaceborne SAR systems, where a large amount of on- board data needs to be stored and then transmitted to the ground.
SOPHOS introduces a novel performance-optimized block-adaptive quantization (PO-BAQ), which extends the concept of the state-of-the-art BAQ and allows for an optimization of the resource allocation by controlling the resulting SAR image degradation. Since quantization errors are significantly influenced by the local distribution of the SAR intensity, such an optimization is achieved by exploiting the a priori knowledge about the SAR backscatter statistics of the imaged scene. Given the severe constraints imposed by the downlink capacity, the approach and analyses presented in this study help to optimize the overall global performance for a given downlink budget and, in this way, to increase the acquisition capability of the system allowing for more continuous observation.
SAR image formation excellence: Although SAR image data are often required in near-real time (NRT) (e.g. ship monitoring), present day SAR sensors do not allow the processing of the large amounts of SAR data on board and image products are delayed due to downlink limitations. Processing the data on-board would require approximately 200 GFLOP (giga floating point operations) and would at first multiply the data amount by a factor of 4 (hardly feasible and not beneficial). However, for most NRT application purposes “filtered” data are being used and in this case the data amount is reduced to less than 100 Mbyte (more than one order of magnitude less compared to the initial raw data volume). “Filtering” the data is application specific and user dependent and can be either a multilooked image, detected ships or just predefined regions of interest from a complete image.
Sufficient computational power on-board will allow for the generation of NRT products of reduced data amount compared to the SAR raw data. This will save downlink capacities and will open the possibility to transmit/broadcast final data products directly to the end-users and therefore increase the quality of service.
Within SOPHOS the basic algorithm for the conventional and wide-spread stripmap SAR mode will be optimized, validated and benchmarked on the target SpaceCloud platform.
SpaceCloud the next generation payload processing solution
The need for on-orbit processing is growing in all areas of space based remote sensing. In the SOPHOS project the SpaceCloud solution designed and manufactured by Unibap will be used as a beyond-state-of-the-art solution that carries the needed processing architectures to perform on-orbit processing of high data volumes by carrying a combination of different computing architectures. This heterogeneous compute solution contains a AMD GPU V1000 SoC as the main compute node. This is complimented by different augmenting compute offerings like PolarFire FPGA from Microchip and, Intel Movidious MyriadX to make up the full solution.
The SpaceCloud OS and Framework is in combination with the different available hardware a solutions that enables a new way of processing on orbit. The SpaceCloud OS is available in standard version and as an OS with real-time capabilities to progress and develop to a level matching terrestrial Cloud compute solutions.
SpaceCloud OS holds a number of different functionalities designed based on well-known tools for developing processing applications on ground. They have been adapted for use in space through a long development process supported by the Swedish National Space Agency (SNSA) and the European Space Agency (ESA) to ensure stability of the system and tools when brought into the space environment. This means that well known tools used today for application development and orchestration are moved onto the space segment and enabling the next generation of intensive data processing in space e.g. SAR applications.
SpaceCloud comes with a number of different tools integrated for use out-of-the-box that are carried into space. This includes Tensorflow, ROCm, Docker, AES encryption and Intel SDK for full utilization of their MyriadX architecture. In addition to this SpaceCloud carries an x86 CPU architecture that allows for use of larger software that has been developed to run on terrestrial servers this is the geospatial analysis tools like ENVI from L3Harris that holds tested algorithms for SAR, Electro Optic and LIDAR analytics along with digital elevation maps and data preparation functions.
The SpaceCloud Framework provides a user friendly interface for programmers with defined functions like resource allocation for different applications and resource sharing, Unibaps Generic Sensor Interface (gSIF) that is designed to map different peripheral sensors/devices that needs to be accessed and communicated with from SpaceCloud and is intended for re-use. There is a function for isolation between different users of SpaceCloud so multiple clients can access and use and use the same hardware as one would see on terrestrial Cloud solutions.
Next Generation Mass Memory Module
Current state of the art
SOPHOS aims to advance the state of the art in mass memory systems by developing a new, small form factor mass memory module incorporating the latest advances in processing and storage technology.

The Mass Memory Module (MMM) to be developed will use high speed serial links to eneable low pin count, high data rate interfaces for very high data rate applications such as space-based SAR. This high speed link can support exception data rates to and from the Payload Processing Module (PPM). This bidirectional link allows the MMM to enable retraining of AI/ML models in the The MMM will utilize the modern NAND Flash modules with synchronous interfaces to provide high storage speeds, while also being able to be upscreened to ensure survivability and reliable performance in the harsh radiation environment in space. A high end FPGA is at the heart of the Mass Memory, with firmware based on heritage DSI flight systems, to provide a high-performance and reliable system. European sourced componets will be examined to provide the best possible solution for use in the European market. The hardware developed for SOPHOS will be based on COTS components, but will be possible to produce in a space qualified version with qualified components for the most demanding missions. Scope for this project is to establish a TRL6 version of this Next Generation Mass Memory Module, providing several scalability options in terms of speed, capacity, storage technology, etc.
Ground Segment
Preliminary design: The following architecture is foreseen for the SOPHOS demonstrator.
The EGSE / SAR Instrument Emulator will allow to command the payload processing module and mass-memory unit and to monitor all the necessary parameters for their functional testing. The EGSE shall ensure the telemetry (TM) and telecommands (TC) communication with the payload subsystems through various Ethernet links simulating the S/C platform (e.g. On Board Computer).
The SAR Instrument Emulator will transit synthetic science data (i.e. SAR raw data) via LVDS links (e.g. Camera Link or similar interfaces) to the payload-processing module simulating a SAR Instrument. The EGSE and SAR Instrument Emulator will be controlled from the Ground Control System via TM/TC communication through LAN connection receiving telecommands to control their operation, and sending telemetry for monitoring their status. The EGSE will also forward S/C telecommands to the payload processing and mass-memory modules and receive and forward S/C telemetry from these modules to the ground segment for processing and analysis, simulating the role of the platform.

The Ground Control System will simulate the ground segment transmitting telecommands and receiving telemetry to/from the EGSE to command and control the payload-processing module, the mass-memory unit and the instrument emulator. The Ground Control System will also include a unit receiving the science data from the mass- memory (simulating a ground telemetry receiver) and will perform decompression, decryption and storage for later offline analysis and visualisation. The Ground Control System will include the appropriate control and data software (Central Checkout System like), that will provide the required control and monitoring during end-to-end tests and measurements. The other part of the software will give access to received science data through a visualisation tool to verify the proper operation of the payload processing and mass memory unit by visualisation and analysis of the received SAR data.