2. The basic idea : Miniaturized and low cost camera with advanced features High resolution (4K and up) cameras are most often of a big size – usually have a form of professional, cinematographic equipment The concept of miniaturization (i.e. fitting all within a small case, may lead to a new set of possibilities).

3. The electronics market / state of the art Video sensors of 8 megapixels and above have left professional domain and go into a new areas, strongly connected with the handheld market (examples: Apple™ products, the stunt/entertainment camcoders like GoPro™). New generaton of tiny video screen sensors, with improved dynamic range / signal to noise ratio are yet available By combining the state of the art Technology, with possibilities of last generation digital circuits (most useful in this case are FPGAs) it should be possible to create a unique device with high-end features

4. The market situation (cont.) Single, small components, with good price/quiality ratio and powerful enough to build the miniaturized hi-res camera combined together within single PCB project may give fascinating results, keeping the price low. There were no approaches of releasing a miniaturized camera with professional features. The closest try may be observed at the field of Closed Circuit television systems, or entertainment / stunt area ( like GoPro ), but compression is a basic goal there. By putting it all together, the concept of proprietary camera project begins to be sensible.

5. Technical goals To build a small device, with proffesional features (like HD-SDI video interfaces for uncompressed signal transmission) To achieve at least 4K (or - in prototype version - near 4K) resolution To keep the final price of the project low, tuning it to a best possible capabilities.

6. Most significant costs and estimated price, based on local distributor offer Total: below 2000 USD

7. Possible use cases for the product MICROSCOPIC VIEW Small size and reduced weight Guarantees easier combination of camera with specialized equipnent 3D RECORDING With miniaturized design it is possible to achieve a real 3d recording, with lens-to-lens distance, adequate to a distance between human eyes QUALITY CONTROL For industrial field, where high resolutions and miniaturized design is of a great concern

8. Industrial research MIPI / LVDS sensors with hi-res capabilities, and reduced count of differential links are yet available. A closer look at the documentation shows, that the video acquisition based on proprietary p-cores should be possible. Review od HD-SDI standard, possible ways of signal transmission (internal ROCKET I/O serializers of FPGA device / external chips dedicated to video transmission) led to the first stage of the designing work.

9. Design and the prototype Two variants: 8 and 10 megapixel models, with promising BSI-2 technology of new generation sensors from OmniVision™ - for research and final product (respectively) Multilayered PCB projects, with respect of signal integrity terms, able to convey signals of houndreds of MHz’s, mostly in double data rate technology

10. Processing board as a base, with well defined sensor interface - multiple resolution versions achievable, without any significant changes / hacks to the main board. JTAG Streaming header GPIO connector HD-SDI Serializer output The modular model

11. Internal architecture inside out Configuration subsystem, based on a proprietary bus, similar to AXIS model proposed by the ARM corp. Different mechanisms/interfaces achievable through a unified address space, for easier software development Fast video acquisition p-core, proprietary HD-SDI controller ADDRESS BUS ADDRESS Decoder Sensor config module Serializer Config module Video Acquisition block DATA BUS Top level block …

12. Effects of work Acquisition IP-core Demosaic Block Bayer pattern to YCbCr Buffering block HD-SDI streamer Streaming mechanism through about 400MHz/DDR lanes is robust and stable Demosaic block (reconstruction of missing components in bayer pattern), gives good results. Proprietary HD-SDI streamer for SMPTE-292 1.5G link works fine in all domains (streaming/veryfying/scrambling)

13. Video issues Missing parts in product documentation led to a big difficulties in proper set up – a single sensor has a wide register address space. Most of the registers have „debug” or even unknown status, but with significant impact to the streaming behavior. Reference (test) pattern, generated by internal logic of the sensor, with correct decomposition of color components RGB values histogram

14. Video issues Real video stream with insufficient sensor configuration Misconfiguration was confirmed by vendor engineers, but selected product (8M sensor) is no longer available, due to the EOL, hence the support for that product is much weaker than for other ones. RGB histogram

15. Selected model limits, plans for the future Despite high, bandwidth for video streaming is limited to about 8 Mbps (1 Mbps per single LVDS in case of using internal serializers of SPARTAN 6 products) MIPI controller implementation should give better results Connecting the base board with ASP-C sensors is possible, but limited (high-end sensor usualy sends the video stream through multiple LVDS lines). Changes lead to a misconception, i.e. big, expansive and messy project, but in case of further OV sensor issues, it will have to be taken under consideration