HSCAM NANO — Vision System on Chip
Prototype — ready for pilot projectsHSCAM NANO
A camera built around a Vision System on Chip (VSoC): the sensor chip itself acquires and pre-processes images, and a microcontroller with NPU on the camera adds neural-network inference. Built on the patented VSoC technology invented by Fraunhofer IIS. Currently in prototyping and evaluation.
- VSoC processing frequency
- 13 kHz full-frame
- VSoC resolution
- 2 × 1,008 × 1,008 px
- On-chip processing latency
- < 1 ms
- Power consumption (whole camera)
- < 1 W
Where this stands today
HSCAM NANO prototypes are ready to be used in pilot projects — not yet a general, off-the-shelf catalog product. The photography on this page shows real prototype and chip-level hardware, not a finished production unit.
On-sensor processing
Image acquisition and processing on the sensor chip
The VSoC acquires and processes images at up to 13 kHz full-frame. It processes image data as fast as a high-speed camera but does not output it as fast: at high rates the camera reports results, such as object coordinates, instead of full frames.
The underlying, patented VSoC technology was invented by Fraunhofer IIS. HSCAM designs the camera and system around it.
Capabilities
What runs where
Each capability is tagged with the part of NANO it belongs to: the VSoC sensor chip, the additional microcontroller with NPU, or the camera as a whole.
VSoC chip
On-sensor processing
3 independent ASIPs (SIMD, LCTRL, GLB) govern acquisition, analog processing and system communication on the chip itself.
VSoC chip
Dynamic ROI
The region of interest is selected on the chip.
VSoC chip
Image-based triggering
Triggering can be derived directly from image content on the chip.
Microcontroller + NPU
Edge AI
An additional microcontroller with an NPU on the camera runs neural-network inference, with no host PC required. Used in the droplet-inspection research setup.
Whole camera
Minimal-bandwidth output
The camera can output object coordinates only, instead of full frames.
Whole camera
Standalone operation
Can run without a host PC through the M8 connector (RS232, GPIO, optional power supply).
Application example — lab test setup
3D scanning via laser light-section
In this bench setup, a line laser projects a straight line across the object; a NANO prototype, mounted on the optical rail above it, captures how that line deforms across the surface. The aim is to turn that deformation into a height profile directly on the VSoC — a classic laser-triangulation (light-section) approach to 3D scanning, being tried here at prototype stage.
Output
Coordinates and events, not necessarily full frames
NANO can report results — object position, size and a detection flag — instead of full frames, so only a small amount of data leaves the camera.
| Field | Description | Output rate |
|---|---|---|
| Object X, Y | Center-of-gravity coordinates | 10 kHz |
| Object size | Bounding box dimensions | 10 kHz |
| Detection flag | Object present / absent | 10 kHz |
| ROI trigger | Region-of-interest events | Event-based |
The output rates above are those of the result fields. The VSoC's full-frame processing frequency is 13 kHz.
Camera interfaces
- USB — full configuration access and image streaming, with Python & C++ APIs (Windows/Linux)
- M8 connector — RS232, GPIO and optional power supply (24 V) for autonomous (standalone) operation
- Optional — proprietary high-speed link to an FPGA processing platform, for high-speed readout
Specifications
Camera, microcontroller and VSoC specifications
NANO has three levels, and each table says which one it describes. Chip-level figures describe the VSoC sensor chip.
Whole camera
| Specification | Value |
|---|---|
| USB interface | Configuration and image streaming; Python & C++ APIs (Windows/Linux) |
| M8 connector | RS232, GPIO, optional power supply (24 V) |
| Standalone operation | Autonomous, without host PC (via M8 connector) |
| Power consumption (whole camera) | < 1 W Measured for the complete camera over USB at 5 V, across different resolutions, frame rates, exposure settings and modes (0.67 to 0.92 W). |
| High-speed readout (optional) | Proprietary link to an FPGA processing platform |
Microcontroller and NPU
| Specification | Value |
|---|---|
| Component | Microcontroller with integrated NPU, on the NANO camera module |
| Function | Neural-network inference without a host PC |
VSoC sensor chip
| Specification | Value |
|---|---|
| VSoC processing frequency | 13 kHz full-frame |
| VSoC resolution | 2 × 1,008 × 1,008 px |
| On-chip processing latency | < 1 ms Latency of the on-chip processing; a conventional camera pipeline typically needs 10–100+ ms. Not the end-to-end latency of the camera output. |
| Sensor technology | 180 nm CIS (monolithic) |
| Chip size | 11 × 13 mm (143 mm²) |
| Pixel size | 8.75 × 8.75 µm |
| Processing power | 117 GOps @ 60 MHz |
| PE architecture | 1,024 × 8-bit SIMD processing elements |
| Dynamic range | 59.9 dB |
| On-chip memory | 128 kB DRAM |
| Chip-level digital interfaces | LVDS (500 MHz DDR), SPI, JTAG, 16× GPIO |
Sources
Academic references
The underlying VSoC chip technology, peer-reviewed and published by the Fraunhofer IIS/EAS team that invented it.
-
13000 FPS Vision System-on-Chip with Mixed-Signal Compressed Sensing
J. Döge, C. Hoppe, P. Reichel, N. Peter, A. Reichel, C. Skubich — Fraunhofer IIS/EAS, 2021
DOI: 10.24406/publica-fhg-413485
View original source -
A Novel Vision-System-on-Chip for Embedded Image Acquisition and Processing
J. Döge — Fraunhofer IIS/EAS (presentation), 2018
DOI: 10.24406/publica-fhg-402420
View original source -
Low-Latency Image Acquisition and Processing with a Programmable Vision-System-on-Chip
J. Döge, C. Hoppe, A. Reichel, P. Reichel, N. Peter, H. Priwitzer — IEEE International Symposium on Circuits and Systems (ISCAS), 2018
DOI: 10.1109/ISCAS.2018.8351843
View original source -
Live Demonstration: Hitting a Table Tennis Ball with a Programmable Vision-System-on-Chip
J. Döge, C. Hoppe, A. Reichel, P. Reichel, N. Peter, H. Priwitzer — IEEE International Symposium on Circuits and Systems (ISCAS), 2018
DOI: 10.1109/ISCAS.2018.8351794
View original source -
A High Speed Asynchronous Multi Input Pipeline for Compaction and Transfer of Parallel SIMD Data
C. Hoppe, J. Döge, P. Reichel, P. Russell, A. Reichel, P. Schneider — IEEE International Symposium on Asynchronous Circuits and Systems (ASYNC), 2018
DOI: 10.1109/ASYNC.2018.00027
View original source
Related
Elsewhere on the site
Discuss a NANO pilot project
NANO prototypes are ready for evaluation and pilot projects. Tell us your application, the speed of what you need to detect or track, and what the camera should output.
Helpful, but not required
- What your process does and what goes wrong, or what you cannot see today
- Speed or cycle time, if you know it
- Measurement or trigger signals that exist on the machine
- Environment: temperature, lighting, mounting space
contact@hscam.com · hscam.com