FAQ - Computer Technology

Which one is faster - PC or DSP-computer?

This question cannot be answered in general. Overall every computer can run C++ programs. Nevertheless, architcture and compiler are different. Many image processing algorithms work optimally with DSP. According to comparative tests the DSP is in average as fast as a 3,5 GHz pentium or 2 GHz dual core prozessor. But if it comes to criteria like robustness (without ventilator, no hard drive) and scalability (high speed applications) the DSP-computer is much better. If more than one computer is used, another advantage becomes clear: due to their energy efficiency, it is easier to scale DSPs. For example, the VisionBox Quad. Its performance of 32.000 MIPS cannot be placed in a small and airless case of PC technology. In summary: As a solution for the system the DSP-computer is a vantage - but not regarding the performance of the computer.

Are DSP-computers common in the branch?

Many solutions in image processing work with machines based on DSP - predominantly Texas Instruments processors. Many provider of vision sensors use DSP. DSP can also be found in programmable Intelligent Cameras. As well, DSP works with challenging image processing solutions. In neighboring branches one cannot manage on without DSP - like Intelligent Cameras for security technology or driver assistant systems in the automotive industry. From that moment, when image processing has the character of a serial production, the embedded processors becomes relevant. The VisionBox offers a platform of solutions to the developer of machine vision.

Is embedded cheaper than PC?

At first sight, embedded seems to be more expensive than PC. But two aspects relativise this point of view: For reproducing PCs, permanent manpower is needed. In case of a new driver, a new motherboard, or an update of the operating system - highly trained employees have to monitor the PC application constantly- even if the image processing program has not been changed. Embedded solutions are better and reproducible for many years. The user has more time for his actual image processing tasks. High class machines have to work. Service in the time of warranty are performed for free. PCs are often vulnerable and can be changed by the user. Embedded systems are relieable and run around the clock. In the case of mass customization, the hardware can be reduced to a minimum and costs decline enormously. Because of this, an accurate cost listing should be made before forming an opinion about costs.

Can the memory be enlarged?

The VisionBox is very flexible and in contrast to other embedded computers available with up to 1 GB memory (Type Quad). This is the current limit. For your calculations you have to take into account that, unlike a PC, it is not the case that a high percentage of the memory has to be used for the operating system. 500 MB of 512 MB working memory can be used for image memory (Type MSP). A XGA-resolution camera can memorize up to 600 images. The VisionBox MPS has a 512 MB memory, the VisionBox Quad a 1.024 MB memory.

Why no PC?

PCs are often used for industrial image processing - and vision sensors are mostly embedded processors. Many of these are DSPs. PCs are common and we got used to the weaknesses they implicate. Many programs work with PCs. However, users like to handle one machine only. Often, PCs are used due to a lack of alternatives. There is a a wide range of optimized hardware for machine vision. The engineer thinks about local and intelligent components: E.g. CAN-interface, PLC, sensor with Ethernet-access and touch panel. This is the category of embedded VisionBox. It is the goal to sell machines which are absolutely reliable for many years.