Major Problems/Issues in Developing a Space Reconfigurable Computing System and

Potential Solutions

Jeremy Ramos

Honeywell DSESjeremy.ramos@honeywell.com

Ramos 2 <papernum>/MAPLD 2005

Major Problems/Issues Non Technical Problems/Issues

Culture Resistance to change Microprocessors and ASICs are kings of their own domains with a significant foothold and

organizational structure to back them up. Some misunderstanding of what RC is and what it can do for missions Solution: Get a good marketing engineer and give her a large expense account

Human Resources Not many people are trained in the area of RC or have the necessary skills to design,

develop, and implement RC systems and applications Solution: Select One

a) train hardware engineers to be RC engineersb) train software engineers to be RC engineersc) Wait for the right people to develop

Technology funding for RC Not much technology funding is available for space RC. In order to overcome “first adopter” reluctance, more research needs to be funded to

demonstrate successful deployment of RC in space Solution: See the solution for culture issue above

Lack of Standards Need to develop standard RC languages, tool flows, interfaces, libraries, and

architectures. What is the equivalent of an ISA (instruction set architecture) for an RC system? Solution: Find a PhD candidate and convince them this is a worthwhile dissertation topic. Solution: Develop an industry consortium to establish standards based on research and

practical experience.

Ramos 3 <papernum>/MAPLD 2005

Major Problems/Issues Technical Problems/Issues

Lack of RC Devices The market is lacking true RC device offerings. FPGAs are a good start, but they are truly not RC devices. Rather they are the best thing engineers could

find to implement RC systems. Solution: Develop designs for RC devices that take advantage of RC research performed in the last 15 years.

Fault Tolerance Available devices do not address fault tolerance Effective fault tolerance design could solve problems with soft errors.

Natural Radiation Effects Total Ionization Dose for available devices is sufficient for short duration missions. Higher TID needs will

require microelectronics process changes. SEEs are a major issue for COTS devices and current solutions consist of “patches” that increase system

complexity and inefficiency. Power and Efficiency

Power constraints in space applications demand high efficiency from microelectronics. FPGAs are very inefficient with significant idle resources.

Floating Point Available devices do not support floating point. Most space applications either require or demand floating point algorithm implementation.

Packaging Conflict between the need for high pin count devices and the reliability of those packages for space

applications Solution: Need to foster continued development of standardized components with high-speed serial

interfaces -- RapidIO, etc.