Talk:MM breakthroughs needed
From Wise Nano
MironCuperman 00:16, 15 September 2007 (CDT)
I suspect some error control in final products would follow existing semiconductor error control: embedded test procedures, redundancy, fast-failure and decoupling. Other error control would be mechanical in nature. In particular:
- Break the design into small enough units such as to keep the unit error rate low
- Design units so that if one fails, it stops working and decouples (rather than delivering bad input to other units) (fast-fail, decoupling)
- Have sanity checks within the units that output an alert flag if the unit has failed (embedded test)
- Have a bus (similar to the JTAG standard) to collect flag status for monitoring the system (monitor)
- Have redundant units to compensate for unit failure (redundancy)
- Design structural items so that a radiation hit doesn't deform them outside of tolerance
Error rates due to radiation are covered in the Nanofactory paper.
For example, if building a motor:
- Design the enclosure and supports so that they don't deform much if hit by radiation
- Have multiple sub-motors, and a decoupling mechanism if a motor gets out of spec
- Have a flag output from failed motors for monitoring and post-mortem analysis. The flag can be a mechanical rod that gets pushed out and is sensed by a monitoring computer.
- transfer the resulting force with multiple thick output shafts
- Have enough motors to allow for accumulated unit failures over the projected lifetime of the product
Another general observation: although you have the capability to build atomically precise, you shouldn't depend on atomic precision when you don't have to.
