VLSI%DEC-MARLBORO@sri-unix.UUCP (10/24/83)
From: John Redford <VLSI at DEC-MARLBORO>
A little while ago I heard an interesting talk given by Don Pederson,
chairman of the EE and CS departments at UC Berkeley. His subject was
"VLSI in perspective", but some of his comments on how technologies can
fail are relevant to what we've been discussing on SPACE.
Integrated circuits have been around now for some twenty years. A
terrific variety of technologies have come and gone. Pederson listed
some of the ones that he considered to be failures:
- Silicon on Sapphire MOS promised much greater speeds than regular MOS by
reducing the parasitic capacitances on the chip.
- Charge-coupled devices were to be the first means of getting 64K bits
of RAM on a chip.
- Magnetic bubbles were going to give us non-volatile storage without
moving parts.
- Josephson junctions switched orders of magnitude faster than regular
transistors. However, IBM has recently cancelled their research program in
them (or so Pederson claimed).
When he looked back on what things failed he noted a number of common patterns:
- Oversell on advantages. Josephson junctions were going to blow all other
computer technologies out of the water, even though they needed liquid helium
cooling, were difficult to fabricate, and difficult to connect to room
temperature peripherals.
- Wrong market choice. Why use magnetic bubbles when floppies are so cheap?
They were going for a replacement market rather than generating a new one.
That meant that the existing technology had a vast lead on them.
- Soon, major problems emerged with the new technology. It turned out to be
difficult to fabricate complex chips with silicon on sapphire MOS, because
the sapphire substrate was full of defects, and expensive to boot. Because
these technologies were out of the mainstream of research, not much effort
could be focussed on solving their particular problems. If you have a thousand
people working on improving a mainstream technology, and ten working on this
specialized one, then the specialized one is quickly left behind.
The final result is slow death. Bubbles still have proponents, as does
silicon on sapphire, but they are clearly dropping out of the race. The
market is a moving target; if you shoot for a narrow market niche it will
be gone by the time your R and D is finished.
Let's judge, say, asteroid mining by these criteria. Sure, there are
potentially vast mineral resources out there (oversell on advantages).
But products like steel are bulky and not particularly valuable
(wrong market choice), and already well served by Earthside industries
(led by an existing technology). The means for getting the material
back is not developed (major problems soon emerge), and only NASA and
its Russian equivalent can work on them, whereas everybody in the
world is working on regular steel production.
A better alternative would be advanced materials processing. Go for the
ceramics and alloys that can only be made in space, but can open up
entirely new product possibilities. Glass, for instance, is brittle
because of surface defects. Fiberglass gets its strength because the
fibers can be made very cleanly. In weightlessness, large pieces of glass
could be cast without ever touching the crucible walls. Their strength
and toughness could approach that of metals, while keeping their
high temperature properties and transparency. Let's go for the radical,
imagination-stretching applications, and not try to attack the turf of
old, entrenched industries.
John Redford
DEC-Hudson
--------VLSI%DEC-MARLBORO@sri-unix.UUCP (10/25/83)
A little while ago I heard an interesting talk given by Don Pederson,
chairman of the EE and CS departments at UC Berkeley. His subject was
"VLSI in perspective", but some of his comments on how technologies can
fail are relevant to what we've been discussing on SPACE.
Integrated circuits have been around now for some twenty years. A
terrific variety of technologies have come and gone. Pederson listed
some of the ones that he considered to be failures:
- Silicon on Sapphire MOS promised much greater speeds than regular MOS by
reducing the parasitic capacitances on the chip.
- Charge-coupled devices were to be the first means of getting 64K bits
of RAM on a chip.
- Magnetic bubbles were going to give us non-volatile storage without
moving parts.
- Josephson junctions switched orders of magnitude faster than regular
transistors. However, IBM has recently cancelled their research program in
them (or so Pederson claimed).
When he looked back on what things failed he noted a number of common patterns:
- Oversell on advantages. Josephson junctions were going to blow all other
computer technologies out of the water, even though they needed liquid helium
cooling, were difficult to fabricate, and difficult to connect to room
temperature peripherals.
- Wrong market choice. Why use magnetic bubbles when floppies are so cheap?
They were going for a replacement market rather than generating a new one.
That meant that the existing technology had a vast lead on them.
- Soon, major problems emerged with the new technology. It turned out to be
difficult to fabricate complex chips with silicon on sapphire MOS, because
the sapphire substrate was full of defects, and expensive to boot. Because
these technologies were out of the mainstream of research, not much effort
could be focussed on solving their particular problems. If you have a thousand
people working on improving a mainstream technology, and ten working on this
specialized one, then the specialized one is quickly left behind.
The final result is slow death. Bubbles still have proponents, as does
silicon on sapphire, but they are clearly dropping out of the race. The
market is a moving target; if you shoot for a narrow market niche it will
be gone by the time your R and D is finished.
Let's judge, say, asteroid mining by these criteria. Sure, there are
potentially vast mineral resources out there (oversell on advantages).
But products like steel are bulky and not particularly valuable
(wrong market choice), and already well served by Earthside industries
(led by an existing technology). The means for getting the material
back is not developed (major problems soon emerge), and only NASA and
its Russian equivalent can work on them, whereas everybody in the
world is working on regular steel production.
A better alternative would be advanced materials processing. Go for the
ceramics and alloys that can only be made in space, but can open up
entirely new product possibilities. Glass, for instance, is brittle
because of surface defects. Fiberglass gets its strength because the
fibers can be made very cleanly. In weightlessness, large pieces of glass
could be cast without ever touching the crucible walls. Their strength
and toughness could approach that of metals, while keeping their
high temperature properties and transparency. Let's go for the radical,
imagination-stretching applications, and not try to attack the turf of
old, entrenched industries.
John Redford
DEC-Hudson
--------
--------REM%MIT-MC@sri-unix.UUCP (10/27/83)
From: Robert Elton Maas <REM @ MIT-MC> You may be correct. Perhaps at this time we should concentrate on novel applications (x-ray telscopes, true zero-gee science-fiction movies, ...) and existing applications that are much too expensive when done on Earth (pure pharmaceuticals, ...) and leave the more mundane things (bulk platinum from the Moon, energy from the Sun beamed or fuel-delivered to Earth, ...) for later after we've established a foothold in space.