New subject: Things I'd want on a desert island (was Re: Transistors...)

24 Apr 2009

Jim Battle <frustum at pacbell.net> wrote:

...
  Johnny Billquist wrote:
 ...
  > It isn't difficult, and it isn't 
 > expensive (well, compared to a transistor I guess it might be, but we're 
 > still talking close to no money).   
 Maybe it is possible to build a transistor that has poor gain, low 
 bandwidth, and poor stability, probably like the first transistors made.

 Jeri Ellsworth built some FETs and simple gates at home -- but she said 
 it took two years to figure out how to make it work, and she needed some 
 equipment (like vacuum chambers) that few people have at home.  It was 
 far from simple.

 http://trail.motionbased.com/trail/episode/view.do?episodePk.pkValue=8067618

 Yes, that is a FET, not bipolar.  If you have some links for making 
 simple bipolar transistors, please supply some.  It will make for some 
 interesting reading.

 I recall reading how IBM spent a lot of money and a lot of brainpower 
 building their own transistors back in the early days.  If it really was 
 all that simple, I doubt that IBM would have wrestled with it as much as 
 they did. 
I purposely picked bipolar since they're easy. I haven't looked at what 
it would take to make a FET. Don't think it's that relevant to reply 
with stuff about FETs to a post about bipolars. :-)

...
   > All it
requires is two diodes. That's 
 > all it is, really.  
 If you wire two diodes in series (PN->NP), it isn't the same as a 
 transistor (PNP), at all. 
It is. A PNP is just the same thing as a PN-+-NP.
The things left that vary is the areas between the PN and NP layers, 
which would be equal if you just jused two diodes, but usually are 
dissimilar in a transistor. That can be solved by using different 
diodes, or putting two in parallel.

What else? Yeah, the amount of doping is also relevant. All this ends up 
to is that different transistors have different specs (just as different 
diodes do). You'd hardly replace a transistor with a couple of diodes on 
a card anyway, since it would be so much trickier soldering wise, and 
possible space wise anyway, not even considering characteristics. But 
that don't change the fact that you get a working transistor with just 
two diodes.

...
   > A
transistor (and now I'm just talking about bipolar transistors, since 
 > it's the easiest, and they are very common) is extremely simply to 
 > understand and build.  
 I've heard of people disassembling a diode and making a point contact 
 transistor, but that is already leveraging a lot of technology that put 
 the $.10 diode in your hand to begin with. 
Well, I did say that you'd have to use diodes. If you consider it 
cheating to use diodes, then I plead guilty. It's very difficult to make 
a transistor if you're only allowed to use copper, aluminium and silicon.

However, all this was just an exercise in sidetracking. The real point 
was that something built with transistors can be repaired by just 
replacing the transistor. A very standard, small and cheap component 
that you probably still can get a hundred years from now. A graphics 
card will never be at that situation. If it's broke, you replace it. Ten 
years from now, that will not even be an option, since those graphics 
cards are no longer manufactured, and even the bus which they interface 
to is probably obsolete, meaning no other graphic card can be found 
either. And it is not possible to repair it, since it's just a bunch of 
highly intergrated ICs on it, which you cannot obtain, nor diagnose 
which one of them actually is busted. They are also close to impossible 
to remove.

	Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol