flat assembler
Message board for the users of flat assembler.
Index
> Heap > balanced ternary computing 
Author 

revolution
I think that before you start designing the hardware and selecting the transistors you will need to define the set of functions you need to support. Boolean functions are only good for binary, what is the ternary equivalent? Once you can answer that then you can decide whether or not CMOS (or another) topology is suitable.
Last edited by revolution on 15 Dec 2011, 15:12; edited 1 time in total 

15 Dec 2011, 02:12 

shoorick


15 Dec 2011, 10:34 

Matrix
MCD wrote: As I went on to a chase for unusual number systems, I encountered the balanced ternary system  a positional system using 3 different figures: 1 0 and 1, and the best of it is that numbers don't need a separate sign symbol to denote negative numbers!... ok, so how is that superior to a binary system? btw.: i'd like to have a "16384 level primitive 3d storage matrix", so i can store 128 TB of data on my micro SDXWC type keychain pendrive (size = 15x11x1mm) shoorick wrote:
haha, that triangle PCB reminds me of a fluxcapacitor 

15 Dec 2011, 14:58 

cod3b453
I think {1,0,1} probably could be implemented in CMOS using polarity and gates similar to bidirectional thyristors; though my guess is this is more silicon than the two bits you could have to represent the sign bit I have no idea how such a system would work but it's likely there'd be as many tricks you could use.


17 Dec 2011, 20:53 

edfed
build the 1,0,1 logic is not a problem at all. it is exactlly the same as the 0/5V and 12/+12V binary logic, mixed with full and halves bridge gates. it is easy, and very powerfull.
then, how to encode numbers with trinary? what would be the result of 1 and +1, etc, etc... first, define the arithmetic, second, define a way to use it in our real world (decimal domain, using multiplications, additions, divisions, pi, sinus, etc...), and then, we wlll design and print the circuit. 

18 Dec 2011, 02:12 

malpolud
Yeah, revolution is right, the most important thing is the logic. Encoding numbers with potentials under and over 0 is easy, 1 is low, 0 is middle, 1 is high
SHL1 = multiplying by 2 00010001 *2 = 00100010 but what happens when we SHL trenary? The same thing but the SHLed number tripples 1 1 1 1 0 0 1 1 = 0d44 will become 1 1 1 0 0 1 1 1 =0d132 About boolean logic: maybe some simple bit logic could be converted into a comparison logic? 1 = 1  true 0 1 = 0  false 1 (more) 0 = 1  false 1 (less) Of course a single byte will contain much more information than a binary byte, but the question is if trenary logic can be used more efficiently than binary by electronic circuits? 

18 Dec 2011, 23:43 

Matrix
You might be better off defining a trinary system with values [02]
(octal is [07]) A trinary bit can contain zero=0, half/floating=1, hi=2 Advantage is: it uses space more efficiently Disadvantage is: floating level is slower to handle, and noise immunity is lower. Same logic can be extended to any amount of data in a single bit, 16 bit information stored in an analog way needs to have 2^16 levels with good signal to noise ratio. 

19 Dec 2011, 15:26 

< Last Thread  Next Thread > 
Forum Rules:

Copyright © 19992020, Tomasz Grysztar.
Powered by rwasa.