Author Options:

How/What would make terenary (trinary) computers and above more advantageous? Answered

I was wondering if going above binary computing would be very advantageous. I was listening to a professor's lecture at a summer camp at a university, and he talked about making computers that used light changing colors and lasers instead of only electricity. I figured that something like that would use terenary computing or above. Can you make a computer using non-binary from parallel circuits? If anyone can make a computer using terenary code or above, please post an instructable on it. A light computer would be cool too.


This is a very intelligent discussion, but the general idea I see is that it's not so great (now). Go ahead and read it all if you're curious about it.

About a decade ago I realised a concept I call "dualstream trinary." Since chips work on 0 and 1 in binary why not have these representing a base 3 trinary?- Well quite simply you can't represent every number value- but eureka if you run 2 streams of trinary that are added together you can represent every number. And given the groundbreaking work done in parallel distributive processing by such companies as Starwave corp. for the SETI@home program the mathematical principles are almost there. Where dualstream trinary comes into its own is in the much smaller bitstreams needed to process very large numbers. Please considor this my copyright on the design and thank you for allowing me to put the idea out there.


Parallel binary circuits would just be binary. We already do that.

Ternary and the like has been looked at many times, and the costs/complexities tend to outweigh the possible gains. The design is considerably more complex, which means you can't fit as many gates into a given space and have trouble running them as fast; electrical noise has a much greater risk of introducing errors (and serious computer design does have to deal with noise created by the circuits themselves as their power draw changes, as well as noise from outside), and generally there have been more disadvantages than advantages. Of course that might change if someone comes up with a really clever design, but so far...

Optical might reduce some of ternary's issues, but that is unproven as yet. I suspect it won't reduce them enough.

Even quantum computing has been being handled as "qbits", where the analog-ish component is the probability of one choice versus another.

For a better answer on "why binary", you might want to check whether MIT has put... I don't know what the current class number is; I knew it as 6.032... onto the Open Courseware servers. That's a wonderful class; it starts from "Here's the transfer curve of a transistor -- hey, it inverts! -- hey, we can put another input on it and get a NAND gate! -- but if we want that to run fast, we need to keep the transistor from going into saturation..." and builds up from there to digital logic, computer system design, and the implications on computer languages, one level of abstraction at a time and showing why those abstractions where selected. Other choices could be made in some of those cases, but there were good reasons the field evolved as it did.

I'm not very experienced in electronic engineering, but couldn't you make a computer that works by one circuit distributing commands to other circuits and making a system that mimics an above-binary system. I'm not in college if you're thinking that. I'm an accelerated 8th grader (half in 9th), but I dropped in on University courses a few times.

Yes, you can mimic a trinary or quaternary system by using two wires per trit. Heck, you can do it in software if you want to. But once you're emulating it, you've lost any advantage of going that way except the pedegogical one of experimenting with how it might work if it was practical.

(Actually, emulated quaternary winds up looking almost exactly like binary, since the two bits of data are spread across... uhm... two bits.)

It is possible but infinitely more complex to have a computer step above binary since right now transistor logic is based on input output of base 2. They either conduct, or they don't. On or Off. How would you even represent a third number? more wires? Sorry, but that isn't trinary, its multiple bits of binary. Where would that benefit?

As for quantum computing (switches that operate on photons of light instead of electrons) , that's far and above my comprehension -- but something involving a qubit, a packet of information that can be, and is both a 1 and 0 at the same time. The mind-blowing part is, if it is both, then the time it takes to make a calculation is no longer relevant, and theoretically, all calculations would be instant. Again, I do not profess to even thoroughly grasp the underlying physics to quantum computing, but I know once it happens, it will be to calculation what landing on the moon was to the steam locomotive.

Light computing and quantum computing aren't the same. Quantum computing carries out its processes inside atomic or molecular structures. Surprisingly, these have been made at a functional level. They're just not advanced enough to perform the full demands of modern computing. Light computing is a whole different subject where light is used in a normal circuit style. This concept was meant to reduce the problems of electronic noise, overheating, circuit aging, and many more problems in electronic computing. Lasers, fluorescent chemicals, and fiber optics would last much longer and eliminate these problems. Light computing is currently being developed with fluorescent chemicals that work like common electronic components. I don't know if someone has made such a device, but it is very intriguing. Instead of binary, a light computer could work in many different bases: off, infrared, red, green, blue, violet, ultraviolet, and as many recognizable wavelengths between.

In addition to the feature size issues which Orksecurity has pointed out, tristate logic (go look it up) is much more sensitive to errors: setting a single threshold (think of TTL logic) allows you to easily and readily distinguish states wth very low error rates.

Tristate logic requires two separate thresholds, with one state defined by the "in between thresholds" condition. Since the full range of signals (lowest voltage to highest voltage) is determined by the power supply, this means that your signal states are either closer together than in binary (and hence more sensitive to confusion == error), or else you need to use a higher voltage supply (which means more power, more heating, etc.).

If you do some research on your own, you will be able to find all of the information you need on what tristate computing has been implemented (and there are good applications for it), as well as why it has not been deployed more broadly.

When you've finished that research, why don't you come back and tell us all what you learned?

Its not a big deal. You use ground as one state, and +/- signal as the others. The noise margin is relatively easy to handle. The equivalent of Boolean Algebra in three states must be interesting though.

Yeah, but to get the same signal separation as existing 5V TTL, you need a 10V power supply (to get -5, 0, +5 signals). With "standard" 5V supply, and hence the same current/heat load, your signals are at -2.5, 0. and +2.5V.

Ternary logic is definitely interesting, and somewhat confusing. However, if you have some experience with databases (or even with Microsoft Excel), it should sound familiar: #REF and UNKNOWN are concrete examples of ternary logic.

Given that the processor on which you are typing now runs on ~1.8V.....

:-) The argument holds in any event. Existing designs are made so that voltage fluctuations do not cause state confusion (that is, the half-separation between states (whether 0.9, 2.5, or whatever) should be larger than the credible voltage fluctuation due to noise.

If you reduce that separation by a factor of two, then you increase the probability that a given fluctuation can cause confusion.

That's why logic has a "noise margin" - but you knew that.

Yes, indeed! And that's why I'm making my point. If you add levels without changing the voltage range, then you're reducing that noise margin.

And just running on higher voltages isn't as straightforward as it sounds. More heat to be dissipated, different transistor designs (remember, you don't want to drive the transistors into saturation or you'll lose a lot of speed!)...

It's worth looking at. It hasn't usually been a good answer.

Trinary systems are intrinsically more numerically efficient ways to represent numbers in a more compact configuration than Binary,

To answer Frollard, you have +ve, Gnd and -ve as your reference levels.