Balanced Ternary vs Binary

Binary won for reasons of 1940s engineering, not because two is the right number. This page sets the two systems beside each other honestly — including the places where binary is genuinely better and the cost of ever changing.

The Two Systems

Binary is base 2 with the digit set {0, 1}. Each position is a power of two and every digit adds its place value or contributes nothing. Negative numbers need a convention on top — sign-and-magnitude, ones' complement, or in practice two's complement.

Balanced ternary is base 3 with the digit set {−1, 0, +1}. Each position is a power of three and a digit may add its place value, subtract it, or contribute nothing. Negative numbers need no convention at all: they are already representable, and the representation of every integer is unique.

Where They Differ, and by How Much

42, and Then −42

In binary, 42 is 101010. To get −42 in an 8-bit two's complement register you invert every bit to 11010101 and add one, giving 11010110 — an operation with a carry chain, and a representation whose meaning depends on knowing the register is signed.

In balanced ternary, 42 is 1TTT0, where T is −1: that is 81 − 27 − 9 − 3. To get −42, flip every trit: T1110, which is −81 + 27 + 9 + 3. No carry, no register convention, no ambiguity about whether the value is signed. The balanced ternary calculator will do this for any number you give it, and show the carries.

Where Binary Genuinely Wins

Noise margin per state. Splitting a fixed voltage swing three ways leaves less room between states than splitting it two ways. Every electrical ternary logic family ever built has paid for its third state in margin, and that — not the mathematics — is why ternary hardware stalled. It is also the reason maniTLab's device work is photonic and phase-based rather than a third voltage level.

Seventy years of everything else. Fabrication, EDA tools, standard cell libraries, verification methodology, compilers, operating systems, file formats, network protocols and the training of every engineer alive assume two. Being right on radix economy does not buy any of that.

Bitwise work is genuinely binary. Masks, hashes, checksums and cryptographic primitives designed around bit operations do not become cheaper in base 3; some become harder. A ternary machine that has to talk to a binary world needs an interface — which is why one of the twelve patents is exactly that: Thatte9, a binary-ternary interface circuit.

So Why Would Anyone Switch?

Nobody switches for radix economy. The argument only becomes interesting when the underlying device is naturally three-state — when representing three values costs no more than representing two, because the physics already offers three. A carbon nanotube conducting one way, not at all, or the other way is that device, and encoding two states on it would be throwing a third away.

That is the whole thesis of this lab, and it is what the twelve patent specifications cover, from the device up through gates, a processor, memory, security, a compiler and a microkernel. Setun, built at Moscow State University in 1958, proved balanced ternary works in hardware; what it did not have was a naturally three-state part to build on.

Longer treatments: what balanced ternary is and how it works, the argued version in Why Balanced Ternary? The Case Against Binary, and the terms used across this site in the glossary.

Full Technical Brief Available Under NDA

Simulation data, quantum transport results, fabrication specifications, and complete patent claims are shared under mutual NDA only.

Follow maniTLab

New research, patent updates and essays, as a feed your reader collects for you. No list to join, no address to hand over, nothing to unsubscribe from.