I've been brushing up on semiconductor physics, and I've been wondering if there has been any research into constructing multistate transistors with resonant tunneling diodes (or devices like them)?
I think that in theory it shouldn't be more difficult to construct RTDs with several zones of negative differential resistance than those with a single zone.
See attached current-volage diagram, where the red curve is the behavior of a RTD, and the black curve is what I propose might be a building block for multistate transistor-like devices (transistor is probably a misnomer, but I think they could be coaxed to do the same job).
Okay. I'm not 100% certain it's possible to make all the components required with this technology. It may or may not be doable. It was just an idea I had when I heard of some of the properties of quantum resonance devices. I will continue examining the possibility though.
Hey!! That's something new I am seeing first time....GoodLuck with your research. And if you get some time please post a little more detailed explanation.
I haven't found anything conclusive, but there seems to be a lot of articles that suggest that it's at least possible to construct multistate memory cells with this technology.
Personally I think that this approach doesn't have any advantages over regular discrete (binary) electronics. Another words: it is too complex (and expensive). But if we will support balanced ternary directly (as negative current for -1, ground for 0 and positive current for +1) we may have not so complex schematics that may be simpler and cheaper than discrete electronics for the same amount of processing information.
P.S. I almost installed my second desktop computer that is Power Macintosh G4 Cube and may start working on big-endian adaptaion of your code.
And it mentions the possibility of using RTDs to implement multi-state logic, and also describes an interesting RTD and FET hybrid that allows for the multi-state logic while still being relatively cheap to produce.
I also stumbled across mentions of memristor latches being able to, to a higher degree than transistor logic at any rate, support multi-state logic.
Do you know how does modern flash memory work? There are each single transistor can hold up to 6-8 bits. Logic part can read transistor state and detect how many electrons it has. E.g. 5000 electrons is logical 1, and 30000 is logical 0. Controller can perform non destructive reading of the transistor state.
Personally I don't care about "multivalued" logic - it is out of scope of my interests. Balanced ternary is much better because it is optimal (or at least as much optimal as possible).
But ternary logic, balanced, non-balanced, and other ways of doing it, is a subset of multivalued logic. So advances in implementing multi-valued logic at large is almost certainly helpful for ternary logic as well.
eudoxie wrote:
But ternary logic, balanced, non-balanced, and other ways of doing it, is a subset of multivalued logic. So advances in implementing multi-valued logic at large is almost certainly helpful for ternary logic as well.
Basically "binary logic" also may be part of "multivalued logic" family, because it has more than 1 number of levels
Heh, that's technically true, but it's almost never used in that sense, since single-valued logic is so trivial it's useless, as the only available logical operation is unity; therefore, multi-valued logic is almost always taken to logic with 3 or more logic states.
