Ternary Computing and its impact

So one thing I discovered while looking into the Soviet Union is that in 1958 that Setun was made. One of the world's only Ternary Computers. What this means that instead of using 0s and 1s, Setun used -1s, 0s, and 1s.

This got me thinking what does this mean for computers if computers were built based of of balanced Ternary how does that change computing? Are there any developments that are slowed or increased? From what I can tell this hasn't been that explored, so I am curious what ideas might come about with this focus and development?
 

BigBlueBox

Banned
This question is perhaps better suited for a physics and mathematics forum. That being said, there’s a reason why binary is dominant: because it’s easier. Sure, there are theoretical advantages to ternary logic, but when it comes to implementing it in hardware binary is just simpler, and simpler is better. A binary transistor counts everything below a certain voltage threshold as zero and everything above that threshold as one. With ternary getting involved there would need to be two separate thresholds. Finally, a ternary computer is not “50% more efficient”. It can store numbers efficiently than than binary because 3 is closer to e (~2.718) than 2, but that’s not enough to overcome the disadvantages. If ternary computers ever make a comeback it would be in some exotic computing technology of the distant future like optical or quantum computers, not the silicon transistor computers we all know and love.
 
adding 50% more efficiency into computing means more leeway for bad code so you'd see faster software innovation

Hmm, so you think we could see Moore's Law happening at a faster rate earlier perhaps?

This question is perhaps better suited for a physics and mathematics forum. That being said, there’s a reason why binary is dominant: because it’s easier. Sure, there are theoretical advantages to ternary logic, but when it comes to implementing it in hardware binary is just simpler, and simpler is better. A binary transistor counts everything below a certain voltage threshold as zero and everything above that threshold as one. With ternary getting involved there would need to be two separate thresholds. Finally, a ternary computer is not “50% more efficient”. It can store numbers efficiently than than binary because 3 is closer to e (~2.718) than 2, but that’s not enough to overcome the disadvantages. If ternary computers ever make a comeback it would be in some exotic computing technology of the distant future like optical or quantum computers, not the silicon transistor computers we all know and love.

I am more looking at what would have happened if the Soviets decided to adopt Ternary in the 60s following the work on Setun as part of say their research into making OGAS a reality, or as much as a reality as the 20th century can allow for and the impact this would have on computer development. :)
 
No change to moore's law, but expect faster innovation as far as what's done with the code. More countries with videogame industries, perhaps.
 

trurle

Banned
So one thing I discovered while looking into the Soviet Union is that in 1958 that Setun was made. One of the world's only Ternary Computers. What this means that instead of using 0s and 1s, Setun used -1s, 0s, and 1s.

This got me thinking what does this mean for computers if computers were built based of of balanced Ternary how does that change computing? Are there any developments that are slowed or increased? From what I can tell this hasn't been that explored, so I am curious what ideas might come about with this focus and development?
This was basically a technological deadlock - because although base of 3 is more advantageous being closer to e=2.71828, all advantages are erased by reduced noise margin of voltage-level approach as chip integration level increase.
You get +50% of data and +300% of noise at the same power for Setun-styled base-3 computation.
The only semi-realistic way to use base of 3 so far is base-3 PSK keying of data in data-bus (simple binary is OOK keying). This mean +50% of data and +33% of noise compared to base-2 PSK (differential signalling), netting some advantage. Unfortunately, the actual circuit to transmit base-3 PSK data (0 degrees, 120 degrees and 240 degrees) do include ring oscillator which had significant technological problems early on (to the point of deeming ring oscillator a "forbidden circuit" due frequently unpredictable behavior).

IOTL, the data bus evolved from binary (OOK) to differential binary (base-2 PSK), and wireless datalink skipped directly to base-4 PSK.
 
Last edited:

Ramontxo

Donor
Just to let you know that I have find this so interesting that I have posted a link of this thread to some friends...
 
So one thing I discovered while looking into the Soviet Union is that in 1958 that Setun was made. One of the world's only Ternary Computers. What this means that instead of using 0s and 1s, Setun used -1s, 0s, and 1s.

This got me thinking what does this mean for computers if computers were built based of of balanced Ternary how does that change computing? Are there any developments that are slowed or increased? From what I can tell this hasn't been that explored, so I am curious what ideas might come about with this focus and development?

The real problems for any PoD to do with Soviet computing are:

1) The Soviets really struggled to manufacture enough electronics
2) The Soviets produced software at a lower rate than the US

So if they go for ternary computing, I just don't see how they can get the critical mass behind it to really compete. As I understand it, ternary is harder to program in than binary, so the Soviets have even less software, and without software their computers really aren't too useful.

So unfortunately, this probably makes things worse for the Soviets.

fasquardon
 

trurle

Banned
What if ECL Logic is used?
Used that with the Sperry Mainframes for speed and noise control
ECL have better noise margin compared to TTL or CMOS due low impedance of output nodes, and very small power supply noise level. ECL will be obsolete by 1980 though, regardless of ternary or binary architecture. Power dissipation of ECL was too high. Last technology iteration (IIL) has improved power dissipation of ECL, but it finally lost to CMOS by 1990 due larger chip area (simple NAND2 IIL had 7 components, each of them thrice in area of MOSFET, while CMOS NAND2 had only 4 MOSFETs.
Overall, some early ternary ECL/IIL computers are possible, but i doubt they will evolve into PC.
 
Overall, some early ternary ECL/IIL computers are possible, but i doubt they will evolve into PC.
Early is what you want. ECL gives you an edge from the late '60s to the mid '80s, where you also get the advantage of higher clock speeds besides a lot less noise.

It would give them decent hardware, but really, the only 'Killer app' to come out of the USSR was Tetris. The Software is what really sells computers, even in Big Iron days.
 
The Software is what really sells computers, even in Big Iron days.

Sure, the reason why the Soviets invested so much in industrial espionage against IBM was because they wanted to get access to all the software that came with having the most popular machine of the day (the System/360 and System/370) while it's kinda borne some fruit today, with tech support for the remaining System/360 and 370 machines done in Russia by an outfit that also supports the Soviet era clone (the ES EVM), so far as I have read, it was a real failure for the Soviets. Their clone took years to get to work, came out just as the 370 was becoming obsolete and was different enough that System/360 and 370 software didn't run well on the Soviet machine.

All in all, it doesn't seem to have paid off the resources invested in it, and actually led to the Soviets falling even further behind.

Early is what you want. ECL gives you an edge from the late '60s to the mid '80s, where you also get the advantage of higher clock speeds besides a lot less noise.

Gotta say, this would be some really cool hardware. Not enough of it and not enough software for it sure, but really cool nonetheless.

Overall, some early ternary ECL/IIL computers are possible, but i doubt they will evolve into PC.

To be fair, the Soviets don't really need PCs. Microchips are super useful and all, even when not used in PCs, but what the Soviets really need is just enough reliable electronics and the software for it. It doesn't have to be as good at what the West has, it just needs to be plentiful.

But I'm not sure what kind of PoD would lead to plentiful Soviet ternary computers. Unless it involves a vastly less damaging Barbarossa.

fasquardon
 

trurle

Banned
To be fair, the Soviets don't really need PCs. Microchips are super useful and all, even when not used in PCs, but what the Soviets really need is just enough reliable electronics and the software for it. It doesn't have to be as good at what the West has, it just needs to be plentiful.

But I'm not sure what kind of PoD would lead to plentiful Soviet ternary computers. Unless it involves a vastly less damaging Barbarossa.
Actually a fine alternate history book series "Not late yet" (sorry, in Russian only) on exactly topic above can be found here:
http://samlib.ru/d/dmitriew_p/

The story centers around modern data network technician falling to Soviet Union of 1965. The main character is focusing on prematurely bringing electronics revolution (and in some parts succeeds). Unusually for Russian novels of this sort, views seems not to be politicized, and i rank it high in plausibility of storyline.
 
Last edited:
But I'm not sure what kind of PoD would lead to plentiful Soviet ternary computers. Unless it involves a vastly less damaging Barbarossa.

For soviet computers, I was thinking a longer Khrushchev rule followed by Alexie Koysgin leadership and his various reforms. Part of that would include a focus on computerization as plans to make OGAS more of a reality and less of a theory. Would this be a plausible scenario for this?
 
Ternary logic controllers are being used in the latest trackside railway signalling applications in the UK for certain detection inputs. The computers that process the data however are still binary.
 
For soviet computers, I was thinking a longer Khrushchev rule followed by Alexie Koysgin leadership and his various reforms. Part of that would include a focus on computerization as plans to make OGAS more of a reality and less of a theory. Would this be a plausible scenario for this?

Longer Khrushchev rule could well lead to higher R&D spending in the latter part of the 60s and into the 70s, yes.

Kosygin's reforms could turn out to be a bit of a mixed bag. Yes, he had some good ideas, but he also bought into many of the bad ideas that led the European satellites into disaster. But then again, the satellites were able to borrow from the West and were much smaller economies, so the West (well, the Western Europeans actually) may not be able/willing to lend the amounts of money that would allow the Soviet Union to get into the same over-consumption trap as Poland and Romania did. So maybe things work out in the end?

Also, the big issue with the whole OGAS idea is that the Soviets weren't in a position to computerize anywhere near to a significant number of their factories. The problem is not desire for computers, it's supply of computers. (Also, OGAS isn't magic, while it might serve a useful purpose, due to disrupting the corrupt patronage network, it's gonna go wrong in lots of interesting ways. Possibly, it could go wrong badly enough to be a net drain on the Soviet system.)

The story centers around modern data network technician falling to Soviet Union of 1968. The main character is focusing on prematurely bringing electronics revolution (and in some parts succeeds).

So how does the main character get enough electronics?

fasquardon
 

trurle

Banned
Also, the big issue with the whole OGAS idea is that the Soviets weren't in a position to computerize anywhere near to a significant number of their factories. The problem is not desire for computers, it's supply of computers. (Also, OGAS isn't magic, while it might serve a useful purpose, due to disrupting the corrupt patronage network, it's gonna go wrong in lots of interesting ways. Possibly, it could go wrong badly enough to be a net drain on the Soviet system.)


So how does the main character get enough electronics?
He was driving on the way to home after dismantling the obsolete local network. Therefore, fell back to 1965 together with jeep loaded with several crates of assorted electronic components, some of them primitive enough to be reverse-engineered back in 1965.

Actually in novels by Dmitriew you can find a severe beating of OGAS concept too. According to him, problem was not only meager supply of computers, but sabotage/corruption by users too. As long as OGAS have non-computerized components, these components will be easily sabotaged/corrupted.

I tend to agree corrupted patronage network was a problem, and likely it was much larger problem than you estimate.
 
Last edited:
So one thing I discovered while looking into the Soviet Union is that in 1958 that Setun was made. One of the world's only Ternary Computers. What this means that instead of using 0s and 1s, Setun used -1s, 0s, and 1s.

Also note that internally, Setun' used two binary bits to encode the trits (with 11 combination unused). This is not that different from BCD, but while BCD has straightforward applications and is rather intuitive, binary-coded-balanced-ternary has not. And BCD became obsolete once binary sizes and performance grew up enough to make conversions to/from decimal trivial.

So at the best, I guess it would be obsoleted with going down the BCD route, but even faster. It might linger a bit in communication lines (negative, zero, positive voltage), but it still makes electronics more complicated.
 

trurle

Banned
Also note that internally, Setun' used two binary bits to encode the trits (with 11 combination unused). This is not that different from BCD, but while BCD has straightforward applications and is rather intuitive, binary-coded-balanced-ternary has not. And BCD became obsolete once binary sizes and performance grew up enough to make conversions to/from decimal trivial.

So at the best, I guess it would be obsoleted with going down the BCD route, but even faster. It might linger a bit in communication lines (negative, zero, positive voltage), but it still makes electronics more complicated.
Triple branching (which is plausible with ternary computer) have good applications for sorting and database management (fast search). Together with data lines application sutability, ternary computing may linger in the phone and data switching hubs even after becoming obsolete for general applications.
 
Last edited:
Top