> During 25 years of operation no failures of the system were noted when working in control systems. By the production volume it is unrivalled among space computers.
Not only still in production, but still in production with no failures reported. This is beyond amazing. Triple redundancy and hardware majorization will surely help you greatly, but no failures across those numbers is just mind-boggling.
The system is essentially three identical computers wrapped in a control system. Hardware majorization means the computers have to agree on the computed result, that is, each computes the output from the given inputs independently and the results are then compared. This mitigates errors stemming from any kind of hardware faults, except for the control system itself.
.. unless of course the standard application for the device is in safety-critical/life-protection systems, in which case "no failures noted" better be honest.
Why? They try to keep all information secret, especially the one that make USSR looks bad. For example, they denied existence of Level 6 nuclear accident for 30 years [0].
If the nuclear accident which had 10,000 people evacuated was suppressed and denied for 30 years, why would you trust them to publish any negative information about military computer?
The first launch was Soyuz MS-01 on July 7, 2016 aboard a Soyuz-FG launch vehicle towards the ISS.
Soyuz MS Improvements
The new computer (TsVM-101), weighs one-eighth that of its predecessor (8.3 kg vs. 70 kg) while also being much smaller than the previous Argon-16 computer.
The new modernized Soyuz TMA-M[citation needed] series was developed and built by RKK Energia as an upgrade of the baseline Soyuz-TMA. Thirty-six obsolete pieces of equipment have been replaced with 19 new-generation devices and the vehicle's total mass has been reduced by 70 kilograms (154 lbs).[3] In particular, the reliable but heavy (70 kg) Argon digital computer[4] and analogue systems, which had been used on Soyuz ships for more than 30 years, has been replaced with a new digital computer, the TsVM-101, and digital avionics.
Two flight development flights were launched: Soyuz TMA-01M on Oct 7, 2010 and Soyuz TMA-02M on Jun 7, 2011.
I suspect this is due to the small thermal window of (then state of the art) germanium transistors which was still in widespread use. It wasnt until the mid 1960s that the silicon transitor overtook germanium in terms of speed. And the USSR was many years behind the West, in terms of materials & design.
Interesting related fact. The thermal operating range of germanium transistors was a concern even in the music world. The "Fuzz Face" pedal which helped give Jimi Hendrix his distinctive distorted guitar tone was built with two germanium transistors in a very simple circuit (< 5 components if I remember correctly). The distortion produced by this pedal was well known to change on stage, under hot lights or in cold basements, etc. To make matters worse, it was also sensitive to the supply voltage, so a fresh 9v acid battery would sound different than a dying battery. (ironically, the dying battery sounded better to most ears. The circuit only drew around 30 mV so a near-dead battery could still last a long time.) As a result, it was not unusual for working musicians to travel with several fuzzfaces and a pocket full of batteries, trying a few at soundcheck to see which sounded the best.
Correct, component base lagged behind in USSR and quality assurance was tricky. Even in 1980s there were 5 rejection classes of bipolar transistors compared to just two in the West.
I counted 3-4 years R&D cycle prior to 1974 and then subtracted a few years to account lag in the technology. However for high profile projects like these it wasn't uncommon to simply source the quality Western components.
One of my coworkers in the 1990s was an older guy who was developing systems like that in early 1980s. A lot of stories, not sure how apocryphal. Once they had problems with one of the copied Motorola chips produced in Estonia. The deadline was looming, so they tried to source original chips. However that chipset was discontinued for years at that point with no stock anywhere, neither original Motorola nor compatibles. So they had to restart production on one of the fabs in Asia, using a KGB shell company to handle transaction.
Given the iffy translation it’s not clear if that it for the unit or the ICs hardened systems like this are usually thermally stabilized to ensure that all components operate within their temperature range and that there is no large temperature variance between redundancies.
The power consumption (280W) for the weight (70kg) seems rather low. What sort of power consumption did similar era computers, that took up a room, have?
It didn’t took a room it’s a flight computer iirc it’s still in use on the Soyuz.
And it sounds about right for a 70’s era hardened military computer.
The US used 5 AP-101s for the shuttle they where about 22kg each and the power figures I could find range from 370w to 480w.
Bare in mind that there is a good chance that a large part of the power consumption could have went into thermal stabilization rather than computation given that the system power consumption of say an Apple II was around 5w.
No need for it in the crew compartment, overall it’s likely worth bringing and additional 70kg of equipment back than a computer that can likely won’t survive the landing.
Is there anything particular you would like to know? I've heard a little about Setun from my parents, but nothing too exciting, I can ask around if you're interested.
For a little context: my mother studied cybernetics and father studied electrical engineering in USSR in 70s. They still have their college textbooks lying around, so I can try to dig something up for you from that era. But keep in mind, even in 70s it was obvious that USSR was decades behind USA. Heck, my first algorithm textbook was my parents' 2nd edition of Knuth's TAOCP (interestingly, my parents combined have 1 copy of 1st volume, 0 copies of 2nd volume, and 2 copies of 3rd volume).
DSSP, what’s so special about it? I sense LISPers and Forthers sense of “enlightenment” (which I have not obtained). But what is so special about DSSP that it’s considered a generalization of Forth?
Setun:
just about everything! I read it failed lately because bureaucrats stymied engineer’s creativity to do things the “Western” way (it seems this slowed the Soviet’s bomb since their scientists figures a better way to do something)
How were logic gates defined ? (Ie what type of ternary logic was used)
What were the voltage levels to represent the states?
I wanted to note that language's name in Russian (РАЯ, RAYA) is most likely a play on words. It is an abbreviation of phrase "Extensible Adaptive Language" and at the same time it is a female name. And Google Translate mistakenly translates is as "paradise", so one should read this header:
> Developed adaptive language PARADISE dialogue programming system PRSP
as "Extensible adaptive language RAYA for Dialogue (Interactive) Programming System (DSSP)". I think "Dialogue" here means something like REPL, where one can input commands and expressions and see the result.
I looked through Google Translate copy of the article about RAYA and DSSP - it is somewhat readable. Here are some words that machine translation gets wrong:
- PRSP, PSP, DPF -> incorrect translation for DSSP (Dialogue (Interactive) System for Structured Programming)
- paradise, heaven -> RAYA
- Developed adaptive language -> Extensible Adaptive Language
- "is known as descending programming" -> "is known as a waterfall model"
- "or as letters of the I / O alphabet" -> probably means "as letters for Input/Output of text"
- entered into -> introduced, added
- "Data types and their associated constraints and checks can be entered into problem-oriented language extensions." -> "can be introduced (added) by ... extensions"
- team -> command
- vertex -> top (of the stack)
- sub-top, sub-row, sub-vertex, sub-bottom, bottom, sub-bush, subspace, subshift -> second value from the top of the stack (I am surprised how many ways Google has to translate the same word)
- "assignment to her team! X value of the last (top) item" -> "assignment a top value from the stack to it using command !X"
- "Signal the end of the input and the command processor" -> Signal of the end of input and the command to processor ...
- binary additional code -> two's complement code
- "The byte output command to the terminal issues a letter on the screen..." -> "TOB", the command for byte output ...
- "present a modifiable factor of the variable K"-> represent a modifiable factor with a variable K
- The PRSP processor distinguishes words from the first seven letters, identifying them by polite comparison -> The DSSP processor uses only first seven characters of the word, identifying words with letter-by-letter comparison
- monologue -> "monoword", single-word command
- verbose commands -> multiword commands
- pre-team -> special command prefix, which authors call "pre-prefix", like "BYTE" before "VAR"
- The remainder is always divisible. -> The remainder has the sign of divisible
- subloviera, sublovier, subcouples, subloader, sublorer -> sub-dictionary (don't know where Google took those words, I doubt they exist)
36 comments
[ 1.0 ms ] story [ 79.9 ms ] thread> Speed: addition - 5 ms, multiplication - 45 ms.
The times are in microseconds, not milliseconds.
> Время выполнения операций (мкс): сложения - 5, умножения 45.
мкс = микросекунды
Solid engineering, I guess.
> During 25 years of operation no failures of the system were noted when working in control systems. By the production volume it is unrivalled among space computers.
Not only still in production, but still in production with no failures reported. This is beyond amazing. Triple redundancy and hardware majorization will surely help you greatly, but no failures across those numbers is just mind-boggling.
If the nuclear accident which had 10,000 people evacuated was suppressed and denied for 30 years, why would you trust them to publish any negative information about military computer?
[0] https://en.wikipedia.org/wiki/Mayak#Kyshtym_disaster
If its good enough for NASA, its good enough for me. Or are you suggesting that NASA also fell for the scam?
The first launch was Soyuz MS-01 on July 7, 2016 aboard a Soyuz-FG launch vehicle towards the ISS.
Soyuz MS Improvements
The new computer (TsVM-101), weighs one-eighth that of its predecessor (8.3 kg vs. 70 kg) while also being much smaller than the previous Argon-16 computer.
And from https://en.wikipedia.org/wiki/Soyuz-TMA :
The new modernized Soyuz TMA-M[citation needed] series was developed and built by RKK Energia as an upgrade of the baseline Soyuz-TMA. Thirty-six obsolete pieces of equipment have been replaced with 19 new-generation devices and the vehicle's total mass has been reduced by 70 kilograms (154 lbs).[3] In particular, the reliable but heavy (70 kg) Argon digital computer[4] and analogue systems, which had been used on Soyuz ships for more than 30 years, has been replaced with a new digital computer, the TsVM-101, and digital avionics.
Two flight development flights were launched: Soyuz TMA-01M on Oct 7, 2010 and Soyuz TMA-02M on Jun 7, 2011.
Interesting related fact. The thermal operating range of germanium transistors was a concern even in the music world. The "Fuzz Face" pedal which helped give Jimi Hendrix his distinctive distorted guitar tone was built with two germanium transistors in a very simple circuit (< 5 components if I remember correctly). The distortion produced by this pedal was well known to change on stage, under hot lights or in cold basements, etc. To make matters worse, it was also sensitive to the supply voltage, so a fresh 9v acid battery would sound different than a dying battery. (ironically, the dying battery sounded better to most ears. The circuit only drew around 30 mV so a near-dead battery could still last a long time.) As a result, it was not unusual for working musicians to travel with several fuzzfaces and a pocket full of batteries, trying a few at soundcheck to see which sounded the best.
I counted 3-4 years R&D cycle prior to 1974 and then subtracted a few years to account lag in the technology. However for high profile projects like these it wasn't uncommon to simply source the quality Western components.
One of my coworkers in the 1990s was an older guy who was developing systems like that in early 1980s. A lot of stories, not sure how apocryphal. Once they had problems with one of the copied Motorola chips produced in Estonia. The deadline was looming, so they tried to source original chips. However that chipset was discontinued for years at that point with no stock anywhere, neither original Motorola nor compatibles. So they had to restart production on one of the fabs in Asia, using a KGB shell company to handle transaction.
But they don't mention anything else besides that about stability. No environmental hardening or else.
And it sounds about right for a 70’s era hardened military computer.
The US used 5 AP-101s for the shuttle they where about 22kg each and the power figures I could find range from 370w to 480w.
Bare in mind that there is a good chance that a large part of the power consumption could have went into thermal stabilization rather than computation given that the system power consumption of say an Apple II was around 5w.
Edit: Found a nice source for a lot of information including comparision between the Argon and the TsVM: http://www.bis-space.com/belgium/wp-content/uploads/2015/05/...
The new module is 8.3kg and consumes 46 watts of power.
- Setun, a ternary (as opposed to binary) computer
- DSSP, a Forth like language that users seem to gush about in the same manner Lispers gush about LISP.
Info/history on both those systems are increadibly difficult to obtain in English!
For a little context: my mother studied cybernetics and father studied electrical engineering in USSR in 70s. They still have their college textbooks lying around, so I can try to dig something up for you from that era. But keep in mind, even in 70s it was obvious that USSR was decades behind USA. Heck, my first algorithm textbook was my parents' 2nd edition of Knuth's TAOCP (interestingly, my parents combined have 1 copy of 1st volume, 0 copies of 2nd volume, and 2 copies of 3rd volume).
Setun:
just about everything! I read it failed lately because bureaucrats stymied engineer’s creativity to do things the “Western” way (it seems this slowed the Soviet’s bomb since their scientists figures a better way to do something)
How were logic gates defined ? (Ie what type of ternary logic was used)
What were the voltage levels to represent the states?
What were the performance specs?
What did it do well? What did it do poorly?
Can it be re-implemented in CMOS?
Setun emulator, quick start and manual: http://trinary.ru/projects/setunws/
DSSP doc: http://trinary.ru/kb/d62d8074-50eb-422e-b571-fe2a9ab80584
There is even pdf with list of all Setun machines produced.
Google Translate should be enough to get through presented materials.
> Developed adaptive language PARADISE dialogue programming system PRSP
as "Extensible adaptive language RAYA for Dialogue (Interactive) Programming System (DSSP)". I think "Dialogue" here means something like REPL, where one can input commands and expressions and see the result.
I looked through Google Translate copy of the article about RAYA and DSSP - it is somewhat readable. Here are some words that machine translation gets wrong:
- PRSP, PSP, DPF -> incorrect translation for DSSP (Dialogue (Interactive) System for Structured Programming)
- paradise, heaven -> RAYA
- Developed adaptive language -> Extensible Adaptive Language
- "is known as descending programming" -> "is known as a waterfall model"
- "or as letters of the I / O alphabet" -> probably means "as letters for Input/Output of text"
- entered into -> introduced, added
- "Data types and their associated constraints and checks can be entered into problem-oriented language extensions." -> "can be introduced (added) by ... extensions"
- team -> command
- vertex -> top (of the stack)
- sub-top, sub-row, sub-vertex, sub-bottom, bottom, sub-bush, subspace, subshift -> second value from the top of the stack (I am surprised how many ways Google has to translate the same word)
- "assignment to her team! X value of the last (top) item" -> "assignment a top value from the stack to it using command !X"
- "PRSP processor", "PSP processor" -> DSSP processor
- Keyboard numbers -> Numbers, typed at keyboard
- "Signal the end of the input and the command processor" -> Signal of the end of input and the command to processor ...
- binary additional code -> two's complement code
- "The byte output command to the terminal issues a letter on the screen..." -> "TOB", the command for byte output ...
- "present a modifiable factor of the variable K"-> represent a modifiable factor with a variable K
- The PRSP processor distinguishes words from the first seven letters, identifying them by polite comparison -> The DSSP processor uses only first seven characters of the word, identifying words with letter-by-letter comparison
- monologue -> "monoword", single-word command
- verbose commands -> multiword commands
- pre-team -> special command prefix, which authors call "pre-prefix", like "BYTE" before "VAR"
- The remainder is always divisible. -> The remainder has the sign of divisible
- subloviera, sublovier, subcouples, subloader, sublorer -> sub-dictionary (don't know where Google took those words, I doubt they exist)
I hope it helps to understand the text.
(shoutout Artiom)