“It’s a piece of junk. The fuel system leaks all over the place! It’s a piece of junk!”
“Always with the negative waves, Moriarty. Always with the negative waves!”
Okay, they were discussing a Tiger, but it was really a T-34 in drag.
Kindasorta. They built one in the late 50s for the *Seawolf *which had leakage problems, then scrapped the whole idea as being too complicated to operate and maintain and too prone to failure, with little upsides to speak of compared to a regular reactor.
I took that to mean the first actual weapon (not a prototype) was a Soviet made one.
It’s the “when it worked” part that was the problem. A quarter to a third not being operational was the norm, not something that was an issue with early production models at Kursk. At Kursk:
Operational rates with later models: Eastern Front:
That’s a 29% non-operational the day before a surprise offensive when the Germans had been husbanding their reserves, its not like they were suffering a high non-operational rate from recent action. This is what it looked like when they were suffering from recent action, at the end of the Ardennes Offensive, a 66% non-operational rate:
I also recall reading that the Tiger was very hard to repair in the filed-to change out an engine was a huge task (in a garage)-out in the open, damn near impossible. Everything a read about the T-34 was that it was very repairable , and spare parts were plentiful.
I know I am treading a fine line but I think the Tiger was far more advanced than the T34. However, the T34’s were able to be built in such overwhelming numbers and were so rugged they didn’t need to match the Tiger in capability.
(There is also an argument that the Tiger was terrible for Germany because it tied up so many resources).
Also, if the somewhat naive British Govt of the day had not given the blueprints of the Nene engine to the Soviet Union, would the Mig 15 even have happened?
Define “advanced”. Is a tank that cannot cross bridges, or travel along a road without irremediably fucking it up, or travel very long across broken country either “advanced” ? Is a tank that suffers so many losses to routine, self-induced mechanical failure than combat damage “advanced” ?
The T-34 was a simpler design, perhaps - but it takes a very smart person to design a simple machine that just… works. And works. And works. All the way to Berlin.
The Tiger was “advanced” in a way a very large Rube Goldberg device is. Overdesigned, overcomplicated with little thought given to practical realities, ultimately lost sight of its point sometime during development. As well, its construction relied on a quality of materials and fine, precision engineering that Germany simply wasn’t able to provide any more.
One of the nice stories from the development of the 747 is the design of the undercarriage. During the design process it became apparent that the undercarriage would have to be made of titanium in order to meet the weight goals. Problem was, Boeing didn’t have the expertise. So a deal was struck at the Paris airshow for a technology swap with the Russians. Boeing swapped swept flexible wing design for titanium expertise. (Cite: Clive Irving - Wide Body: The Triumph of the 747)
This is a curious deal - it is hard to imagine that the Lockheed Skunkworks didn’t know a great deal about titanium, but that was all funded from the military budget, and the US has rather strict rules about the transfer of technology from a militarily funded project to a civilian one.
But no doubt, the Soviets were very good at titanium.
ISTR that the Soviets were actually ahead of the West on titanium forgings and the like during much of the Cold War.
A lot of this was because they had their own native sources of titanium, whie the US did/does not. So we developed a range of very advanced steels and aluminum alloys, and they just worked with what they had.
Think about it- titanium alloys were generally Space-Agey stuff reserved for the SR-71 and things of that nature in the US during the Cold War, while in Russia, they made submarine hulls and other much more mundane stuff out of it.
Here is the best article I can find on it. Many of the claimed advantages seem to be a function of its instruction set architecture, not it being ternary. It looks to me that it was kind of inherently microprogrammed.
I wouldn’t put that much credence in their claims of superiority. One of the biggest problems with the Russian papers I’ve seen submitted was the almost total lack of understanding of where the West was in computer design - not the fault of the authors, but of the system.
If ternary logic had ever been adapted for computer systems in the West, it would no longer be used. Now that we are down to 1.0 volts or so, distinguishing two states is hard enough.
I don’t see any support for better accuracy, by the way. I don’t even know what that means. Improved reliability perhaps, not any evidence for that either.
There’s a method of logic gate construction called triple modular redundancy.
Simply put, you design your electronic circuits at the gate level, but you use a gate library that specs out 3 physical gates for every logic gate on your circuit, and you have majority gates on the output lines.
This obviously increases the number of physical transistors needed by a factor of 3-4 or so, and slows it down because of additional propagation delays (since all the signals have to go through a majority gate in between every step, so it’s probably going to run at half speed at best)
Anyways, the setups to do this in ternary logic might be physically smaller or use shorter logic paths, since you are getting more done in each stage (since a trinary output obviously contains more information, so you need less majority gates)
Oh. I didn’t mention why you go to all the trouble of doing this. This is how you make radiation hardened processors and FPGAs.
[quote=“smiling_bandit, post:6, topic:717167”]
That said, the Soviets were clearly competitive in the field of aerospace technology. Their planes were top-notch.
[QUOTE]
Forgive me for no cite, but well do I remember reports when in the late 70’s or early 80’s, the first combat between US fighters (F14 or F16 IIRC) and MIGs. They were Israeli versus some other country (Syria?); both were earlier models than the latest US or Russian planes, but were contemporaries. All MIGs were destroyed before visual contact.
I guess the MIGs could outfly the US planes and might have won a dogfight, but at least with the Israeli electronics (no doubt latest and probably as good as US, maybe better), it was no contest.
At least, that was the word on it at the time. Perhaps someone here has some actual facts to relate here!
Russian papers are political documents. They were virtually required to comply with party line, which isn’t conducive to science. I had a friend who reviewed a number of Russian papers in his field of metallurgy, and he reported that the vast majority were complete crap. On the other hand, you can’t have a successful rocket program with bad materials science, so perhaps the published work wasn’t representative.
Either accuracy doesn’t count, or it’s an accuracy/speed tradeoff. That is, a larger modulus for stored numbers (let’s call them “words”) yields higher accuracy, but that can be compensated for by combining multiple words to represent a single number – which slows down operation.
A larger modulus provides higher single-word accuracy for non-integral results (fixed point or floating point).
Strictly speaking, within its memory limits, one computer can’t be “more accurate” than another, just faster at a given accuracy.
I haven’t seen TMR gates mentioned for decades. Yes I know what it is, I’ve reviewed books on reliability in this sense. I wonder if anything has been made with TMR gates - besides the problems you mention they would suck up power. I have worked with some national labs working on ICs for various highly reliable applications and they sure don’t use them. TMR is typically done at the module level.
However, even if you did use them, it would fall into the reliability bucket, not the accuracy bucket.
One thing I wonder about is whether there is a three state Boolean algebra. And I don’t see a lot of majority gates these days.
There used to be a multivalued logic technical committee in the IEEE Computer Society. Not sure if it is still in existence, but I haven’t seen much from it lately.
My friends in defense were really, really interested in rad hardened hardware. But they did it though design and process methodology. They had their own fab, and licensed processor designs which they built themselves. No TMR gates involved, I assure you.
Plus you can do a good job with less expensive solutions. There was an interesting project at Stanford where they sent a cpu board up on a satellite that flew through regions with lots of nasty particles. They left it unshielded, and were able to demonstrate that they could get good results through software techniques.
BTW there was a flurry of interest a while ago about redundant flops to deal with cosmic ray issues. it was shown that you really didn’t have to worry about gates since there is a very small probability that a hit on a gate would cause a temporary bit flip when the output value was critical and at a time when it would affect a storage element. Flops obviously would cause a problem for a much longer period. However, I’m unaware of this being a problem so far, even with 24 nm geometries. Maybe some day, but I’ll be retired by that point.
This was after the fall of the Soviet Union. Never even saw a submission before that.
I’ll have to check but I think there were people from the Soviet bloc (but not Russia) at an architecture workshop I went to in Austria in 1980. But they had slightly more opportunity to travel than the Russians.
No, there was the French Schneider CA 1, the SOMUA S35 , etc.
As for the Tiger I
Joe-4 wasn’t a “real” thermonuclear weapon, in that it was a layer-cake/sloika design, which is a sort of hybrid, as opposed to the classic Teller-Ulam/Sakharov design, which the US did pioneer.
They were quite innovative when it came to listening devices and other eavesdropping devices. They were the ones who came up with a bug planted in an ambassador’s shoe heel. In another case, they managed to plant a bug in the United States of America seal on the front of the ambassador’s podium.
One of the more ingenious devices was planted in a goodly number of IBM Selectric typewriters. For some reason, our embassy thought it was a good idea to farm out typewriter repair/maintenance to a local vendor. The Selectric had an aluminum bar inside the machine that was in proximity to where the ball struck the paper. The KGB electronics folks figured out how to implant a keystroke sensor into a similar metal bar and replaced the original with it. When plugged into a grounded outlet and turned on, each key stroke emitted a unique signal that could be picked up downstream.
The only way it was discovered was because the Soviets had replaced the aluminum bar with one made of steel. Somebody discovered, during a routine check, that a magnet stuck to the supposedly-aluminum bar, which lead to the discovery of the electronics inside.
This was during World War II, not the Cold War, but it’s interesting (if true). The story is that Churchill visited Moscow in 1942 and was very impressed with the mixer taps that let you get hot and cold water in any proportion from one spigot. In the UK, the two-tap system was still in use. I don’t know when mixer taps became widespread in the US. But if the story is to be believed, Soviet plumbing technology was at least ahead of Britain’s for a time.
http://www.xilinx.com/support/documentation/application_notes/xapp197.pdf
It’s actually pretty easy to do, relatively speaking. You can use the same FPGA synthesis tools, you just use either special FPGAs that have the TMR implemented in the hardware itself, or you use synthesis libraries that will spec out your design by adding the additional logic automatically.
By using FPGAs for this, in theory you can launch the same chip in your satellite that is used in millions of terrestrial uses, except that the chip in the satellite has the additional gates for each operation. This makes the chips cheaper and easier to obtain.
Also, you can build lots of different satellites using the same few chips for all the digital logic tasks.
The big advantage over this versus trying to do it in software is that there are vastly more failure modes in software - a corrupted bit could affect the instruction pointer in many different ways, sending execution off somewhere unknown.
FPGA logic can be constructed in ways where this can’t happen - you don’t implement a general purpose CPU, you construct a machine that is only capable of very specific tasks. You do not have an instruction pointer - instead, discrete areas of circuitry run in parallel with each other perpetually, and data just flows from one region to another. Much easier to check the data flows for errors this way.
The Mars Science Laboratory rover (Curiosity) uses this.