Dreaming up a new defence contractor

Dreaming up a new defence contractor

Summary: A bureaucracy extending far into the industries affected and having roughly the splendid indifference to mission and tax dollars otherwise characteristic of nationalised industries like the British and Canadian national health services.


On August 22, 1963 a guy named Joe Walker flew the Bell X-15 experimental spaceplane to 354,200 feet - a record not equalled until Brian Binnie flew SpaceShipOne to 367,442 feet in October of 2004.

What happened at NASA in between was basically that the national urgency put into the moon missions forced NASA to choose the heavy lift rocket approach over continuation of the space plane program - with the long term consequences we see today: a classic case of not having the time to do things right, and then getting trapped into spending fifty years trying to pretend that a successful short cut based on doing more of what you know how to do can substitute for real research and development.

Look at NASA today and what you see is a titanic struggle between past winners who claim the right to repeat the mistakes of the past and a much smaller group of entrepreneurial scientist/engineers who're trying to get the spaceplane program back on track. These are the people using SpaceShipOne's success as both the hammer and the anvil for their attacks on NASA's established bureaucracy -and getting no where fast because that bureaucracy extends far into the industries affected and has roughly the splendid indifference to mission and tax dollars otherwise characteristic of nationalised industries like the British and Canadian national health services.

The field of anti-submarine warfare offers a remarkable parallel to this: in which a detour undertaken for the best of reasons came to dominant subsequent planning to the point where it became impossible to return to the original long term path. There is no Ansari-X prize for submarine detection, but a private company able to do this significantly better than the U.S. Navy might well kick start a similar internal conflict -one that, in the  end, can only be won by proponents of the long term right answer.

The right answer comes from a simple observation: a submarine moving under water leaves a three dimensional wake - and that wake distorts the surface in predictable ways even if the sub is crawling along amid the rocks 900 feet below the surface of a Norwegian Fjord.

In other words: don't bother listening for the thing, because the other guy can always make his subs quieter - but look for its surface trace because the ability to do that well keeps the other guy's submarines in port, where they waste his resources.

So how? Well it's actually a relative simple (cough) photogrammetric application - one for which Altivec equipped parallel machines like IBM's cell processor are almost perfectly suited. The processing volume is high, the best and most timely data sources aren't directly available to civilian companies, but the job itself is not that hard: take a real time surface image, subtract surface differences that are both chaotic and consistent across the image, and look for the characteristic "embossed i" the submarine's wake leaves as it interferes with other surface motion.

Organisationally this is a matter of having the right people, with the right access, make the right pitch - technically it's a matter of racking up a supercomputer of whatever size is needed for the demonstration work by assembling enough Mercury blades (sold by IBM as the QS-20). customising some fairly standard codes, and getting false positive recognition down to some manageable level.

Okay, it's not a weekend's work, but it could make it possible to follow every Chinese or other suspect sub through every step of every sortie - and that's a big money contribution to U.S., and therefore world, security.

Topics: Nasa / Space, Health, IBM, Telcos

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Pattern recognition fails when the pattern is changed.

    Streamers can confuse some tracking missiles.

    Suppose something similar here, a way to "blur" the surface signal. That could be a disguise, could be used to send the hunters in the wrong direction.

    A surface pattern isn't inherent in the existence of a submarine, so it can be camouflaged successfully.

    That kind of confusion of symptom for disease has caused many projects and programmers to confuse workarounds for cures.
    Anton Philidor
    • First there has to be a pattern...

      ... that can be recognised. Murph's idea would work in an ideal world in which the surface of the ocean is completely flat. As it is the surface motions caused by wind and tide mean that the "noise" is considerably larger than the "signal". Anton said

      [i]"Suppose something similar here, a way to "blur" the surface signal. "[/i]

      Nature has already achieved that for you. Another consideration is that the ocean is not a homogenous mass. The water separates into layers of isothermal water. This "thermocline" is what submarines use to hide from sonar detection. You can count on the layers to foul up Murph's "surface hump" especially if the subnmarine is running deep.

      Murph - you are trying to use a sledgehammer to crack a nut here. Remote submarine detection is already carried out by a number of methods including magnetometers, thermal imaging and (at night) wake detection from bioluminence. The thermal imaging can be done from satellites (and has been since the 1980s) and relies on the fact that the submarine generates heat as it moves through the water and this warmer water rises to the surface were it can be detected. No supercomputers required, just a sensitive detector in orbit.
      • Agreed

        The topic may be lively, but it stops at bportlock's mention of isothermal layers. He could have also mentioned the SOSUS networks.

        If you want to get the interest of the US or Canadian navy, figure out how to stop super-cavitating torpedoes. The Russian "skvall" torpedo--a case of original Russian R&D, travels submerged at 300 kts+--and that's the declassified early version. They're meant to be carrier killers.

        • The shkvals ....

          ... are unsteerable in the normal sense, but since they are supposed to deliver a nuclear payload accuracy isn't that important.

          When one goes off the "hump" in the water should be detectable by one of Murph's proposed systems.

          • Nuclear or not...

            At thoss speeds they can develop non nuclear variants. If an Akula gets, say, within 2 miles of a carrier group and fires a modern shkval, calculate the response time the carrier has. Also, this shkval "1.0" was a late seventies creation, do you really think there's been no advancement since then? There were rumours of a 500-600 kt variant. Hell, when the Iranian navy test fires one, you know the Russians have better domestic models. They don't commonly export their best hardware.

            The steering issue was a big R&D focus--it's more than theoretically possible by extending the cavitation field outwards and using vectoring thrust.

            Besides, you wouldn't need a Murph proposed heavy compute system--the shkvals are rocket propelled--a noise maker by any standard (Hmm, perhaps too loud--saturatingly loud).

            Anyway, it's moot-because it's 10:30 ET and no one is going to read this.

          • The point I was trying (and failing) to make was ....

            ... it does not matter how fast it goes if the d*mn can't find or hit its target. A carrier is big, but in the context of the ocean, it is a tiny target, especially if aimed at from 10 or 12 miles away.

            Given the noise it make the sub will be "blind" after firing.
          • Also....

            [i]"If an Akula gets, say, within 2 miles of a carrier group and fires a modern shkval"[/i]

            Then the ASW people should be taken out and shot! At two miles, even the soviet equivalents of the Mk38 ADCAP would be hard to dodge.
        • SOSUS = Saturn (NT)

          • That (presumably) tongue-in-cheek comment ....

            ... was lost on me.

            SOSUS is (was?) a series of underwater hydrophones, mostly based in the Atlantic to detect Soviet subs attempting to enter the Altantic via the GIUK gap (that's Greenland-Iceland-UK in civvie-speak).

            Saturn is of course a large, ringed planet - so no connection there!

            Or is it a mathematical equation

            SOSUS = Saturn(NT)

            would mean that if planet Saturn took Windows NT and ran a transform on it, the result would be an underwater defence network?

            Wow! Murph - that is profound. I didn't see it at first.
          • ok -lets assume you're serious

            The NASA illustration used in the blog said that they choose heavy launch rockets over the spaceplane because that's what they
            knew how to do. The launcher was named the Saturn Rocket,

            SOSUS is an underwater acoustic detection network. SOund detection is what the navy knew how to do when they sidetracked track detection.

            hence sosus = saturn - a short term solution undertaken for good reasons but which eventually blocks further progress until some external force kicks it out of the way.

            so, you were kidding, right?
          • I prefer my version! (nt)

  • Surely the concentration on rocketry

    Was precisely because of its military uses.

    The Russians apparently had a lot of problems persuading the Kremlin to divert funds to satellites when they were principally interested in ICBMs.

    And didn't Werner von Braun title his autobiography "I Aim for the Stars (but sometimes I hit London)"?
    • No - and yes because causes can become effects

      I believe that the choice came down to:

      1) we have a firm deadline and a known budget;
      2) we have a proven launch technology that will get
      us there - but is expenseive and limited
      3) we have some engineers who think they have a better idea - and mostg of us think so too - but it hasn't been proven
      So go with the least short term risk: the technology that's proven - even if self limiting in the end.

      Of course the reason it was a proven technology was ICBM development.
  • Using a cluster with AltiVec...

    Would probably not be that great of any idea. The signal-to-noise ratio is going to be awful. Assuming you can extract "large object moving underwater" from "generally chaotic ocean surface," you still have to separate a whale from a submarine.

    So you're going to have to scan an awful lot of surface really, really fast. Fortunately it's trivially parallelizable. But given the size of the oceans, I think number of processors required will far exceed the threshold where custom ASICs give you far more bang for your buck and you can use something cheaper as a general-purpose processor to report back the results.

    Never mind building a sensitive enough sensor. I imagine we're talking about ripples here, not waves. Put it on a plane and you'll need too many of them, put it on a satellite and you have to be that much more sensitive.

    Interesting idea, though.
    Erik Engbrecht
    • COTS + SPU + Altivec = cheap, fast, flexible

      Not having done it I can't be sure - but my guess is that the eight way SPU parallelism in cell would work extremely well to filter input to the altivec on the master processor to allow code written for cell to out-perform any other commercial off the shelf (COTS) technology.

      COuld someone build a custom machine? Sure, but that would trap the code into one hardware generation - and I rather doubt it would be either faster or cheaper. Remember, part of the "moral of the story" here is that people should avoid getting trapped into dangerous behavior by reliance on methods tied to software tied to hardware - this story just re-invents yesterday's COBOL story, this time tied to tomorrow's defence problem.
      • Tied to hardware

        Cell would definately be an interesting option. My other proposal was going to be DSPs chained together or high-end FPGAs. The thing is general purposes processors are very slow for highly optimizable tasks like pattern recognition.

        Using an ASIC doesn't necessarily mean you are any more tied to hardware than you are with PPC or Cell. You can always fabricate essentially the same design for a smaller, faster, less power hungry chip.

        I imagining dedicated compute cluster(s) and special purpose sensor arrays. If you were doing sensor fusion from an only generally characterizable set of sensors, then a more general purpose architecture would make more sense.
        Erik Engbrecht
      • Have you checked out the new Power 6's?

        4.5Ghz at half the power of Power 5+.
        Roger Ramjet
        • Isn't it 4.7? - but way cool anyway

          and part of the move backto the megahurtz wars. (Intel's next generation will start at 4Ghz too).
  • Ice Station Zebra

    Just hide under the ice - that will negate this technology. And with all of you people poo-pooing global warming - this technique would be available forever!
    Roger Ramjet
    • But the magnetometer option still works...

      ... as all submarines have steel hulls - except the soviet Alpha class which were titanium. Interestingly enough (given yesterday's postings) the Alphas were nuclear boats with a liquid metal coolant - reportedly either sodium or lead - which were sealed in production and never refueled. This made them very small with a 20% higher power rating than a PWR reactor.

      On the good news front - the last Alpha went out of service in the 90s so the magnetometer option is looking good!