Tech Thursday: Billions & Billions

Oscar Gordon

A Navy Turbine Tech who learned to spin wrenches on old cars, Oscar has since been trained as an Engineer & Software Developer & now writes tools for other engineers. When not in his shop or at work, he can be found spending time with his family, gardening, hiking, kayaking, gaming, or whatever strikes his fancy & fits in the budget.

Related Post Roulette

45 Responses

  1. Kolohe says:

    I ordered a Shamrock chocolate shake earlier this week from my local McD’s, but they didn’t give me the fancy straw.Report

  2. Michael Cain says:

    Re CNG-fueled delivery trucks… All of the brown UPS delivery trucks that I see in my part of metro Denver are CNG-fueled. They are noticeably cleaner and quieter than the diesel-fueled trucks of similar size that FedEx runs.Report

    • Mo in reply to Michael Cain says:

      I think most airport buses are CNG powered. Though part of that is due to infrastructure. If there’s one place near you that has a CNG gas station, it’s the local international airport.Report

    • Oscar Gordon in reply to Michael Cain says:

      I like CNG fleets. Quieter, less stinky, and just economically smarter. Sure, it only works if you have a main hub (or 2) and the fleet never gets too far from the hub(s), but when it comes to a fleet working exclusively in a metro area, it works really well.Report

  3. dragonfrog says:

    Smarter cars – What a wonderfully American approach. Faced with the fact that the road as built kill people when they make the kinds of driving mistakes humans inevitably will, the Dutch think “we must design the roads so that mistakes won’t lead to so many crashes, and the ones that do happen will be less deadly,” and the Americans think “Robots!”

    • Kim in reply to dragonfrog says:

      not precisely true. the americans thought “no cars at speeds above 20mph in cities” until a very very large lobbying effort by the car companies.Report

    • Oscar Gordon in reply to dragonfrog says:

      I’m seeing more and more ideas like this getting implemented in the US, but it’s a slow process, and hard to do in a lot of urban areas.Report

      • dragonfrog in reply to Oscar Gordon says:

        Any major changes to widespread infrastructure are going to be slow, just because there’s so much of it.

        In Edmonton it seems the actual building part isn’t the limiting factor for the pace of change – it’s changing people’s minds to accept that the ultimate goal shouldn’t be enabling people to drive as fast as possible, while grudgingly allowing somewhere for weird non-drivers to get up to their perverted ugly walking.Report

    • DensityDuck in reply to dragonfrog says:

      “Faced with the fact that the road as built kill people when they make the kinds of driving mistakes humans inevitably will”

      Well, kill people who aren’t in cars but insist on being on roads designed for use by cars, sure.

      People get run over by trains but the response is somehow not “WE NEED TO STOP BEING SO TRAIN-CENTRIC, WE NEED TO WEAN OURSELVES OFF OUR DEPENDENCE ON TRAINS, SHARE THE TRACKS”Report

      • Oscar Gordon in reply to DensityDuck says:

        To be fair, trains are limited to the tracks, and can not maneuver beyond that constraint, so the ability for a train operator to make a mistake is smaller (although the consequences when they do can be considerably greater).

        Ships & boats make for a better comparison (rules of right of way are clear and easy to follow, and if all else fails, the laws of raw tonnage still apply).Report

        • dragonfrog in reply to Oscar Gordon says:

          Actually @densityduck ‘s attempt at a counterexample makes a great example of what they’re arguing against.

          Train infrastructure is designed separated from infrastructure for other modes of transportation.

          A driver can make all kinds of mistakes of inattention, and they will not end up colliding with a train because the road went underneath the tracks not straight across. Level crossings have loud bells, flashing lights, barrier arms, all of which activate long before the train arrives – you’d have to be missing several senses not to know a train is coming in plenty of time.

          In town, there are often foot bridges over the tracks rather than ever expecting people on foot to interact with the tracks.

          In the countryside, the right of way around a train track is cleared of trees quite a ways back – if you’re walking across the tracks, you can look left and right and generally see for a kilometre or so each way.Report

        • Last week a driver in Denver apparently took advantage of the train’s limited maneuverability to commit suicide…Report

      • dragonfrog in reply to DensityDuck says:

        Um. Guess you didn’t watch the video, huh.

        The first main point of “systematic safety” is separation of incompatible types and speeds of traffic – which is one of the things that allows more inevitable human mistakes to happen without serious consequence. You can build a road for exclusive use by cars, but then you need to build infrastructure for other incompatible modes of transportation, with a suitable buffer to separate it from that road. Or you can build a road for multiple uses, which, yes, does mean building it to slow cars the heck down.

        Since the roads in the US are (1) largely designed only with motor traffic in mind, and (2) largely the only transportation infrastructure available for all modes of transportation, your insist on being on roads designed for use by cars basically translates to insist on existing.

        Which, yes, I do insist on existing, I absolutely insist on it.

        Here’s the thing though – I absolutely do not insist on using roads only designed for cars. I argue against it every chance I get. I would very much rather use some combination of segregated infrastructure and streets designed to slow all traffic enough to make mixed use safe.Report

  4. Francis says:

    Space elevators? I confess that I still don’t understand the underlying concept. Principal questions:

    1. Are we talking about an elevator or a climbing rope? What motor could possibly lift a 22,000-mile long cable? Or is the idea for the elevator pod to grab onto the cable and pull up, using some kind of peristaltic motion? In either case, where does the lifting energy come from?

    2. Why the need to extent a cable outward? Why not just have equal mass above and below the geostationary orbit? Just build outward extensions of the space station.

    3. Whether the model is elevator or climbing rope, what force is generating the orbital velocity? As best I can tell, the idea is that the cable will “pull” the pod forward to match the station’s orbital velocity. But then don’t you just need to lift the propellant necessary to re-accelerate the station?

    4. The results of failure are interesting, to say the least. Ever since Mr Otis, terrestrial elevators have stopped when the cable failed. What’s the backup system when something goes wrong half-way to the space station?Report

    • Oscar Gordon in reply to Francis says:

      1) It’s a climbing rope. The cable(s) are fixed, one end at the earth, another at a station at geosynchronous orbit. The elevator cars climb the cables (there are numerous mechanisms available for climbing a cable). The power to do so would likely be in the cable itself, either from a ground or orbital power supply, or the natural electrical potential such a cable could generate just by passing through the earth’s magnetic field.

      2) The outward mass is necessary to balance the center of gravity of the station, which must always remain at geosynch. So you need mass beyond geosynch to balance out not just the mass of the cable below the station, but also of the cars riding the cable (i.e. as the cars ascend and descend, the counter balance mass would be reeled in or out as needed to maintain orbital station keeping.

      3) Earth’s rotation is maintaining the orbital velocity. It’s akin to swinging a yo-yo around. There is no propellant, just mass and it’s position relative to other things (and the energy needed to change the positions of the mass). Orbital mechanics and satellite dynamics are a bit counter-intuitive until you’ve learned to to the math. Now, occasionally the geosynch station would need to expend reaction mass to correct for variations, but the required mass could either be collected while in orbit (see Bussard Ramjets and magnetic collection), or, more likely, it would just be resupplied via the elevator cars. Additionally, most minor corrections that could not be accomplished via gyros or mass positioning could be done via ion drives. You’d save the use of large impulse reaction drives for emergency adjustments.

      4) This is why you have engineers. You wouldn’t have just one cable, you’d have a huge array of cables. For instance, you could have 7 main cable runs, a center run (which is the original) and 6 perimeter runs in a hexagonal pattern. Each run would consist of multiple cables of sufficient strength that a large number of them would need to be damaged before a car was in danger (and the car would climb on all the cables in a run at the same time). Additionally, the cars themselves would be equipped with cable maintenance systems that would scan each cable during movement and either flag damage for a specialized maintenance car, or the car itself would have the necessary equipment on board to make minor repairs. Finally, each car would be, in effect, a habitat module, with a recovery system (like parachutes), so if a cable snapped while in atmosphere, the car would use the parachutes to land, and if in orbit, ideally the geosynch station would be able to dispatch rescue craft.Report

      • Aaron David in reply to Oscar Gordon says:

        “This is why you have engineers”

        So! Just as I suspected. A jobs program in disguise!


        • Oscar Gordon in reply to Aaron David says:

          Yes! It is! And one with a potential benefit far beyond the typical “dig a hole, fill it in”.Report

          • Aaron David in reply to Oscar Gordon says:

            I have actually been in love with the idea of a space elevator for at least 30 years now, so this stuff is really cool to me. Also, I am the son of a scientist so its kinda natural to love this stuff.Report

          • I have claimed for years, when people complain about the number of engineers China and India are graduating from their colleges, that the scary thing is not the number of graduates, it’s that China and India are finding ways to put those graduates to work in engineering jobs. Sure, some of them will turn out to be effectively make-work jobs. But some of them will make real progress on problems.Report

      • Francis in reply to Oscar Gordon says:

        Thanks! I learned some new things today.

        Point 3 is really counter-intuitive to me. On the one hand, rockets that deploy a payload into geo-synch are huge. But once you dangle a rope down and connect geo-synch to Earth surface, you can use the mass of the earth to provide the acceleration. Cool.Report

        • Oscar Gordon in reply to Francis says:

          Well, the rotation of the earth, anyway. If the earth didn’t spin on its axis, it wouldn’t work. And that spin just keeps the cable tight, it doesn’t move the cars.

          So we couldn’t do this on, say, the moon, since the moon is tidally locked to earth. Except that we could, but for a completely different reason (re: Lagrange Point 1).Report

          • Francis in reply to Oscar Gordon says:

            If the earth didn’t spin, wouldn’t the geostationary point be the Sun?

            But returning to my original question, if the spin isn’t moving the car what provides the change in orbital velocity from .4 km/s (Earth’s surface at the Equator) to the 3.07 km/s that the space station is traveling at?

            (All figures per Wikipedia, so I may be completely misunderstanding.)Report

            • Oscar Gordon in reply to Francis says:

              No, the geostationary point of a tidally locked body is the Lagrange Point 1 (L1), which is the point between the two bodies where the force of gravity balances out. For the sun & the earth, that is about 1.5M km away.

              The cable is accelerating the car to orbital speed, much in the same way your car is accelerating you to highway speed. The acceleration would happen very, very slowly (a trip to the station would probably take anywhere from 8 hours to a full day), but a cable in tension can provide a lateral force.

              The complete elevator system is a massive balancing act of forces to keep everything stable. There is a considerable amount of natural stability in the system, once established, but it would be subjected to a near constant set of perturbations, most of which would be expected and compensated for as a matter of course. The rest would have to be detected and addressed very quickly by automated systems, which we have gotten pretty good at (no way you park a telescope in orbit and get clear pics of the universe without being real good at keeping her steady), although such a system would be an order of magnitude more complex.

              Good thing we’ve gotten real good at building fast computer systems.Report

              • Francis in reply to Oscar Gordon says:

                Thanks again.

                The micro-gravity cable production facility up in orbit is going to be an interesting place. Does anyone have even a back-of-the-envelope calculation on the tonnage to be rocketed to orbit before a first cable can start carrying supplies?Report

              • Oscar Gordon in reply to Francis says:

                Do we produce cable in orbit, or haul it up? Kinda depends if we have a ready source of raw materials in orbit and the equipment to turn it into cable is light enough to make it worth putting up there.

                A 1 cm diameter cable that long would weigh on the order of 6300+ kgs (6.3 metric tons). A Delta IV rocket can put 28000 kgs into LEO, which is the major effort. Once in orbit, the amount of reaction mass to get it to geosynch will depend on how quickly you want it there.Report

              • Do we produce cable in orbit, or haul it up?

                So long as we don’t outsource it to Viacom.Report

    • veronica d in reply to Oscar Gordon says:

      This is the math I love!

      Honestly, this is not new, inasmuch as antenna arrays have long been computer designed, and thus end up with weird asymmetries that serve to mitigate interference patterns. (Or so I gather. I’m way better at the math than the physics.) That said, with more CPU comes better, more resilient designs.Report

      • Oscar Gordon in reply to veronica d says:

        At work, we own a software package that does optimization, and we’ve modified to work with our flagship package. It’s a lot of fun to run it against hull or airframe designs to see what kinds of shapes you get.

        It also helps to have a room of clusters to run those simulations on.

        The trick is usually, “OK, now we got this optimal design, how the hell do we manufacture it?”Report

        • veronicad in reply to Oscar Gordon says:

          @oscar-gordon — Which brings up what is new about these techniques: these are generative models, which means the “search space” is not over the “product specs.” Instead, it is over something rather like the instructions on how to build the finished product. This is pretty cool.

          I don’t know all the details, but I would expect they are using robust optimization techniques, which basically say, “When you score a candidate solution, assume that there is n% of tolerance in the finished product, and score it thusly.” There are a couple approaches. One is simply run a few monte carlo simulations of the expected process. The other is to run a kind of min-max on the thing: find the best parameters assuming a worst case of “within tolerance” on the output.

          As I said, this is my favorite math!Report

          • Oscar Gordon in reply to veronicad says:


            It makes my inner structural engineering geek all giddy.

            Besides, it gives the industrial and manufacturing engineers something new to do.Report

          • Morat20 in reply to veronicad says:

            I’m awful at theoretical math, but I understand genetic algorithms quite well (and I’m okay with neural nets, but have not played with them much. I’m a little fuzzier on instance based learners, which I skimmed over the theory for and have never used), and I absolute love to use them.

            I got a Master’s degree (paid for by my company) in CS entirely to get to play with them while feeling semi-productive.

            Strangely enough, I seem to have an actual talent for fiddling with fitness criteria and recognizing sticky spots. (I currently have a half-finished project designed to ‘smooth out’ already explored local minima, but the string-comparison algorithm I’m using to compare chromosomes is not as effective as I like. This is where being good at math would help. I’m effectively smashing down already discovered
            ‘mountains’ on the fitness landscape, rendering that area flat so the algorithm can expand outwards, but the smashing effect is still leaving a few isolated hills in that area)Report

            • veronica d in reply to Morat20 says:

              The theory can help, sometimes, but then so often the problem you’re solving is pretty “non analytical” anyhow, and thus the only thing theory tells you is your problem doesn’t fit into the theory and none of the proofs apply.

              Like, if you have a nice convex quadratic problem, then sure, theory abounds.

              You probably don’t have a nice convex quadratic problem — although the theory might teach you how to use a convex dual problem to build a branch-and-bound approach — or whatever. But in the end you’re probably doing simulated annealing anyhow, and it is all empirical and guesswork and who the fuck knows.

              Anecdote: I’ve actually listened to a machine learning expert who works on real, very-large noteworthy systems. He explained why, when he fits his model, he just runs a few hundred iterations of a crappy gradient descent. He doesn’t care about convergence or conditioning because, in his experience, a really good optimizer just overfits his data and everything changes every five minutes on the Internet anyhow — so whatevs. Get some fuzzy results, some cross validation, and push!

              His product makes $$$$$.

              (You can probably guess.)

              On the other hand, if you’re designing antenna arrays or whatever, then you’ll want some theory and proofs.

              I like theory on its own terms. Math is its own reward.Report

  5. Michael Cain says:

    Re the renewable rubber for tires… It’s an exciting time to be doing work in catalytic chemistry, particularly the things that nano-structures seem to be capable of. Worth pointing out that the nano-structure stuff is all an outgrowth of technology developed to give consumers more and more processor cycles (and stored bits) at a constant price.Report