Shop OBEX P1 Docs P2 Docs Learn Events
EEWeb article on deterministic real-time systems — Parallax Forums

EEWeb article on deterministic real-time systems

LeonLeon Posts: 7,620
edited 2011-06-13 00:50 in General Discussion
Digi-Key's EEWeb (http://www.eeweb.com/) has an article on deterministic real-time systems by that other purveyor of parallel processing devices. It's the first in a series. Parallax Semiconductor should do something similar - Digi-Key distributes products from both companies. I gave them a plug in a comment to the article.

Comments

  • potatoheadpotatohead Posts: 10,261
    edited 2011-06-10 08:39
    IMHO, that short piece is a great positioner. Agreed that Parallax should be expressing similar things. The philosophy behind both devices is important, and contrary to a lot of established norms. To me, that's the most useful part of short briefs like this. Tech detail is light, but that doesn't matter. The purpose of that kind of higher level "customer education" is to validate the NEED to consume the details. Nicely done, in this case.

    That basic need to validate consuming new information is common to highly differentiated technology solutions. And I know, this post is full of what Heater would call "speaking MBA". Guilty as charged. However, I do that kind of thing regularly, and am just contributing on that basis. The effort is necessary, or large chunks of potential adopters of the tech won't go and take a look.

    (and they need to take a look)
  • Heater.Heater. Posts: 21,230
    edited 2011-06-10 08:59
    "Speaking MBA"

    Chuckle. No, I thought you summed it up nicely.

    Or am I just getting used translating MBA to English:-)

    Anyway, Parallax could also benefit from putting out such articles. Indeed they might benefit from that article indirectly.
  • potatoheadpotatohead Posts: 10,261
    edited 2011-06-10 09:22
    I think they will, particularly with Leon planting the seed there.

    Well, I get tagged for MBA speak at times. No worries, because I find the whole thing humorous. Really good tech should just get used, but... it doesn't largely because producers of not so good tech want their share of the pie too, so it follows that we have to endure that stuff. Dollars often trump reason, it seems. That's the humor in it for me anyway.

    And because of that, there is always this "wall" between the business element of things, and the technical. That's funny too, because it's all far more productive when there is synergy there, not barriers. Nothing to be done, of course. The camps are such as they are. It does however, pay well, to be able to deal with both, and since I need to buy toys, I do.

    I have since added, "I love it when you speak MBA" to my playbook to great effect so far! Classic :)
  • LoopyBytelooseLoopyByteloose Posts: 12,537
    edited 2011-06-12 02:28
    A few years back, I got into reading on RTOSes and the topic is quite intriguing. It seems the airlines ticketing network is one of the most successful RTOSes, Sabre.

    At some point, there is a significant divide between Real time processing of a dedicated microcontroller and Real time processing within an operating system. It is not necessary to have an OS and in many ways it can create a substantial burden on both the programmer and the goal.

    Linux has some very good RTOSes that are free and thus a good place to start out. Maybe an old PC can be adapted to be a dedicated RTOS machine for development purposes.

    http://en.wikipedia.org/wiki/List_of_real-time_operating_systems

    On the other hand, Deterministic Real Time Systems (generally without the OS) are very much what the Parallax is all about.
  • LeonLeon Posts: 7,620
    edited 2011-06-12 02:40
    Adam Dunkels' Protothreads is an interesting approach:

    http://www.sics.se/~adam/pt/

    It works with anything for which there is a C compiler, including a PC. I've tried his code on various MCUs.

    He has expanded the idea into a full-blown RTOS.
  • Heater.Heater. Posts: 21,230
    edited 2011-06-12 02:53
    Yep I was looking forward to getting proto threads running on zog.

    Anyway, what does one mean by "real time"?

    When you are predicting the weather you want the prediction for tomorrow today. The day after is to late.
    When you are recieving bits over a serial line you have to deal with them at line speed.
    It's all about timing guarantees not actual speed.
    Different operating systems, or even no operating system, are applicable in different cases.
  • LoopyBytelooseLoopyByteloose Posts: 12,537
    edited 2011-06-12 22:53
    What is real-time? That is a good question as the academics have divided the topic between 'hard' and 'soft' real time. It gets a bit dicey to try to figure out what they are talking about and what the real outcomes will be.

    Most importantly, it seems that when you need a deterministic result (where the time taken to do something is absolutely clear), you have the kind of real-time systems that can generate good video or adequately control motion.

    What is NOT real-time? Waiting for your document to be printed on a shared printer and in an office printer server.

    In other words, it is all about getting the needed response in time. I suppose if the task is complex, there are elements that can tolerate a delay while there is one critical path through the process that must adhere to a deadline.
  • HollyMinkowskiHollyMinkowski Posts: 1,398
    edited 2011-06-12 23:27
    Loopy, how's this for needing real-time response and determinism.
    Programming a system that uses radar target data to enable the
    destruction of RPG rounds before they strike an armored vehicle.
    While simultaneously determining the exact spot where the RPGs
    were fired from....whew You can't just trust to luck and hope that
    you have enough processor speed to enable the tracking to be 100%
    ....you just gotta be certain!

    http://www.youtube.com/watch?v=-2IqZhonKzU&feature=related
  • LoopyBytelooseLoopyByteloose Posts: 12,537
    edited 2011-06-13 00:33
    Yep, that seems right. Real Time deterministic processing is more of a state machine than an OS for such endeavors and with military application one absolutely must know that one is faster than what is being shot at them.

    Parallel processors, such as the Propeller seem better adapted to such needs than anything with an interrupt driven, threaded approach.
  • LeonLeon Posts: 7,620
    edited 2011-06-13 00:50
    That system probably uses a hardware approach with FPGAs for the time-critical stuff, rather than a processor.
Sign In or Register to comment.