13 Replies Latest reply on May 25, 2015 10:13 AM by FerryT

    ViperOS plans

    FerryT

      I want to do 1MBd serial transmissions that are started by a timer with 10us accuracy. This can't be done from linux but probably could be offloaded to the MCU.

       

      However, currently the MCU doesn't support 1MBd transmission with parity enabled.

       

      I'd really like to know where the MCU with it's RTOS is heading:

      - will the configurability of the UART be expanded

      - are we going to see the sources opened

      - is there a road map of any kind

        • 1. Re: ViperOS plans
          PabloM_Intel

          Hi FerryT,

           

          Currently there are no plans to expand the configuration of the UART and the sources won’t be opened (and it is unlikely that this will change any time soon). Regarding the road maps, we can't disclose information on this matter.

           

          Regards,

          PabloM_Intel

          • 2. Re: ViperOS plans
            FerryT

            That is unfortunate. Current functionality exposed by the MCU will imho not be enough for serious applications.And without MCU the Edison losses its real time appeal.

            • 3. Re: ViperOS plans
              odevices

              I guess that puts the last nail in the coffin for any possibility of using the Edison in a serious real-time embedded application.  It seems to me that Edison is fated to just be a headless linux box.

              • 4. Re: ViperOS plans
                FerryT

                A very cool linux box though. We attached a usb ethernet board, so that creates wired ethernet + 4 usb ports. Very cool. Need to disable wifi for it to work though.

                 

                Now we are investigating to get RT communication at 1Mb/s. We found that the RTS/CTS implementation in the HSU is HW controlled vs. SW for many other linux uart drivers. That is cool, at it means you can time transmits accurately by having the MCU control the CTS line of the uart, but have the data by actually transmitted to the UART by the CPU (at high speed).

                 

                Still, going to be hard to handle reception errors in RT from linux and the MCU now is almost undocumented, the RTOS API limited and no way to expand that (without reverse engineering the binaries).

                 

                So either we need to add an external MCU (sigh) or wait for documentation land.

                • 5. Re: ViperOS plans
                  nniles

                  There are other ways to get RT besides an MCU.  One of the Intel guys posted about the progress he had made on building a Xenomai kernel for the Edison (Re: Bitbaking Edison Realtime Xenomai Kernel).  This at least suggests (I'm not really sure how formal the relationship is with the various Intel-named posters) that they are looking into more general RT solutions, and might suggest that something like this is on their radar for an upcoming release.

                   

                  Realtime is really a question of how much latency/jitter you can tolerate anyway.

                   

                  Having access to the MCU is a big step in the right direction, and I am seeing some pretty tight timings with I2C and the MCU-host IPC, although the latter seems to have a large overhead (950+ us just to send 1 byte, ~1.2 ms to send ~150 bytes).

                   

                  I just think it's really bad policy for them to play their cards so close with respect to the roadmap, given that this product is marketed to hobbyists.  It seems like a good way to turn people off.  There's got to be a way to keep a competitive edge and still keep customers intrigued.

                  • 6. Re: ViperOS plans
                    odevices

                    @nniles, 950ms to send 1 byte?!  Do you mean 950us, perhaps?

                     

                    @FerryT, It's definitely a cool linux box.  However, I was under the impression that the Edison was targeting low to medium volume products in addition to the hobbyists.  This is where I am feeling disappointed.  It took me a non-trivial amount of work to get from the sparkling the-sky-is-the-limit Edison product descriptions encountered at the beginning, to sifting through the documentation, reading about people's experiences, watching videos, lurking on the forum reading threads to figure out that in the end Edison is not ready (yet?) for product development. Shame on me for getting so excited in the beginning. 

                    • 7. Re: ViperOS plans
                      nniles

                      odevices: Oops, yes, fixed.

                      • 8. Re: ViperOS plans
                        Frederick Blais

                        odevices,

                         

                        I hear ya. Maybe its not quite ready yet for product development, but I prefer being able to play with it now than wait until next year for a finished product. People from Intel look like the are working hard to get the stuff done. I just hope that they will continue to push software updates. The hardware is pretty solid.

                        • 9. Re: ViperOS plans
                          FerryT

                          At my company we don't see the Edison as a hobbyist device, but as a basis for industrial automation products. And attribute the lack of documentation of the MCU, alfa RTOS, industrial temperature specs, to a hasted market introduction.

                           

                          If it is really targeted to hobbyists that would mean someone at Intel has gone mad. No, can't believe that.

                          • 10. Re: ViperOS plans
                            nniles

                            FerryT,

                             

                            No one will argue that the rollout of the Edison wasn't rushed, to be sure.

                             

                            I don't know about you, but the URL I went through to get to these forums was maker.intel.com.  To my mind, maker and hobbyist are very nearly synonymous, at least in this context.  That doesn't mean it doesn't have serious potential, but I certainly hope the rollouts for their flagship products are at a completely different level...

                            • 11. Re: ViperOS plans
                              FerryT

                              Maybe you are right, that's what it means.

                               

                              Until all the hardware is documented, it is suitable for 'makers'. And until intel can provide support for this, it is supported through the 'community'.

                               

                              Not that I think 'maker' 'hobbyist' and 'community' are bad words. (We (my company) are in the business of industrial automation. Here many products are manufactured at small volumes, maybe even smaller than some 'hobbyist' products.)

                              • 12. Re: ViperOS plans
                                odevices

                                I certainly don't mean to use hobbyist as a bad word and I don't think it has anything to do with volumes.  It's a matter of consequences, i.e. the consequence of failure to meet specifications or deadlines.  For a hobbyist, failure has no consequence except as a learning experience, for a professional it means the loss of a client's trust which I'm sure you all know.  So instead of bringing hobbyists/maker culture into the discussion, I guess what I mean is that the Edison has not achieved the maturity where you can stake your business' reputation on it.

                                • 13. Re: ViperOS plans
                                  FerryT

                                  Yeah, that's why this thing is still considered experimental by us.