1 4 5 6 7 8 231 Replies Latest reply on Sep 25, 2012 5:22 AM by Roger Wolff Go to original post
      • 75. Re: RG1 1.8v regulator
        PeteL

        Question is - were the Rubycon's real !

         

        P

        • 76. Re: RG1 1.8v regulator

          Pete,

          You wrote:

          "There is an implication in posts that this is responsible for something else to do with USB but to date no info is forthcoming? (Stop press - just seen that other thread - will go and look later)."

           

          I'm quite sure that there is no established connection at this point

          between this 1.8v power issue and any USB or other functional problems.

          Users with hot lan chips have been having USB problems, but so have

          users with cooler lan chips, so the USB problems could very well be due to

          other causes.

           

          SMSC has said "don't do that", but they haven't said what will happen

          if you do.  We've seen pretty convincing evidence that the lan chip

          will get hot, but don't know if it will malfunction, or cause a malfunction

          on the other devices connected to 1.8v, or whether its expected lifetime

          will be shortened.

           

          you wrote:

          "only a small population (sub 0.01%) are reported to run unbearably hot (although more may, just not flagged up because it does not concern their owner - they are just having fun with it). "

           

          I'm not sure what you are basing your statistics on.  Is it the return rate?

          I suspect that the rate is higher than 0.01%.  If the rate was that low,

          then certainly there would be no hesitation to announce that due to a

          design defect, a very small number of boards have chips that run blisteringly

          hot, and any user who is unlucky enough to have gotten one is welcome

          to exchange it.

           

          you wrote:

          "We tried to do the best job possible with the limited support and resources that the early phase project had."

           

          I don't think anyone would deny that you did a fantastic job with the limited

          resources you had.  But I think jamodio's point is that releasing schematics

          prior to production would not have cost anything, and could have resulted

          in great savings by uncovering such errors before mass production.

           

          I am not a hardware guy, but I am quite surprised to see that hardware

          schematics aren't clear about the direction of power flow.  It is a bit

          ironic that the beta board had a string of decoupling capacitors that

          should have been connected to 1.8v, but wasn't, and the production

          board has a similar string of decoupling capacitors that shouldn't have

          been connected to 1.8v, but was.  But there is nothing in the schematics

          to show which pins on the ICs have power going in, and which have

          power coming out.  So it is very difficult to check the schematics to

          find these kinds of errors, where components are either not connected

          to any source of power, or are connected to more than one source.

           

          At this point we are completely in the dark about what hardware revisions

          are contemplated, other than Eben's mention of some unspecified pcb

          change for FCC/CE residential compliance.  Hopefully that will change.

           

          Eben said he wanted to fix the FCC/CE issue prior to the educational

          release.  Since the 2012/2013 school year is about to start, the timing

          seems really odd not to have fixed that by now.  The timing also seems

          really odd not to have published the user's manual by now.  Amazon is

          showing a projected date of 16 October.  So are you aiming for the

          2013/2014 school year?  If so, I'm quite sure you will need 512MB ram

          to be competitive.

          • 77. Re: RG1 1.8v regulator
            jamodio

            First of all thanks for being here and for following up. I really appreciate the work you have done and I truly believe that some of the mishaps from the RPF are not your own making.

             

            Connecting the two 1V8’s is a sub optimal piece of design (you can look that up in the engineering translation dictionary - starts with a 'c' ends in ‘up’), and I didn't think that was of debate.

             

            Well, you may call it "sub optimal design" but it is actually an error, and in my dictionary it starts with 'f' and ends in 'up' ;-)

             

             

            The design did go through extensive review within the 'inner circle' of supporters who 'know' and have worked with the chips before especially the BCM. Both prototypes and pre-prototypes had this same connection and that of course made it more difficult. I also suspect, had there not been some reason to go and look, it would have never been identified. Just for the record, your technical input is appreciated, but rather than just implying - you could have just asked? Maybe you did earlier and I missed it.

             

            Obviously the process didn't work, and as you clearly know in the previous prototype it was reported that various power connections were missing, kind of a surprise since part of the 'inner circle' was apparently involved in the design of the BCM SoC chip. Perhaps the 'inner circle' has a very small radius and some of the supporters actually "don't know." Not just me but many other asked while before the boards went to production for schematics/gerbers and the only we obtained was a crop showing a psu section. I reported the problem as soon as the schematics were made public, and it didn't took too much know how, just reading the SMSC datasheet to figure what each pin was used for, something that we don't even have for the SoC part.

             

            And about asking, I asked what else is on "the list", no response yet.

             

              We tried to do the best job possible with the limited support and resources that the early phase project had. (We were going to make 5000-10,000 beta boards max). None of the early stress testing revealed the issue and only a small population (sub 0.01%) are reported to run unbearably hot (although more may, just not flagged up because it does not concern their owner - they are just having fun with it). Of the returns I have for analysis, I have still have not found a real ‘steamer’. These could be in part due to poor underfill or even a short/defect elsewhere on top of what we are asking the chip to do.

             

            The foundation should have put they arrogance away and ask for help and additional support, there has been a large group of people willing all the time to cooperate, and they are still out there but the RPF attitude has been always "what we did is perfect and we know everything." I'll not trust any number, percentages or analysis derived from them given that there are no public numbers about how many boards have been manufactured, how many have failed, how many have been sold. how many have been shipped, how many have been returned, etc, and there is no formal or reasonable system to track complains/failures/fixes.

             

            There is an implication in posts that this is responsible for something else to do with USB but to date no info is forthcoming? (Stop press - just seen that other thread - will go and look later).

             

            I'm not sure if there is a direct connection with several of the problems reported with USB. Certainly the entire power architecture does not help, but there are some hardware/firmware issues related to USB where things are not working as expected and somebody is now reading the Verilog files for that piece of silicon on the SoC discovering some limitations and other stuff hidden behind the obscurity of the drivers.

             

            I know that everybody have tried to do their best, but recognition for a successful endeavour comes from producing positive results and not from the effort put to get them.

             

            My .02

            -J

            • 78. Re: RG1 1.8v regulator
              PeteL

               

              So I guess it is not worth me publishing here what I know and what I can measure?

               

              I do take issue with your final comment - it has been a success so far - with some measure of "success failure" thrown in.

               

              I've been talking to users who just have one or two and are over the moon with them (faults issues and all) and they tell me that they have already learnt so much.

               

              My overall positive view may yet be proved wrong, but as I said right at the outset it isn’t perfect, never will be - we just have to remember why we are clearing the swamp!

               

              Off to watch the Olympics on TV

               

               

              Pete

              • 79. Re: RG1 1.8v regulator

                I am not a hardware guy, but I am quite surprised to see that hardware

                schematics aren't clear about the direction of power flow.

                In many years of working in the subcontract hardware assembly business I've seen hundreds of schematics from all sorts of companies, from the biggest names to tiny one man outfits and I've rarely seen anything indicating power direction.

                This sort of thing wouldn't have been a problem in years gone by as it was rare for IC's to have internal regulators and often power was supplied from an off board supply so it was obvious.

                Things change, technology gets more complex, leaving room for ambiguity and errors to creep in.

                 

                IME Petes schematics are pretty good, they lack some things you'd normally find on much larger schematics like a cross reference of refdes and signals to page and location, but for four pages most of that stuff isn't really necessary.

                • 80. Re: RG1 1.8v regulator

                  lol.. nope.. it's like one of those russian dolls you keep removing another layer until you get to the 0201 smt cap somewhere inside

                  • 81. Re: RG1 1.8v regulator

                    So I guess it is not worth me publishing here what I know and what I can measure?

                     

                    I'd certainly appreciate anything you're willing to share Pete

                    • 82. Re: RG1 1.8v regulator
                      John Beetem

                      PeteL wrote:

                       

                       

                       

                      I remember saying at the outset that Pi would never be perfect - just doesn't happen in engineering there is always something to be optimised, improved.

                      There is a story about a craftsman who made beaten copper doors for the Tsar of Russia's throne room.  He would take a large sheet of copper and beat it over and over with a hammer, creating beautiful swirling patterns.  Someone asked him "how do you know when you're done?"  He answered: "It's never done.  I just keep hitting it until they take it away from me."

                       

                      Engineering is like that -- especially software engineering.

                       

                      Thank you for being open to suggestions, Pete!

                      • 83. Re: RG1 1.8v regulator

                        Pete,

                        >So I guess it is not worth me publishing here what I know and what I can measure?

                         

                        You are being way too defensive.  Nobody is blaming you for the design error.

                        If I understand your earlier comment correctly, the error was copied from the

                        alpha board that you didn't design yourself.  And I don't think it matters much

                        to hear that you have found a happy customer.  On this forum and the RPi

                        forum, happy customers are hugely outnumbered, not that that matters either.

                         

                        It would be very nice for you to publish what you know and what you

                        can measure.  However, regardless of what further information you can

                        provide, I think it is clear enough at this point that the lan chip is not

                        supposed to be blisteringly hot, as people have been complaining about

                        in vain for months, and it is essential to promptly tell those customers what

                        their remedy is, rather than what they have been told so far, which is that

                        hot chips are normal in PCs.

                         

                        Going forward it is clear enough that the foundation has limited resources

                        for debugging their hardware and software, and would greatly benefit from

                        a bit more openness with regard to schematics, and a bit more cooperation

                        with those who point out problems.

                        • 84. Re: RG1 1.8v regulator
                          tmackay

                          Hi Pete,

                          PeteL wrote:

                           

                          I see the issue with 'higher' temp but now I get the feeling you are looking to fix something else with this mod?  Can you confirm?  There is mention of the USB issue - point me at it please and anything else you think is connected.

                          I'm a big fan of reliability and uptime, after a few kernel panics I was hoping that this might help with stability, it didn't seem to though. Still worth a try. I can trigger a kernel panic by using nfs pretty reliably. Software upgrades seem to have helped significantly, and development is ongoing. It is a pretty harsh test, blasting udp packets from a gigabit nic to the Pi with all error correction done in software. I've since not been able to get it to crash repeatably with other (more polite) tcp based protocols.

                           

                          I'm also interested in increasing power efficiency, as I plan to build battery powered robots and data loggers.

                           

                          So no, I personally don't have a specific hardware issue.

                          • 85. Re: RG1 1.8v regulator
                            tmackay

                            The more I think about it, I think I like Tooms idea better. By replacing the 1.8V reg with a slightly higher switcher, you have the added benefit of bypassing the internal linear regulator of IC3 and saving even more power. It has been stated that the LAN chip was not indended to be driven from these pins, but I don't think that's going to be any worse than the current situation. It obviously "works" in this configuration.

                            (edit) If only the 1.8V rail was exposed on the GPIO this power saving "feature" could be a plug-in solution. They really are quite clever, the other SMSC chips with documented "internal regulator disable mode" probably cost more. How's that for spin?

                            • 86. Re: RG1 1.8v regulator

                              This sort of thing wouldn't have been a problem in years gone by as it was rare for IC's to have internal regulators and often power was supplied from an off board supply so it was obvious. Things change, technology gets more complex, leaving room for ambiguity and errors to creep in.

                               

                              Selsinork,

                                Thanks for the explanation.  It makes sense. 

                              Here I've assumed that at least since Intel's fdiv bug, hardware

                              guys have had the advantage over software guys because they

                              use mostly formal verification techniques where we rely mostly

                              on ad hoc testing.

                                I couldn't figure out at first why it was difficult to verify that

                              all the components on the RPi board were properly connected

                              to power, which I assumed to be a pretty fundamental property

                              to be sure was verified, until I saw that power flow isn't

                              specified on the schematics.

                                It makes one wonder what is meant by the claims that the RPi

                              design was carefully checked.  How can you check something

                              that isn't specified, especially when the datasheets for the ICs

                              are either non-existent or ambiguous themselves?

                                As technology gets more complex, ambiguity has to be reduced

                              so checking can be increased, just to keep reliability from getting worse.

                              • 87. Re: RG1 1.8v regulator
                                morgaine

                                The engineering process after initial release is about identifying existing shortcomings and evaluating alternative solutions for future releases, and doing this iteratively for each new version or model.  It is not about pointing fingers at engineers (which is completely irrelevant and unhelpful), and it is not about those engineers reacting defensively when problems are identified.

                                 

                                Reacting defensively makes it hard to analyse technical problems because it personalizes them, and when the defensive reaction is as extreme as on the Foundation forums and leads to messengers being banned, the engineering process is completely stillborn.  We're free of the latter here, but please let's remain free of the former as well.

                                 

                                PeteL wrote:

                                 

                                I do take issue with your final comment - it has been a success so far - with some measure of "success failure" thrown in.

                                 

                                Pi Model B is most definitely not an engineering success.  Engineering success is a fairly measureable quantity related to how much of the engineering involved was actually successful, ie. how much of the product works as expected.  (And "success failure" is not a term any engineer should ever utter.)

                                 

                                If any other computing device had been found to have such a catalogue of compatibility problems when used with totally class-compliant USB devices, it is highly likely that it would have been recalled immediately and withdrawn from sales pending redesign.  The Pi is setting an example of USB incompatibility such as has not been seen since the first year of release of the USB spec and initial devices.

                                 

                                These USB problems appear to be in part fallout from least-cost BCM2835 SoC design and its USB  driver, and in part a consequence of excessively restrictive power design, but the board is a unit and its USB functionality is an extremely important part of its operation.  That key functionality has major faults.  It is not an engineering success (the only kind that concerns us here), because it is failing to operate as expected, and there are countless examples of how it is failing spread across the planet and reported on both RPF and Element 14 forums.

                                 

                                I'm going to assume that the objectively incontrovertible fact of widespread failures is accepted, because anything else would not be a professional engineering response given the evidence.  The big task then becomes to identify the reasons and find solutions, and that is why we are here, in my opinion.

                                 

                                Unless we can get past acceptance of the problems though, there is little prospect of finding solutions.

                                 

                                Regarding how the 1.8V design might affect LAN9512 operation, I have no evidence that they are related either, but I am just hoping that they are related in some way that we cannot currently see because otherwise it seems likely that the Pi's huge USB faults are with us to stay.

                                 

                                Morgaine.

                                • 88. Re: RG1 1.8v regulator
                                  jamodio

                                  Pete anything you are able and wish to share will be very valuable, I can say that so far I've not seen anything being shared yet...

                                   

                                  I don't feel qualified to state if the mission of RPF has been or will be a success or not, my comment was referred to situations where people say "we tried hard," "we did our best."

                                   

                                  Nobody will deny to acknowledge and appreciate the efforts but what it counts at the end are results.

                                   

                                  At one time I had over 100 engineers working in my team and I never accepted that something that was accomplished by "doing their best" couldn't be done better and more efficiently, because as others said engineering is a cumulative process where we constantly learn and apply the new knowledge to do things better, and more efficient, and which will never be perfect.

                                   

                                  I'm not a fanboi or a detractor of the R-pi, I still consider that it was a great idea and something worth to create, I'm not angry and I'm still using the 2 boards I purchased and planning to still work on some projects with them, now would I recommend it to local schools?, the answer at this time is NO.

                                   

                                  BTW kudos to the UK and London for putting such a great 2012 Olympics.

                                   

                                  Regards

                                  Jorge

                                  • 89. Re: RG1 1.8v regulator
                                    morgaine

                                    Everyone doing their best is assumed.  It has no bearing on whether a product works or not, except possibly if the testers are ineffective and hence undermine everyone else on a project.  If anyone else in an engineering team fails, it should be caught at test to within certain limits (not 100%).

                                     

                                    But more importantly and relevantly, that issue is not part of the engineering process.  That process at the core of our profession concerns the objects being engineered, not the people.

                                    1 4 5 6 7 8