23 Replies Latest reply on Aug 8, 2012 5:52 PM by jamodio

    expert opinion needed on Raspberry Pi power issue

      The Raspberry Pi has an LDO voltage regulator to produce 1.8v. 

      However, a thermal image of the board, see post #6 on this thread:

      http://www.element14.com/community/thread/18991

      shows that this regulator, RG1, (the largest blue rectangle in the

      upper-left quadrant) to be cool, while the hottest chip is the LAN chip

      (SMSC LAN9512).  It turns out that the LAN chip has an internal

      1.8v regulator, which user jamodio noticed is apparently connected

      to the 1.8v power plane where it normally would be connected

      only to decoupling capacitors, not the power plane.  User selsinork

      verified that if RG1 is removed completely from the board, the

      board still works, apparently drawing all its 1.8v power through

      the LAN chip.  User beetem, who first noticed the cool RG1,

      has commented on possible problems with having two LDO

      regulators operating in parallel like this.

       

      So is this really a problem?  If so, how serious?  Is there an

      explanation for why some boards run significantly hotter than

      others?  How to get element14's attention?

       

      The latest thread is here:

      http://www.element14.com/community/thread/19236?start=15&tstart=0

       

      SMSC datasheet, showing VDD18CORE connected to decoupling

      capacitors in Figure 2.2, page 16:

      http://www.smsc.com/media/Downloads_Public/Data_Sheets/9512.pdf

       

      Raspberry Pi Schematic 1.0, where top left of page 3 of 5 shows VDD18CORE

      connected to +1.8v in addition to decoupling capacitors.

      http://www.element14.com/community/solutions/5952

       

      jamodio's original 10 June 2010 report (post #137)

      http://www.element14.com/community/message/53280

        • Re: expert opinion needed on Raspberry Pi power issue
          michaelkellett

          I've seen the other threads discussing this subject and I think that what you need is not so much an expert opinion as an informed opinion - from somebody at SMSC who really understands the chip.

          My generalised expert opinion is that (to paraphrase form a 1970's Nat Semi data book) if you operate outside specifications then all bets are cancelled.

          SMSC is in the process of being acquired by Microchip which might well eventually improve the availability of detailed information but the takeover is being held up by legal nonsense (IMHO it's nonsense) so right now I expect they are all worrying about their futures.

           

          I can think of ways to measure the power dissipated by the SMSC chip but you would need a supply of hot and cold Pi's.

           

          Why not try selsinork's mod on a hot Pi and see if it gets better.

           

          Michael Kellett

          • Re: expert opinion needed on Raspberry Pi power issue
            PeteL

             

            Mike Powell pinged me this thread directly and asked if I could respond, first apologies dropping of the radar (is it really that long!).  I have been looking at board bugs and this one is on my list, notes in my lab so more tomorrow.

             

            All - it would be really helpful if you could call out your serial numbers, it would allow me to find one in my stash with (hopefully) the same batch of parts, then we can get test aligned.

            Pete

             

              • Re: expert opinion needed on Raspberry Pi power issue

                Hi Pete,

                 

                Serial number from the SoC or the label on the back of the board ?  The one I started modifying is labelled FN120711618 & SoC serial 000000001395836f not sure which number is more useful.

                  • Re: expert opinion needed on Raspberry Pi power issue
                    PeteL

                    Hi Selsinork

                     

                    The FN number is the one, unfortunately my stash of boards runs out at FN120651352 so if the serial number are 'really' sequential you are 60K boards later than mine. [In reality we get a hole when the month changes so it's hard to predict exactly.] I'll call in some more testing samples from our partners.

                     

                    The warmest SMSC I have is 54C at T-ambient of 29C which is a deal cooler than the 61C reported.  There must be a corner case where RG1 is regulating low and the SMSC regulating high (as already suggested).  Time to ping off RG1 on a few more boards and see if I can kill one or get it to misbehave.

                     

                    On the case....

                     

                    Cheers

                     

                    Pete

                  • Re: expert opinion needed on Raspberry Pi power issue
                    jamodio

                    PeteL wrote:

                     

                     

                     

                    Mike Powell pinged me this thread directly and asked if I could respond, first apologies dropping of the radar (is it really that long!).  I have been looking at board bugs and this one is on my list, notes in my lab so more tomorrow.

                     

                     

                     

                     

                    It would be extremely interesting to see what else is on the "list"

                     

                    -J

                  • Re: expert opinion needed on Raspberry Pi power issue
                    jamodio

                    We would be able to help more and provide better feedback if RPF decides at some moment to make public updated

                    versions of the schematics and Gerber files.

                     

                    Also it will be extremely helpful if RPF/distributors develop/put together a ticketing system and knowledge base, this

                    thing about tracking bugs/errors, on forums or emails lists does not scale and does not provide any basis for statistics

                    or analysis, it is just disorganized techno gossip.

                     

                    My .02

                     

                    -J