1 2 3 Previous 231 Replies Latest reply on Sep 25, 2012 5:22 AM by Roger Wolff

    RG1 1.8v regulator

      Ok, so in a different thread I threatened to remove RG1 and do some current measurements on it's output after seeing those thermal images that show it's not generating any heat...

       

      Well, I did it tonight. Some photos here: https://picasaweb.google.com/selsinork/RPi18v

       

      The jumper pins in the output let me either just put a jumper on and verify the Pi boots ok, or wire a multimeter in series to get some current readings.

       

      The results were interesting to say the least. I had to go back and check I was reading the multimeter correctly, that it wasn't broken etc.

       

      On initial power up I see a negative current for a second or so which then reverses to about 0.5mA (yes half a milliamp, that's not a typo) for a few seconds while we get the first sd-card accesses. Once we're booted and sitting at the login prompt the current reading fluctuates from around 0.001mA to maybe 0.04mA. 

       

      I'm using the 40mA range on a decent Fluke multimeter, so I've no reason to doubt the results. There's obviously going to be some inaccuracy down at that level due to length of meter leads etc, but the result is fairly clear.  You'll understand why I was checking the meter was working and I was reading it correctly though

       

       

      So from there onto the next test, lets try completely disconnecting RG1 and see if the Pi boots while using the LAN9512 1.8v 'output'.  Yes it does! 

       

      I think that's reasonably good indication that jamodio got it spot on, the lan9512 shouldn't be connected to the 1.8v plane and it's heat problems are going to be largely due to supplying current on it's 1.8v filter pin that it was never designed to do.

       

      So anyone willing to pull RG1 off a Pi and verify my results ?

        • 1. Re: RG1 1.8v regulator
          morgaine

          Wow.   That's worrying.

           

          PS. It would be nice if you could include the images here directly, thumbnails at least.  Picasa is somewhat annoying because of its multiple required Javascript sources, and Google is developing quite a habit of turning off free services anyway ... as well as becoming evil. :-P

          • 2. Re: RG1 1.8v regulator
            Nate1616

            Thanks for starting to test this theory out.  I would really like to see the results of removing the RG1 off the Pi.  To bad i dont have a Pi laying around to do this do.

            • 3. Re: RG1 1.8v regulator
              tmackay

              When my other Pis arrive I might be tempted to break out the soldering iron. In the mean time, I'm doing a bit of reading.

               

              Looking at figure 2.2 of the 9512 datasheet (http://www.smsc.com/media/Downloads_Public/Data_Sheets/9512.pdf) would it be possible to cut the 3.3V input to the internal core regulator and have the core logic powered by an external (more efficient) 1.8V source? I'm not sure if this chip will run that way or how oversimplified this diagram is but it might be easier and more power efficient than trying to cut the link to the external 1.8V plane.

               

              I'm guessing those VDD33IO pins do more than just feed into internal regulators though, or the chip would just run off 1.8V to start with.

              • 4. Re: RG1 1.8v regulator

                I think it may be time to contact SMSC, and find out whether it's OK

                to use the LAN9512 chip's 1.8V "output" pin to power other devices

                in parallel with another LDO regulator.

                 

                The datasheet, p. 41, at:

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

                shows power dissipation of 763mW.  But that probably doesn't account

                for this scenario.  SMSC's tech support is available at:

                 

                https://www2.smsc.com/mkt/intforms.nsf/faemail

                • 5. Re: RG1 1.8v regulator

                  Wow.   That's worrying.

                  Well yeah.. but I think we've suspected what's going on since jamodio pointed out the error on the schematics.

                  PS. It would be nice if you could include the images here directly, thumbnails at least.  Picasa is somewhat annoying because of its multiple required Javascript sources, and Google is developing quite a habit of turning off free services anyway ... as well as becoming evil. :-P

                  My reason for using an external host for the pictures is fairly simple. Most forums I try to upload to seem to have ridiculous limits, like max size of 10k and a .JPG will upload but not a .jpg or .patch rendering the facility rather pointless. I also dislike the forums where you have to register and login to see the images, or where the column width is enforced and the image gets cropped. Finally I'm usually going to share these things with several different groups of people and a central point makes it easier.  I accept you may not agree with my reasons, but they are what they are.

                   

                  So, if you can suggest someplace that meets the following: 1. not javascript encrusted, 2. not part of the evil empire, 3. free, 4. reasonably permanent,  I'd be more than happy to use it instead.

                   

                  Just trying it here, not too bad with a 2MB size limit, but still scales down the image with the associated loss of detail.  So I'll at least try uploading stuff here too.

                  IMG_1909.JPG

                  • 6. Re: RG1 1.8v regulator

                    I'm guessing those VDD33IO pins do more than just feed into internal regulators though, or the chip would just run off 1.8V to start with.

                    Yeah, you'd have to assume they're used for 3.3v IO pins back towards the SoC at the very least. Not really clear if the internal regulator is from a single pin, or from many, or which one. So not clear if it's easier or not.

                    It's looking more like we need someone with a model A or who is prepared to pull the LAN9512 of the board to be able to measuer the RG1 current that way.  While I'm probably prepared to try that, I'm not there just yet

                    • 7. Re: RG1 1.8v regulator

                      I think it may be time to contact SMSC, and find out whether it's OK

                      to use the LAN9512 chip's 1.8V "output" pin to power other devices

                      in parallel with another LDO regulator.

                      If it was acceptable to use it as an output I'd expect to see something in the DC specs to tell you how much current it can supply.  I'd also suspect that running it in parallel with another regulator isn't going to be a good idea regardless.

                      • 8. Re: RG1 1.8v regulator

                        Just to expand on my original observations, the negative current I see on the output of RG1 seems to be due to it outputting 1.79v for a few seconds on initial powerup before stabilising at 1.804v. While the LAN9512 appears to power up at 1.805v and stay there.

                        • 9. Re: RG1 1.8v regulator
                          morgaine

                          selsinork wrote:

                           

                          So, if you can suggest someplace that meets the following: 1. not javascript encrusted, 2. not part of the evil empire, 3. free, 4. reasonably permanent,  I'd be more than happy to use it instead.

                           

                          Just trying it here, not too bad with a 2MB size limit, but still scales down the image with the associated loss of detail.

                           

                          Thanks.

                           

                          Aye, it's a problem indeed.  I think I'm going to use this "ARM revolution" to detach myself as much as possible from the various evil empires by hosting resources of various kinds myself on cheap and low-power ARMs, and removing them from the 3rd party sites.  It'll kill several birds with one stone.  It even makes sense in colos, because you can run a whole pile of ARMs within the typical least-cost rack power allocation.

                           

                          Back to running my own servers like in the early days before we succumbed to the dark side.

                           

                          Morgaine.

                          • 10. Re: RG1 1.8v regulator

                            Selsinork,

                             

                            >If it was acceptable to use it as an output I'd expect to see something in the DC specs to tell you how much current it can supply.  I'd also suspect that running it in parallel with another regulator isn't going to be a good idea regardless.

                             

                            right.  But I think in order to actually get this fixed it's going to take more than suspicions that there's a problem. If SMSC can be prompted to issue a clarification, then Element14 might take notice, and if they do, then maybe RPF might take notice.  Otherwise it looks like they will go on indefinitely saying the chip may be running a little warmer than expected, but it's not hurting anything.

                            • 11. Re: RG1 1.8v regulator

                              Well IMHO the simple fact that you can take RG1 off the board and have the Pi still work is good enough proof that there's a problem.

                               

                              I don't expect the RPF to listen, we've seen what happens to anyone suggesting there's a design flaw over on their forum. I think the best we can hope for is that Pete is still watching here and might take a look.

                               

                              Unless you have some decent technical contact within farnell/element14 who can objectively assess what we've found and then arrange an official approach to the RPF through avenues not available to us I don't see how to do much more.

                              Even with an official statement from SMSC you still need to get past the PR dept and get it into the hands of an engineer who's prepared to listen.

                              • 12. Re: RG1 1.8v regulator
                                John Beetem

                                People have reported very different IC3 (LAN9512) temperatures at the RasPi forum, which many ascribe to uncalibrated "digital" temperature sensors and I've suggested is due to how well IC3's ground pad is soldered.  I wonder if some of this variance is caused by different RG1 threshold voltages, where some RasPis generate 1.8V using RG1 and others using IC3, depending on who manufactured RG1?

                                 

                                Does anyone know if IC3's 1.8V is generated linearly or by a switching regulator?  If it's a switching regulator you should see ripple on a 'scope.

                                • 13. Re: RG1 1.8v regulator

                                  hard to be 100% sure, the datasheet isn't the best, and there's some stuff that coule be a switching regulator on the schematics..  but down at 20mV on the scope and I'm not seeing the sort of thing that's characteristic of a switching regulator, so I'd go with the core regulator being linear.

                                  • 14. Re: RG1 1.8v regulator
                                    tmackay

                                    I couldn't resist a little hardware hacking on a Saturday morning.

                                     

                                    My LAN9512 isn't running hot, only luke warm I would say, but I can guarantee a system hang by copying a file from an nfs mount to a usb flash drive in 10-20 seconds. USB keyboard is a little glitchy too. Tried 10W iPad charger and Dell monitor with integrated USB hub for power also without keyboard over ssh. That's motivation enough for me to tinker.

                                     

                                    Carefully, I cut the circuit traces from pins 15 and 38 on LAN9512 and the 1.8V side of C29 to the vias. I then very carefully soldered some wire between these three points. This more closely resembles how the circuit should be (minus the 100nF caps). (The 9512 reference design says 4u7 should be by pin 38 btw) See photo:

                                     

                                    IMG_0976.JPG

                                     

                                    Measurements taken before show 1804mV on both RG1 and C29. After modification I have 1797mV on RG1 and 1809mV on C29, so I am confident the regulators are now isolated. The effectiveness of my filter circuit is a little more dubious.

                                     

                                    Didn't fix the system hang copying over LAN to USB, but I can copy from LAN to the SD card (didn't think to test that previously). At least my Pi still works.

                                     

                                    Now that VDD18CORE is isolated from RG1 I can measure the current drain on RG1. More results to follow.

                                     

                                    [edit] After lifting RG1 and placing my exceedingly cheap multimeter in series with the output I measure about 20mA briefly, then 80mA briefly then 60-70mA for a while while booting then finally settling down to 50.5mA at the login prompt (presumably, no video attached, only power). Would have been nice to take a measurement before, I know, not very scientific of me.[/edit]

                                    1 2 3 Previous