Hi Nikolaus, The story reads like a novel, with another cliffhanger every time:-)
Keep up the good work! Kind regards, Ed On Sat, 2010-11-13 at 08:50 +0100, Dr. H. Nikolaus Schaller wrote: > Hi all, > time for another update. > > Am 06.11.2010 um 12:35 schrieb Sylvain Paré: > > > > > Thanks for the news > > > > > > > > Yesterday I received the second board where we > > > did populate the TPS65950 (power controller). > > > > > > The results of measurements are: > > > * the 32 kHz RTC clock is operating > > > * when inserting a battery, most voltages are > > > available as expected > > > > > > Bad news: > > > > > > * we have a short circuit on the 1V8 rail > > > > > > I have spent most of the night and this morning to > > > track this down. It appears to be a solder short circuit > > > under the TPS chip (a 0.4 mm pitch BGA). > > > > > > So I am currently sitting at our SMD rework company > > > and looking over the shoulder of the CTO who has > > > a lot of experience. Unfortunately I can't make > > > photos. > > > > > > We already have unsoldered the TPS chip (that needs > > > a really sophisticated machine) and the short has > > > disappeared. > > > > > > The next step is to solder it back again and do the next tests. > > > > > > If that works, i.e. we get all voltages from the power controller, > > > the 26 MHz oscillator should also start working. > > > > > > If that is ok, the OMAP and the POP memory will be soldered. > > > Maybe we manage to get it today. Then, we can see if the CPU > > > is doing something. > > > > Current status: the 1.8V is now working but the VDD2 (1.2 V) not. > > The TPS chip aborts the power up sequence early. > > We could solve that by adding a jump start resistor. > > > > Rene will upload some photos of the board and we will > > > post a link. > > > > Here: > > http://download.goldelico.com/gta04/images/DSC00671.jpg > > http://download.goldelico.com/gta04/images/DSC00672.jpg > > Then, we soldered the OMAP3530 and Pop Memory chip and > connected RS232: > > http://download.goldelico.com/gta04/images/DSC00679.jpg > > It did identify itself as "40W" on the RS232 but did not boot beyond > that. Fortunately the OMAP has a ROM bootloader (which generates > the 40W sequence) and we could use it to download a special > second stage boot-loader that has a simple commandline console > on RS232. Note that it must fit into 64k SRAM built into the OMAP > chip. So we are back to C64 times :) > > With this tool, we could identify that the (external) SDRAM has a > stuck-at-1 error on a single bit-line. > > So we expected another soldering issue and had the OMAP+Pop > replaced on Monday. But the error pattern remained the same! > > Unexpectedly, we were able to put the board into a 3D X-Ray machine > on Tuesday (during Electronica Fair). Here are machine & results: > > http://www.yxlon.com/y.Cheetah > http://download.goldelico.com/gta04/images/20111109%20Xray/ > > Fortunately it is *not* a soldering issue. And likely not a PCB production > issue. > > Yesterday late at night we found the problem. It is a bad BGA ball assignment > in our component library. VDDS_MEM (1.8V) is assigned to B17 instead of > B18. And, B17 is data line 14 of the SDRAM and should have been left NC. > Therefore we have tied one data line to a high-level... > > I think we can fix that for our samples but it needs some time and we > run out of components for the samples. Especially memory chips have > quite a long lead time. For the series version this will be very easy to fix. > > Nikolaus > > > > _______________________________________________ > Openmoko community mailing list > [email protected] > http://lists.openmoko.org/mailman/listinfo/community > _______________________________________________ Openmoko community mailing list [email protected] http://lists.openmoko.org/mailman/listinfo/community
