From: Milan WWW Pikula To: Alexander Oelzant , Agenda Devel Subject: [agenda-dev] Upgrading memory in Agenda VR3 List-Archive: Date: Fri, 4 Jan 2002 14:39:49 +0100 (CET) Hi all, This is a story about how hard and how easy a memory upgrade (to 16 MB) can be. Let me start with few bits from a history: The first man, who succesfully upgraded his Agenda was Alexander Oelzant. He did experience a need for an old PMON installed (thus he cannot use the newest sdram kernel, which is much better with 16MB of ram), and had no other big problems, except from the display, which works only when insmoded, and some strange suspend problems. His board is REV.10. I tried to do the same with board REV.9. (halted the server, desoldered a chip from my dimm, soldered it onto my board, started, saw tripple flash)... and nothing. It did not work. Few months passed as I tried almost anything to get it working, without an oscilloscope or proper soldering equipment... Alexander even sent me his spare memory modules, I've been looking at the disassembly of pmon (and later, to the pmon sources), at the specs of the components... Well, now it works, and without any of Alexanders problems. So what to do to extend your ram? Of course, you really don't want to extend your RAM, unless you're crazy or an electrician; you do it on your own, I am NOT responsible for any damage; by opening your agenda you also void your warranty etc. 0. read this list with extra care and do the preparation mentioned in 5.2 and elsewhere. open your agenda and take a deep breath. find the memory chip, CPU and get familiar with the small stuff ;) (smd) around them.. 1. get 64M SDRAM DIMM with 8 chips, or 128M SDRAM DIMM with 16 chips (8 on both side) or 32M SDRAM DIMM with 4 chips. Look up the chip name on google to get pinout and compare it with the pinout of, for example, EliteMT M12L64164A (the chip which Alexander used) or SpecTek 524LKT-8A. In fact, if the shape looks the same, address pins of chips are connected on the DIMM and 4 corner pins (+3 and ground) are connected with bold wires, chances are the chips are the same. If you've got a broken module, you risk some extra soldering. 2. desolder one chip from the DIMM. If there are chips, on which there is a paper label, try to start with some other chips. You'll slightly improve your chance to get a working chip from the broken dimm module. Find CS (chipselect) pin (19) and straighten it. 3. solder the module onto the original chip (if you are doing this for the first time, this will be almost the worst part of whole transaction;) Double check the GND and +3V pins. GND is the only pin, which MUST be connected. 4. connect the chipselect (ram:19) to the SDCS1# (cpu:34) pin of processor (yes, this is the worst part...:) If your solderer is as thick as 8 pins of CPU, you'll often do a nasty spots of solder. Be VERY CAREFUL when using hairy wire to get the solder out. Use some colophony or whatever instead. The (excuse me for my english) evacuator isn't usable here. If some hair from the wire catches the pin from CPU, you're likely to destroy the pin like me. If you will do so, you're done and you have 8meg Agenda forever. Are you REALLY mad? Ok, then follow my recipe: the core of cpu is in the middle, and wires leading to pins inside the processor are wide enough. Just carefully create the pit in the cover to find the wire (I'll have the photos some day;). 5. mystery. At this time, you're done at hardware level, but, for your big surprise, it doesn't work. There are 3 main reasons, why: 5.1. the led does not flash. check for short-circuits anywhere and try again. btw, it's good to check this after each major change in the hardware. 5.2. PMON does not start. timing problems. the led flashes, but nothing else happens. This is the reason, why only SOME pmon-s work, and only on SOME boards. I modified the newest SDRAM PMON to detect only 7 megs of ram (so it does not touch the second ram during startup). You need this one to boot with SDCS1 connected. Disconnect the chipselect (and optionally place some small resistor ~ 330 ohm between the ram chipselect and +3V) and flash my pmon. Or better - flash it BEFORE the whole experiment. 5.2. PMON starts, but linux does not. You are not a taiwan-tsu champion, and the RAM is not properly soldered. Use the 'mt' command to test the memory at a0800000 and above; solder the pins again and again. 6. enjoy.. I had to do one restore_defaults, but I'm not sure whether this is a problem of jffs when the RAM size changes, or my mistake during the half-day full of PMON/SDRAM/16M flashes. PMON says '7 megs of ram'. that's OK, it's hardcoded limit. Kernel says '17 megs, but will use only 16'. that's OK, it's mirrored memory from the first chip, which is the reason why we force only 16 megs of ram. Links :) 1. Alexander Oelzant s post about memory upgrade, including many valuable links http://lists.agendacomputing.com/pipermail/agenda-dev/2001-June/003876.html 2. My page with patched pmon binary http://www.terminus.sk/~www/agenda/ Best luck with upgrading, Milan Pikula -- Milan Pikula, WWW. Finger me for Geek Code. http://fornax.elf.stuba.sk/~www, www@fornax.elf.stuba.sk .. dajte mi pevnu linku a pohnem zemegulou .. _______________________________________________ agenda-dev mailing list agenda-dev@lists.agendacomputing.com http://lists.agendacomputing.com/mailman/listinfo/agenda-dev