Successful upgrade of 3B1 from 1MB to 2MB

Peter Fales pfales at ttrde.UUCP
Thu Jan 5 01:42:39 AEST 1989


After much trial and tribulation, I finally managed to get my UNIX-PC upgraded
from 1MB to 2MB on the system board.  I thought others might be
interested in hearing the procedure.

First, the obligatory warnings:  DO NOT ATTEMPT THIS UNLESS YOU ARE
REASONABLY COMPETENT WITH A SOLDERING IRON.  This is a major modification
to the board and there is a chance of damaging the board beyond repair.
(Especially with the limited diagnostic tools provided with the PC.)  Later,
I will discuss some of the problems I had and how I solved them, but
I hereby disclaim responsibility for anything that might happen to your
system as a result of these procedures.

MATERIALS REQUIRED:
-------------------

36 	256KB - 150 nanosecond DRAMS
36	16 pin sockets (optional but recommended)
27	0.1 microfarad bypass capacitors, 0.3 inch lead spacing
4	74F258N multiplexors
4	33 ohm resistor packs (these are 8 pin packages containing four
		resistors, each one connected to two adjacent pins inside
		the package.  You can also use 16 discrete 33 ohm resistors,
		that's what I did.)
1	Roll of good quality rosin core solder
1	Grounded tip, fine point, soldering iron
1	"Solder sucker" for removing old solder from the board
1	Stiff bristle brush (e.g. toothbrush) for cleaning solder splashes

PROCEDURE:
----------

1)	Use the soldering iron and "solder sucker" to remove the solder
	from all the holes where the DRAM sockets, bypass capacitors,
	multiplexers, and resistor packs will be installed.  I did most
	of this work from the top side of the board, without removing
	the motherboard.  Most of the holes could be de-soldered 
	easily this way.  A few however, particularly the ones connected
	to power and ground, may be more stubborn.  I just left the
	ones that gave me trouble until later when the system board is
	removed from the computer.

2)	Power up the system and verify that everything still works.

3)	Remove the system board from the computer.  I discovered after
	the fact that this is best done by removing all the screws holding
	down the board, as well as the two screws at the sides and the
	two screws on the bottom of the metal  back plate (the one that
	the RS-232 and printer connectors are bolted to.)  The metal
	back plate should be left fastened to the back end of the system
	board.

4)	Finish desoldering any holes that were not desoldered in step 1.
	This is can done by heating the solder side of the
	board, and using your other hand to suck the solder from the
	component side of the board,  while using your third hand to
	hold the system board itself. :-)

5)	Install the 36 DRAM sockets in the system board.  The only thing
	to watch out for here (and in the remaining steps) is that some
	of the pins protrude a good distance from the solder side of the
	board.  When the board is bolted back into place, these pins can
	bend and touch something they are not supposed to.  (This happened
	to me.)  If I was going to do this procedure over again, I would
	clip all the leads off flush with the board (or bend them flat)
	before soldering to keep the sharp points from protruding too far
	from the board.

6)	Install the 27 bypass capacitors at the appropriate locations,
	observing the same precautions as in step 5.

7)	Install the four 74F258N multiplexors and the associated resistor
	packs at locations 12A, 11B, 11C, and 11D, again observing the
	precautions in step five.  As previously noted, I had trouble
	getting a hold of the resistors packs, so I just used sixteen 33 ohm
	resistors and it seems to be working fine.

8)	If you are really paranoid, you may want to connect the board
	back into the system to verify that it still boots
	and runs as a 1 MB system.  If not, just plunge ahead.

9)	I am aware of two types of motherboards.  One is designed for
	the 512K system board only, and can be identified by three pins
	near 12A labeled E1, E2, and E3.  I understand, though I have
	not done it, that this board can be upgraded to a 2MB system by
	removing all 72 64K DRAMS and replacing them with 256K DRAMS.  This
	jumper determines whether 64K or 256K DRAMS are being used.

	The second type of board can be configured for either 512K, 1MB,
	or 2MB.  This is what I have.  There are two banks of jumpers/resistors
	near 12B and 11H.  If you have a 1MB system, these should be
	configured as follows.

	1 MB system:

	o--R 145--o   1K resistor		o---JR1---o   Empty
	|						  |
	o---JR5---o   Empty			o---JR7---o   Empty
							  |
	o--R 144--o   Empty			o---JR8---o   Jumper
	|					
	o---JR 6--o   Jumper			o---JR9---o   Empty
						|
						o--JR10---o   Jumper

	Converting to the 2MB system involves moving JR10 to JR9, removing
	R145, and installing a jumper JR5.

	2 MB system:

	o--R 145--o   Empty			o---JR1---o   Empty
	|						  |
	o---JR5---o   Jumper			o---JR7---o   Empty
							  |
	o--R 144--o   Empty			o---JR8---o   Jumper
	|					
	o---JR 6--o   Jumper			o---JR9---o   Jumper
						|
						o--JR10---o   Empty

10)	Install the 36 DRAM chips in their sockets.  Make sure that
	pin 1 is in the same orientation as the other chips on the board,
	and that no pins are bent or broken.

11)	Carefully inspect the board for any bent pins, solder bridges,
	or unsoldered pins.  The toothbrush is handy at this point for
	cleaning off dust and solder splashes.  A magnifying glass may
	be useful as well.  Re-solder any suspect connections.

12)	Install the system board back in the computer.

13)	Boot the diagnostic disk and look for 2MB on the system board.
	Run the memory diagnostics, and wait for successful completion.
	Congratulations, you have done it!!

DO YOU REALLY WANT TO DO THIS?
------------------------------

	After installing the new RAM, there does seem to be a noticeable
	improvement in performance, especially in using the User Agent
	which is one of the slowest things to begin with.  I also noticed
	a dramatic improvement in the time required to run "compress"
	on large files.  Unfortunately, one thing I forgot to do was
	run some sample benchmarks on both the old and the new system
	in order to get some real numbers to compare.  If anyone else does
	this upgrade, they may want to run some tests and post their
	results.  I may do this myself if I get real ambitious, but I 
	have a lot of other things to do first.

THE WARNINGS AGAIN:
-------------------

	THIS PROCEDURE HAS WORKED SUCCESSFULLY ONCE, BUT COULD BE HAZARDOUS
	TO THE HEALTH OF YOUR SYSTEM BOARD.  I CAN TAKE NO RESPONSIBILITY
	FOR ANYTHING THAT MIGHT HAPPEN TO YOUR SYSTEM OR ITS
	PERIPHERALS.

----------------------------------------------------------------------------
The following sections should be of little interest to most people, but
are included to describe some of the debugging problems I had.
----------------------------------------------------------------------------

IN CASE OF TROUBLE:
-------------------

	If you are like me, things will not go this smoothly.  When I
	had carefully done everything listed above and turned on the 
	computer, it just sat there with a blank screen looking at me.
	Oh S--t.

	The following is not intended to be an exhaustive guide to 
	trouble-shooting a 7300, but gives some tricks I found that made 
	it possible to debug my system.

	If you turn on the system, and nothing appears on the screen, look 
	at the LEDs in the corner of the system board.  If two of them are on
	it means that the built-in memory diagnostic has failed, and there
	is still hope.  If you see some other pattern, some other damage
	has occurred to the system and you are on your own as far as getting
	it repaired or replaced.

	First, put the jumpers back in the 1 MB configuration and boot
	the system.  If this does not work the problem is either with the
	original RAM or the data lines common to both banks and should narrow
	the problem down a bit.

	A data sheet for the 256K DRAMS is probably handy at this point.
	The memory on the board is divided into 4 banks of 18 chips
	each.  These banks are selected using the RAS lines on the
	DRAMS.  You can verify this with a continuity tester, noting that
	all the RAS lines in each of the four rows of two chips are connected
	together.  Banks 3 and 2 are the original DRAMS, while banks 1
	and 0 are the newly installed devices.

	The jumpers at 11H control the RAS decoding and are wired like this:
	
				-----------------
	+5 Volts o--R 145--o----| A1		|
			   |	|		|----- RAS3
	  LA2    o---JR5---o	|		|
				|		|----- RAS2
	+5 Volts o--R 144--o----| A0		|
			   |	|		|----- RAS1
	  LA1    o---JR6---o	|		|
				|		|----- RAS0
				| Decoder	|
				-----------------

	In the 2MB configuration, address lines 2 and 1 are used
	to select one of the four RAS lines.  In the 1MB configuration
	only LA1 is used, and A1 is pulled high.  This ensures that
	only RAS3 and RAS2 will ever be selected.

	By looking at the figure above, you can see how it would be 
	possible to move the jumpers to connect the 5 volt pullup to A0, 
	and LA1 to A1.	This will still be a 1MB configuration, but uses 
	RAS3 and RAS1.  In other words, it uses 18 of the original DRAMS and 
	18 of the new DRAMS.  Try booting the system in this configuration.

	You should also be able to see from the above figure how to
	connect LA1 to A1, and Ground to A0.  This will select only
	RAS2 and RAS0, in other words the other half of the original
	RAMS and the other half of the new RAMs.  Try booting the system
	in this configuration.

	If you are lucky, one of these configurations will boot, and 
	one won't.  This will narrow the problem down to one bank of
	devices.  Some careful work with a continuity tester should
	isolate the open or shorted connection.  You may also want
	to try swapping the upper and lower banks of the new DRAM to
	determine whether the problem follows the devices or the
	sockets.  (Aren't you glad you used sockets?)

	This was how I found the problem on my board.  It turns out
	that one of the address lines was shorted to ground, pin 1 of
	one chip had bent over and come in contact with pin 16 of the
	chip next to it which is ground.  Once I fixed this, the system
	worked beautifully.

-- 
Peter Fales			AT&T, Room 2F-217
				200 Park Plaza
UUCP:	...att!ttrde!pfales	Naperville, IL 60566
Domain: pfales at ttrde.att.com	work:	(312) 416-5357		



More information about the Unix-pc.general mailing list