Here are the empirically determined SIMM codes for the HP LJ5: (o for open circuit, g for grounded jumper)
Pin 70 69 68 67 ----------------------------------------- o o o o treated as an empty slot o o o g 53.00.02 error o o g o 53.00.02 error o o g g 53.00.02 error o g o o 53.14.03 error o g o g 16 Meg RAM o g g o 2 Meg RAM (=LJ5 70ns) o g g g 4 Meg RAM g o o o 53.14.03 error g o o g 16 Meg RAM g o g o 53.14.03 error g o g g 4 Meg RAM (=LJ4 80ns) g g o o 53.14.03 error g g o g 16 Meg RAM g g g o 53.14.03 error g g g g 4 Meg RAM
Since there are several codes that have the same RAM size, it's possible that they correspond to different RAM speeds or other differences that might matter. I connected my $129 16MB 70ns 32-bit-wide 72 pin SIMM in the g o o g pattern, and so far it seems to work OK.
If anyone has more info, such as speed or other characteristics that ought to match the ambiguous codes, I'd be very interested. Some of these codes correspond to the RAM/ROM PostScript SIMM, but those seem to be driven from a logic array and might have more complex behavior. In particular, no static code was interpreted as such a PS SIMM, but that might be because it already found one in another slot.
With a duplexer, the LJ5M barely functioned with the original 6MB RAM, had to run it at 300 dpi or it would just not bother to duplex. Borrowing 4M from my old LJ4M raised it to 10M, which raised the performance ENORMOUSLY. 16M raises it to 22M, which is surely major overkill, but for $129+tax at Fry's, why go for less.
So far the software on my Mac is refusing to control the duplexing features, e.g. the "which edge is the binding edge" choices are always disabled, but turning it on at the printer works (unselectively). But maybe this is due to some beta printing environment stuff I'm using.