I finally gave up on the OEM TI 32k RAM PEB card. It just wasn’t reliable … it passed all diagnostics that I could find, but it wouldn’t pass the acid test — function-6 GROM verification in Gazoo’s excellent Extended BASIC Suite v2.7.
I’ve previously detailed attempts to build a working in-console 32k expansion. The 8-bit expansion, although popular, is too invasive for my tastes; the 16-bit expansion killed my main board for reasons as yet undetermined and has even more wires than the 8-bit version.
Since I was resigned to an 8-bit design, and I had a spare PEB proto board from Ksarul, I decided to build an in-PEB replacement using a static RAM that used +5VDC only (instead of the OEM dynamic 4116s that need +12, +5, and -5).
Luckily, Thierry had already done the design work which can be found here. There are a few things that need changed or clarified to make this go, though.
First, the address connections to the RAM chip are A1/3/4/…/15 (circuit side) to A13-A0 (RAM chip). Note that A2 is not connected, because it’s used with A0 to derive A14. This is not explicitly stated in the instructions.
Second, it’s easier to connect A0 and A2 twice to the four-input AND gate rather than wire two of the inputs to +5VDC. Those signals are buffered; even if you’re using TTL parts instead of CMOS, you’re still not going to have a problem with fanout.
Third, Ksarul’s PEB board expects a non-inverted input to drive the board activity LED. Thierry’s design uses an inverted input (the RAM chip select) to drive the LED. There’s three ways to work around this: rework the board to put the LED in parallel to the 2N3904 rather than in series between Vcc and the transistor collector, or add a transistor to invert the input along the lines of the first option, or add a 74xx04 to invert. I had a spare 74HCT04, so I opted for the third option. Connect RDBENA* to pin 1, connect “LED DRVR” to pin 2, and you’re good to go. Ignore the gunk from RAM !CS through a ‘125 to the LED.
(note: on Ksarul’s PEB board, RDBENA* is isolated from the common PEB RDBENA* by the 74xx125 buffer. It’s safe to tap the signal from this location; you won’t be picking up signals from anything else in the PEB)
(further note: the PEB board handles all of the ‘125 bits for RDBENA* and the ‘245 !CS. The connection that would have been !CS from the RAM chip to CS* should now be from RAM !CS to the board’s RDBENA* pad; the board’s ‘125 will handle the rest. You can ignore the bottom part of the schematic that involves OE*, the ‘125, and RDBENA*)
Fourth, Ksarul’s PEB board has a slight error. The direction pin for the 74xx245 data buffer is connected to MEMEN*, but it should be connected to DBIN. Fixing this requires cutting the trace and running a jumper wire, but it’s not difficult. The LED polarity as silkscreened on the PCB is reversed, too … the flat side should be on the bottom, not the top.
Fifth, there should be a 4.7k ohm resistor between G1 (pin 6) of the 74xx138 and Vcc. You can also opt to connect G2B (pin 5) directly to ground (I did).
Don’t forget to put a 0.1uF decoupling cap between the Vcc and ground pins on every chip.
Anyway, that’s it. It’s trivial to build a PEB-based replacement for the 32k RAM expansion if you have a Ksarul PEB proto board. I highly recommend doing this if you have a few free hours; the power savings alone are worth it.
I may make Eagle schematics and a board file available for download in the near future.