The AC Specifications of the C390 data sheet include these notes:

3. All signals characterized with load capacitance of 80 pF except Port 0, ALE, PSEN, RD and WR with 100 pF.
4. Interfacing to memory devices with float times (turn off times) over 25 ns may cause bus contention. This will not damage the parts, but will cause an increase in operating current.

Now I am not sure precisely what "All signals characterized" means - how much margin if any is there around these values? At what point do things not meet the spec if the specified parameters are exceeded? Is this a typical or maximum value? Our own testing here shows that severe failures (won't run at 40 MHz) occur on a C390 with more than about 130 pF (max values) of load on P0. A system with 130 pF on P0 worked OK under most circumstances. Adding a calculated 50 pF max of other CMOS devices caused serious problems. Note that 130 pF max is about 93 pF typical, not including board capacitance. So the 100 pF spec is pretty close to a "do not exceed" value, it would seem.

To make these calculations, I added up the data sheet typ and max input and output capacitance of TINI, our uCAN2 board (uses a C390), and two prototypes of our SBX1 (not the current SBX2!) card. We tested on several uCAN2s with different crystals. For example, 29F series flash devices have a Cin of 4/6 pF typ/max and a Cout of 8/12 pF.

Adding up the loads shows that TINI itself has (typ/max) 42/60 pF on D[0-7] and 28/40 pF on A[0-8].

TINI with a Socket+ e50 has 84/120 on D[0-7] and 56/80 on A[0-8]. And in fact there have been some reports of apparent load-related problems on a fully populated e50 board.

The above do not include board capacitance, just the data sheet values for the chips on the boards.

<Disclaimer> and a plea to not flame me: I know that the issue of TINI bus loading is a touchy subject and has not been (yet) officially documented by Dallas. I'm trying to not make any wild or irresponsible claims here. I'm trying to be careful what I say and can back up my statements with data. Dallas should be the official source of information, and I am not speaking on their behalf. We are designing with TINI so had to do our best to figure these things out until such time as there are official TINI specifications. Part of our support of these products is trying to be honest with customers about what will and won't work, to the best of our knowledge. </Disclaimer>

If you walk up to an LCD frame and touch it, you could crash the TINI with an ESD surge to the data bus. I've done this on similar '8051 with LCD systems', haven't tried to do it on TINI. On our commercial and custom systems we make sure to ground the frame of the LCD well, and often cover the display with a window or recess the metal bezel behind a frame so that it can't be easily touched. Same thing applies to keypads.

In my opinion, for all these reasons, an LCD interface on TINI should be buffered from the TINI signals.

We are putting our money where our mouth is - the SBX2 card has a Xilinx CPLD which buffers all the signals, presenting only one CMOS load to TINI. It also does the timing for the LCD, so that TINI can execute a memory mapped write at full speed, and then goes on its way. The LCD
cycle timing is slower than a full speed TINI cycle, plus then the LCD takes typically 20-40 usec to process a character, there's no need to make TINI wait around for this. SBX2's CPLD deals with the LCD slowness (slow relative to TINI and most high speed controllers). SBX2 is intended to work with an LCD on at least a 12" ribbon cable.

One other note about LCDs - they are VERY high impedance devices. Also the LCD assembly can be permanently damaged by washing, especially if any minute amounts of contaminants are washed into the LCD module. Soldering to the LCD should be done very carefully. We use special "no clean" flux solder. I've seen poorly soldered and cleaned LCDs grow whiskery looking structures around the connector pins after several months, and these have
enough conductivity to affect the operation. LCD data for 20x4 is online here. http://www.systronix.com/access/lcd.htm

Your mileage may vary, these are my opinions and may not be shared by others.