The TTL input is an emitter, as is any bipolar device input of this class, and that's what ultimately pulls up the inputs in the general case with bipolar logic. LS logic bufers the inputs, thereby reducing the input current required to drive the input, but still sourcing current, albeit through a diode. The pullup on the diodes' anode is nominally 24K-ohms, according to a databook I've checked, and that's what sets the input current. Apparently, some LS inputs are PNP buffered and therefore require only that one drive the base of a PNP that, in turn, drives the diode mentioned above.

Because I've become slightly obsessed with this particular interoperability issue, I've fiddled quite a bit with the three MCU types that I want to understand and here's what I've found.

The HMOS MCU's seem to have different oscillator requirements, though the datasheets don't agree with what I've seen so far. The AMD parts seem to "like" the arrangement where the X1 pin is driven with the external oscillator, and the X2 pin driven with the complement of that, i.e. an inverter is connected between the two, with its input at X1 and output at X2.

While the pullup resistors on P0 and P2 seemed to have some impact in other arrangements, when this clock driver arrangement is used, the AMD parts appear to work normally. The Intel parts, which, according to their datasheet, should "like" this clock driver scheme, seems to prefer having X1 grounded and only X2 driven. The "jury is still out" on that, however, as it occasionally fails to run properly in that configuration. I'm not sure that it's just the oscillator, but I have observed that the output levels from the MCU are considerably higher than what's demanded by the AC latches I'm using. The MAXIM/DALLAS parts seem to like the same arrangement, as one would expect, as it's what their datasheet recommends.

Frankly, I'm puzzled that the clock generator has as much impact as it does, when in a configuration not necessarily the one of choice according to the individual parts' datasheets. I know I'm still not "there" because neither of the "classic" parts works properly with the internal crystal oscillator, irrespective of whether the pullups are provided. I know this seems to be tangling up seemingly unrelated details. That's one thing that has me puzzled.

I'm not going to toss out these old parts, since I never know when I'll have to replace one in an old board, but I doubt I'll use them for applications where the Maxim/Dallas parts will work.

RE