Question about weird thing in the MMU schematics

I personally welcome all the contributions here, as I've learned a tremendous amount, and I think it is useful to document all of this stuff in one place, even if that may seem cluttered.

I'm going to say again that I'm tremendously excited by the possibility that someday, maybe even soon, replacements for the MMU and IOU may be available.

S.Elliott wrote:

wonky NPN transistor...

I'm sure you're aware, but that transistor is just an inverter.

jeffmazur wrote:

S.Elliott wrote:

wonky NPN transistor...

I'm sure you're aware, but that transistor is just an inverter.

That's right, but that transistor's presence is strikingly peculiar from a design standpoint.

• The GLU is working from the same information that's available to the MMU and IOU.  The existence of the OR-gate between the MMU and IOU suggests designers were constrained from modifying the internal logic by which the MMU emits SELIO, and constrained from modifying the internal logic by which the MMU responds to SELIO.  Maybe they initially wanted to keep the MMU and IOU unmodified between the //e and //c?  Maybe silicon had already been manufactured?  Or maybe there was a development cost that made it more economical to insert a gate than to modify the MMU or IOU?  Regardless, the OR-gate is a minor kludge and it's not illogical.
• The GLU is identified as a PAL created entirely for the Apple //c, so its internal logic and output logic levels should have been trivially customizable.  It should have been trivial to choose a logic-level for IOUHOLE (pin 22) that could connect directly to the OR gate.  The NPN transistor enables the circuit to function correctly despite that inverted logic-level, but it would have been illogical to choose that logic-level.  Unless they were constrained at that design stage.

IOUHOLE polarity.png

Some additional evidence that these pins were involved in late, messy revisions: the Apple IIc Reference Manual shows disagreement about GLU pin assignments between  fig 11-8  vs  table 11-10  vs the  schematic.

Apple IIc fig 11-8 vs table 11-10.png

Obviously there are always constraints on the circuit designers.  So it's worth watching for inconsistencies and superficially-illogical design choices like these because they offer clues.  Elsewhere in their circuits, Apple tended to use bipolar transistors for power-supply switching rather than as logic.  (eg: Apple's Disk II, Parallel Interface, Centronics Interface, and their original Serial Interface all used bipolar transistors for switching-off current to power-hungry bipolar ROMs when they were idle.)

S.Elliott wrote:

wonky NPN transistor...

jeffmazur commented:

I'm sure you're aware, but that transistor is just an inverter.

Uncle Bernie's comment on 'wonky transistor logic':

Using single transistors to 'fix' issues in digital logic was commonplace back then, long before the SMD 'bird feed' single gates were available. A transistor can make a gate with up to 2 inputs. For instance, if you want a gate with the function Y = !A | B, then you tie A to the transistor base and B to the transistor emitter of a NPN, otherwise known as an inverter if emitter tied to '0' = 'GND'. So from one transistor you can make a two input OR gate with one inverted input.

Cool trick, that.

What is not so cool is that this type of 'logic' akin to RTL (Resistor Transistor Logic) does not have well defined logic thresholds, and this is where it may get wonky. Furthermore, high ohmic base resistors make that 'gate' pityfully slow and the ON/OFF delays may differ by a lot.

In the semiconductor industry, RTL was replaced by DTL, which was replaced by TTL for good reasons.

But few people know that the Apollo Guidance Computer was based on RTL ICs, and these daredevils flew to the moon with it, and even came back alive and well ! Design of rad-hard / space faring ICs is a very special topic. Not sure how many of the failed  (and unmanned) missions to the Moon or to Mars did fail due to cosmic rays acting on 'modern' ICs. But I digress. Just wanted to say that the good ol' RTL logic can be made to work reliably enough for mission critical systems. Project Apollo proved that. So why should it be 'wonky' in an Apple IIc ? Where this stuff gets 'wonky' is when longer PCB traces are involved. But as long as the transistor is close enough to the source of the input signal, it will work.

- Uncle Bernie

In that case, it would be helpful if the OP would add: "differences between //e vs //c" to the topic subject so we can find this discussion again.