This article aims to document some of my findings while researching the inner workings of my D816 PABX. While several versions exist for all those models, the general operational principles are, more or less, the same.
Those units have a lot in common and the schematics I could find seem to confirm they're iterations of the same project, even though TA1232 has important differences, like not supporting digital proprietary phones.
The CPU is a classic 68k chip. My unit, particularly, uses a TMP68301AF-12 CPU. According to the datasheet, it's based in the 68HC000 core, produced by Toshiba. It's a 16-bit processor, with 24 addressing lines and 16-bit wide bus.
Newer models (-3 onwards) sports 12 MHz CPU's. I haven't been able to find out if newer models are using different CPU's or are just clocking higher. However, if the service manuals are of any indication, all those models seem to be sharing the same CPU model. The replacement part in all three manuals has the same name: PQVI68301FCG, strongly suggesting this is the same 68301 as we see above. Also, we got exactly the same pinout, albeit slightly different connections:
Of course, those service manuals seem to be corresponding to their earlier revisions. So, while supplementary information doesn't show up, we can only speculate on how different, exactly, 12 MHz models are. But at least we can see three different models sporting almost identical hardware.
This PABX is, in fact, a simple, embedded m68k computer. Memory and peripherals are addressed directly, there's no MMU. Also, albeit the system has a 16-bit wide bus, it looks like only the memory chips are using all those lines (see diagram below.) All other devices I saw were using only eight lines of the data bus. Addressing lines seems also to go under-used for most devices, e.g. expansion cards using A1-A5, or even the Caller ID card, using A1 and A5, ignoring all others.
Those systems seem to be typically equipped with four RAM and two ROM chips. TA1232's Service Manual seems to suggest they might not be always populated. On my TD816BX, they are. Since this is a purpose built embedded computer, rather than a general computer, we got SRAM chips soldered to exactly the number of lines they need, without room for larger chips.
I couldn't ascertain why the heck we have two small 32K chips as the first two SRAM IC's, although I find quite interesting the fact that the configuration dump does result in a file that's approximately 64 KB in size, and I remember reading somewhere that those systems hold the configuration in the SRAM's. How this is holding up after all those years scapes me, but the battery I thought to be specific to the RTC may have something to do with this.
As this investigation progresses, I might transfer this section to it's own article, since it's a bit of a rabbit hole.
The digital PABX's (KX-TD) got the KX-TD193 or TD193X as their standard Caller ID add-on card. This part features pretty standard FSK modems that output serial date for two serial-to-parallel converters, that then feed data to the system bus.
Funnily enough, there's a second card for KX-TD systems called KX-TD193DX, that also does DTMF (hence the extra «D» in the name, I guess), and this card also happens to be nearly identical to the TA-123293 card, which also had DTMF decoding facilities. The only significative difference I could spot was an extra IC, which communicated with TA1232 via four extra pins that KX-TD's don't have.
However, for some reason my monkey brain can't process, Panasonic decided that this board, that was manufactured at least as early as 1994, wouldn't be supported by KX-TD system until later revisions, late in the 90's. This is weird. This is wasteful. But it's also how capitalism works, some would say.
So, as currently, I'm fighting capitalism in a infinitesimal scale: a ridiculously pointless exercise to reverse engineer those boards and understand how different they are. I am, however, a salaryman, and also a lot of other alter egos, each trying to fight for my already limited time, so I can't promise anything. Somebody once told me the world is gonna roll me we have to under-promise and over-deliver. Here we over-promise and under-deliver. Get used to it.
Anyway, the first thing to note is that half of the board does basically the same thing: we still got four FSK modems and two serial-to-parallel converter IC's. According to the TA123293's schematic1), the DTMF IC is connected in parallel, completely independent to the FSK processing circuitry from the TIP and RING to the the addressing/data lines. So it's really two cards in one.
Though I haven't thus far been able to track down any schematics for KX-TD193(X) (at least not for a price I'm prepared to pay, i.e. > $0), it seems that the addressing circuitry differs quite a bit from what we have on the DX/123293 version.