Richard Atkinson

Mailing list post regarding V364 / MOS 8706

From: Richard Atkinson
Date: 2001-03-22 11:17:38

It's been a while since I posted to either list, but I have some news
which may be of interest to many of you. For the past month and a half I
have been in Austin, Texas working for a tech company and I took the
opportunity to visit Bo Zimmerman.

Bo and I worked on his V364 among other things and we took some recordings
of the speech chip words. The results are at on his V364
page. Whereas the Magic Voice has 235 words, the V364 has 261 words and a
modified version of Daniel Spiteri's Magic Voice word list is available on
Bo's site. The V364 has the same words as the Magic Voice (with the
exception of 3 less duplications) and some extra words at the end, mainly
BASIC 3.5 function names. Some of them were nearly unintelligible, so we
were fortunate that the BASIC 'SAY "WORD"' construct also works. In this
way we were able to determine all of the uncertain words. Listen to 'go'
and 'hide', for example. Finally, 'grey' is replaced by 'pink' as the word
grey is not used in the 264 series colour palette (being a shade of white)
whereas pink is.

The ROM has been on Funet for some time, thanks to Bo's efforts when he
first obtained the machine. It ought to be possible to get the Magic Voice
to say the extra words, using excerpts from the V364 speech ROM. The V364
speech hardware uses the same Toshiba PARCOR speech synth chip (T6721A,
datasheet on Funet) but a more integrated arrangement of glue
hardware. The speech ROM is in a socket to the left of the BASIC ROM and
it occupies slot 3 LOW in the ROM banking scheme. The empty socket to the
left of that is for slot 3 HIGH. There is a single custom ASIC for the
glue logic to the left of that, so the complete listing of V364 28 pin
chips from left to right in the line goes 8706R0 (custom ASIC), empty
(slot 3 HIGH), speech ROM (slot 3 LOW), BASIC ROM (slot 0 LOW), KERNAL ROM
(slot 0 HIGH), 3 PLUS 1 ROM (slot 1 LOW), 3 PLUS 1 ROM (slot 1 HIGH), PLA.

This ASIC replaces the 6525 TIA, 40105 fifo and gate array inside the
Magic Voice. It has four registers and banks in at $FD2x, which is the
unused PLA decoded address range (function 0) on all other 264 series
machines. There are probably four mirrors of the ASIC registers; we did
not check for this functionality at the time. I took the opportunity to
obtain this ASIC's pinout, as a first step towards reverse-engineering the
hardware and maybe one day making a speech attachment for a plus/4 out of
a modified Magic Voice. Here are the results of the investigation:

U27 8706R0 (1284) Custom speech glue logic ASIC

1 RESET# (cpu) 28 +5V
2 IRQ# (cpu) 27 D0 (cpu)
3 R/W# (cpu) 26 D0 (t6721a)
4 phi 0 (cpu) 25 D1 (cpu)
5 $FD2x# (cpu) 24 D1 (t6721a)
6 A0 (cpu) 23 D2 (cpu)
7 A1 (cpu) 22 D2 (t6721a)
8 ? ? 21 D3 (cpu)
9 EOS# (t6721a) 20 D3 (t6721a)
10 APD (t6721a) 19 D4 (cpu)
11 phi 2 (t6721a) 18 D5 (cpu)
12 DI (t6721a) 17 D6 (cpu)
13 DTRD (t6721a) 16 D7 (cpu)
14 GND 15 WR# (t6721)

There was one pin which did not seem to be connected to anything else on
the V364 motherboard; pin 8. Whether this pin is NC internally or an
unused function I am not sure; multimeter testing of this pin on the
working V364 showed it constantly at 0V, but I cannot say for sure there
was no signal on it. In any case, any function it might have would be of
academic interest. The next phase of V364 hardware investigation will
involve oscilloscope and (hopefully) logic analyser testing, but this
could be a long time in the future.

The next part of the overall project will start to determine what function
the bits of the 4 registers have; and hopefully come up with a way to
attach a Magic Voice (or parts of a Magic Voice) to another 264 series
machine in a V364 software-compatible way. This may be of interest to
emulator authors. Has anyone tried the V364 ROM in an emulator recently?


© 2001 R.G. Atkinson.

The author has asserted the right to be identified as the author of the original works.