Re: Magic Voice Schematics

From: Grósz Attila (
Date: 2005-04-12 23:56:18

Hi Richard,

Nice to see you back :-)

I still haven't given up on emulating the Magic Voice and the 
Toshiba 6721A in general.

A while ago, I have successfully extracted (some of) the PARCOR parameters from the WAV samples you guys have extracted from a 364. I used this Japanese open source toolset:

I have also sort of successfully resolved the interfacing between the
speech chip and the ROM program which I assume mostly analogous in the
two hardware. Based on the Toshiba data sheets available on Funet (BTW, where are the other pages?) I have noted the power up state of the speech chip:

Synth condition set
- filter stages : 10
- repeat available : yes
- frame length : 20 ms/frame
- bit rate : 96 bits/frame

This already gives a few hints about what sort of parameters we could expect...

Bearing in mind that Richard Wiggins was the one who designed the
Speak n Spell toy at TI and later headed Commodore's short liven Speech
Division, I assume these two US patents are also handy when trying to
figure out how those ROM parameters were fed to the lattice filter in
the LSI:


Use to grab them. I did so, too, but it is
really _tough_ read considering I am not an engineer. Besided it is several dozen pages long. It does shed some light though and I hope it will get brighter with time :-)

Richard Atkinson wrote:

>I've had a go at hand-disassembling the first part of the Magic Voice
>code, using information from the schematics. The code starts up in Ultimax
Great! I have attempted making a CRT image of it with Markus Brenner's 
MCRT but I failed
miserably. Would someone else try that who is more literal with the C64 
than I? :-)

>logic. I put my commented disassembly up on the web:
On its way.... :-)

>The code allows a few educated guesses at what some of the control signals
>do. PB6 (connected to pin 9 of the gate array) seems to be used to select
I do have some guesses at the C364 part maybe we could compare them once.

>think the GAME# function on the gate array should be an output not an
>input. The unmarked signals (pins 7, 8 and 9) all appear to be inputs.
Is there more info available over the MOS8706R0 ? Data sheets perhaps? 
Has someone access to ? They seem to have the T6721A as well, by the 
way... but it is not free.

>Will try disassembling the code copied to $C000-$C5FFF tonight.
Eager to hear about the results!


       Message was sent through the cbm-hackers mailing list

Archive generated by hypermail pre-2.1.8.