Sega G80 HW FAQ 1.11 released

A new version of the HW FAQ is up on www.spies.com. This
version has the register descriptions for the 4 player
I/O expander and has a few other Eliminator related
corrections.

Once I get the Universal Sound Board documented, the
FAQ should be just about finished..

To: vectorlist@goonsquad.spies.com
From: Clay Cowgill <clay@supra.com>
Subject: Re: Speech board idea...
Sender: owner-vectorlist@goonsquad.spies.com
Precedence: bulk
Reply-To: vectorlist@goonsquad.spies.com
Status: RO

>> This Star Trek/Space Fury/Etc. Speech board has been bothering me since
>> those SPO-250 speech chips seem impossible to find. Have any of you tried
>> playing with the ISD (Information Storage Devices) Single Chip Voice
>> record/playback chips?
>
> Are these the same ones that they sell at Radio Shack?

I think so. I don't know the "duration" of the radio shack ones.
Apparently there's a few different sampling speeds that you can by. I saw
a little chip-on-board one at RS the other day-- I think that's the little
brother of the 2500 series.

> One thing I notice about your scheme is that it is redundant. If
>the PC samples are good enough, why don't you just Hex dump them (The
>digital sequences) to an EPROM. Run these puppies through a DAC and a
>continuous-time lowpass filter, and you've got your speech! Use one of
>Crystal's DACs and your analog signal out will sound as good as your
>digital data (Had to get my Crystal plug in :) )

Well, the only problem was cost/complexity. A PIC and an ISD chip is
simple and relatively cheap. The way I was thinking of doing "sample
playback" was to take some kind of MCU, somekind of DAC, and a bunch of
EPROM. I don't know how much actual "time" is required for sampling, but I
figured maybe 20 seconds per game? So... 20 seconds * 8Khz sampling =
~160Kbytes multiply by number of games with speech... Gets pretty big.)

Thing is, a cheap 8-bit MCU probably only has 16 bit addressing, so you
bank select to get at more samples, and you use a 27C040 for storage, glue
a DAC on, probably a register to get the speech commands from the G-80
backplane... It starts getting big/expensive.

> Of course, you need a controller for this whole thing, but I'll
>bet your PIC will work just fine for that. You just need to clock the
>data out of the EPROM when you get some sort of enable signal. Which
>enable signal you get determines the address of the EPROM. You know your
>sampling frequency when you make the digital dumps from your PC.

The gotcha is the available I/O's. You can have a little MCU and parallel
load a bunch of binary counters to run the address lines to the EPROM, and
then just use an R-2R ladder for the D/A, but those counters are a pain to
wire up, and you'll probably use something like 74x193's, so it'll take 4
or 5 for a 27c040...

> Probably the reason why you didn't think of this in the first
>place was cost....you assumed that DACs cost more than those Voice
>Record/Playback ICs, but I'm not sure that that's true.

I'm willing to spend more (like the ISD chips) for a simple implementation
(PIC + ISD chip). I'm just not keen on wiring up a full address bus,
databus, etc. if I can avoid it. ;-)

> Look into it....Other than than that, your idea sounds pretty
>cool. Maybe people will actually start buying those Radio Shack chips now
>(Those seem like the only Radio Shack "special" items that have been
>around for years!)

Well, let's look at some criteria:

1) we're only dealing with voice-band signals
2) we need a fairly large amount of storage-- "seconds"-wise
3) low parts count is a plus, as is small PCB space to lay it down
4) fewer wires to hook up is good
5) output filtering is good
6) "bulk" programmability is good

The ISD fit most of these pretty nicely-- built in filters, "compressed"
storage, good duration... Of course having to "load" it from an analog
device is a pain.

Normally I'd just say go with an EPROM and raw 8-bit samples with an R-2R
output ladder and a filter on the output. The main problem (in my mind) is
the amount of sampling time. Anyone actually *know* how long all the
samples in Star Trek take when played back to back? (Al? ;-)

(And before anyone laughs at my R-2R network for D/A... This is voice band
audio, and the human ear isn't going to hear much distortion from
mis-matched resistors. Remember the "source" material is pretty weird
sounding in the first place... *laugh* An R-2R network is only about
$0.10-$0.25. Rolling Thunder uses an MCU driving a pair of R-2R networks
directly off I/O ports for it's stereo sampled audio...)

Another approach could be something like an OKI 6295 with an external ROM.
It uses ADPCM for compression, but retains the "play sound #xxx"
functionality.

Yet another approach would be an Oki 6585 which is just basically a ADPCM
decoder. Pair that up with a MCU and a (smaller at least) external ROM for
the samples. Both of these chips are used pretty frequently in video
games, BTW. (well, look at pretty much every JAMMA game from 1988-1995)

Ok, back to work now. :-)

-Clay

Clayton N. Cowgill Engineering Manager
_______________________________________________________________________
/\ Diamond Multimedia Systems, Inc. clay@supra.com
\/ Communications Division http://www.supra.com/
Received on Wed Jun 25 14:00:20 1997