Sierra AGI Ports to the CoCo 3, Any Sound Improvement?

[Great Hierophant]

(Cross-posted from here) The Tandy Color Computer 3 received official releases of King's Quest III and Leisure Suit Larry, and they look and run almost identically to their IBM PC counterparts. Because the CoCo series only have a 6-bit DAC, the music output from a CoCo is PC-speaker quality on the official ports. Also, outputting sound to the DAC is rather CPU intensive so the official AGI engine for the CoCo never plays sound while animation occurs on the screen. Even an IBM PC can manage animation and sound at the same time.

While Sierra only officially ported these two games, the CoCo community has ported the rest of them and made them all friendlier to larger storage devices than the official 157.5KiB CoCo disk format. A cursory review of these ports suggest that they simply used the official engine and tweaked it to work with the data files from the other games' PC versions and called it a day.

But has there been any more ambition than this? Tandy released the Speech/Sound cartridge which contains an AY-3-8913 sound chip. The 8913 is quite capable of doing justice to the three voice music and sound effects that the AGI engine supported. The music and sound effects were originally designed for the SN76496, which the 8913 can eclipse in almost every way. Unfortunately, Sierra could not use it because of a design flaw : the cartridge fails to work in the CoCo's 3's high speed mode, being designed for the CoCo 2 which only supported low speed mode. However, the Speech/Sound cartridge can be modified without too much difficulty to work with the CoCo 3's high speed mode.

The Tandy CoCo 3 would make a great, fairly compact AGI adventure game playing device if this one remaining flaw was addressed. Has there been any effort to do so?

[linville]

The Game Master Cartridge technology includes the SN76489, which is compatible with the SN76496 from the Tandy 1000 line. Anyone with a "Master Edition" release of Fahrfall has this hardware.

Someone familiar with the encodings of the Sierra game files should have minimal problems unpacking the sound data for each game. The frequency value for each note should require a scalar conversion to match the 4MHz clock on the GMC hardware. This might be done as a processing step when converting the game or it probably could be done "on the fly" with acceptable results -- YMMV, of course.

Anyway, no need to engage the programming horrors of the Speech/Sound Pak! Fahrfall or (potentially another GMC-based cartridge) has all the necesary hardware. The rest, as they say, is a Simple Matter Of Programming... ?

[Great Hierophant]

The Game Master Cartridge sounds like a nearly ideal solution, although the use of a 4MHz clock over a 3.58MHz clock is a bit of a head scratcher, especially as the SN76489 is not great when it comes to the lower sound frequencies it can produce (109Hz being the lowest at 3.58MHz and 122Hz at 4MHz) and most other systems which used the SN76489 ran it at 3.58MHz.  

Edited September 23, 2020 by Great Hierophant

[linville]

I'm not really interested in an argument about the GMC design choices. Nevertheless...

• using an NTSC clock versus the 4 MHz clock only yields 1 or maybe 2 more usable notes at the low end
• many songs can simply be shifted up an octave during processing
• original music can either shft up an octave or simply avoid the missing lower notes
• using the lower frequency clock yields a more notes with frequency errors
• the 4MHz oscillators were somewhat easier to source (not as much demand in the HD era for NTSC oscillators?)
• the BBC Micro used 4 MHz, and most conversion tools seem to support BBC

Finally, the cards can be built with NTSC frequency oscillators (or anything else desired). If someone worked on the CoCo3 game conversions but didn't want to do the frequency conversions, we could figure-out a way to make NTSC-clocked hardware available to interested users.

[Great Hierophant]

I thought that the CoCo 3's cartridge slot outputs a 1.79MHz signal, so could a clock doubling circuit (or a clock quadrupling circuit for the CoCo 1 &2) be used to bypass the issue of needing a 3.58MHz crystal oscillator? 

[linville]

That would trade an oscillator for extra circuitry and it would make the sound output (and therefore the software to drive it) dependent on whether or not the "HI-SPEED poke" is active at any given time. It's hard for me to see how that would be better.