How the sound works

A whole second processor, for music

There is a second Z80 on this board, at 11F, running at the same 3MHz as the main one. It has its own program ROM at 9F and its own 2K of RAM at 8F, and it shares no memory whatsoever with the game.

Devoting an entire CPU to sound sounds extravagant until you consider the alternative. Music needs servicing at precise, frequent intervals; the game logic needs to run at whatever pace the action demands. Doing both on one processor means the music stutters when the screen gets busy - which is exactly what happens on home computers of the era. A second Z80 was cheaper than the engineering required to make one processor do both jobs well.

The connection between them is a single byte. The main CPU writes a command number to C800; the latch at 8J holds it; the sound CPU reads it at address 6000 in its own, completely separate address space. That is the entire interface. No shared RAM, no handshake, no reply.

The one-way design has a real consequence: sound can never block the game. If the sound program crashes the music stops and play continues, because nothing the main CPU does ever waits on an answer. The main CPU can also hold the sound CPU in reset through bit 4 of the control register at C804 - the machine can switch its own musician off and on.

The two processors share exactly one byte. The main CPU writes a command number to C800 and carries on; the sound CPU reads it at 6000 whenever it next looks. There is no shared memory, no handshake and no way for sound to answer back, which is precisely why a sound fault can never hang the game.

Two YM2203s, which is more than it sounds

The sound chips are a pair of Yamaha YM2203s at 9C and 11C, clocked at 1.5MHz - the 12MHz crystal divided by eight, like everything else.

Each YM2203 is really two instruments in one package: three channels of FM synthesis and three channels of square-wave PSG. Two chips therefore give six FM voices and six PSG channels, which is a substantial palette for 1985 and explains why the Commando theme has both a melodic line with real timbre and a hard percussive backing underneath it.

The chips also carry the sound CPU's timers, which is how the music keeps tempo without any external timing hardware.

One small trap for anyone tracing this: sheet 4/8 numbers the DACs against the chips in a crossed order - 8C-2 belongs to chip 1 at 11C and 8C-1 to chip 2 at 9C. It is noted in the model because it is exactly the sort of thing that costs an hour on the bench.

The last analogue circuit in the machine

Sheet 4/8 is the only part of Commando that is not digital, and it reads like a different era of engineering because it is one.

Each YM2203 sends its FM output out as a serial stream to a companion YM3014 DAC. Its three PSG channels leave as raw analogue and are summed passively with resistors. The combined signal passes through LM324 op-amp stages for buffering and gain, then a mixing pot at VR2 and a volume pot at VR1, and finally into an HA1368 power amplifier that drives the cabinet speaker directly.

The drawing carries a piece of history in the title block: the mixing resistors were revised from 33K to 56K at the factory. Somebody built this, listened to it, decided the PSG channels were too loud against the FM, and changed three resistors. That revision is on the schematic and it is on the board.

The only part of the machine that is not digital. Each YM2203 hands its FM voice to a companion DAC while its three square-wave channels leave as raw analogue, and the two are summed with resistors before any amplification happens.

The drawings behind this story

Each one opens the interactive schematic, where every chip and every net on the sheet can be traced pin by pin.

Every chip in this story

18 chips across 2 sheets. Click any of them to open the schematic with that chip selected. A few appear on more than one drawing, because Capcom split some packages across sheets.

← back to the dashboard