This is an OPL-compatible driver for Yamaha's YMF825 FM synthesizer chip (also known as SD-1). It allows playback of OPL2 or OPL3 music on SD-1 by translating OPL register writes to their SD-1 equivalents.
See this video for a demonstration of OPL PC game and tracker music played back on SD-1: https://youtu.be/BEgAx0jngKQ
This repository consists of:
hardware/- breadboard wiring diagram and schematic for a USB to SD-1 interface based on Raspberry Pi Picofirmware/- firmware for the above devicevgmplay/- a Python program that plays back .VGM, .VGZ and .DRO files via a connected USB to SD-1 device
The firmware is written in C++ using the Raspberry Pi Pico SDK. After installing, build the firmware as follows:
$ cd firmware
$ mkdir build
$ cd build
$ cmake ..
$ make
The included vgmplay program requires Python 3 and pyserial. I have used it successfully on Linux and macOS, and it might also work on Windows.
Install pyserial as follows:
$ pip3 install pyserial
To play a file:
$ cd vgmplay
$ python3 vgmplay.py -d <serial port device> <path to .vgm/.vgz/.dro file>
<serial port device> will be, for example, /dev/ttyACM0 on Linux or /dev/tty.usbmodem1234561 on macOS.
The SD-1 has a number of significant differences to the OPL series that prevent a direct 1:1 mapping of registers:
The OPL3 has up to 18 2-operator voices, or up to 6 4-operator voices. Two 2-op voices are combined to make each 4-op voice. Most tunes will use 16 voices or fewer, but may still play sounds on voices 17 and 18. Therefore, the translator allocates SD-1 voices dynamically to OPL3 voices as required. In the event that all 18 2-op voices are used, the least recently allocated voice is re-assigned and some notes may be cut off.
The OPL3 registers contain a bank of FM parameters for each voice, and the sound playing on a particular voice can be modified simply by writing to the registers.
On the SD-1, FM parameters are not associated directly with any particular voice. Instead, there is an array of 16 banks of tone data that must be written to sequentially, and a tone within this array is selected when a note is played.
As the SD-1 has 16 voices and 16 tones, the translator simply reserves a tone for each voice. However, the SD-1 tone data must still be written sequentially starting at the first position in the array - therefore the OPL registers are cached in memory, and all tones converted and re-written up to the location of the changed tone, any time an OPL register changes.
The SD-1 uses a sample rate of 48000Hz, compared to the OPL3's sample rate of 49716Hz. As a result, by default the same block and frequency numbers produce a lower pitched note on the SD-1. The translator sets the global frequency multiplier register of the SD-1 to compensate for this difference.
The OPL3 has an 'envelope type' parameter that determines whether a note is held at its sustain level while the key is on, or whether it transitions immediately from the decay to release phase. The SD-1 instead has a sustain rate parameter that determines the rate at which the level is reduced during the sustain phase. The translator sets the sustain rate either to the release rate or 0 according to the value of the envelope type parameter, to mimic the OPL behaviour.