Yamaha FS1R Comes to Software with FSVR Emulator

via musica.studio

ALL 88 ALGORITHMS
Every FM algorithm from the hardware, with the voiced and unvoiced operators the FS1R is known for.

THE FACTORY BANKS BUILT IN
1408 voices [patches], 384 performances and 90 formant sequences read out of the ROM, so the patch browser needs no dump of your own.

FSEQ PLAYBACK
The formant sequences that made the FS1R sound like nothing else, with every loop mode, the start delay and MIDI clock sync.

THE FRONT PANEL IS THE GUI
Down to the cursor stops and the controller set, read out of the firmware’s own screen tables rather than off a photo.

VST3, CLAP AND STANDALONE
Windows, macOS and Linux, built on every push, with no installer and no dependencies.

The FS1R is a 1998 rack synth built around the YMP706, a tone generator Yamaha never documented and used in exactly two products. It was made for two years, it sounds like nothing else, and a working one now costs more than most people want to spend on a box with a two-line display.

In 2025 Zhiyuan Wan (rgwan) took the hardware apart and published the result as rgwan/fs1r_firmware_RE: the SH7044’s internal flash read out through a hooked UART, the 2 MB EPROM and the PLG150-DX ROM dumped, the board’s own UART traffic captured, and a Ghidra project with the boot, loader, flash and LCD code already named. FSVR is built on that work and is a collaboration with rgwan, who runs the debug monitor and the measurements on a real unit while the engine is calibrated against what comes back.

It is a rewrite rather than an emulation of the board. The note-on path, the 192.3 Hz tick pipeline, the parameter conversion tables and the 88-algorithm routing table are read straight out of the decompiled firmware and ported to C++. Nobody has register semantics for the two custom chips, so those are models, and every constant behind them sits in one namespace in the source where calibrating against a recording of a real unit is a single table edit.

It is a work in progress and it is honest about that. Against rgwan’s recording of the FS1R playing its own demo songs, the median envelope correlation is 0.978 and seventeen of the 29 frozen capture files agree with the unit inside half a decibel. What still differs is mostly the effects, and the repository ranks the remaining gaps with the numbers behind each one.