PC Engine / TurboGrafx-16 core for Sipeed Tang FPGA boards, integrated with TangCore (BL616-based ROM loading, joypad and on-screen display).
▶ Video: pcetang on the Tang Console 60K · Download: latest release (needs the firmware v0.2.1+) · Support: Ko-fi
HuCard and CD-ROM² games boot and play on Tang Console 60K — exact-locked HDMI (720x480, CEA 480p timing) needs monitor set up to correct 4:3 aspect, PSG audio, a DS 2 controller and an in-game OSD. CD games run from real CHD images served over UART, with CD-DA music and ADPCM voices. Most CD games are already full playable. backup-RAM saves persist on the SD card. SuperGrafx works (some titles may still be imperfect). Arcade Card games boot and play, not perfect yet — tested games reach gameplay; some glitches to fix.
Primer 25K and Nano 20K build clean but have never run a game yet — support is on the way through an external MCU to give them the storage path they lack; wiring in progress). Don't buy those boards for this core today except if you are developer !!!
This is a port, and the interesting parts were written by other people:
- The PC Engine RTL is srg320's TurboGrafx16_MiSTer.
- The ROM-loading / joypad / OSD layer is nand2mario's TangCore and nestang.
The work here is the port itself: Gowin BSRAM/SDRAM mapping, the HDMI output path, the BL616 companion protocol, and a from-scratch SCSI/CD bridge that serves CHD images over UART. GPL-3.0 throughout — see THIRD_PARTY_LICENSES.md for file-by-file provenance.
| Console 60K | Primer 25K | Nano 20K | |
|---|---|---|---|
| Builds clean (gw_sh, 0 errors, 0 timing violations) | yes | yes | yes |
| Bitstream loads, core runs on real hardware | yes | yes (HDMI sync locked) | never loaded |
| HuCard game runs on real hardware | yes | no — no storage path | no — no storage path |
| Video / audio / controllers on real hardware | yes | no | no |
| CD-ROM²: system card boots off a CHD | yes | no | no |
| CD game playable, with CD-DA and ADPCM | yes | no | no |
| SuperGrafx | works (some titles may be imperfect) | compiled out | no room |
| Arcade Card | boots and plays, not perfect yet | no room | no room |
| Backup-RAM saves (persist on SD) | yes | no | no |
Arcade Card games are not perfect yet. Garou Densetsu 2 and World Heroes 2 have minor graphic glitches in gameplay, undiagnosed. Sapphire plays but its audio does not reach HDMI capture devices, and it freezes at the level-1 boss (the CD music keeps playing).
Two players: switch Multitap on. It is off by default, like on MiSTer. Open the in-game OSD, go to Options and set Multitap to On (needs firmware v0.2.1). The setting is saved on the SD card.
Use DualShock 2 pads. The USB gamepads tested so far are not recognised. I need to do more testing.
Cheap HDMI capture devices (UVC) can have issues with timing Testing on several PC monitors through HDMI (Iiyama PC monitor + cheap HDMI portable monitor) works okay. Tested also on real Sony Bravia 4K TV, no issue at all. Some HDMI fixes are under investigation.
Primer 25K and Nano 20K still play nothing, for want of a storage path — see above.
Requires the Gowin toolchain (gowin-edu, not gowin-pro — the latter segfaults
inside its own libgwsyn.so on these designs):
gw_sh build_console60k_cd.tcl # also build_primer25k_cd.tcl, build_nano20k_cd.tcl
Simulation (GHDL). The unit suites take seconds and need no ROM or disc image:
sim/cd/run_cd_bridge.sh # 18 cd_bridge checks, incl. the 11 real boot commands
sim/cd/run_scsi_phase.sh # SCSI phase lines + a 2048-byte burst at real MCU pace
The full-system boot simulation needs a system card and a sector slice, and is much faster on GHDL's LLVM backend than on mcode (about 3.5 s versus 30 s of wall clock per simulated millisecond):
scripts/cd_toc.py game.chd > toc.txt
chdman extractcd -i game.chd -o g.cue -ob g.bin
scripts/cd_slice.py g.bin --lbas 3588-3600,3620-3632,11940-11975,12070-12078 > sec.hex
TOC_FILE=toc.txt SECTOR_CNT=80 sim/cd/run_cd_boot_llvm.sh syscard3.pce sec.hex
See the header of run_cd_boot_llvm.sh for the three things ghdl-llvm needs before it will
start at all.
The BL616 MCU firmware lives in a separate repo,
rtissera/firmware-bl616 — a fork of nand2mario's
firmware-bl616. Use v0.2.1 or later (see its releases). Stock TangCore firmware has no PC Engine support of any kind — its
cores are NES, SNES, GBA, Mega Drive, Master System and PC/XT. The fork adds:
core/pce.cpp— HuCard (.pce,.sgx) loadingcore/pcecd.cpp/.h— the whole CD-ROM² side: TOC upload, the sector server that streams CHD sectors over UART, CD-DA delivery by DMA and the ADPCM feed- a HID descriptor-parser fix for controllers that use extended (4-byte) usage items
So a bitstream from this repo running against stock firmware loads nothing at all — no HuCard, no disc. The two repositories are one system and have to be built together.
This is a one-person project with real costs: FPGA boards, capture and test gear, and a lot of engineering time. Ko-fi donations and hardware donations are both welcome and go straight into the next round of work — the Arcade Card fixes, Primer 25K / Nano 20K via the external MCU, and more.
- Ko-fi: ko-fi.com/rtissera
- Hardware (Tang boards, dev kits, test equipment): get in touch through a GitHub issue or Ko-fi.
Development was partly AI-assisted (Claude, under my direction). Saying so here rather than burying it: the interesting question about a project like this is not whether a model was involved, it is whether the claims hold up.
This is NOT a purely vibe-coded project and the methodology from start to this first release has been real hardware testing, hardware simulation and exercizing against both reference HDL implementation and software emulators.