NORbert: chasing a ground bounce through the quad I/O address phase

Mon, 20 Apr 2026 00:00:00 +0000

Found a weird bug in NORbert last week. Quad I/O reads (0xEB, the 1-4-4 mode) driven by an FT4222 would silently return 0xFF at specific SPI clock frequencies. 1-1-1, 1-1-2, 1-2-2 and 1-1-4 all worked fine. 10, 15 and 20 MHz failed; 12 and 30 MHz worked. Same data, same emulator, only the SPI clock changed.

Peak "this is supposed to be the fun hobby project" energy.

The symptom

Flashprog verify run at 10 MHz, quad mode:

Fpga on Arthur's blog

NORbert: chasing a ground bounce through the quad I/O address phase

The symptom