🔱
[BOOT] one board, every radio OK
T-EMBED CC1101 PLUS · 5 RADIOS ONBOARD
04 // FIVE AUDIT PASSES

47 defects went in the bin

Bringing the port up meant tearing through the entire firmware, adversarially, one domain at a time. The crashes were never the interesting part. The interesting part was everything that ran, reported success, printed a rising counter, and did absolutely nothing.

$ audit · sub-ghz | wifi | ble | nrf24+ir | sd+nfc+boot
confirmed defects ......... 47  each verified against real code
dead protocol decoders .... 3   a comparison that could never pass
features unreachable ...... 15  bound to keys this board lacks
bus-destroying paths ...... 2   menu items that killed the display
RESULT: FIXED, FLASHED, VERIFIED ON HARDWARE

Four worth telling

Three decoders were dead code

The pulse-width comparison took the absolute value of one side but not the other. Every sync test uses a negative width, so the check could never pass. Princeton, CAME and NICE silently fell through to a generic decoder that reported invented protocols for real fobs.

SUB-GHZ

A counter that counted nothing

Samsung BLE packets declared one more byte than they contained, so conformant phones dropped them. The on-screen counter still climbed, because it counted the controller accepting the buffer rather than anything reaching a target.

BLE

The antenna was never switched

This board routes its sub-GHz front end through a band-select network. The pins were declared and never driven, so the radio sat on whatever filter it powered up in and sensitivity collapsed across most of the range.

RF PATH

A fifth of every waveform

Captured .sub files store 512 pulses per line, far more than the parser buffer. Continuation chunks were silently discarded, so replay keyed the radio with roughly 20 percent of the signal.

REPLAY