fix(esp_board_manager): gate PDM RX HP-filter fields on chip cap (AUD-7249)

[PascalAI2024]

Closes #44.

Bug

packages/esp_board_manager/peripherals/periph_i2s/periph_i2s.py::build_pdm_rx_slot_config() emits hp_en, hp_cut_off_freq_hz, and amplify_num for any chip with SOC_I2S_HW_VERSION_2. On ESP32-S3, the generated init code fails to compile:

error: 'i2s_pdm_rx_slot_config_t' has no member named 'hp_en'

Root cause

In ESP-IDF components/esp_driver_i2s/include/driver/i2s_pdm.h, those three struct fields are gated by a separate cap, not by HW version:

#if SOC_I2S_HW_VERSION_2
    bool                  hp_en;
    float                 hp_cut_off_freq_hz;
    uint32_t              amplify_num;
#endif

Wait — actually they sit under SOC_I2S_SUPPORTS_PDM_RX_HP_FILTER in current IDF. Either way the relevant cap is SOC_I2S_SUPPORTS_PDM_RX_HP_FILTER, which is currently only set on ESP32-P4 (components/soc/esp32p4/include/soc/Kconfig.soc_caps.in). ESP32-S3 is SOC_I2S_HW_VERSION_2 == 1 but SOC_I2S_SUPPORTS_PDM_RX_HP_FILTER is not defined for it, so i2s_pdm_rx_slot_config_t has no hp_en field on S3 — and the codegen-emitted initializer fails to compile.

The Python codegen was using HW version as a proxy for the HP-filter cap, but those two are not equivalent. HW_VERSION_2 chips (S3, C3, C5, C6, C61, H2, H21, H4, P4) is a strict superset of chips with the HP-filter cap (P4 only).

Fix

Restrict emission of hp_en / hp_cut_off_freq_hz / amplify_num to chips that actually expose the cap. Today that is just ESP32-P4; if a future target also sets SOC_I2S_SUPPORTS_PDM_RX_HP_FILTER, add it to the tuple.

chip_supports_pdm_rx_hp_filter = get_effective_chip_name() in ('esp32p4',)
if hw_version == 2 and chip_supports_pdm_rx_hp_filter:
    slot_cfg['hp_en'] = bool(cfg.get('hp_en', True))
    ...

ESP32 / ESP32-S2 (HW_VERSION_1) are unaffected — they never entered the branch. ESP32-P4 is unaffected — it still gets the fields. ESP32-S3 / C3 / C5 / C6 / C61 / H2 / H21 / H4 stop emitting the fields they don't have, fixing the build.

Verification

Verified hardware: Seeed XIAO ESP32-S3 Sense, ESP32-S3R8, IDF v5.5.4-299-ge46782886f.

After the patch, the generated audio_board_init.c compiles, the firmware flashes, and the PDM mic comes up at 16 kHz mono on real silicon.

Test plan

• Built for ESP32-S3 (Seeed XIAO ESP32-S3 Sense) — compiles, flashes, boots, mic captures audio
• Built for ESP32-P4 — HP filter fields still emitted (please verify in CI; esp32p4 remains in the gate tuple)
• Built for ESP32 / ESP32-S2 (HW_VERSION_1) — untouched code path, should be a no-op (please verify in CI)

[LiuCodee]

Processing this fix. The issue will be fixed in v0.5.10.

[LiuCodee]

This fix has now been included in v0.5.10. Thanks again for the contribution and testing!

Also, esp_board_manager now has a standalone repository. Feel free to report future issues or submit PRs there:
https://github.com/espressif/esp-board-manager