Ft8 integration; round 2 - #17
Conversation
|
I get the following error when I try to compile. Is there a missing file that defines a new data type? |
I just configured GitHub to attempt to compile a pull request before allowing it to merge. This will help test for cases like this for future PRs (don't think the rule will run against this PR which was submitted before the change) |
…ensy core mods to usb_audio.cpp/h
# Conflicts: # code/src/PhoenixSketch/Loop.cpp # code/src/PhoenixSketch/SDT.h
Upstream renamed BUFFER_SIZE to USB_BUFFER_SIZE in SDT.h to avoid a namespace conflict with another library, and updated DSP.cpp's non-FT8 code accordingly. This left three references in the FT8 additions (PlayUsbBufferPreVol's test tone generator and ReadMicrophoneBuffer's FT8 sample path) still using the now-undefined BUFFER_SIZE macro.
Ft8UsbBridge_DrainToUSB() was the only project code calling a core_mods-only function (usb_audio_push_block) -- confirmed by searching src/ for it and the other three core_mods additions (usb_audio_pop_receive_block, usb_audio_tx_has_room, usb_audio_manual_transmit), none of which are otherwise referenced. core_mods/usb_audio.h now defines T41_CORE_HAS_USB_AUDIO_PUSH_BLOCK next to the push_block declaration, so project code can detect at compile time whether the patched core is installed. Anyone re-copying this header into a Teensyduino install must bring this #define along -- applied here to both the repo copy and the installed copy at AppData/Local/Arduino15/packages/teensy/hardware/avr/1.62.0/cores/teensy4/. When the sentinel is absent, Ft8UsbBridge_DrainToUSB() drops the buffered RX samples instead of calling the missing function, with a #warning at the top of the file explaining why. FT8 TX (PC -> radio) is unaffected since it never depended on the core patch. compile.yml gains a second, independent job (compile-usb-audio) that builds with usb=serialmidiaudio against CI's stock-installed core, so the FT8 code path actually gets compiled and the new gate gets exercised instead of silently skipping (the existing usb=serial2 job never defines USB_AUDIO/USB_MIDI_AUDIO_SERIAL, so it excludes this code entirely and proves nothing about it).
… just T41_USB_AUDIO
usb_audio.h wraps AudioInputUSB/AudioOutputUSB themselves in
#ifdef AUDIO_INTERFACE, which boards.txt only defines for USB Type =
Audio / Serial+MIDI+Audio. T41_USB_AUDIO (Config.h) is independent of that
board menu selection, so a usb=serial2 build previously still tried to
declare AudioInputUSB g_usbIn -- compile failure ("does not name a type"),
plus several call sites (Ft8UsbBridge_SetTxActive at
MainBoard_AudioIO.cpp:379/402, Ft8UsbBridge_Init at
MainBoard_AudioIO.cpp:570/Globals.cpp:518, Ft8UsbBridge_DrainToUSB at
MainBoard_AudioIO.cpp:575) called into functions that wouldn't even be
defined in that configuration -- link failures waiting to happen.
Restructured Ft8UsbBridge.cpp: the RX FIFO plumbing, Ft8UsbBridge_PutRxSamples,
Ft8UsbBridge_DrainToUSB (its only interface-dependent line already gated by
T41_CORE_HAS_USB_AUDIO_PUSH_BLOCK), and Ft8UsbBridge_SetTxActive never
touched AudioInputUSB/AudioOutputUSB in the first place, so they're now
defined unconditionally -- fixing every mismatched caller above without
touching the callers themselves. Only Ft8UsbBridge_Init/Ft8UsbBridge_GetSamples
and the AudioInputUSB/AudioRecordQueue/AudioConnection objects they depend on
stay behind the strict #if defined(T41_USB_AUDIO) && (defined(USB_AUDIO) ||
defined(USB_MIDI_AUDIO_SERIAL)) guard, with safe stubs (silence/no-op) in
the #else so callers don't need to know which USB Type is selected.
Audited every reference to AudioOutputUSB, AudioInputUSB, usb_audio_, and
Ft8UsbBridge_* under src/: CAT.cpp/Loop.cpp/MainBoard_DisplayHome.cpp only
call the always-defined GetFt8Mode()/SetFt8Mode(), and MainBoard_AudioIO.cpp's
Q_usbOut_L/R/usbRxGainL/R (AudioPlayQueue/AudioAmplifier) don't require
AUDIO_INTERFACE at all -- none of those needed changes.
|
All checks now green. Since your comment: the compile error was FT8 code gated only on T41_USB_AUDIO (Config.h) without also checking the board's USB Type provides the audio interface — fixed so all configurations build: USB Serial (FT8 compiled out), Serial+MIDI+Audio on stock core (RX-to-PC path stubbed with a #warning — see core_mods/README for why the 2-file core patch is needed), Serial+MIDI+Audio with core_mods applied (full FT8, verified on air). Also added a second CI job compiling the USB-audio configuration — happy to drop it if you'd rather keep compile.yml minimal. Stock-core feasibility analysis is in docs/EXPERIMENT_LOG.md. Ready for review. |
|
Has a spectral analysis been done on the TX signal? Just listening to the signal or making successful contacts isn't sufficient. It doesn't make sense that a transfer from the PC at 48kHz to the T41 at 192kHz without modification will yield a pure FT8 transmission. The smoking gun is that modification is needed from the T41 to the PC. |
|
I have not done a spectral analysis. i'll look into that |
3 participants
Added Oliver's suggestions to the code.