vladimir/cudaSDR
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
Files
6ca18411cf942651f6a29518607a2613023cf2e5
cudaSDR/Source/WORKLOG_PROTOCOL2_2026-04-13.md
T
vladimir 6ca18411cf protocol 2 fix
2026-04-14 22:06:31 +03:00

12 KiB
Raw Blame History

cudaSDR Protocol 2 Worklog

Date: 2026-04-13

This file records the Protocol 2 related work completed in this repository so future sessions can quickly restore context.

Implemented Earlier In This Session History

  • Added separate openHPSDR Protocol 2 transport implementation:
    • src/DataEngine/cusdr_protocol2_io.h
    • src/DataEngine/cusdr_protocol2_io.cpp
  • Added separate Protocol 2 discovery implementation:
    • src/DataEngine/cusdr_discoverer_P2.h
    • src/DataEngine/cusdr_discoverer_P2.cpp
  • Wired Protocol 2 into the existing project flow without replacing Protocol 1.
  • Updated DataIO so it delegates network start/stop and receive path to the separate Protocol 2 module when the selected device reports protocolVersion >= 2.
  • Added Protocol 2 fallback search after Protocol 1 and fixed the Search button path so manual search also tries Protocol 2.
  • Added initial RX-only Alex support for Protocol 2:
    • General packet enables Alex.
    • High Priority packet sends Alex 0.
    • Filters switch automatically by band/frequency using existing Alex settings.

Implemented Today

1. Device capability model for Protocol 2

  • Extended TNetworkDevicecard in src/cusdr_settings.h with Protocol 2 specific fields:
    • adcCount
    • ddcCount
    • dspClockHz
    • iqFormatFlags
    • phaseWords
    • hasAlex0
    • hasAlex1

2. Separate board profile helper

  • Added:
    • src/DataEngine/cusdr_protocol2_profile.h
    • src/DataEngine/cusdr_protocol2_profile.cpp
  • This helper maps Protocol 2 board ID to default hardware capabilities and human-readable model name.
  • It also provides a summary string for UI device lists.

3. Discovery improvements for Protocol 2

  • Updated src/DataEngine/cusdr_discoverer_P2.cpp to:
    • use board profiles,
    • fill the new capability fields in TNetworkDevicecard,
    • parse extra discovery reply fields from the documentation:
      • byte 20: number of implemented DDCs,
      • byte 21: frequency vs phase word,
      • byte 22: supported IQ/endian formats.
  • Updated src/DataEngine/cusdr_discoverer_P1.cpp to initialize the new fields to safe defaults for Protocol 1 devices.

4. Protocol 2 initialization based on actual device profile

  • Updated src/DataEngine/cusdr_protocol2_io.cpp:
    • fixed General packet bit handling,
    • enabled phase-word mode where required,
    • selected IQ format only when the discovery flags say it is supported,
    • used discovered/default ADC count in the DDC Specific packet,
    • routed DDCs across ADCs,
    • limited receiver count to hardware DDC count,
    • used board DSP clock to compute phase words,
    • preserved RX-only behavior.

5. Data engine behavior for Protocol 2 devices

  • Updated src/DataEngine/cusdr_dataEngine.cpp so Protocol 2 devices:
    • are treated as modern integrated hardware instead of legacy Mercury/Penelope stacks,
    • automatically clamp receiver count to discovered DDC count,
    • publish Hermes firmware field from the Protocol 2 discovery result for existing UI/status paths,
    • enable Alex UI only if the selected Protocol 2 board profile says Alex is present.

6. UI modernization for Protocol 2 devices

  • Updated src/cusdr_networkWidget.cpp so the device list shows a summary instead of only IP address.
  • Updated src/GL/cusdr_oglDisplayPanel.h and src/GL/cusdr_oglDisplayPanel.cpp so the top status display uses the real Protocol 2 model name instead of always showing Hermes.

7. Documentation-driven corrections made today

  • Re-read local documentation:
    • doc/openHPSDR Ethernet Protocol v4.3.docx
  • Corrected Protocol 2 board naming and assumptions from the document:
    • board 1: Hermes / ANAN-10/100
    • board 2: Hermes / ANAN-10E/100B
    • board 5: Orion MkII / ANAN-7/8000DLE
    • board 10: Saturn / ANAN-G2
  • Corrected one important capability assumption:
    • board 2 no longer auto-enables Alex RX support, because the document indicates no Alex receive filters for that board.

Build Status

  • Full build completed successfully:
    • make -j4
  • Output binary:
    • bin/cudaSDR

Additional Fix Applied Later Today

Protocol 2 stop path hardening

  • Investigated the stop path after a reported crash when pressing Stop with a Protocol 2 device.
  • Confirmed that the shutdown flow was effectively hitting Protocol 2 stop twice:
    • first through DataEngine::stop() -> DataIO::networkDeviceStartStop(0),
    • then again through DataIO::stop() -> Protocol2DataPath::stop().
  • Updated src/DataEngine/cusdr_protocol2_io.cpp so that:
    • Protocol2DataPath::stop() only sends the final run = 0 High Priority packet if the path was actually running,
    • networkDeviceStartStop(0) returns immediately if Protocol 2 is already stopped,
    • closeSockets() explicitly disconnects timer and socket signals before deletion.
  • This removes the most obvious duplicate-stop race and reduces the chance of a timer/socket callback firing during teardown.
  • Build was re-checked after this fix:
    • make -j4
    • success

Still needs runtime verification

  • This stop-path fix is compile-verified but still needs validation on real Protocol 2 hardware.
  • If the transceiver still does not stop cleanly, the next thing to verify is whether the specific board expects more than a single High Priority run = 0 packet before socket teardown.

Additional Investigation And Fixes After Emulator Testing

1. Cross-thread DataIO control calls

  • Emulator testing showed that Protocol 2 startup and shutdown were still unreliable.
  • Investigation found that DataIO is moved to its own QThread, but DataEngine was still calling these methods directly from another thread:
    • sendInitFramesToNetworkDevice(...)
    • networkDeviceStartStop(...)
    • stop()
  • Updated src/DataEngine/cusdr_dataEngine.cpp to route those calls through:
    • QMetaObject::invokeMethod(..., Qt::BlockingQueuedConnection)
  • This was done so the Protocol 2 sockets and timers are manipulated from the DataIO thread they belong to.

2. Additional Protocol 2 stop crash findings

  • User logs then showed:
    • QObject::killTimer: Timers cannot be stopped from another thread
    • QObject::~QObject: Timers cannot be stopped from another thread
    • DataEngine:: data IO thread not yet finished...
  • This indicated there were still object-lifetime issues during teardown, not only packet-order issues.

3. DataIO destructor bug in Protocol 2 mode

  • Found a concrete bug in src/DataEngine/cusdr_dataIO.cpp:
    • in Protocol 2 mode, m_dataIOSocketOn can be true while m_dataIOSocket is never created,
    • but DataIO::~DataIO() still dereferenced m_dataIOSocket.
  • Fixed destructor logic so the legacy socket is only closed/deleted when the pointer is valid.
  • This was a direct segfault candidate in Protocol 2 shutdown.

4. Sound output teardown issue

  • Found that DataIO::stop() was stopping and deleting CSoundOut while Protocol 2 shutdown was already in progress.
  • Adjusted src/DataEngine/cusdr_dataIO.cpp so:
    • DataIO::stop() no longer deletes the sound object,
    • CSoundOut shutdown is deferred to DataIO::~DataIO().
  • This was done to reduce cross-thread/timer teardown problems during the live stop sequence.

Current status after these fixes

  • Build re-checked successfully after each change:
    • make -j4
  • Protocol 2 stop handling is more robust than before, but user testing still shows there may be at least one remaining shutdown/lifetime issue.
  • Most likely next area:
    • final deletion order of DataIO / Protocol2DataPath and their child QObjects after m_dataIOThread->quit().

Known Limitations / Remaining Work

  • Protocol 2 implementation is still RX-only.
  • Protocol 2 Alex support currently sends Alex 0 only.
  • Boards with Alex1 support such as Orion MkII / SATURN are not yet fully mapped to the second Alex register according to Appendix D.
  • Full XML or extended hardware-description discovery modes (board 254/255) are not implemented.
  • Wideband Protocol 2 path is still not integrated.
  • Some UI sections still conceptually assume old Metis/Hermes naming internally even though the visible device/model handling is now better.

Good Starting Points For Next Session

  • If continuing Protocol 2 work, read these files first:
    • src/DataEngine/cusdr_protocol2_profile.cpp
    • src/DataEngine/cusdr_discoverer_P2.cpp
    • src/DataEngine/cusdr_protocol2_io.cpp
    • src/DataEngine/cusdr_dataEngine.cpp
    • src/cusdr_networkWidget.cpp
    • src/GL/cusdr_oglDisplayPanel.cpp
  • For documentation reference, use:
    • doc/openHPSDR Ethernet Protocol v4.3.docx
  • Highest value next task:
    • implement proper Alex1/BPF2/RX2 handling for Protocol 2 boards that support a second Alex register.

Additional Fixes On 2026-04-14

1. Protocol 2 control-timing corrections

  • Re-read the local protocol document and verified the command timing requirements again:
    • High Priority / C&C must be sent at least once per second,
    • 100 ms is the recommended periodic rate,
    • packets should also be sent when relevant data changes.
  • Updated src/DataEngine/cusdr_protocol2_io.cpp so Protocol 2 now reacts immediately to settings changes instead of relying only on the 100 ms heartbeat:
    • center/VFO frequency changes,
    • current receiver changes,
    • band changes,
    • Alex config/state changes,
    • sample-rate changes,
    • DDC dither/random changes,
    • number-of-receivers changes.
  • General, DDC Specific, and High Priority packets are now refreshed according to the type of change, and DDC sockets are rebuilt if the active RX count changes.

2. Protocol 2 wideband support integrated

  • Re-read the Wideband sections of doc/openHPSDR Ethernet Protocol v4.3.docx.
  • Implemented Protocol 2 wideband receive support in:
    • src/DataEngine/cusdr_protocol2_io.h
    • src/DataEngine/cusdr_protocol2_io.cpp
  • Added Protocol 2 wideband socket handling:
    • bind UDP wideband receive socket on base port 1027,
    • receive raw ADC wideband packets,
    • track per-port sequence numbers,
    • reassemble a full wideband block into io->wb_queue for the existing wideband DSP/display pipeline.
  • Updated Protocol 2 General packet wideband fields from the document:
    • bytes 21..22: wideband base port 1027,
    • byte 23: enable bit for ADC0 wideband,
    • bytes 24..25: 512 samples per packet,
    • byte 26: 16-bit samples,
    • byte 27: 20 ms update rate,
    • byte 28: packets per frame from Settings::getWidebandBuffers().
  • Added live handling for widebandDataChanged(...) so Protocol 2 can enable/disable wideband by resending the General packet while running.
  • When wideband is disabled, the partial Protocol 2 wideband accumulator is cleared and the wideband spectrum display is reset.

3. UI wideband button now controls the actual stream

  • Investigated the existing UI and found the Wideband button only showed/hid the dock widget.
  • It did not previously toggle the actual hardware wideband data stream.
  • Updated src/cusdr_mainWidget.cpp so the Wideband button now:
    • sets widebandData,
    • sets widebandStatus,
    • shows/hides the dock.
  • Updated dock visibility handling so manual close/open remains synchronized with both UI state and data-stream state.
  • Also removed the old behavior where the button stayed disabled merely because widebandData was off in the saved settings.

4. Protocol 1 live wideband toggle aligned with UI

  • Updated src/DataEngine/cusdr_dataEngine.h and src/DataEngine/cusdr_dataEngine.cpp so widebandDataChanged(...) now reissues the legacy start command while the data engine is already running:
    • 0x03 with wideband,
    • 0x01 without wideband.
  • Protocol 2 keeps its own separate live wideband handling inside Protocol2DataPath, so the legacy restart command is skipped for Protocol 2 devices.

5. Build re-check

  • Rebuilt after the wideband work:
    • make -j4
    • success

Current status after wideband integration

  • Protocol 2 now has:
    • periodic and change-driven command refresh,
    • integrated wideband receive path,
    • UI-triggered wideband enable/disable behavior.
  • This is compile-verified.
  • Runtime validation on real Protocol 2 hardware is still required for:
    • actual wideband packet cadence,
    • live on/off switching while running,
    • board-specific wideband constraints such as fixed packets-per-frame behavior on some hardware.
Reference in New Issue View Git Blame Copy Permalink