Extends the firmware's 0xFD vendor feature report from 32 to 44 bytes
to carry the three host -> dongle -> BT trigger counters added in
5da1be4 (host_out02_total / host_out02_trig_allow / host_out02_to_bt).
mic_diag.sh's bt-trace decoder reads them and prints a one-line
verdict so the user can triage issue #3 in one terminal command
instead of switching to the OLED Diagnostics screen mid-game.
Verdicts encoded:
host02 > 0, trig == 0 -> host driver never sets the trigger
Allow bits; not a firmware problem.
trig > 0, tx < trig -> trigger Allow bits arrive but the
speaker-active gate in main.cpp
swallowed them.
full chain populated -> dongle delivered, controller didn't
actuate; Sony BT-side limit.
The 0xFD report ID stays undeclared in the HID descriptor — Linux
hidraw ioctls don't enforce that and WebHID never sees these reports
(config goes through 0xF6). bt-trace's IOCTL_SIZE bumped to 45
(1 byte for the kernel-prepended report ID + 44 bytes payload).
Docs:
README.md - new "Diagnostics & debug tooling" section walks
through the four mic_diag.sh subcommands; previously
the script was only mentioned inside
BLUETOOTH_AUDIO_NOTES.md.
README.md - Diagnostics-screen description updated for the
scrolling row-list + cross-link to bt-trace.
CLAUDE.md - "Where features live" gets a host-side-diagnostics
entry explaining how to extend the 0xFD payload and
wire it into bt-trace's decoder.
CHANGELOG - records both the firmware extension and the README
discoverability fix.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Issue #3 (Few things — Triggers and Screen) reports that adaptive
trigger tension is missing through the dongle. The send_trigger_effect
preset cycle on the OLED Trigger Test screen works fine, so the BT
forwarding path is sound — meaning the loss is either at the host
(driver / Steam Input / game not emitting trigger output) or in our
forwarding of the host's 0x02 output report. Adds counters so we can
tell which.
Counters in main.cpp's tud_hid_set_report_cb 0x02 handler:
- host_out02_total — every 0x02 OUT report seen
- host_out02_trig_allow — those where the host set
AllowRight/LeftTriggerFFB
(valid_flag0 bits 2 & 3)
- host_out02_to_bt — 0x02 reports forwarded as BT 0x31
Diagnosis matrix:
total grows, trig_allow flat → host driver never sets the trigger
Allow bits = host-side issue
trig_allow grows, to_bt lags → we're dropping (audio path stealing
the BT pipe while speaker is active)
all three grow proportionally → forwarding works; controller is
receiving but not actuating
= Sony / DS5 firmware limit
Same flash gets a scroll-only Diagnostics screen — the old layout
was already at the 6-row ceiling and three new counters push it over.
Rows are now a single switch-formatted list (10 rows), 5 visible at
a time, D-pad up/down scrolls. Caret glyphs at the right edge mark
"more above/below". Read-only screen so no cursor (unlike Settings).
Order is by triage relevance: Uptime, BT state, host02, trig/tx,
BT31 in/s fit on the default view; audio + parked-mic counters are
below the fold.
flush_fb split earlier (flush_fb_raw + flush_fb) is unaffected.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces the single-tier auto-dim (5 min, contrast 0x10) with a three-tier
ladder that actually protects the OLED:
- 0–2 min idle: full brightness per bright_idx (unchanged).
- 2 min: clear the panel and render a 2×2 "breathing" dot (1 s on / 1 s
off) that walks through 8 evenly-spaced screen positions every ~30 s.
Drops contrast to 0x01. The dot is the indicator that the dongle is
alive; rotating its position prevents any single pixel from
accumulating wear.
- 15 min: SH1107 display-off (cmd 0xAE). Panel fully dark.
Wake on KEY0/KEY1, controller pair (BT-connect rising edge), or any
controller input report change.
Why this shape: bench-tested 0x10 and 0x02 contrast values — both
looked nearly identical to full brightness because the SH1107 contrast
register has a heavily non-linear perceptual curve on this Waveshare
panel. The only reliable per-pixel "dim" available is to render fewer
pixels. The breathing dot lights ~4 of 8192 pixels half the time,
which is roughly a 1000× drop in cumulative current vs the bright
Status screen.
flush_fb split into flush_fb_raw + flush_fb so the dim-tier renderer
can push the framebuffer without the chrome arrows (no navigation
target while sleeping).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
In-progress work on DualSense microphone capture over BT. Mic-add tap
itself is disabled (was decoding standard input bytes as Opus and
producing INT16_MIN garbage on the USB IN endpoint) but everything
around it is wired and ready to re-enable once we identify the actual
mic transport.
Firmware:
- src/audio.cpp: Opus decoder on core0, mic_fifo queue, audio_loop
mic-in path with decode + mono->stereo + tud_audio_write. Decoder
init in audio_init() (creates 48kHz mono OpusDecoder).
- src/audio.h: exports mic_add_queue() + per-frame diagnostic
accessors (audio_mic_frames, last_decoded, last_want, last_wrote,
last_toc).
- src/main.cpp on_bt_data(): BT-side instrumentation — counts every
INTERRUPT input report, tracks min/max length, OR mask of byte[2],
most recent non-0x31 report ID, hex prefix of last 0x31/other/any
frame, full content of the longest 0x31 frame seen. Mic-tap call
itself stubbed behind `if (false)` pending the real detector.
- src/state_mgr.cpp: state_init_data byte 6 (VolumeMic) 0xFF→0x40
(was out of range), byte 9 (MuteControl) 0x0F→0x00 (clear all
PowerSave bits — AudioPowerSave was muting DSP).
- src/cmd.cpp: two new vendor feature reports — 0xFD returns 32-byte
diagnostic state (counters + prefixes), 0xFE returns the longest
0x31 frame in full (up to 80 bytes). Both queryable via
/dev/hidraw on Linux from the host script.
- src/oled.cpp: Diagnostics screen shows TOC + decode result + USB
wrote/want bytes for live BT-side visibility.
Host-side:
- scripts/mic_diag.sh: subcommands `status`, `capture [secs]`,
`watch`, `bt-trace`. The bt-trace subcommand reads the 0xFD
feature report via hidraw ioctl, decodes counters + recent
prefixes, computes per-second rates. Drastically cuts iteration
time — no OLED relay or per-test flash cycle needed.
Findings to date:
- Upstream/mic's mic-flag bit ((data[2] >> 1) & 1) does NOT match
this DS5 firmware; bit 1 of byte[2] is NEVER set. Bit 0 is the
standard input report type indicator, not a mic tag — confirmed
by stick-bytes appearing as our supposed "Opus prefix".
- DS5 sends both report ID 0x01 and 0x31 over BT; the longest frame
is a standard 79-byte 0x31 input report with sticks/IMU/touchpad
but no audio bytes appended.
- Conclusion in progress: the DS5 firmware on this controller is
not currently streaming mic over BT at all, even with
AllowAudioControl=1, VolumeMic=0x40, AudioPowerSave=0,
MicMute=0. Next investigation step: compare against a USB-mode
DS5 to see what a real mic stream looks like at the UAC1 layer.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Audit revealed the Status screen hardcoded "v0.6.0" while v0.6.1 and
v0.6.2 had already shipped — the version string was scattered and
fell behind every release. Centralizing:
- CMakeLists.txt: target_compile_definitions adds
FIRMWARE_VERSION="${VERSION}" so the existing -DVERSION=$tag flag
passed by release.yml propagates into C++ as a compile-time macro.
Local builds without -DVERSION get "dev" — a deliberate "this is
not a release build" tell on the Status header.
- src/oled.cpp render_screen() now renders
"DS5 Bridge " FIRMWARE_VERSION via string-literal concatenation. The
only place a version literal exists in source is now the macro at
the link layer.
- CLAUDE.md gets a "Versioning — single source of truth" section
documenting the release ritual (CHANGELOG bump → tag → push → gh
release create → CI builds + propagates) and naming the
WEB_REPO_DISPATCH_PAT follow-up that's still pending.
Web preview is updated separately to read firmware-latest.json at
runtime — same single source of truth, different distribution path
(the GitHub API).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Refactor the Pico-OLED-1.3 button model and visual layout:
- KEY0 single press = "next screen" (double-click reboot path removed —
it was easy to fire by accident while paging quickly).
- KEY1 short press = "previous screen" on every screen. The old
contextual cycle on Trigger Test (cycle trigger preset) and Lightbar
(cycle lb_mode) moves to controller buttons:
* △ rising edge on Trigger Test → cycle trigger preset
(re-applies via send_trigger_effect)
* R1 rising edge on Lightbar → cycle lb_mode
△ on Lightbar stays as "save current RGB to favorite slot 0" (existing
UX), so R1 is used there instead.
- KEY1 long press = cycle brightness — unchanged.
- KEY0 + KEY1 held simultaneously for >= 1 s → watchdog_reboot. Pre-empts
per-key handlers so a chord cancels any armed single press. DS5
PS+Mute hold-2 s remains the headless backup.
Visual layout:
- New draw_button_chrome() paints '>' at (0, 8) and '<' at (0, 49) on
every frame from flush_fb(), pairing the on-screen labels with the
physical KEY0/KEY1 buttons on the left edge of the device.
- Every render_screen_*() shifts main content right by kContentX (6 px)
to clear the chrome strip. Stick boxes, L2 column, L1 bar, and D-pad
center on the Status screen also shift right by 6 to avoid the '<'
glyph painting inside the live left-stick area.
- "K0=next K1=back" footers removed from all screens (the chrome strip
is now the affordance). Trigger Test footer = "Tri=cycle"; Lightbar
footer = "R1=mode"; Slots/Settings keep their contextual hints.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
New CPU/Clock OLED screen (kScreenCpu), between Diagnostics and BT
Signal in the K0 cycle. Shows:
- Set - configured system clock (SYS_CLOCK_KHZ overclock target)
- Real - clk_sys from the RP2350 frequency counter, measured once on
screen entry and cached (it is fixed at boot; avoids a per-frame
multi-ms stall)
- Vcore - core voltage read back via vreg_get_voltage()
- Temp - on-die sensor via cpu_temp_raw_smoothed()
render_screen_cpu() is noinline like the other render functions (Thumb
literal-pool reach). Hardware-in-the-loop verified on Pico 2 W +
Waveshare Pico-OLED-1.3.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Add cpu_temp_raw_smoothed() (cmd.cpp/cmd.h): 256-sample average over a
slow EMA (alpha=0.15) so the RP2350 on-die temperature converges to the
true value instead of chasing per-sample 12-bit ADC noise (~1 LSB =
0.47 C). Single source of truth, ADC initialised in one place.
Also export CPU/Clock telemetry on HID feature report 0xfc (set_khz,
cached real_khz, vreg code, ADC ch4 raw). The handler is dormant: live
read over WebHID is not feasible (Chrome requires the report declared
in the HID descriptor, and declaring the OLED Edition vendor reports
breaks DualSense enumeration on Windows). It is harmless and left for a
possible future non-WebHID consumer. Link hardware_adc.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The low-battery blink would stay frozen mid-cycle (or briefly resume
during reconnect retries) after a controller died and lost connection,
because the stale-report check stopped toggling but never turned the
LED off, and the cached interrupt_in_data[52] still read low.
New `battery_led_on_disconnect()` clears blink state, forces the LED
off, and zeros last_report_us so the stale-check early-return blocks
any new blink until a fresh 0x31 report arrives on the next connection.
Called from bt.cpp's HCI_EVENT_DISCONNECTION_COMPLETE handler.
Stale-check in the tick also now forces LED off when it fires while a
blink was in progress (defense in depth for unclean disconnects).
Same bug exists in upstream awalol/DS5Dongle (their battery_led.cpp is
byte-identical to ours pre-fix) — Sura Academy reported it in their
Discord. Plan to send this back as a PR after we validate it locally.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Build now compiles. audio.cpp on top of upstream's state_mgr-based
audio_loop adds three additive pieces:
1. VU peak meters for ch0/1 (spk_max) and ch2/3 (hap_max) tracked
per-sample in the existing loop; written back to g_peak_spk /
g_peak_hap after. Accessors decay 12.5% per read so the OLED
VU bars fall back naturally over a few frames.
2. Monotonic byte-flow counters: g_usb_frames++ after tud_audio_read
(the OLED Diagnostics screen + web emulator compute USB-aud/s
and BT-0x32/s as delta over time). g_bt_packets++ after each
0x36 bt_write.
3. Audio Auto Haptics DSP — 1-pole LP + envelope follower +
modulation + soft-clip per loteran/DS5Dongle 5d6bc2f, with our
added 4th "Fallback" mode that fires only after the game's
native haptic path has been silent for ~1 s (NATIVE_SILENT_TIMEOUT
= 100 audio_loop calls). Preserves native HD haptics (Spider-Man
Remastered) while filling in for games that send no haptic data
(Ghost of Tsushima). Modes: 0=Off, 1=Fallback (default), 2=Mix,
3=Replace.
Crucially the state_mgr-based state_set(pkt+13, 63) call is left
untouched — that's upstream's correct replacement for our prior
state_data restore hack, and it's load-bearing for speaker output.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
bt.h: add OLED accessors (bt_is_connected, bt_get_addr, bt_hci_err_count)
and 6 multi-slot accessors (bt_get_slot, bt_set_slot, bt_forget_slot,
bt_wipe_all_slots, bt_slot_occupied, bt_slot_get_addr).
bt.cpp:
- #include "slots.h"; add g_current_slot + update_discoverable()
helper that gates gap_discoverable_control on slots_any_empty().
- 3 OLED accessor impls + 6 slot accessor impls.
- bt_init: slots_load() + g_current_slot from Config_body.current_slot,
then update_discoverable() instead of bare gap_discoverable_control(1).
- HCI_EVENT_INQUIRY_RESULT: slot-ownership filter (skip devices owned
by other slots; if our slot is occupied, only accept matching addr).
- L2CAP_EVENT_CHANNEL_OPENED HID_CONTROL: slot_assign on empty slot
(Phase G auto-bond pattern).
- L2CAP_EVENT_CHANNEL_OPENED HID_INTERRUPT: replace upstream's
gap_*_control(false) pair with update_discoverable() — they wanted
"non-discoverable once connected"; we want "non-discoverable once
all 4 slots are full". Strictly looser, which is correct since
empty slots need to accept new pairings.
- Inquiry-complete + disconnection-complete handlers similarly
swapped to update_discoverable().
Upstream's state_mgr hooks (state_init / state_set / state_update
calls in L2CAP_EVENT_CHANNEL_OPENED and elsewhere) are preserved
as-is — they're the upstream rumble-fix infrastructure.
Build still fails on oled.cpp references to audio_peak_* / counter
accessors that Phase E adds.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three additions on top of upstream's v0.6.0-hotfix main.cpp:
- #include "oled.h" + oled_init() after audio_init/state_init,
oled_loop() in the main while(1).
- PS + Mute hold-2-seconds combo at the top of interrupt_loop()
for headless soft-reboot via watchdog_reboot(0,0,0). Bit pattern
(interrupt_in_data[9] & 0x05) is rare in normal gameplay so the
long hold prevents accidental triggers.
Upstream's state_init() call sits right between audio_init() and our
new oled_init() — preserved as-is. The state_mgr → 0x31 output report
flow in tud_hid_set_report_cb is upstream's rumble path; not touched.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Build-system, config, usb, and CI wiring on top of upstream's
v0.6.0-hotfix tree. Build will still fail until Phase C+D land bt.h
accessors that oled.cpp references; that's expected mid-rebase.
CMakeLists.txt: add src/oled.cpp + src/slots.cpp to add_executable,
link hardware_spi, restore OUTPUT_NAME ds5-bridge-oled.
src/config.h: append the 4 OLED Edition Config_body fields
(current_slot + auto_haptics_enable/gain/lowpass). Preserves
upstream's field order for the first 8 fields.
src/config.cpp:
- speaker_volume validity default -100 dB → 0 dB (footgun fix)
- 4 new validity guards for the appended fields with sensible
defaults (Fallback / 100 % / 160 Hz / slot 0)
- config_default() impl (was declared in upstream's config.h since
v0.5.4 but never defined; the OLED Edition's "Reset to defaults"
Settings item calls it)
src/usb.cpp: drop the UAC1 SET_CUR Volume → flash-config sync that
let PipeWire/Pulse silently override the user's saved Spk Vol on
every device reconnect. Borrowed from loteran/DS5Dongle commit
03fa1e4.
src/usb_descriptors.cpp: restore iSerialNumber = STRID_SERIAL.
Upstream's commit e79c762 zeroed it to work around SpecialK (#32),
but that broke Windows device identity (#100) — users lost
per-device volume / app preferences when moving USB ports. The
OLED Edition prioritizes Windows device-identity stability for
the broader user base over SpecialK compat for a narrower one.
.github/workflows/{build,release}.yml: artifact filenames use
ds5-bridge-oled.uf2 (and -debug / -picow variants) per the
OUTPUT_NAME rebrand.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Upstream only checks left stick X axis, so pressing buttons, using
the d-pad, or moving the right stick won't prevent disconnection.
Extends the same dead zone pattern to cover both sticks (all 4 axes),
both button bytes, and d-pad/misc bytes.
The low-battery indicator is critical info — keep blinking even when
the user has the LED disabled (Speaker mute toggle). Steady-state
LED still respects the flag: when battery recovers or charging
starts, the LED returns to off if the flag is set, otherwise to the
existing connected = solid-on state.
(cherry picked from commit 775346cd29d9a3c566309ec7d29db80f0b6979f1)
Blink the Pico onboard LED at 1 Hz when the connected DualSense
reports PowerPercent <= 1 (i.e. <= 10%) and PowerState == Discharging.
Source data is byte 52 of the BT 0x31 input report, already copied
into interrupt_in_data; no new BT parsing required.
The new module owns the LED only while it is actively blinking; it
detects controller disconnection via stale-report timeout and steps
out, so bt.cpp's existing connect/disconnect LED handling stays in
charge in all other states. Honors disable_pico_led.
Gated by -DENABLE_BATT_LED=ON (default). With the option off, the
source file is not compiled and behavior is identical to upstream.
CI gains a compile-only check for the OFF flavor.
(cherry picked from commit 2f8ea73c9fb695e24e7cc3329db7cb925e82e1c9)
When the inactive_time configuration value exceeds 35, the expression
`get_config().inactive_time * 60 * 1000 * 1000`
overflows the maximum int value due to implicit integer conversion.
