Skip to content

TaskList.md

Latest commit

 

History

History
556 lines (487 loc) · 33.2 KB

TaskList.md

File metadata and controls

556 lines (487 loc) · 33.2 KB
Raw

TODO

Priorities:

  • P0: Correctness or data-integrity issue; address before the next release.
  • P1: Reliability, lifecycle, or API-contract issue; address in the next development cycle.
  • P2: Improvement or cleanup with lower immediate risk.

Implement Error Fixes

P0 - Correctness and Data Integrity

  • Preserve RX callback data lifetime. Invoke onDataReceived() before clearing the NimBLE characteristic value, or copy the received value into stable storage before passing it to user code.
  • Make flush() safe in PumpMode::Task. Do not run pumpTx() concurrently from the caller and the TX worker; wake the worker and wait for the queue/in-flight frame to drain within the existing timeout.
  • Correct RX overflow accounting. Calculate how many existing bytes will be overwritten before calling RingBuffer::push(..., true), update rxDrops, and call onRxOverflow() with the actual number of discarded bytes.

P1 - Reliability and API Contracts

  • Make TX task creation failure switch pumpMode to Polling so update() can continue transmitting. Keep the warning consistent with the resulting state.
  • Correct the MTU API contract. setPreferredMTU() sets the preferred local MTU; a server cannot initiate renegotiation for an existing connection.
  • Keep preferred and negotiated MTU state separate. getMtu() reports the negotiated value, while getPreferredMTU() reports the local preference.
  • Add centralized begin() failure cleanup. Clear the custom GAP handler and active instance, deinitialize partial NimBLE state, release callbacks/resources, and leave the object reusable after any failed initialization step.
  • Own the dynamically allocated RX and TX characteristic callbacks and release them during end() and failed initialization. NimBLE-Arduino does not delete characteristic callback objects.
  • Avoid overriding the application's global logger level in begin(). A level set through logSetLevel() by the sketch or another library must remain effective; document that setLogLevel() controls the shared logger.
  • Record lastTxUs at the actual task-mode transmission attempt, after any pacing delay, so consecutive notifications respect sendIntervalUs.

P2 - Improvements and Cleanup

  • Revisit two-PDU transmission in Mode::Fast when NimBLE-Arduino supports it. Re-enable the two-PDU limit in computeTxChunkSize() and keep ATT/L2CAP header, MIC, computeSendIntervalUs(), and PDUS_PER_WINDOW calculations consistent.
  • Remove or replace the deprecated no-op NimBLEService::start() call.
  • Resolve warning-enabled build findings: unused variables and deprecated increment operations on volatile-qualified fields. Use explicit synchronization rather than treating volatile as thread safety.
  • Review lifecycle resets so end() and clearStats() consistently reset all documented counters, retry state, pending state, locks, and buffers.

Test and Validate the Library

P0 - Regression Tests for Correctness Fixes

  • Verify onDataReceived() receives byte-for-byte correct data after the characteristic value is cleared, including maximum-size writes.
  • Fill the RX ring, force overwrite, and verify retained data ordering, rxDrops, and onRxOverflow(lost) exactly match the discarded byte count.
  • In task mode, write continuously while calling flush() and verify there are no duplicate, missing, reordered, or corrupted notifications.
  • Force TX task creation failure and verify the library enters polling mode and continues transmitting when update() is called.

P1 - Build and Automated Validation

  • Compile every example for esp32:esp32:nano_nora with explicit local UUtzinger_RingBuffer and UUtzinger_logger dependencies and distinct build paths.
  • Compile with warnings enabled and treat new library warnings as regressions.
  • Compile at least one example with DEBUG defined and one without it.
  • Verify all logger levels (DEBUG, INFO, WARN, ERROR, NONE) and confirm application-level logSetLevel() and logSetOutput() survive begin().
  • Add host-testable coverage for chunk sizing, pacing calculations, watermarks, retry transitions, timeout rollover, and lifecycle state resets.
  • Add repeated begin()/end() and failed-begin() tests, checking that the object can be reused and that callbacks/tasks/resources do not leak.
  • Validate preferred versus negotiated MTU behavior before connection, during a connection, after peer negotiation, and after reconnect.

P1 - Hardware and Fault-Injection Validation

  • Exercise Polling and Task pump modes in Fast, Balanced, LowPower, and LongRange modes with sustained bidirectional traffic.
  • Validate Security::None, JustWorks, and PasskeyDisplay, including AlwaysOpen, Windowed, and BondedOnly pairing policies.
  • Force EMSGSIZE with a negotiated MTU of 23 and confirm txChunkSize decreases or clamps to MIN_CHUNKSIZE, timing is recomputed, probing stops, and cooldown is set.
  • Force isolated and sustained congestion, then confirm bounded retries, cooldown, controlled last-known-good escalation, recovery toward the pacing floor, and eventual disconnect behavior for non-recoverable failures.
  • Disconnect during queued and in-flight TX. Confirm txState returns to Waiting, pendingLen is cleared, locks are reconciled, and advertising restarts correctly.
  • Measure notification spacing in task mode and confirm no interval is shorter than the configured/computed pacing floor after scheduler delays.
  • Run long-duration throughput and reconnect tests while monitoring heap usage, task handles, callback allocation, RX/TX drops, and error counters.

Diagnose BLE Text Truncation

Problem Statement

Large terminal-style output over BLE can be truncated or misaligned when a sketch prints long help menus or logs through Print::print() / Print::println(). This was observed with MAX30001G_BLESerial: a help-menu line can be followed immediately by Mode changed to... without the expected line ending.

The original library-level failure was that BLESerial::write() could return a partial count or 0 when txLocked was set or the TX buffer was congested. Arduino Print callers did not retry those bytes. Reliable bounded writes now address that path.

Current test evidence:

  • USB baseline is complete: END_OF_TEST bytes=21509 lines=160 checksum=64C22A3F and USB_WRITE_SUMMARY requested=21562 short=0.
  • After the reliable-write fix, the independent Android BLE Serial terminal reaches END_OF_TEST with short=0; the original mixed-capture interleaving is therefore most likely in the desktop terminal capture path.

P0 - Reproduce and Isolate

  • Add a deterministic BLE text-stress example or test command that emits at least 8 KB, exceeding the current 4 KB TX buffer.
  • Include numbered lines, fixed-width fields, explicit line endings, total byte count, line count, checksum, and a final sentinel such as END_OF_TEST bytes=<n> checksum=<hex>.
  • Send the identical text over USB Serial as a baseline to prove the source text and formatting are complete before BLE transport.
  • Capture BLE output with the normal terminal client.
  • Compare USB and BLE captures for byte count, line count, checksum, sentinel, missing line endings, and reordered or merged lines.
  • Capture with a second independent BLE UART client after the library fix if the normal terminal still shows anomalies.

P0 - Targeted Library Fix

  • Fix println(const char*) so it returns the actual byte count and does not report +2 unless both CR and LF were accepted.
  • Confirm println() never appends CR/LF after a partial payload write unless the remaining payload bytes have been accepted or the bounded write timeout has expired.
  • Confirm flush() drains queued bytes and pending in-flight notifications in both task and polling modes, remains bounded, and returns after disconnect.

P1 - Reliability Fix

  • Make default Print::write() behavior reliable and bounded for terminal/log output, likely by routing write(uint8_t) and write(const uint8_t*, size_t) through writeTimeout() with a default timeout.
  • Update writeTimeout() to wake/pump while waiting for TX space and to return only the bytes actually queued before timeout or disconnect.
  • Ensure short writes increment TX drop/timeout diagnostics consistently instead of silently losing bytes.
  • Add configurable timeout accessors, such as setWriteTimeoutMs(uint32_t) and getWriteTimeoutMs(), or reuse an existing timeout setting if one is appropriate.
  • Preserve an explicit non-blocking API for high-rate streaming, such as writeNonBlocking(...), tryWrite(...), or documented writeReady() usage.
  • Document the difference between reliable bounded Print writes and the explicit non-blocking streaming path.

Omitted

  • Broad TX enqueue tracing is omitted because the stress test and BLE stats already identify library-level TX drops.
  • Extra diagnostic counters are omitted for the first fix because existing stats already report TxDrops, congestion, and software errors.
  • MAX30001G-specific reproduction is omitted before the library fix; the deterministic BLESerial stress test isolates the same failure mode without sketch-specific buffering or command routing.

P1 - Verification After Fix

  • Repeat the deterministic BLE text-stress capture and verify byte count, line count, checksum, and final sentinel match USB.
  • Verify MAX30001G help-menu output is complete over BLE and is not truncated before follow-up command responses.
  • Verify high-rate streams still work when using the explicit non-blocking API.

P1 - Compile Validation

  • Compile all BLESerial examples for esp32:esp32:nano_nora with explicit local UUtzinger_RingBuffer and UUtzinger_logger dependencies.
  • Compile MAX30001G_BLESerial against the local BLESerial, UUtzinger_logger, UUtzinger_RingBuffer, and UUtzinger_MAX30001G checkouts.
  • Confirm no LOG_LEVEL_* redefined warnings return.

Acceptance Criteria

  • Large BLE text output preserves line endings, table text, byte count, checksum, and final sentinel.
  • Print::print() and Print::println() over BLE are reliable for normal terminal/menu/log use.
  • Non-blocking high-rate streaming remains available through an explicit API.
  • Diagnostics distinguish delayed writes, timeout writes, dropped bytes, high-water locks, notify failures, and flush timeouts.
  • MAX30001G_BLESerial help menus and command responses are complete over BLE.

Improve Link Adaptation Reliability

Observed comprehensive-test evidence:

  • The example reached an enqueue rate of 732 fixed 36-byte lines/s, or 26,352 bytes/s. This is accepted-to-buffer throughput, not confirmed over-the-air throughput; use the Bytes TX delta for delivered throughput.
  • Isolated raw RSSI excursions (-108, -54, -105 dBm) drove repeated 2M/1M changes despite the filtered average.
  • Repeated ENOMEM responses escalated the last-known-good pacing interval from about 10.4 ms to more than 40 ms while the TX buffer still had free space.

Post-fix hardware evidence:

  • Fast mode negotiated 2M PHY, MTU 517, and a 7.5 ms connection interval. The comprehensive stream sustained an enqueue rate of 1,751 fixed 36-byte lines/s.
  • Stats reported 13,849,675 transmitted bytes with TxDrops=0, Timeouts=0, and final send=min=lkg=1968 us despite 21,398 congestion status events. Persistent congestion produced bounded backoff and later recovered to the pacing floor.

P1 - Implementation

  • Require sustained RSSI evidence before changing PHY. Use consecutive filtered-RSSI decisions so one raw RSSI excursion cannot trigger a 2M/1M or coded-S2/S8 transition.
  • Apply separate upgrade and downgrade confirmation criteria and a post-change holdoff. Clear pending decisions on reconnect, mode change, disconnect, and a failed or externally initiated PHY update.
  • Keep mode boundaries explicit: Fast may adapt only between 2M and 1M, LongRange only between coded S=2 and S=8, and Balanced/LowPower remain on 1M.
  • Treat isolated ENOMEM responses as transient congestion. Return to the exact confirmed last-known-good interval, using the connection-event share only as a hard fallback when no confirmed interval exists.
  • Escalate the last-known-good interval only when congestion persists across a defined observation window or coincides with repeated high-water TX-buffer pressure. Require one second for general persistence or at least eight high-water events over 250 ms, and cap congestion backoff at four times the computed pacing floor.
  • Define recovery after congestion: require a stable success window, probe in bounded steps, and return to the mode/link pacing floor after a confirmed PHY, MTU, DLE, or connection-interval improvement.
  • Distinguish the negotiated DLE maximum TX time from estimated packet airtime in status and log output so values such as 17,040 us are not reported as actual 2M packet airtime.
  • Rate-limit repeated congestion warnings while retaining counters and state transition logs needed to diagnose backoff and recovery.

P1 - Hardware Verification

  • Measure sustained throughput before and after the adaptation changes using the Bytes TX delta; report enqueue throughput separately.
  • Exercise Fast, Balanced, LowPower, and LongRange modes when supported.
  • Verify Fast changes only between 2M and 1M, ignores isolated RSSI outliers, does not oscillate near a threshold, and recovers its pacing floor after a sustained RSSI improvement.
  • Verify LongRange remains coded and changes between S=2 and S=8 without constraining DLE to the initial coding scheme or reacting to isolated outliers.
  • Verify Balanced and LowPower remain on 1M with their requested connection intervals, latency, MTU, and power settings.
  • At a stable PHY and connection interval, observe intermittent and persistent ENOMEM; verify isolated events hold the confirmed LKG, persistent congestion produces controlled escalation, and successful traffic recovers to the floor.

Prevent Recursive BLESerial Diagnostics

Problem Statement

When MAX30001G_BLESerial routes the shared logger to ble, BLESerial's own TX adaptation messages can be written back into the same TX queue that is currently being drained. In the malformed ? help capture, an internal Probe 1948 accepted message appears inside a help row, two complete rows are missing, and the following row is joined without its expected CRLF.

The capture strongly suggests self-directed BLESerial diagnostics, but does not yet prove that recursion is the byte-loss mechanism. SerialUI normally retains partial BLE lines until EOL and has a 64 KiB input buffer. Manually stopping its receiver is a separate case because unsubscribing can lose upstream output.

P0 - Reproduce and Confirm

  • Record ble.printStats(Serial) immediately before and after each test and compare TxDrops, Timeouts, congestion events, and software errors.
  • At the default INFO level (l 3), issue ? repeatedly and capture the raw bytes, including pauses, injected BLESerial diagnostics, missing rows, merged rows, total byte count, and final CRLF.
  • Repeat the same sequence at l 2. This suppresses BLESerial INFO diagnostics while leaving the sketch's unconditional LOGln() help output enabled. The compact command documented by the sketch as l2 currently sets level 0 because its parser reads the value starting at character three.
  • Repeat with an independent BLE UART client to separate BLESerial behavior from SerialUI rendering and notification handling.
  • Confirm whether each reported stop was a passive output pause or a manual SerialUI receiver stop; do not combine those cases in the result.
  • Add a deterministic BLESerial regression case that routes logSetOutput(ble), triggers TX adaptation diagnostics while sending a payload larger than the TX ring, and validates byte count, checksum, line endings, and final sentinel.

Confirmed P0 Evidence (2026-08-01)

  • A raw Linux BlueZ/D-Bus notification capture reproduced the same missing rows and merged CRLFs without SerialUI and without manually stopping reception.
  • A complete help response is 6043 bytes with 79 CRLFs and SHA-256 2008900b170658850fe422b9ee968ac3ec20ba1ed5ab78ed57b47fb0f583d1b5.
  • At INFO, three stats-correlated responses changed Bytes TX by 5711, 6092, and 5877 bytes. The first added 431 TxDrops and 20 timeouts; the second added no drops but transmitted 49 bytes beyond the help payload; the third added 216 drops and 10 timeouts.
  • At WARN, the response changed Bytes TX by exactly 6043 while TxDrops remained 647 and Timeouts remained 30.
  • These controls confirm the loss occurs in the BLESerial/logger path, not in SerialUI. Suppressing INFO is a valid temporary workaround, but the library still needs to prevent any level of internal diagnostic from targeting its own BLE transport.
  • The BLESerial_text_stress selflog command reproduces the recursive path deterministically with a 21,605-byte payload. Before the fix it added 84 TxDrops and 6 timeouts. With dedicated diagnostics it sent the same 21,605 bytes with zero drops/timeouts; its transient ENOMEM warning appeared only on USB Serial. The BlueZ capture contained the final checksum sentinel and BLE_WRITE_SUMMARY ... short=0.
  • With dedicated diagnostics, five MAX30001G ? responses at INFO increased Bytes TX by exactly 6043 bytes each (30215 total). TxDrops=0 and Timeouts=0 remained unchanged. Probe start and acceptance messages appeared on USB Serial but did not change the BLE payload count. A persistent BlueZ subscriber received the help notifications from beginning through the final CRLF.

P0 - Dedicated Internal Diagnostics

  • Add one library-wide diagnostic helper that writes to a dedicated Print destination, defaulting to Serial, without reading or changing the application's global logger destination.
  • Apply the dedicated route to all BLESerial diagnostics, including GAP/characteristic callbacks, TX/RSSI tasks, probe success/start messages, congestion recovery, timeout handling, and disconnect/error paths; fixing only the observed probe message is insufficient.
  • Keep application writes through Print, LOG(), and LOGln() unchanged. Global logSetOutput(ble) remains valid for application output.
  • Allow Serial1, Serial2, and application-provided Print implementations as BLESerial diagnostic destinations, but reject the BLESerial object itself without changing the existing destination.
  • Remove the translation-unit #undef LOG* wrappers and route every BLESerial-owned diagnostic explicitly through the dedicated helper.
  • Avoid temporarily swapping the shared logger output because callbacks and application tasks may log concurrently.
  • Ensure dedicated diagnostics do not recurse, block the TX worker, mutate TX queue contents, increment application TxDrops/Timeouts, or alter link adaptation state.
  • Document the dedicated-routing behavior and new diagnostic API in API.md, README.md, and CHANGELOG.md.

P1 - Automated and Compile Validation

  • Add focused coverage for global application logging to BLE while BLESerial diagnostics remain on their dedicated non-BLE Print; verify self-selection is rejected.
  • Force an ERROR path and verify it cannot recursively fill the queue during timeout or disconnect handling. WARN routing was hardware-verified with a transient ENOMEM event during selflog.
  • Re-run the deterministic text-stress test and all examples in polling and task modes, with DEBUG both enabled and disabled.
  • Compile all BLESerial examples for esp32:esp32:nano_nora with warnings enabled and explicit local logger and RingBuffer dependencies.
  • Compile MAX30001G_BLESerial against the local BLESerial, UUtzinger_logger, UUtzinger_RingBuffer, and UUtzinger_MAX30001G checkouts; confirm no new warnings or LOG_LEVEL_* redefinitions.
  • Publish UUtzinger_logger 2.2.1 before BLESerial 1.3.2. The BLESerial metadata already declares UUtzinger_logger (>=2.2.1).

P0 - Hardware Acceptance Criteria

  • Repeated ? output is byte-for-byte complete over BLE at l 3: no missing Er/En rows, merged Eq row, injected BLESerial diagnostic, or missing CRLF.
  • The same capture passes in SerialUI and an independent BLE UART client.
  • TxDrops and Timeouts do not increase during a completed help response.
  • BLESerial diagnostics still reach a separately configured non-BLE sink, while self-directed diagnostic configuration is rejected.
  • High-rate non-blocking streaming, adaptation, pacing, and congestion recovery behavior remain unchanged.

Implementation Order

  1. Run the l 3/l 2 control and preserve before-fix stats and raw captures.
  2. Introduce the centralized self-routing guard and migrate every internal log call to it.
  3. Add regression coverage for self-directed and separate logger outputs.
  4. Run warning-enabled BLESerial and MAX30001G builds plus offline stress tests.
  5. Repeat the hardware captures and close the acceptance items only from byte-count/checksum/stat evidence.

Evidence Capture

The first command completed without a visible pause. The second paused twice and shows both missing content/CRLF and an injected BLESerial diagnostic:

[INFO] BLESerial: Client e0:d4:64:23:f6:1c subscribed (notify) 6e400003-b5a3-f393-e0a9-e50e24dcca9e.
================================================================================
| MAX30001G ECG and Bio-Impedance Program                                      |
| 2026 Urs Utzinger & GPT                                                      |
================================================================================
| GENERAL COMMANDS                       | DATA COMMANDS                       |
|----------------------------------------|-------------------------------------|
| ?: help screen                         | z: toggle data reporting on/off     |
| s: show current settings               | z[s|b|a]: USB, BLE, both, off       |
| h: run health check                    | z[S|B|A]: USB, BLE, both, on        |
| i: print device info                   | c: reset sample counter             |
| r: print all registers                 | f: FIFO reset                       |
| t: print status registers              | p: print config registers           |
================================================================================
| PERSISTENCE                            | RAM SNAPSHOT                        |
|----------------------------------------|-------------------------------------|
| Ps: save NVS preferences               | (: save volatile register snapshot  |
| Pl: load NVS preferences               | ): restore volatile register snap   |
| Pd: print NVS preferences              |                                     |
| Pc: clear NVS preferences              |                                     |
|========================================|=====================================|
| OPERATION MODES (auto-stop previous)   | START/STOP                          |
|----------------------------------------|-------------------------------------|
| m1: ECG mode                           | .: toggle start/stop                |
| m2: BIOZ mode                          | >: start measurement                |
| m3: ECG + BIOZ mode                    | <: stop measurement                 |
| m4: ECG signal calibration             |                                     |
| m5: BIOZ signal calibration            |                                     |
| m6: BIOZ internal impedance            |                                     |
| m7: BIOZ external impedance            |                                     |
| m8: BIOZ impedance spectroscopy        |                                     |
|========================================|=====================================|
| ECG SETTINGS                           | BIOZ SETTINGS                       |
|----------------------------------------|-------------------------------------|
| Es<n>: speed      (0-2)     Es1        | Bs<n>: speed      (0-1)     Bs0     |
| Eg<n>: gain       (0-3)     Eg2        | Bg<n>: gain       (0-3)     Bg1     |
| El<n>: dig LPF    (0-3,255) El255      | Ba<n>: analog HPF (0-7)     Ba1     |
| Eh<n>: dig HPF    (0-1,255) Eh255      | Bd<n>: digital LPF(0-3)     Bd1     |
| Ee<n>: leads      (2 or 3)  Ee3        | Bh<n>: digital HPF(0-3)     Bh0     |
| Er<n>: R-to-R     (0=off,1) Er1        |                                     |
| En<n>: notch  (0=off,50,60) En0        | Bf<n>: frequency Hz         Bf8000  |
| Eq<n>: notch Q    (1-100)   Eq20       | Bc<n>: current nA           Bc8000  |
|                                        | Bp<n>: phase deg            Bp0     |
|                                        | Bl<n>: lead bias  (0=off,1) Bl1     |
|                                        | Bo<n>: lead-off   (0=off,1) Bo0     |
|                                        | Bw<n>: wires      (2 or 4)  Bw2     |
|========================================|=====================================|
| BIOZ SCAN SETTINGS                     | INTERNAL CALIBRATION SETTINGS       |
|----------------------------------------|-------------------------------------|
| Sa<n>: averages   (1-8)     Sa8        | Cr<n>: internal resistor    Cr1000  |
| Sf<n>: fast mode  (0=off,1) Sf0        | Cm<n>: cal modulation(0-3)  Cm0     |
| Sr<n>: full range (0=off,1) Sr0        | Cf<n>: mod frequency(0-4)   Cf3     |
| Si<n>: source     (0=ext,1=int) Si0    | Ce<n>: ECG sig mode(0/1)    Ce1     |
| Sp<n>: phase rng  (0=full,1) Sp0       | Cb<n>: BIOZ sig mode(0/1)   Cb0     |
| Sh<n>: AHPF mode  (0=dyn,1=fix,2=byp)  |-------------------------------------|
| Sv<n>: fixed AHPF (0-7)     Sv1        |                                     |
| Sx<n>: int AHPF   (255,0-7) Sx255      |                                     |
| St<n>: settle     (1-64)    St24       | Sm<n>: max current nA       Sm96000 |
| Sc<n>: cur settle (1-64)    Sc24       | So<n>: autorange  (0=off,1) So1     |
|                                        | Sy<n>: human-safe (0/1)     Sy0     |
|----------------------------------------|-------------------------------------|
| Kp:    print scan calibration          | Kr:  reset scan calibration default |
| Kg<n>: set global K ppm (1250650)      | Ke<n>: enable correction (0/1)      |
|========================================|=====================================|
| LOG LEVEL                              | SPECIAL                             |
|----------------------------------------|-------------------------------------|
| l0: none (silent)                      | w: software reset                   |
| l1: errors only                        | y: synchronize                      |
| l2: warnings                           | k: clear latched status flags       |
| l3: info (default)                     | a: apply current settings (re-setup)|
| l4: debug (verbose)                    |                                     |
================================================================================

Examples:
  m1       - Switch to ECG mode
  Eg3      - Set ECG gain to 160 V/V (level 3)
  Bf40000  - Set BIOZ frequency to 40 kHz
  .        - Start/stop measurement
  z        - Toggle continuous data display

================================================================================
| MAX30001G ECG and Bio-Impedance Program                                      |
| 2026 Urs Utzinger & GPT                                                      |
================================================================================
| GENERAL COMMANDS                       | DATA COMMANDS                       |
|----------------------------------------|-------------------------------------|
| ?: help screen                         | z: toggle data reporting on/off     |
| s: show current settings               | z[s|b|a]: USB, BLE, both, off       |
| h: run health check                    | z[S|B|A]: USB, BLE, both, on        |
| i: print device info                   | c: reset sample counter             |
| r: print all registers                 | f: FIFO reset                       |
| t: print status registers              | p: print config registers           |
================================================================================
| PERSISTENCE                            | RAM SNAPSHOT                        |
|----------------------------------------|-------------------------------------|
| Ps: save NVS preferences               | (: save volatile register snapshot  |
| Pl: load NVS preferences               | ): restore volatile register snap   |
| Pd: print NVS preferences              |                                     |
| Pc: clear NVS preferences              |                                     |
|========================================|=====================================|
| OPERATION MODES (auto-stop previous)   | START/STOP                          |
|----------------------------------------|-------------------------------------|
| m1: ECG mode                           | .: toggle start/stop                |
| m2: BIOZ mode                          | >: start measurement                |
| m3: ECG + BIOZ mode                    | <: stop measurement                 |
| m4: ECG signal calibration             |                                     |
| m5: BIOZ signal calibration            |                                     |
| m6: BIOZ internal impedance            |                                     |
| m7: BIOZ external impedance            |                                     |
| m8: BIOZ impedance spectroscopy        |                                     |
|========================================|=====================================|
| ECG SETTINGS                           | BIOZ SETTINGS                       |
|----------------------------------------|-------------------------------------|
| Es<n>: speed      (0-2)     Es1        | Bs<n>: speed      (0-1)     Bs0     |
| Eg<n>: gain       (0-3)     Eg2        | Bg<n>: gain       (0-3)     Bg1     |
| El<n>: dig LPF    (0-3,255) El255      | Ba<n>: analog HPF (0-7)     Ba1     |
| Eh<n>: dig HPF    (0-1,255) Eh255      | Bd<n>: digital LPF(0-3)     Bd1     |
| Ee<n>: leads      (2 or 3)  Ee3        | Bh<n>: digital HPF(0-3)     Bh0     || Eq<n>: notch Q    (1-100)   Eq20       | Bc<n>: current nA           Bc8000  |
|                                        | Bp<n>: phase deg            Bp0     |
|                                        | Bl<n>: lead bias  (0=off,1) Bl1     |
|                                        | Bo<n>: lead-off   (0=off,1) Bo0     |
|                                        | Bw<n>: wires      (2 or 4)  Bw2     |
|========================================|=====================================|
| BIOZ SCAN SETTINGS                     | INTERNAL CALIBRATION SETTINGS       |
|----------------------------------------|-------------------------------------|
| Sa<n>: averages   (1-8)     Sa8        | Cr<n>: internal resistor    Cr1000  |
| Sf<n>: fast mode  (0=off,1) Sf0        | Cm<n>: cal modulation(0-3)  Cm0     |
| Sr<n>: full range (0=off,1) Sr0        | Cf<n>: mod frequency(0-4)   Cf3     |
| Si<n>: source     (0=ext,1=int) Si0    | Ce<n>: ECG sig mode(0/1)    Ce1     |
| Sp<n>: phase rng  (0=full,1) Sp0       | Cb<n>: BIOZ sig mode(0/1)   Cb0     |
| Sh<n>: AHPF mode  (0=dyn,1=fix,2=byp)  |-------------------------------------|
| Sv<n>: fixed AHPF (0-7)     Sv1        |                                     |
| Sx<n>: int AHPF   (255,0-7) Sx255      |                                     |
| St<n>: settle     (1-64)    St24       | Sm<n>: max current nA       Sm96000 |
| Sc<n>: cur settle (1-64)    Sc24       | So<n>: autorange  (0=off,1) So1     |
|                                        | Sy<n>: human-safe (0/1)     Sy0     |
|----------------------------------------|-------------------------------------|
| Kp:    print scan calibration          | Kr:  reset scan calibration default |
| Kg<n>: set global K ppm (1250650)      | Ke<n>: enable correction (0/1)      |
|========================================|=====================================|
| LOG LEVEL                              | SPECIAL                             |
|----------------------------------------|-------------------------------------|
| l0: none (silent)                      | w: software reset                   |
| l1: errors only                        | y: synchronize                      |[INFO] BLESerial: Probe 1948 accepted. LKG=1948.

| l2: warnings                           | k: clear latched status flags       |
| l3: info (default)                     | a: apply current settings (re-setup)|
| l4: debug (verbose)                    |                                     |
================================================================================

Examples:
  m1       - Switch to ECG mode
  Eg3      - Set ECG gain to 160 V/V (level 3)
  Bf40000  - Set BIOZ frequency to 40 kHz
  .        - Start/stop measurement
  z        - Toggle continuous data display