Blinky Hawk Firmware Manual

Three states, always live

The firmware loops continuously and classifies the test leads into one of three states:

While charging over USB the normal alerts are replaced by a slow dim red blink (green once the battery reads full), and at power-on the LED blinks green 1–4 times to show the battery level in quarters.

LED & speaker behavior

LED states

Speaker

• Continuity: a short beep the moment contact is made, then (by default) a repeat beep every second for as long as the circuit stays closed (CONTREP/CONTREPMS).
• Voltage: a double beep, once per detection (VOLTPULSES/VOLTREP). Voltage is a priority alert — it sounds immediately even if a continuity beep just fired.
• Both alerts run at the same pitch. The fitted buzzer is only usefully loud near 4 kHz, so that is where both are set; the two alerts are told apart by their pulse pattern (one beep vs two), not by tone. CONTFREQ and VOLTFREQ can still be set independently, but moving either far from 4 kHz mostly just makes that alert quieter.
• Rate cap: no new beep sequence can start within BEEPMIN ms of the previous one, so an intermittent connection can't machine-gun the beeper.
• Boot mute: if the leads are shorted together while the unit powers up, audio stays muted until the next power cycle. Handy for quiet environments; disable the feature entirely with !SET,BOOTMUTE,0.
• Kill it permanently: !SET,BEEP,0 then !SAVE.

The low-power sleep mode

Running flat out, Blinky Hawk draws about 11 mA. Asleep it draws about 2 mA — and because it can still detect while asleep, sleeping is the normal operating state rather than a standby mode. The unit spends almost all of its life there and wakes fully only when it finds something.

What "asleep" actually does

After SLEEPSEC seconds in which the leads have read nothing but OPEN, the firmware parks every load it can switch — the LED rail, the speaker, the battery-sense divider, and optionally the bridge MOSFET — and puts the RA4M1 into Software Standby: CPU and peripheral clocks stopped, RAM retained. A real-time-clock interrupt wakes it every SLEEPTICKMS. Every SLEEPTICKS wakes it runs one cheap probe, and goes straight back to sleep unless the probe finds continuity or voltage — in which case the unit returns to full-rate operation and alerts normally.

Worst-case detection latency is therefore SLEEPTICKMS × SLEEPTICKS.

• The probe is a real detection pass, using the selected DETMETHOD and the live threshold — it is only cheaper, averaging SLEEPAVG readings instead of VOLTAVG and skipping the display debounce.
• Voltage detection is deliberately blunter while asleep. It uses only the wide instant-bypass band (VOLTFAST × REFBAND), never the tight averaged window, so lead noise cannot wake the unit on every tick.
• Plugging in USB wakes it, as does any serial traffic, and it never sleeps while charging or in diagnostic mode. SLEEPSEC,0 disables sleeping entirely.
• A 6 ms dim-blue flash every SLEEPHB ticks says "asleep, not dead". Set SLEEPHB to 0 to suppress it — each flash costs real current.
• It will not sleep in the first 30 seconds after power-on. That grace period is a safety net: sleeping breaks the USB link, so it guarantees a window in which a host can always reach a freshly-flashed unit.

Wake period: SLEEPTICKMS

The wake period is produced by the RTC's periodic interrupt, which only offers a fixed ladder of rates. Whatever you set is snapped to the nearest available rung and the snapped value is written back, so !CFG always reports the rate actually in force:

2000100050025012563311684 ms

The rungs below 125 ms are rounded: 63 is really 62.5 ms (1/16 s) and 4 is really 3.90625 ms (1/256 s). Only the reported latency arithmetic uses the rounded number; the interrupt itself runs at the exact rate.

Wake thresholds

A probe taken out of standby measures a few percent high — the analog path has not fully settled and the rails have just come back up. If a threshold tuned on the bench sits below the sleeping unit's closed-lead reading, continuity will silently fail to wake the unit. The SLEEPTHR00…SLEEPTHR11 keys give each slot its own wake threshold for exactly this case; 0 means "use the ordinary THRESH value for that slot", which is the default.

Commands

Talking to the unit

Connect over USB and open the port at 115200 baud. Any serial terminal works (Arduino Serial Monitor, PuTTY, a Python script). The protocol is line-based text:

• Commands in start with ! and end with a newline. Arguments are comma-separated. Commands are case-insensitive.
• Structured data out starts with $ and is comma-separated — easy for a program to parse. Human-readable status lines are printed alongside.

Configuration commands

Unit serial number

Each unit can carry a serial number — a short identity string (up to 15 characters) written once, usually at first bring-up, and read back on every boot. It is stored in its own EEPROM block, separate from the configuration, so it survives !DEFAULTS, a !SAVE of new settings, and even a firmware update that changes the config layout. Assign one with !SN,<value>; read it with a bare !SN. The current serial number is also reported at the end of every $STATUS line as sn=<value> (empty when unassigned).

Status & diagnostic commands

Low-power commands

Described in full in section 5.

Data lines out

How EEPROM storage works

• Two copies exist: the RAM config (what the unit is using right now, edited by !SET) and the EEPROM config (what it will boot with). !SAVE copies RAM → EEPROM; !LOAD copies EEPROM → RAM.
• Unsaved changes are flagged: the $STATUS line reports dirty=1 when the RAM config differs from what was last saved/loaded.
• Validation: the stored block carries a magic number, a version, and a CRC. If any check fails at boot (first boot, corrupted data, or a firmware update that changed the config layout) the unit falls back to factory defaults and re-seeds the EEPROM automatically — it can never boot with garbage settings.
• Write endurance: the config is only written on an explicit !SAVE (or that first-boot seed). The firmware never writes EEPROM on its own during normal operation, so data-flash wear is not a practical concern.
• Factory reset: !DEFAULTS then !SAVE.
• The serial number is separate: it lives in its own EEPROM block with its own magic/CRC, so it is not affected by !DEFAULTS, !SAVE, or a config-layout change. It is written only by !SN,<value>. See section 6.

Common configuration recipes

Permanently disable the beeper

!SET,BEEP,0
$CFG,BEEP,0
!SAVE
$OK,save

Make threshold slot 00 trigger on very high resistances

!SET,THRESH00,0.02      // smaller threshold = more sensitive
$CFG,THRESH00,0.0200
!SET,THRESHSEL,0        // and make that slot the active one
$CFG,THRESHSEL,0
$DIP,0,0.0200
!SAVE
$OK,save

Shift the resting center after a hardware change

!SET,REFCENTER,-0.005
$CFG,REFCENTER,-0.0050
!SET,REFBAND,0.03
$CFG,REFBAND,0.0300
!SAVE
$OK,save

Quieter, slower continuity reminder

!SET,CONTREPMS,5000     // remind every 5 s instead of every 1 s
!SET,CONTHOLDMS,20      // and shorten the reminder pulse
!SAVE

Detuning CONTFREQ away from 4 kHz also makes it quieter, but it fights the buzzer's resonance rather than working with it — shortening the pulse or slowing the repeat gives a cleaner result.

Check what a unit is currently configured to

!CFG
$CFG,REFCENTER,-0.0100
$CFG,REFBAND,0.0250
… (one row per key) …
$CFGEND

Undo experiments you haven't saved

!LOAD
$OK,load            // RAM restored to the last-saved EEPROM state

Switch to the time-to-return detection method

!SET,DETMETHOD,1        // metric is now recovery time in ms
$CFG,DETMETHOD,1
!SET,THRESH11,0.9       // re-tune the active slot in the NEW units (ms)
$CFG,THRESH11,0.9000
!SAVE
$OK,save

Set up the shipping low-power configuration

!SET,SLEEPSEC,1         // sleep a second after the leads go open
$CFG,SLEEPSEC,1
!SET,SLEEPTICKMS,63     // wake ~16x/sec (snaps to the 1/16 s RTC rung)
$CFG,SLEEPTICKMS,63
!SET,SLEEPTICKS,1       // probe on every wake
!SET,SLEEPHB,0          // no heartbeat flash (it counts TICKS, not seconds)
!SAVE
$OK,save

Find out why a sleeping unit won't wake on continuity

!SLEEPTEST              // leads shorted: what does the sleeping probe see?
$SLEEPTEST,CLOSED,metric=0.3820,thr=0.4500,awakethr=0.4500,…
// If that says CLOSED but a real sleeping unit stays asleep, the
// difference is the battery ground reference, not the probe logic.
// Unplug, exercise the leads, plug back in, and look at the real thing:
!SLEEPLOG
$SLOGSTART,64,412,method=0
$SLOG,0,SLEEP,FLOAT,0.5100,0.4500,…
$SLOG,1,AWAKE,CLOSED,0.3800,0.4500,…
…
$SLOGEND
// SLEEP rows reading high against thr = raise the wake threshold for
// that slot, leaving the awake threshold alone:
!SET,SLEEPTHR10,0.55
!SAVE

Assign a serial number to a new unit

!SN                     // check first — is one already set?
$SN,                    // empty = unassigned
!SN,BH-2026-001
$SN,BH-2026-001
$OK,sn                 // written to EEPROM immediately; no !SAVE needed

Hardware revision 2

One revision 2 unit exists. If you are not holding it, skip this appendix — everything else in this manual describes the V3 board that all current units use. The firmware supports both from a single build, switched by the HWREV configuration key.

What is different

Reading the DIP switches

Two of the four switch positions (on pins D8 and D10) select which stored threshold is active. Each pin has a pull-up and the switch shorts it to ground, so a switch that is ON reads 0 and one that is OFF reads 1. The position code is the two pin readings, first digit = D8, second digit = D10 — which is where the THRESH00…THRESH11 key names come from:

The remaining two switch positions are wired into the analog bridge rather than to the microcontroller: they bypass resistors to set the permanent series resistance. The pairs are silkscreened on the PCB and are moved together — the switches the firmware reads tell it which resistance the other pair just selected.