Skip to content

Power and batteries

This page explains who powers whom (USB, the half's small cell, the module packs, Qi), lists every battery as filed and as marketed, gives the ratings, shows how software reads battery voltages and turns them into percentages, and collects the power behavior users and tool authors run into: sleep, resets, flat modules, recovery mode and cables. The one thing to know: a half runs from USB or from its docked module; its own 50 mAh cell only bridges a module swap, so wireless use depends on the modules' batteries.

At a glance

  • Power flows both ways through the dock: USB charges the module through the half; off USB the module powers the half.
  • Cells as filed: half 50 mAh; Tune 1000 mAh; Touch 700 mAh; Track 600 mAh (2 x 300 mAh). Retail capacities are not confirmed.
  • Ratings: each half 5 V DC 1.5 A; module bases 5 V 500 mA; dongle 5 V 30 mA.
  • fe/1006 reads the half's cell in mV; de/100b reads the module's cell in mV; de/1009 reads the module cell and its charging rail in 0.1 mV.
  • NayaFlow's module percentage is (clamp(mV, 3300, 4200) - 3300) * 100 / 900, truncated, then clamped to 1-100 %: 9 mV per point.
  • Replacing a worn cell: specs, fit criteria, safety and where to buy are on Battery replacement.

The power model

A half runs from one of two sources: a computer USB port, or a docked module with enough charge. Its own small cell only keeps it on while a module is swapped and "is not intended to power normal operation"; it charges while the half draws power DOC45. Power therefore flows both ways through the dock: on USB the half feeds and charges the docked module; off USB the module's pack powers the half DOC45 (dock nets MZ_VBAT/VBAT and USB_5V/POGOPIN_5V; see Module dock).

Wireless use depends on the modules' batteries, since the half's cell is too small; the manual's wireless troubleshooting asks to check that modules are docked and charged, and the Naya Connect FAQ states the same model ("modules contain the batteries that power the ecosystem") DOC INFERRED522. Pass-through charging is a core design element: every module has a Qi coil at its base and charges while docked, mainly when the keyboard is wired; module batteries are not designed to be swapped easily DOC19. In 2024 the vendor reported lower runtime consumption, "nearly eliminated" quiescent draw, better module charging efficiency and a larger Tune battery DOC20.

History: in January 2023 the design had no battery in the keyboard at all; the May 2023 campaign still placed all batteries in the modules while its Risks section mentioned the prototype's "backup battery, BMS"; by June 2023 each half had "a hot-swap battery inside" so modules can be swapped without dropping host connections DOC232418 (Reddit quotes: archived text, not live-verified).

Batteries as filed and as marketed

Device Cell label Filed v1.1.0 manual Website and 2023 campaign Evidence
Each half FH301217, 3.7 V, 50 mAh, 0.185 Wh Li-ion polymer pouch 50 mAh "45mA" (v1.0.6 spec sheet) none in the keyboard (early design) DOC CRL IP2 p13-p14; CRR IP1 p10-p1112
Tune FH 202030, 3.7 V 1000 mAh 3.7 Wh 1000 mAh 1500 mAh 1500 mAh DOC1718
Touch FH364046, 3.7 V 700 mAh 2.59 Wh (a second FH364045 700 mAh cell appears in a teardown photo) 700 mAh (one or two cells, open) 1500 mAh 800 mAh DOC1618
Track QS801630 1S2P, 3.7 V 600 mAh 2.22 Wh, two QS801630 300 mAh cells in parallel 600 mAh 700 mAh 800 mAh DOC1818
Float (never shipped) - - - 1500 mAh DOC18

The half's cell is about 3 x 12 x 17 mm by its size code (inferred), taped with its protection board, on a three-wire lead (red, white, black) to a receptacle at VBAT2 beside an NTC pad DOC INFERRED. The test reports declare the half's two power sources as "1# Supplied from battery. Model: 301217, DC 3.7V 50mAh 0.185Wh" and "2# Supplied from PC USB port, DC 5V" DOC BLE reports p10 (text layer verified 2026-09-23)12. The manual's "45mA" is a typo or a minimum rating; the cell and the reports say 50 mAh DOC INFERRED4.

The filed units are pre-production, so retail capacities are not confirmed (details). This differs from naya-create-kb, which gives the Tune's FH 202030 1000 mAh pack to the Touch and the Touch's FH364046 700 mAh cell to the Track1617; check a replacement against the table above. Cell makers are not identified (label prefixes FH and QS only); module retail boxes carry a UN 3480 lithium-ion battery label DOC25. For decoded sizes, connectors, thermistors and replacement cells, see Battery replacement.

Ratings and supply rules

  • Each half's input is rated 5 V DC 1.5 A on its label; the manual's "Input Voltage 4.2V" reads as the cell's full-charge voltage DOC INFERRED24.
  • Module bases are rated Input: 5V⎓500mA (read on the Touch and Track) DOC CRL IP4 p27; CRL IP6 p371.
  • The dongle is bus-powered at DC 5 V (rated 5 V DC 30 mA on its label) and has no cell DOC DG RF report p10; DG LBL3.
  • The manuals' rule: do not plug the keyboard into a wall outlet; charge from a computer USB port ("USB 2.0 5V-1A" in v1.0.6, "USB 3.0 5V-1A" in the v1.1.0 module manuals) DOC46.

Charging a module

A module charges in the dock when the keyboard has power, and on a Qi charger whether docked or not; halves charge from USB, and docked modules charge from the half DOC4. Inside each module the dock 5 V and the Qi receiver's output pass through Schottky diodes into an SGM41523 single-cell charger; the Qi receiver (MT5705) is a 5 W WPC receiver DOC INFERRED1 (the topology is inferred; the schematic is confidential).

Reading battery values

The keyboard never pushes battery values; NayaCore polls each half about every 6 s MEASURED STATIC (NayaCore's own log timing and its set_handshake_frequency example "6000")9. Percentages are computed on the host from millivolts, not sent by the device MEASURED.

Command What it reads Reply (site byte convention) Units Evidence
fe/1006 (NayaCore: GET KB BATTERY LEVEL) the half's own cell; NayaFlow shows it as "Internal Battery Voltage (mV)" 00 <mV hi> <mV lo>, for example 00 0f f5 = 4085 mV mV, big-endian STATIC MEASURED (owner's board, 3.41.0, 2026-09-11)9 (also reported by nayactl12)
de/100b (GET PRECISE BATTERY LEVEL) the docked module's cell 00 <mV hi> <mV lo> <valid> (valid 00 = good) mV, big-endian MEASURED (owner's board, 3.41.0, module 2.3.3, 2026-09-07)1315
de/1009 (GET BATTERY) the module's cell and its charging rail (USB or Qi input) 00 <b0> <batt hi> <batt lo> <rail hi> <rail lo>, for example 00 00 a2 37 c8 f6 = 4.1527 V battery, 5.1446 V rail 0.1 mV, big-endian MEASURED13
  • de/100b is the cell, not the rail. Back to back on the same modules: Track de/100b 4152 mV versus the de/1009 battery field 4142.3 mV, and Tune 4236 mV versus 4257.2 mV; the rail read 4.58-5.14 V on USB MEASURED (owner's board, 3.41.0, module 2.3.3)13. See Transport for the byte convention and Modules for the commands.
  • Module firmware differences. 2.1.2 does not implement de/100b (no payload and about 1.09 s of timeout on every request); de/1009 works on 2.1.2 and 2.3.3 MEASURED (a second board, 3.28.7, module 2.1.2, 2026-09-19). The nayactl maintainer reports that 2.2.2 does not reply to de/100b at all and that de/1009 works there REPORTED15.
  • Units. Feeding the millivolt value into tenth-of-a-millivolt math shows every module at 1 %; tools must use the right unit per command (fixed in nayactl pull request 5) MEASURED15.
  • Readings. Half cells: 4.152 V and 4.066 / 4.094 V (right / left) on the owner's halves, 4.21 V on a half after recharge MEASURED (owner's board, 3.41.0; a second board, 3.28.7). How high a half's cell sits on USB depends on the keyboard firmware: after 30 minutes on USB with the readings flat, a board on 3.28.7 held its halves at 4193 and 4206 mV and a board on 3.41.0 at 4067 and 4091 mV MEASURED (owner's boards, 2026-09-26), so a reading near 4.1 V on 3.41.0 can be a full cell; the newer firmware most likely stops charging earlier INFERRED. The nayactl maintainer reports 4.173, 4.176 and 4.194 V on full halves and 2.771 V on a drained right half that tools show as 1 % REPORTED121415. Modules: full modules 4.12-4.26 V; a Tune on module 2.1.2 read 4000 mV through de/1009 MEASURED13; a module at 1 % read 2.8311 V with its USB rail at 4.8706 V (nayactl thread) REPORTED14.
  • Noise. Read repeatedly over 30 minutes flat on USB, each half's cell reading spread 8-15 mV, under 2 points on the vendor's 9 mV-per-point scale MEASURED (owner's boards, 3.28.7 and 3.41.0, 2026-09-26). The nayactl maintainer saw a full keyboard cell swing between about 90 and 100 %, and tools read it five times and take the median REPORTED STATIC1312. How steady the reading is on battery has not been measured.
  • Host facts. NayaCore tracks per device InternalBatteryVoltage, ModuleBatteryVoltage, ChargingSource (Battery or USB_Qi) and USBVoltage; the protocol also has a category 0xCA named IC_CHARGER with no known subcommands STATIC9. nayactl labels a module's rail "Qi" below 4.5 V and "USB" at or above (a tool heuristic, thresholds unverified) STATIC12.

Rails at a glance on a healthy board on USB: module cell about 4.1-4.26 V full (de/100b, mV; de/1009 on older module firmware); module charging rail about 4.6-5.1 V on USB (de/1009 second field, 0.1 mV); half cell about 4.1-4.2 V (fe/1006, mV), the top depending on the keyboard firmware (see Readings above) MEASURED.

Presence versus voltage

Take presence from the handshake (de/1001), not from a voltage: with both docks empty, both halves answered de/100b with 00 00 00 00, a zero reading that still carries the valid flag MEASURED (owner's board, 3.41.0, 2026-09-01). de/1001 replies 00 01 <addr> for a docked, booted module MEASURED (owner's board, 3.41.0). A docked module that has not booted (at about 0-1 % on pogo power) stays dark and answers as 0xF0/0xF1 with firmware 0.0.0, the same reply an empty dock gives MEASURED (a second board, 3.28.7, 2026-09-19); the nayactl maintainer reports that it boots and lights at about 1 % REPORTED14. See Module dock and Modules.

Percentages

NayaFlow's module percentage is (clamp(mV, 3300, 4200) - 3300) * 100 / 900, truncated, then clamped to 1-100 %, from the de/100b millivolts STATIC (NayaCore 6.11.0, the function feeding connectedModuleBatteryLevel)9. That is 9 mV per point; any valid reading at or below 3309 mV shows 1 %, and an invalid reading shows "?". NayaFlow computes no percentage for the half's own cell; it shows that cell in millivolts STATIC9.

Point Source Reading NayaFlow showed Formula gives
a Tune on module 2.3.3 MEASURED13 4228 mV 100 % 100 %
the same Tune MEASURED INFERRED13 4127 mV 92 % 91 % (the displayed reading was probably at least 4128 mV)

NayaFlow 1.20.0's notes acknowledge a 5-10 point fluctuation in module percentages DOC10. nayactl uses the same linear 3.3-4.2 V scale as NayaCore and floors the result STATIC12. OpenFlow uses NayaCore's formula for modules, so a module reads the same percentage there as in NayaFlow; for the half's own cell, which NayaFlow shows only in millivolts, it uses a lithium-ion resting-voltage curve scaled so that the level the keyboard firmware charges to reads 100 % (see Readings above) STATIC26.

Flat modules and module recovery mode

Long docking can drain a module until its battery protection makes it unchargeable ("Battery Zero"); the vendor answered with firmware and hardware changes and a recovery mode, and the manual advises undocking modules for long storage DOC2157. Off USB a docked module powers its half, and the drain overnight can be large: in the naya-create-kb maintainer's published overnight dumps, off USB, a docked Touch fell from 4184 to 3920 mV and a docked Track from 4166 to 3714 mV (98 to 68 % and 96 to 46 % on NayaFlow's scale) while the half cells held near 4.10 V REPORTED (raw data checked)27.

Module recovery mode and the indicator

Module recovery mode is the key action MODULE_FORCE_CHARGING (NayaCore action 401; key record type 0x06). The v1.1.0 manual puts it on LA5 of the System layer ("Module Force Charge" on Windows, "Module Recovery Mode" on macOS). NayaFlow's tooltip: it recovers a module "from a critically drained battery", the module does not work meanwhile, and a keyboard restart turns it off DOC STATIC MEASURED 59 (stock System layer record 06 with 401 on the owner's board, 3.41.0, 2026-09-08).

Measured behavior: recovery mode acts only on the half whose key is pressed (each half needs its own key), puts that half's LEDs out, force-charges that bay, times out on its own, and a restart clears it; it revived an unresponsive Tune and Touch MEASURED (a second board, 3.28.7, module 2.1.2, 2026-09-19; only this firmware) (details).

Because module firmware 2.3.3 disabled the battery blink, the manual's "blinking red = battery low" may not appear on current module firmware INFERRED10. One module was seen repeating three red breaths and one green blink; its meaning is unknown MEASURED (owner's board, Touch docked right, 2026-09) (details). The indicator table is on Manuals.

Recovery mode

Recovery mode stops the module working until it times out or the keyboard restarts; use it only for a module that does not respond, as the manual says. It writes nothing to the keymap. Tested by us on 3.28.7 only.

Sleep and idle

The manual gives sleep after 1.5 min and deep sleep after 10 min (deep sleep drops Bluetooth); NayaFlow 1.25.1 defaults to 90 s idle and 300 s sleep DOC STATIC459. The device takes three timeouts; NayaCore names them idle_time_ms, sleep_time_ms and sleep_battery_time_ms, and the third stayed 30 s in every one of our captures MEASURED STATIC (owner's board, 3.41.0, 2026-09-11). A board not yet written by NayaFlow held 90 000 / 600 000 / 15 000 ms, which matches the manual's 1.5 and 10 minutes REPORTED (raw data checked)27. The firmware refuses values under 30 s and treats 0 as off MEASURED. Details on Settings and timing.

The LED idle timeout did not run with the cable in and USB output selected, and ran on battery (LEDs off at 90 s, one tap restores); whether the power source or the output mode gates it is not settled MEASURED (owner's board, 3.41.0, 2026-09-11) OPEN. The "LED scan mode" (PWM) setting exists to reduce power use and LED wear; the vendor's 3.31.1 LED PWM control cut consumption at full brightness by about 50 % DOC91011.

Resets

Action A real reset? How to see it Evidence
Power switch OFF, then ON, with USB connected yes: OFF shuts the half off, ON restarts it the half drops off USB MEASURED (owner's board, 2026-09-23); DOC "Create will respect the ON/OFF state even while connected over USB"5
Cold boot: USB out, modules undocked, switches off, then on yes, like any power-on bootloader PID for about 1-1.7 s MEASURED INFERRED
ee/10ae into the bootloader, then SMP os reset on the data port yes; the application is back in about 8 s bootloader PID, then application PID MEASURED (a second board, 3.28.7, 2026-09-19; owner's board, 2026-09-20)
ee/10ce (NORMAL_RESET) yes; the half reboots through MCUboot like every other boot its port drops at once (a transport error 7.7 ms after the command; owner's board, 3.35.4, 2026-09-20); the bootloader PID shows from about 0.7 s to about 2 s and the application is back after about 3 s (left half, 3.41.0, 2026-09-23, OpenFlow's restart probe); from the third restart in a row it needed a cable replug MEASURED (closed question)
One half powers on and re-links yes, for the central: it reboots through the bootloader about 2 s off USB MEASURED (owner's board, 3.41.0, 2026-09-22)

Every boot shows the half's bootloader PID (0x006F left, 0x00D3 right) for about 1-1.7 s (right 0.98 s and left 1.40 s on 3.41.0; 1.6-1.7 s on 3.28.7) MEASURED (2026-09-19, 2026-09-22). ee/10ce goes the same way: the left half left USB at once, showed its bootloader PID from about 0.7-0.8 s to about 2 s after the command, and was back at its application PID after about 3 s. The first two restarts in a row came back by themselves; from the third on the half stayed off USB after the same bootloader pass until its cable was unplugged and replugged MEASURED (owner's board, 3.41.0, Windows 11, 2026-09-23, two runs on different cables and ports). See Bootloader and Recovery.

The half's LEDs can stay lit on its internal cell while it has no data connection to the host, so a lit half is not proof of a working link INFERRED (follows from the power model).

Vendor power fixes by firmware

Firmware Change Evidence
NayaFlow 0.1.0 "Batch 1 / DVT#2" units can charge modules over USB DOC10
keyboard 3.31.1 (announced in beta 1.16.0; the image first ships in beta 1.17.1) LED PWM: about 50 % less at maximum brightness DOC1011
module 2.3.2 Tune LEDs about 20 % less DOC10
keyboard 3.35.4 module battery management about 10 % better DOC10
keyboard 3.39.4 (the beta note says 3.39.3; reached stable with 3.41.0) fixed power cycling on a low internal battery plus a depleted module battery DOC11
keyboard 3.40.4 lower low-battery LED threshold; fixed "Track not keeping the Create awake" DOC11
module 2.3.3 disabled the battery blink because "NayaFlow can read the battery" DOC10

See Firmware versions.

Cables and hubs

The manual's Y-cable carries power and data for both halves to one host port; NayaFlow 1.25.1 warns that Y-cables "are known to cause issues" for updates and pairing and asks for two direct USB-C to USB-C cables DOC STATIC49. A USB hub or dock gave the halves and the dongle power but no data in one session; use a port on the machine for configuration MEASURED (owner's board and dongle, 2026-09-11, single session).

Pre-production runtime claims (2023): "a full work week with LEDs turned up to the max or up to 100 days with LEDs off"; the campaign spec sheet gave 80 days for the Track and Touch and 100 days for the Tune and Float; marketing later claimed a 14-day runtime for the Tune. None is measured DOC241825.

Safety

  • Do not charge from a wall outlet (manuals); the vendor gives no reason, but it is the manual's rule.
  • Never short or bridge the dock contacts: VBAT is the module battery.
  • A drained half cell (the 2.8 V region) is below a Li-ion cell's comfortable range; recharge it promptly INFERRED.

Open questions

  • OPEN Retail module capacities; one or two Touch cells (details, Touch cells).
  • OPEN Whether the power source or the output mode gates the LED idle timer.
  • OPEN What the third timeout field does (always 30 s) (details).
  • OPEN The meaning of the red-red-red-green module blink (details).
  • OPEN The half's power-tree ICs and rail voltages (details).
  • OPEN Recovery mode per side on firmware other than 3.28.7 (details).
  • OPEN How steady the half-cell reading is on battery (on USB it spreads 8-15 mV), and whether the half cells sag overnight only when the modules are undocked (details).

Sources

  1. FCC ID 2BQ4V0825CRL (left half; the module photos are in both filings): exhibits on the FCC record (FCC EAS search, grantee 2BQ4V; mirror fccid.io/2BQ4V0825CRL). Codes are explained on Regulatory records. ↩↩↩↩↩↩↩

  2. FCC ID 2BQ4V0825CRR (right half), including the label artwork; mirror fccid.io/2BQ4V0825CRR. ↩↩↩

  3. FCC ID 2BQ4V0825DG (dongle); mirror fccid.io/2BQ4V0825DG. ↩

  4. Naya Create User Manual Version 1.0.6, FCC ID 2BQ4V0825CRR user manual exhibits (fccid.io); see Manuals. ↩↩↩↩↩↩↩↩

  5. Naya Create User Manual Version 1.1.0 (vendor PDF, 2025-11-10), p4-p5, p11, p20, p23, p25; see Manuals. ↩↩↩↩↩↩↩

  6. Naya Touch User Manual Version 1.1.0 (vendor PDF), p2, p5. ↩↩

  7. Naya Tune User Manual Version 1.1.0 (vendor PDF), p2, p7. ↩↩

  8. Naya Track User Manual Version 1.1.0 (vendor PDF), p2. ↩

  9. NayaFlow 1.25.1 (Windows): NayaCore 6.11.0 strings and disassembly (the module percentage function), the renderer, the settings table and user-interface strings (static reading). ↩↩↩↩↩↩↩↩↩

  10. Vendor release notes, NayaTech/NayaFlow-releases (0.1.0, 1.17.2, 1.19.1, 1.20.0, 1.25.0). ↩↩↩↩↩↩↩↩

  11. Vendor beta release notes, NayaTech/NayaFlow-beta-releases (1.16.0, 1.22.0, 1.24.0). ↩↩↩↩

  12. nayactl, github.com/Qonfused/nayactl (README, cli/status.py, commit 9799866). ↩↩↩↩↩

  13. nayactl, issue #4 (module battery readings from the owner's board on 3.41.0 with modules on 2.3.3, and the maintainer's comments). ↩↩↩↩↩↩↩

  14. nayactl, pull request #2 and its comments (maintainer's board on 3.30.1, modules on 2.2.2). ↩↩↩

  15. nayactl, pull request #5 (the unit fix; maintainer's comments on 2.2.2). ↩↩↩↩

  16. naya-create-kb, hardware deep dive (third party). ↩

  17. naya-create-kb, exhibit inventory (third party). ↩

  18. Kickstarter campaign page with its Specs Sheet and Risks section, naya-create/naya-create (2023). ↩↩↩↩↩↩

  19. Kickstarter update 11, 2024-02-15. ↩

  20. Kickstarter update 15, 2024-09-03. ↩

  21. Kickstarter update 21, 2025-06-16. ↩

  22. Kickstarter Naya Connect FAQ, archived 2026-03-14. ↩

  23. Reddit, vendor comment j34wi1g (2023-01-06); archived text, not live-verified. ↩

  24. Reddit, vendor comment jn3wrmk (2023-06-06); archived text, not live-verified. ↩↩

  25. The vendor's former website, archived by the Wayback Machine (naya.tech captures); cited only, images not reproduced. ↩↩

  26. OpenFlow: its battery percentages (modules on NayaCore's formula; the half's cell on a lithium-ion curve, from the release after 0.5.0). ↩

  27. The naya-create-kb maintainer's published captures and dumps (an overnight pair of device dumps of 2026-09-16 with modules docked, and USB CDC captures of 2026-09-15 before NayaFlow wrote any timeouts); raw data decoded by us, never copied. ↩↩