prettycoolboards

Appearance

Architecture ​

Status: rev A in design. The schematics are captured and ERC clean; nothing has been laid out, built or measured. Every current below is a datasheet typical or an estimate, never a measurement.

This page explains how the caliper is put together and why: the system blocks, the choices that were made (and the alternatives that lost), and the power budget. The architecture of record is SPEC; where this page and SPEC disagree, SPEC wins. The circuit details and part numbers are in Schematics and boards.

1. Overview ​

A 150 mm caliper whose slider carries one 62 × 34 mm, 4-layer board. The slider reads its position from an AS5311 magnetic linear encoder over 2 mm-pole magnetic tape glued into the beam, and, as an experiment that can be cut out with jumpers, from an 8-phase capacitive scale (a passive 200 × 7 mm PCB bonded into the beam). An ESP32-S3-MINI-1 runs everything: the reading on a 1.14 in colour IPS display, three buttons, an RGB LED and a piezo, a reading log in flash, BLE (HID keyboard + GATT) and Wi-Fi. A LIS2DW12 accelerometer wakes it on pickup. Power is a 1S 500 mAh LiPo charged from USB-C by a BQ25180, regulated by a TPS63802 buck-boost, and measured by a MAX17048 fuel gauge.

2. Block diagram ​

3. Decisions and the alternatives that lost ​

3.1 MCU and radio: ESP32-S3 vs nRF52840 + nRF7002 ​

ESP32-S3-MINI-1-N4R2 (chosen)nRF52840 module + nRF7002 Wi-Fi companion
RadiosBLE 5 + 2.4 GHz Wi-Fi in one module, PCB antenna, certifiedBLE from the nRF52840, Wi-Fi 6 from a second chip with its own antenna, RF layout and certification
Parts / areaOne 15.4 × 20.5 mm moduleTwo modules (or a module + a QFN with RF matching) on a 62 × 34 mm board that already carries a display, a sensor and a comb
USBNative USB 2.0 FS (CDC + MSC + HID in TinyUSB)Native USB on the nRF52840
Deep sleep7–8 µA (module datasheet v1.7 Table 6-7)~1.5–3 µA System ON idle (nRF)
Active current13–33 mA CPU (modem sleep, 40–80 MHz), 189 mA peak BLE TX at 0 dBm, 355 mA Wi-Fi TX peakMuch lower: nRF BLE TX ~5 mA
Cost (LCSC, qty 1)$5.00~$10 + ~$5 + RF parts

The caliper needs Wi-Fi for the web page and MQTT/HTTP push (SPEC §1), which the nRF52840 cannot do alone. Adding an nRF7002 doubles the RF work on a board whose antenna already has to hang clear of a steel beam. The ESP32-S3 costs more energy while awake, but a caliper is awake for minutes a day and spends the rest in deep sleep, where the difference (~6 µA) is small next to the buck-boost's 11 µA. The S3 also has the PCNT peripheral for the AS5311 quadrature and a dedicated-GPIO bundle for the 8-phase drive. N4R2, not N8R2: the N8R2 is not stocked at LCSC; 4 MB of flash holds the firmware plus a LittleFS log of thousands of 16-byte records.

3.2 Position sensing: AS5311 vs capacitive vs optical ​

AS5311 magnetic (primary)Capacitive 8-phase (experimental)Optical (incremental strip + quadrature sensor)
Resolution0.488 µm/count (4× quadrature), 12-bit absolute within each 2 mm pole pairSet by the ADC/demodulation: ~1/500 of 5.08 mm ≈ 10 µm expected10–25 µm with hobby-grade strips (e.g. 150–180 LPI)
RobustnessInsensitive to coolant, dust, oil, light; 0.3 ± 0.1 mm gapSensitive to moisture and contamination in the 0.15 mm gap, and to the slider's tiltNeeds a clean, sealed path; dust and scratches drop counts
Absolute?Within 2 mm (SSI) plus counting; zero on ZEROWithin 5.08 mmNo
PartsOne IC + magnetic tape ($12.63 IC, tape extra)One op-amp + copper (cheap) but a second PCB in the beam and much firmwareEmitter/detector + a strip + optics alignment
RiskSupply: 80 pcs at LCSC; 16 mA while onUnproven here: it is what commercial calipers use, but their ASICs do the hard partMechanical alignment in a printed housing

The AS5311 is the low-risk route to "electronics are not the limit" (SPEC §1): its resolution is ~20× finer than the 0.01 mm display and it is insensitive to the shop environment. The capacitive scale is built alongside it as an experiment because it is the low-power route (the commercial calipers run for years on a coin cell), and it costs only an op-amp, 18 jumpers/resistors and copper. With R60–R67 and R80 left off it is electrically absent. Optical lost on contamination and alignment in a printed slider.

The AS5311 draws 16 mA (21 mA max), so it is powered only while the caliper is awake (+3V3_SNS). The position is re-established after every wake: the SSI gives the absolute position within the current 2 mm pole, and the firmware keeps the pole count in RTC memory; a slider moved by more than ±1 mm while asleep (wake-on-motion from the IMU starts the sensor) needs re-zeroing. To verify on prototype.

3.3 3V3 regulation: buck-boost vs LDO ​

A 1S LiPo runs from 4.2 V down to ~3.0 V. An LDO (e.g. a 2 µA-Iq RT9080) works while VBAT > 3.3 V + dropout, so it loses regulation for the bottom ~25–35 % of the cell's capacity, exactly when the ESP32-S3's Wi-Fi TX peaks (355 mA, datasheet Table 6-4) pull the cell lower still; the S3 needs 3.0–3.6 V. The TPS63802 buck-boost keeps +3V3 regulated from 1.3 V to 5.5 V in, with 2 A out. The price is quiescent current: 11 µA Iq plus 5.5 µA in its feedback divider (the datasheet limits the lower resistor to 100 kΩ), against ~2 µA for an LDO. That is the largest single line in the sleep budget below, and the reason SPEC's sleep target moved from 25 µA to 35 µA.

3.4 Smaller decisions made during capture ​

  • LIS2DW12 at 0x19 (SA0 high), not 0x18: SA0's internal pull-up cannot be disconnected; grounding it costs 60–170 µA (Schematics §3.6).
  • +3V3_UI switched rail: the ST7789 still draws 15 + 5 µA typ in sleep-in (DM-TFT114-397 §5.3) and a WS2812 idles at hundreds of µA in older revisions, so both lose power in deep sleep. No GPIO is free, so the switch shares SNS_PWR_EN with the sensor rail.
  • WS2812B-2020-V6 (3.3–5.3 V) on 3.3 V: no level shifter and no 5 V rail.
  • Buttons with 100 k pull-ups + 100 nF: no static current, hardware debounce, and they work as ext1 wake sources without keeping the RTC pull-ups powered.

4. Power tree ​

USB-C VBUS ─ SMF5.0A/USBLC6 ─ BQ25180 IN ─┬─ SYS → VSYS ─ TPS63802 → +3V3 ─┬─ ESP32-S3, LIS2DW12, I2C pull-ups, piezo, backlight (FET)
                                          │                                ├─ TPS22917 U5 → +3V3_SNS: AS5311, TLV9062, VREF
                                          │                                └─ TPS22917 U6 → +3V3_UI : ST7789 VDD, WS2812
                                          └─ BAT ─ LiPo (J2) ─ MAX17048 (VDD = CELL)

Charging is 250 mA (≈ 0.5 C for the 500 mAh cell), set over I2C; the BQ25180 power path runs the system from USB while it charges.

5. Power budget and battery life ​

Assumptions: cell 500 mAh nominal, 450 mAh usable (estimate). Battery-side currents at VBAT ≈ 3.7 V; the buck-boost is taken as ~90 % efficient at tens of mA (estimate; TPS63802 efficiency curves), so a 3.3 V load of I draws ≈ I × 3.3 / (3.7 × 0.9) ≈ 1.0 × I from the cell. In deep sleep the converter's Iq dominates its efficiency, so the sleep column lists the quiescent currents directly, plus one line for the conversion loss on the 3.3 V-side sleep loads (ESP32, LIS2DW12, divider: ≈ 14.5 µA at an assumed ~80 % light-load efficiency).

BlockDeep sleep (µA)Idle, awake (mA)Measuring (mA)Basis
ESP32-S3-MINI-182535Deep sleep, RTC peripherals on: 8 µA (Table 6-7). Awake: modem sleep at 80 MHz, 22–36 mA WAITI (Table 6-6) with light-sleep between frames (estimate); BLE connection events add ~1–3 mA average (estimate)
TPS63802 Iq + FB divider16.5––11 µA (SLVSEU9D) + 0.5 V / 91 kΩ
BQ25180 (battery only)3––IQ_BAT 3 µA typ with the watchdog and push-button disabled; 4 µA with the push-button on (SLUSE99C Electrical Characteristics, IQ_BAT). Firmware must set it
MAX1704830.0230.0233 µA hibernate with VRESET.Dis = 1, else 4 µA; 23 µA active (19-6171 Electrical Characteristics, IDD0). Plus ≤ 2 × 0.4 µA SCL/SDA pull-downs (IPD) against the pull-ups
LIS2DW1210.050.05Low-power wake mode ~1 µA (estimate, ST DS11811)
TPS22917 ×2 (off)0.02––ISD,VIN 10 nA typ each (SLVSDW8B; the 0.5 µA figure is the on-state IQ)
Buck-boost conversion loss on the 3.3 V sleep loads≈ 1.7––14.5 µA × 3.3 / (3.7 × ~0.8) − 14.5 (estimate)
AS53110161616 mA typ, 21 mA max (rev 1.13, electrical characteristics); off in sleep
Capacitive AFE01.11.2TLV9062 2 × 538 µA + 16.5 µA divider; TX drive negligible
ST7789 logic088DM-TFT114-397 §5.3; off in sleep (+3V3_UI)
Backlight0510≈ 20 mA at 100 % (15 Ω, VF 3.0 V); 25 % / 50 % PWM (estimate)
WS2812B-2020-V60~1~1Idle current per LCSC listing (to verify); off in sleep
Total≈ 33 µA (≈ 35 µA if the firmware leaves the 4 µA defaults)≈ 56 mA≈ 71 mA

Battery life (450 mAh usable, estimates):

ScenarioAverageRuntime
Deep sleep only (storage, not ship mode)33 µA≈ 13 600 h ≈ 1.5 years (the cell's self-discharge matters more)
Ship mode (BQ25180 BATFET off)~3 µA + pack PCMyears; wake by plugging in USB
Continuous measuring, display on71 mA≈ 6 h
Wi-Fi sync burst (TX ~300 mA peaks, ~100 mA average for a few s)–a few mAh per sync
Typical shop day: 30 min awake measuring, 23.5 h asleep≈ 36 mAh/day≈ 12 days
Light use: 10 min/day≈ 13 mAh/day≈ 5 weeks

What dominates: awake, the ESP32-S3 (~50 %) and the AS5311 (~22 %); asleep, the buck-boost Iq and its divider (50 %) and the ESP32-S3 (24 %). The cheapest rev-B improvement is the auto-off delay (the IMU sees the caliper put down), then a lower-Iq buck-boost (e.g. TPS63900 class, 75 nA Iq, if it can supply the Wi-Fi peaks).

Peak current: Wi-Fi TX 355 mA (802.11b, 20.5 dBm) + AS5311 21 mA + LCD/backlight ~30 mA ≈ 405 mA at 3.3 V, so ≈ 405 × 3.3 / (3.0 × 0.9) ≈ 500 mA from a 3.0 V cell (≈ 350 mA from a full 4.2 V one). That is inside the TPS63802's 2 A rating, but it is about 1 C for a 500 mAh cell (for bursts only) and half of the BQ25180's default 1 A battery discharge limit (IBAT_OCP; the firmware must not lower it to 500 mA). The bulk 22 µF (C6) plus 22 µF at the module ride through the TX bursts.

6. References ​

Hardware design files, docs and generators for prettycoolboards.