Skip to content

module_design_notes.md

Latest commit

 

History

History
620 lines (436 loc) · 31.6 KB

module_design_notes.md

File metadata and controls

620 lines (436 loc) · 31.6 KB
Raw
name YieldFlo module hardware design notes
description Sensor approach, flow measurement method, RPM sensor options, FarmTRX comparison
type project

Packet format (current — PGN 40001, 13 bytes)

Bytes Field Type Notes
0–1 PGN 40001 uint16 LE 0x41, 0x9C
2 Packet type byte 0x01
3–4 Sensor 1 count uint16 obstruction ratio × 1000
5–6 Sensor 2 count uint16 second optical sensor
7–8 Moisture raw uint16 ADC/capacitance reading
9–10 Module RPM uint16 elevator shaft speed
11 Status flags byte bits: S1 OK, S2 OK, moisture OK
12 CRC8 byte sum of bytes 0–11

Potential additions: module temperature (moisture compensation), supply voltage, fault code byte.

Flow measurement — optical sensor

An IR TX/RX pair mounts across the clean-grain elevator housing. The beam passes through the elevator throat:

  • Elevator paddles interrupt the beam periodically at a rate proportional to elevator RPM
  • With no grain, only the paddles block the beam — this is the baseline obstruction ratio (~0.2)
  • Grain on the paddles increases blocked duration → higher ratio
  • Teensy counts blocked ticks vs total ticks per window, reports ratio × 1000

App calculation:

netRatio = clamp(sensorRatio - SensorBaseline, 0)
InstantYield = netRatio × YieldFactor × speed / HeaderWidthM

Similarity to RC flow meter pulse counting

A flow meter (RC module) and the optical elevator sensor are structurally identical:

  • Flow meter: pulses_per_second × volume_per_pulse = flow rate
  • Optical sensor: (blocked_fraction − baseline) × YieldFactor = flow rate

Both are a rate-proportional signal multiplied by a calibration constant. The ratio approach is actually better suited to an elevator than pure pulse counting because it naturally handles variable elevator RPM — if the elevator slows, the ratio stays proportional to grain load without requiring RPM compensation.

The existing clsYieldCalculator code is already structured identically to an RC flow accumulator. Only Calculate() and the packet parser would need changes to switch to a pure pulse-count approach.

RPM sensor — optional vs required

Without RPM sensor:

  • The ratio approach still works
  • The YieldFactor calibration implicitly includes "normal" elevator speed
  • Calibration remains valid as long as elevator speed doesn't vary significantly
  • Suitable when elevator is driven directly from combine (speed tracks engine RPM, varies little)
  • Simpler hardware — one fewer sensor

With RPM sensor:

  • Allows normalizing sensor ratio by actual elevator speed: normalizedFlow = netRatio / (moduleRpm / referenceRpm)
  • Makes YieldFactor independent of elevator speed variation
  • Important if elevator can slip, slow under heavy load, or run at different speeds
  • module_rpm field already exists in the packet and is sent by the module
  • Currently unused in clsYieldCalculator.Calculate() — field is received but not applied

Recommendation: Design firmware to support both modes. If an RPM sensor is connected, the module reports actual RPM. If not, it reports a fixed reference value (e.g. 200) so the app can detect the difference and enable/disable RPM normalization via a settings flag.

Moisture measurement — capacitance sensor

Grain flowing past two electrodes changes the dielectric constant → changes capacitance → ADC reading proportional to moisture. The Teensy converts ADC to moisture % using a crop-specific lookup or polynomial.

Limitations:

  • Temperature dependent — module temperature should be logged for compensation
  • Different crops have different dielectric curves — hence moisture offset is per-crop in YieldFlo
  • Needs to be in the grain stream, not measuring air gaps between paddles

Alternative: NIR spectroscopy — more accurate but much more expensive. Capacitance is standard in entry-level monitors (FarmTRX, Ag Leader Insight).

FarmTRX: Whether FarmTRX uses an elevator RPM sensor is unconfirmed — verify against their installation documentation.

Selected optical sensor — Banner T18-2

Datasheet: p/n 201875 Rev. E (D:\Sync\YieldFlo\PhotoSensor\201875.pdf)

The Banner T18-2 Epoxy Encapsulated Right-Angle sensor family is the selected sensor for the elevator optical measurement. Key specs relevant to YieldFlo:

  • Supply: 10–30V DC — powered directly from 12V combine rail
  • Output response: 1.5 ms ON / 1 ms OFF (opposed mode); repeatability 187 µs — adequate for elevator paddle rates
  • Environmental: IP67/IP68/IP69K, –40 °C to +70 °C, ECOLAB certified — suitable for combine environment
  • Connection: 4-pin M12 quick disconnect

Model selection:

  • Emitter: T18-2NAEL-Q8 (visible red, Brown/Blue power only — no signal wire)
  • Receiver NPN: T18-2VNRL-Q8 — complementary sinking outputs
  • Receiver PNP: T18-2VPRL-Q8-809578 — complementary sourcing outputs — selected variant

FarmTrx uses the T18-2VPRL-Q8-809578 (PNP). Adopting the same part ensures alignment with a proven field installation and simplifies sourcing.

M12 pinout:

Pin Wire Function
1 Brown +12V supply
2 White Complementary output (opposite of Pin 4)
3 Blue GND
4 Black Main output

Complementary dual-output noise rejection

The T18-2 receiver provides two complementary outputs (Pin 4 and Pin 2) that are always opposite states. Using both in firmware gives noise rejection without additional hardware:

  • A valid beam transition flips both signals simultaneously
  • A noise glitch on one wire leaves the other unchanged → reject if both pins read the same logic level
  • A noiseCount metric per reporting window flags marginal alignment or EMI during installation

ESP32 ISR logic:

On CHANGE interrupt from either pin:
  read both pins
  if (pinA == pinB)         → noise, discard
  if (newState == currentState) → duplicate interrupt, discard
  else                      → valid edge, accumulate blocked/clear time

The ratio reported each window is blockedTime / (blockedTime + clearTime), accounting for the partial in-progress segment at snapshot time.

RC16 PCB as module base

The RC16 rate controller PCB (D:\Sync\RATE CONTROL\PCB Design\RC16 v9\RC16.kicad_sch) is a strong base for the YieldFlo hardware module.

Key components:

Component Status Notes
ESP32-WROOM-32U Keep Identical MCU
MCP2562 CAN transceiver + 120Ω + ESD Keep Identical to requirements
XL2596 buck + AZ1117 3.3V LDO Keep 12V in → 3.3V, correct rails
CH340C USB-UART Keep Firmware programming
74HC14 Schmitt trigger (×6 gates) Keep Sensor signal conditioning
BSS138 level shifters (Q1, Q2) Keep/repurpose
PC817 optocouplers (U5, U6, U7) Keep U5=RPM, U6=MainSignal, U7=CompSignal
VNH5200 H-bridge motor driver Remove Not needed — biggest saving

Changes required:

  • Remove VNH5200 and associated motor drive passives
  • Swap output connector pinout for M12 sensor connector
  • Rework PC817 U6/U7 input circuits for PNP sourcing sensor (see sensor interface section below)
  • Add 10kΩ pull-downs on MainSignal and CompSignal nets — new components, replaces NPN pull-ups
  • Add moisture sensor header and RPM sensor header

Sensor interface — T18-2VPRL (PNP) requires PC817 input circuit revision:

Verified from YF1 v2 schematic (D:\Sync\YieldFlo\PCB design\YF1 v2\YF1.kicad_sch). The two optical sensor channels are:

  • U6 (MainSignal / T18-2 Pin 4 Black) — series resistor R8 (270Ω), output pull-up R11 (10K)
  • U7 (CompSignal / T18-2 Pin 2 White) — series resistor R9 (270Ω), output pull-up R12 (10K)
  • U5 (RPM) — series resistor R7 (270Ω), output pull-up R10 (10K) — RPM sensor type TBD

Current NPN circuit (U6 as confirmed in schematic):

+5V ──── R8 (270Ω) ──── U6 Anode (pin 1)
                         U6 Cathode (pin 2) ──── MainSignal ──── [NPN sensor → GND]
U6 Collector (pin 4) ──── R11 (10K) ──── 3.3V / MainPin → GPIO
U6 Emitter   (pin 3) ──── GND

Required PNP circuit (U6 after rework):

MainSignal ──── R_PD (10kΩ to GND)   [pull-down, new component]
MainSignal ──── U6 Anode (pin 1)
                U6 Cathode (pin 2) ──── R8 (560Ω) ──── GND
U6 Collector (pin 4) ──── R11 (10K) ──── 3.3V / MainPin → GPIO   [unchanged]
U6 Emitter   (pin 3) ──── GND                                     [unchanged]

Apply identically to U7/R9/R12/CompSignal.

LED current: (12V − 1.2V) / 560Ω = 19 mA — well within PC817 spec, well above ~1 mA threshold.

No firmware changes needed. Output polarity through PC817 → 74HC14 → GPIO is preserved.

Schematic change summary — per optical channel (U6 and U7):

Item Current (NPN) Required (PNP)
R8 / R9 value 270Ω 560Ω
R8 / R9 position +5V side → Anode Cathode → GND
+5V power symbol on input net Present Delete
GND symbol on cathode/R end Absent Add
Sensor signal connects to Cathode (pin 2) Anode (pin 1)
Pull-down on signal net None Add 10kΩ to GND (new R)
R11 / R12 (output pull-up) Unchanged No change
Collector / Emitter wiring Unchanged No change
Firmware — No change

RPM channel (U5/R7/R10): Same changes required if RPM sensor is PNP. Leave unchanged until RPM sensor is selected.

Signal protection — MainSignal and CompSignal (YF1 v2)

Existing protection (confirmed in schematic)

D6 (SMBJ15CA, SMB) — MainSignal to GND D8 (SMBJ15CA, SMB) — CompSignal to GND

LCSC part: C19077570. Bidirectional TVS, 15V standoff, clamps at ~16.7–24.4V, 600W peak pulse. Shunts high-energy transients from the field harness to GND before they reach the PC817 anode. Bidirectional variant handles both positive and negative spikes.

Proposed additional protection — BAT42 diode clamps

Add a BAT42 clamp pair per channel, directly at the PC817 anode input, for tighter clamping within normal operating range:

MainSignal ──┬──[SMBJ15CA]──GND       (high-energy transients, clamps ~16.7V+)
             │
             ├── BAT42: anode→signal, cathode→+12V   (clamp high to ~12.2V)
             ├── BAT42: cathode→signal, anode→GND    (clamp low to ~-0.2V)
             │
             ├── series resistor (100–470Ω, anode side)   [limits clamp current]
             │
             └──── PC817 anode (U6 pin 1)

Apply identically to CompSignal / U7.

Why BAT42 upper clamp (to +12V): Normal PNP sensor output swings to ~12V — clamping at 12.2V does not affect normal operation but limits overvoltage at the PC817 LED, which has a low absolute maximum reverse voltage (6V).

Why BAT42 lower clamp (to GND): Clamps negative transients to ~-0.2V, protecting the PC817 LED from reverse breakdown. The SMBJ15CA handles large reverse events; the BAT42 provides a tight fast clamp for smaller negative spikes.

Series resistor: A small resistor (100–470Ω) on the anode side limits current through the BAT42s when clamping. The existing R8/R9 (560Ω, cathode side) partially serves this role but is on the wrong side of the LED for anode-side protection.

Layer Device Clamps at Handles
1 SMBJ15CA ~16.7–24.4V High-energy harness transients
2 BAT42 upper ~12.2V Overvoltage above normal 12V swing
2 BAT42 lower ~-0.2V Negative spikes / PC817 reverse voltage

YF1 v1 PCB — confirmed design

Schematic: D:\Sync\YieldFlo\PCB design\YF1 v1\YF1.kicad_sch

J1 connector pinout (DTM13-12PA-R008)

Pin Signal Connects to
1 +12V T18-2 emitter Brown
2 +12V T18-2 receiver Brown
3 GND T18-2 emitter Blue / RPM GND / DS18B20 GND
4 GND T18-2 receiver Blue
5 CompSignal T18-2 receiver Pin 2 (White) → PC817 U7
6 MainSignal T18-2 receiver Pin 4 (Black) → PC817 U6
7 CAN_L CAN bus
8 CAN_H CAN bus
9 RPM RPM sensor signal → PC817 (3rd opto, same circuit)
10 Moisture A Electrode A → ADS1115 AIN0
11 Temperature DS18B20 DQ → ESP32 IO17 (1-Wire)
12 Moisture B Electrode B → ADS1115 AIN1 (differential pair with AIN0)

Moisture measurement is differential AIN0 − AIN1. Both electrode wires run in the same harness — common-mode noise cancels in the subtraction. RPM and DS18B20 GND returns share pins 3/4.

Key design decisions confirmed

  • ADS1115 powered at +5V — better analog range; PCA9306 I2C level shifter bridges ESP32 3.3V I2C to 5V ADS1115
  • DS18B20 on ESP32 IO17 — 1-Wire digital, frees all ADS1115 channels for analog; requires 4.7kΩ pull-up to 3.3V on PCB
  • RPM through third PC817 — consistent isolated input, same circuit as MainSignal and CompSignal
  • Moisture on ADS1115 AIN0 — single-ended; AIN1–AIN3 available for expansion
  • Dual 12V pins — supplies T18-2 emitter and receiver independently
  • Triple GND pins — return paths for three isolated sensor circuits

Firmware notes

  • DS18B20 parasitic vs normal power mode must match PCB wiring — normal mode preferred for reliability
  • ADS1115 ADDR pin sets I2C address — confirm net connection (GND=0x48, VDD=0x49, SDA=0x4A, SCL=0x4B)
  • ADS1115 AIN1–AIN3 unallocated — available for supply voltage monitoring or second analog sensor

Moisture sensor — signal conditioning front end (REQUIRED)

The ADS1115 is a DC/low-frequency ADC and cannot directly measure a 30 kHz capacitive signal. A signal conditioning stage must sit between the electrode and the ADS1115.

Excitation and detection circuit

ESP32 LEDC PWM (30 kHz)
        │
        ├──── 74HC4053D analog switch (synchronous demodulation)
        │              │
        │         Electrode A (P1) ──── grain ──── Electrode B (P2)
        │                                                │
        └──── AD8604ARUZ op-amp stage ◄─────────────────┘
                        │
                   DC voltage (1.5–2V span across moisture range)
                        │
                  ADS1115 AIN0/AIN1 differential

Optimal excitation frequency: 30 kHz — confirmed by MDPI 2024 research via frequency sweep 1–100 kHz. Above 40 kHz the output saturates and resolution is lost.

Key components to add to YF1 PCB:

Component Value/Model Purpose
74HC4053D NXP, SOIC-16 3-channel analog switch — synchronous detection
AD8604ARUZ Analog Devices, TSSOP-14 Quad op-amp — amplify and differential output
LM4040 2.5V SOT-23 Precision mid-rail reference for op-amp
C0G reference cap Adjustable 6–10 pF Nulls no-load capacitance, maximises ADC range

Output voltage span: approximately 2268–3793 mV across 12–26% moisture (from research data). Well within ADS1115's 5V-supplied input range.

Electrode design — coplanar PCB sensor

The moisture sensor electrode is a small PCB that mounts flush with the elevator wall at the cleanout hatch, with grain sliding across its face.

Why coplanar, not parallel plate

A parallel plate capacitor sandwiches grain between two opposing plates — impractical to mount in an elevator. A coplanar design places both electrodes on the same surface. The electric field lines arc from P1, out through the grain above, and back to P2. This fringing field is sensitive to the dielectric constant of grain flowing over the face — no opposing plate needed.

        Grain flow over surface
    ────────────────────────────────
    |  P1 (excitation) | gap | P2 (guard/return) |
    └─────────────────────────────────────────────┘
              FR4 substrate (1.6mm)

    Field lines: arc up through grain from P1 → P2

Dimensions (from MDPI 2024)

Feature Dimension
Substrate FR4, 1.6 mm, 2 oz copper
P1 (excitation plate) 24 mm × 56 mm
P2 (outer guard plate) 30 mm × 60 mm
Guard strip gap 0.254 mm (10 mil — standard PCB process)
No-load capacitance ~8 pF

How the guard electrode works

P2 surrounds P1 and is held at a defined potential (ground or driven). This eliminates edge fringing effects at the boundary of P1 — the field at P1's edges terminates on P2, not on surrounding metal. Without the guard, nearby elevator metalwork would cause readings to vary with installation geometry.

Sensitivity and penetration depth

The fringing field penetrates to a depth approximately equal to the electrode pitch. The 24 mm P1 width gives sensitivity to grain a similar distance above the surface — sufficient for elevator flow depths. Grain with higher moisture content has higher dielectric constant (water ε ≈ 80, dry grain ε ≈ 3–5), so even small moisture changes produce measurable capacitance shifts.

Reference capacitor

A 6–10 pF C0G (NP0) adjustable capacitor in the signal conditioning circuit subtracts the no-load (~8 pF) baseline. This means the op-amp stage amplifies only the change due to grain, using the full ADC range for the measurement window rather than wasting range on the fixed baseline.

Calibration

Output is non-linear. MDPI example calibration curve (soybeans):

M = −0.000009U² + 0.064U − 92.665    (U in mV, M in % moisture)

Crop-specific curves required — fits naturally with YieldFlo's existing per-crop settings.

Temperature compensation

correctedMoisture = rawMoisture + k × (sensorTemp − 15.5°C)

k is crop-specific (~−0.18 %/°C for corn). DS18B20 on IO17 provides sensorTemp. 15.5°C (60°F) is the industry reference temperature.

CAN tapping — OEM moisture and yield as supplementary data

The YF1 already has an MCP2562 CAN transceiver. By passively listening on the combine's CAN bus, OEM moisture and yield data can be decoded as a supplementary source — cross-checking or replacing the YF1's own sensors where OEM data is available and reliable.

ISOBUS DDI — standardised data identifiers

ISO 11783-11 defines standard Data Dictionary Identifiers (DDIs) for harvest data. A Task Controller (TC) node broadcasts these in real time:

DDI Description
271 Moisture content (%) × 100
272 Dry mass per time (yield)
273 Dry mass per area
274 Dry mass flow (kg/s)
275 Wet mass flow
276 Volume per area

These are accessible on the ISOBUS (implement bus, 9-pin Deutsch connector) if the combine's terminal implements TC client. Modern John Deere (GS3/GS4), Case IH AFS Pro 700, and New Holland IntelliView 12 all expose some harvest DDIs.

OEM-specific CAN buses

OEM Bus Notes
John Deere Internal machine CAN + ISOBUS ActiveYield data on internal bus (address 117); some data mirrored to ISOBUS TC
Case IH / NH AFS CAN + ISOBUS AFS partly ISOBUS-compliant; moisture and yield on AFS bus
AGCO Internal + ISOBUS Ag Leader integration exposes data on ISOBUS

Implementation approach

  1. Connect YF1 CAN to combine's ISOBUS via 9-pin Deutsch connector (already standard on most combines)
  2. ESP32 passively listens — does not transmit, no risk of interfering with machine bus
  3. Parse incoming CAN frames for known DDIs (271 moisture, 274 mass flow)
  4. Where OEM data is present and valid, use as CANMoisture and CANYield in the packet
  5. Fall back to YF1 sensor data when OEM data is absent or zero

Value

  • OEM moisture sensors are pre-calibrated and crop-aware — better accuracy than a new custom sensor during initial deployment
  • Cross-checking OEM vs YF1 moisture readings identifies calibration drift
  • On John Deere machines with ActiveYield, the OEM mass flow sensor provides an independent yield channel
  • Zero additional hardware cost — MCP2562 is already present

Caution

OEM internal CAN messages use proprietary PGN/CAN ID ranges and require per-model decoding. ISOBUS DDIs are standardised but only exposed if the combine's terminal implements TC. Passive listening is safe; do not transmit on OEM internal buses.

How to apply: Reference these notes when designing firmware, PCB modifications, or expanding clsYieldCalculator to support RPM normalization.

YF1 v4 — J7 connector pinout (DT13-12PA, confirmed 2026-03-24)

Schematic: D:\Sync\YieldFlo\PCB design\YF1 v4\YF1.kicad_sch

J7 is the main vehicle harness connector (TE Connectivity DEUTSCH DT13-12PA, 12-pin, 90° PCB header).

Pin Signal Notes
1 12V Out Switched 12V to sensors (T18-2 emitter/receiver)
2 RPM RPM sensor signal
3 MainSignal T18-2 primary output (Pin 4 / Black wire)
4 CAN H J1939 CAN High
5 Moisture B OEM sensor Moisture Reference (differential pair)
6 Moisture A OEM sensor Moisture Signal
7 Chassis Ground Shield/drain reference (isolated CHASSIS_GND net)
8 Chassis Ground Shield/drain reference (isolated CHASSIS_GND net)
9 MoistureTemp OEM sensor temperature output
10 CAN L J1939 CAN Low
11 CompSignal T18-2 complementary output (Pin 2 / White wire) — noise rejection pair with MainSignal
12 12V In Supply from combine (12V nominal)

Dual Chassis Ground pins (7 & 8) provide adequate current capacity for shield returns. CompSignal (pin 11) and MainSignal (pin 3) are always complementary — firmware discards edges where both read the same logic level.

YF1 v2 — Moisture daughter board (confirmed approach)

Schematic: D:\Sync\YieldFlo\PCB design\YF1 v2\YF1.kicad_sch

Note: Connector/header numbering below reflects v2. In v4 the signals are routed via J7 (DEUTSCH) and internal headers J5/J6. Update this section when v4 daughter board PCB is designed.

The moisture signal conditioning circuit is not integrated on the main YF1 PCB. Instead, the main board exposes a daughter board header. This isolates the unproven moisture front-end from the stable optical/CAN/RPM circuitry and allows the moisture approach to be iterated independently.

Main PCB provides (2×8 header, 2.54mm or Molex KK)

Pin Signal Direction Purpose
1 +12V → DB OEM sensor supply
2 +5V → DB ADS1115 supply
3 +3.3V → DB Logic/op-amp supply
4 GND — Signal return
5 MoistureA ↔ J1-10 passthrough
6 MoistureB ↔ J1-12 passthrough
7 MoistureTemp ↔ J1-11 passthrough
8 SDA ↔ ESP32 I2C bus
9 SCL ↔ ESP32 I2C bus
10 PWM_EXCITE → DB ESP32 LEDC 30 kHz (custom electrode only)
11 DB_INT → ESP32 ADS1115 ALRT/RDY interrupt
12 CHASSIS_GND — Connector shell / shield drain (isolated)

Daughter board variants

Variant A — OEM sensor (simple): ADS1115 + PCA9306 only. OEM sensor outputs conditioned DC — no excitation circuit needed. Pins MoistureA/B route directly to ADS1115 AIN0/AIN1 differential; MoistureTemp to AIN2.

Variant B — Custom coplanar electrode: Full front-end: 74HC4053D + AD8604ARUZ + LM4040 + C0G trim cap + ADS1115 + PCA9306. PWM_EXCITE drives the synchronous demodulator.

ADS1115 channel assignment (both variants)

Channel Custom electrode OEM sensor
AIN0 Electrode A (demodulated) Moisture Signal (OEM Pin 3)
AIN1 Electrode B (reference) Moisture Reference (OEM Pin 4)
AIN2 unallocated Temperature (OEM Pin 5)
AIN3 unallocated unallocated

Read mode is differential AIN0−AIN1 in both variants — firmware read logic unchanged, only calibration curve differs.

OEM moisture sensor — Case IH 2388 pinout and J1 mapping

The 2388 uses a 6-pin connector (Deutsch DT06-6S on harness side). Sensor is self-conditioned — outputs DC analog, no external excitation required.

OEM Pin Function J1 Pin Notes
1 +12V J1-1 or J1-2
2 Signal Ground J1-3 Analog return
3 Moisture Signal J1-10 → ADS1115 AIN0 DC analog voltage
4 Moisture Reference J1-12 → ADS1115 AIN1 Differential pair
5 Moisture Temperature J1-11 → ADS1115 AIN2 Processed analog voltage
6 Shield / Drain Connector shell See shield ground section below

J1-11 is freed from DS18B20 for this use — DS18B20 is PCB-mounted on IO17 and does not need to exit via J1.

Shield ground — KiCad net and PCB topology (YF1 v2)

The shield/drain wire (OEM Pin 6, or any shielded cable shield) connects to the connector shell / backshell, not to a signal pin and not directly to the GND net.

KiCad implementation

  • Define a separate net: CHASSIS_GND
  • J1 shell mounting pads connect to CHASSIS_GND copper pour (isolated island)
  • CHASSIS_GND ties to GND at one point only — at the main power input connector, via a ferrite bead:
J1 shell  →  CHASSIS_GND pour  →  [ferrite bead 300Ω@100MHz, 0603]  →  GND plane
                                    (e.g. Murata BLM21PG300SN1L)

Single tie-point at power entry is the lowest-impedance, most stable reference point. Tying anywhere else risks a ground loop through the combine chassis.

Why not a signal pin

Crimping the shield into a pin cavity and routing it to GND creates a loop: combine chassis → shield → PCB GND → sensor return wire → sensor → combine chassis. This loop couples chassis noise directly into the moisture signal. The backshell/shell approach avoids the loop entirely.

Backshell

Use Deutsch DTM strain relief backshell (DTHD-24-00 or equivalent) — it has an internal provision for a shield drain wire saddle clamp.

FarmTrx signal tap — shared optical sensor (9070 combine, planned 2026-07-04)

The 9070 has a FarmTrx yield monitor installed: module AG-YMK-2500 (Yield Monitor Plus+, legacy ECU, DB15 connector at head unit in cab). Its optical pair is T18-2NAEL-Q8-809577 emitter + T18-2VPRL-Q8-809578 receiver — the same PNP complementary Banner sensor selected for YieldFlo. Plan: run both monitors from the one sensor via a Y-tap at the DB15 in the cab, using the FarmTrx install as a calibrated reference while proving out YieldFlo.

Why parallel sharing is safe

  • YF1 input per channel = PC817 LED + 560Ω ≈ 19 mA, plus 10k pull-down ≈ 1.2 mA — small fraction of the T18-2 output rating (~150 mA, verify datasheet 201875)
  • PC817 opto isolation: only load is LED between signal and sensor GND — no back-path from ESP32 logic into FarmTrx wiring
  • 10k pull-down improves PNP off-state definition for both monitors
  • Both modules in cab share the Y → common ground automatic

DB15 pinout — NOT public

Installation PDFs (media.farmtrx.com) are image-based, no pinout; confirmed unhelpful. Option: email support@farmtrx.com with part no. Otherwise identify pins by measurement:

Continuity trace (chosen method, unpowered both ends):

  1. Unplug M12 at receiver and DB15 at head unit
  2. Jumper M12 harness-side pin 4 (black) → pin 3 (blue/GND); at DB15 find ~0 Ω pin to known GND = Main
  3. Repeat with M12 pin 2 (white) = Comp
  4. Jumper pin 1 (brown) to find sensor 12V feed — if ECU-switched output (not battery feed), YieldFlo must tap the battery side

Cross-check (powered, after Y wired): cardboard beam-block test — Main ≈12V beam clear / ≈0V blocked, Comp always opposite. Firmware reads HIGH on Main = beam clear (BeamBlocked = (digitalRead(MainPin) == LOW)).

Toggle test variant: with everything connected and key on, block beam — exactly two DB15 pins swap 12V↔0V; HIGH-when-clear = Main → J7 pin 3, other = Comp → J7 pin 11. Probe with head unit connected (its load pins the PNP off-state low); unloaded pins float on a 10MΩ DMM.

Key risk — complementary wire may not reach the cab

If FarmTrx only wired black through the harness, white/Comp never reaches the DB15. Firmware requires both signals (Begin.ino attaches interrupts only when MainPin AND CompPin valid; ISR noise rejection needs the pair). Fallback: tap white at sensor M12 with a splitter, or pull one extra wire cab-to-elevator.

Hardware for the tap

  • DB15 male-female breakout board with screw terminals — probe point during identification, then permanent Y-tap (check FarmTrx shell: 2-row DA-15 vs 3-row HD15 before ordering)
  • M12 5-pin female flying-lead cable (A-coded, mates with 4-pin sensor) for the receiver-end breakout during the trace

Reference measurement

Empty-elevator duty cycle on the OEM install = baseline obstruction ratio. Measured 2026-07-12 via the Y-tap: ~6%, identical at idle and full rpm (rpm-independence is expected — the ratio is pure paddle geometry, width ÷ spacing). The earlier ~20% planning estimate was wrong. FarmTrx exposes its own tare for comparison: app → Device Calibration → Sensor Calibration card, "Calibration Results: X% @ Y Hz" (blocked % + paddle frequency; operator's manual p. 20 example shows 5% @ 18.53 Hz — same ballpark). Cross-check: FarmTrx % vs YieldFlo empty flow % from the same run, FarmTrx Hz vs sprocket × chain-speed math. Then compare FarmTrx yield trace vs YieldFlo SensorRatio/NoiseCount during a harvest pass. NoiseCount near zero = clean shared tap.

References

Sensor hardware

OEM moisture sensors

Aftermarket moisture sensors

Academic research

Community