Pressure CAN node: six pressure sensors and eight inputs on the CAN bus

A board that reads six pressure sensors (four 10 kPa, two 40 kPa, above and below zero) and eight 0–5 V inputs and puts them on the car's CAN bus at 100 Hz, for aero, airbox, turbo-inlet and radiat...

CAN sensor node · rev E

Pressure CAN node

Six pressure sensors and eight 0–5 V inputs on one board. It reads all fourteen and puts them on the car's CAN bus a hundred times a second, so the data logger records aero, intake and cooling pressures alongside everything else.

6pressure channels: four 10 kPa, two 40 kPa
80–5 V analog inputs
100 HzCAN frames at 500 kbit/s
12 Vswitched supply, or USB on the bench
Render of the red rev E board at an angle: six pressure sensors, each with one hose barb, in two rows at the back left, the ESP32-S3 module on the left edge, the second analog block and its connectors at the back right, the analog, USB-C and CAN connectors along the front
Rev E, rendered from the board files. 79 × 70 mm, four layers.

What it measures

Each sensor has one hose barb and reads the pressure there against the air inside the node's box, below it as well as above. Four read to 10 kPa for aero and cooling; two read to 40 kPa for the vacuum a turbo pulls at its inlet. A drop across a part is two channels subtracted. How to fit them.

Aero

Pressure taps on the splitter, underbody, diffuser or wing: how hard the floor sucks at speed and what each surface sees. Dynamic pressure is about 0.5 kPa at 100 km/h and 1.9 kPa at 200 km/h.

Airbox and turbo inlet

Vacuum before and after the air filter, on the two 40 kPa channels. The difference is the filter's restriction, and the second reading is what the turbo sees. A restrictive filter can pull 5 to 7 kPa at full flow, and a turbo inlet more.

Cowl and radiator

Pressure at the cowl and on both faces of the radiator. The drop across the core shows how much air actually goes through it.

Eight analog inputs

For sensors you already have: oil or fuel pressure senders, a spare MAP sensor, ride-height pots, or a tap on an ECU sensor's signal. Each connector carries a switched 5 V supply. Remove its link when tapping a sensor the ECU already powers.

The board

Everything is surface mount and placed by the board house, the pressure sensors included.

Pressure sensors DP1–6WF100DPZ, 10 and 40 kPa, one barb5 V ORLM66100 pair: car or USBStatus LEDOn when every fitted channel readsRESET, BOOT, ZEROZERO zeroes the pressuresESP32-S3Antenna hangs past the edge3.3 V regulatorTLV76733 1 A LDO, room for Wi-FiIsolate and bufferTMUX1511 and TLV9004, AIN1–416-bit ADCADS1115 for AIN1–4Analog inputs AIN1–45 V, ground, 0–5 V signalI2C switchTCA9548A, a branch per sensorAIN5–8 front endSame parts as AIN1–4Analog inputs AIN5–85 V, ground, 0–5 V signal5 V buckLM5164 and its 33 µH inductorInput protectionFuse, reverse diode, TVS12 V inSwitched ignition supplySensor 5 V switchTPS2051B, 500 mA limitCAN transceiverTCAN1042 and ESD clampCAN busBridge TERM for 120 ΩUSB-CProgramming, logs, bench power
Top view, rendered. Open the full-size render.
Close render of the six pressure sensors in two rows of three, each a small black body with one grey hose barb in the middle, with the I2C switch to their right
The six sensors, DP1–3 at the back and DP4–6 in front. Each has one barb; DP5 and DP6 are the 40 kPa parts.

How it works

Power

Switched 12 V comes in through a 1 A fuse and a reverse-battery diode, with a 33 V TVS for load dump. An LM5164 buck makes 5 V. Two ideal diodes choose between that and USB, with the car's side first, so the board runs on the bench from a laptop, stays up when both are plugged in, and the two supplies never feed each other. A 3.3 V LDO runs the logic.

Pressure

All six sensors have the same I2C address, so a TCA9548A switch puts each on its own branch and the ESP32-S3 reads them in turn. A reading only counts once the sensor says its conversion finished, and a sensor that stops answering drops out without stalling the others. ZERO averages 100 readings from every channel and subtracts them. Press it with every port open to the same air before a run.

Analog inputs

Each input goes through a divide-by-five divider (500 kΩ to ground, so it barely loads an ECU signal), an isolation switch that stays open until the firmware is running, a buffer and a 16-bit ADC; AIN1–4 and AIN5–8 have one ADC each, converting side by side. A 12 V short on an input only puts 2.4 V on the switch.

Sensor supply

The 5 V on the analog connectors comes off the buck through a 500 mA current-limited switch. It's only there with the car's power on, never from USB alone, and a fault turns it off.

On the CAN bus

A TCAN1042 transceiver sends five 8-byte frames every 10 ms and a status frame every 100 ms, at 500 kbit/s. The IDs and bitrate are firmware settings. A channel that stops updating, isn't fitted, or reads past its range is flagged invalid in its frame instead of repeating its last value. The 120 Ω terminator is off unless you bridge TERM on the board at the end of the bus.

IDContents
0x520DP1–DP3 in Pa (signed 16-bit), valid bits, frame counter
0x521DP4–DP6 in Pa, valid bits, frame counter
0x522AIN1–AIN4 in mV, 0 to 5,250
0x523Health: protocol version, counter, valid masks, power and zero flags, uptime
0x524AIN5–AIN8 in mV
0x525Status, ten times a second: fitted, overrange, zeroed and fault masks, bus recoveries

Fitting it to a car

Each channel is one hose: a tap where you want the pressure, run back to the node. Every sensor reads its hose against the air inside the node's box, so most of the work is deciding what that air is, and the hose runs.

nodeair4 Cowlbase of the windscreen5 · 6 Airbox2 Front floorunder the splitter3 Diffuserrear of the floor1 Pitot-staticon a mast, in clean airhosestatic to the box
One way to spend the six channels, numbered by sensor. Every tap gets its own hose back to the node in the cabin. The node's box is sealed, with the probe's static line on its one reference barb, so every channel reads against static.

The taps

air0.5–1 mm hole, square tothe surface, clean edgemetal tube glued inhose to the node
Drilled tap A small hole square to the surface with a short metal tube glued in behind it. The edge has to be clean and flush; a burr or a raised lip changes what it reads. Race cars and wind-tunnel models are tapped this way.
airflat disc, port in the middleunder 20 mm across, about 2 mm talltape over the front edge and along the hose
Tape-on pad For paint you don't want to drill. A thin disc with the port in the middle and the hose leaving downstream, taped across the front edge and along the hose. Teams use aluminium ones; a printed one works the same way.
airstatic: to the boxtotal: to DP1ring of side holes: statichole at the tip: total
Pitot-static probe Mounted level, pointing into clean air: a tall mast on the roof or a boom ahead of the nose. The tip reads total pressure and the side holes read static. With the static line on the node's box, DP1 on the tip reads dynamic pressure, and so airspeed.
open end faces across the flowfrontbackcore
Hose end In open spaces, like in front of or behind the radiator, the hose end is the tap. Zip-tie it with the opening across the flow. Pointed into the flow it picks up some dynamic pressure as well.

Plumbing the six channels

SensorRangePortReads
DP110 kPaPitot-static, tipDynamic pressure, so airspeed and headwind
DP210 kPaFront floor tapSuction under the front of the car
DP310 kPaDiffuser tapSuction under the back
DP410 kPaCowl, base of the windscreenWhether the cowl really is high pressure
DP540 kPaAirbox, before the filterDepression ahead of the filter
DP640 kPaAirbox outlet, toward the turboDepression the turbo sees; DP5 minus DP6 is the filter

Each reading is the port's pressure minus the box's. A drop across a part takes two channels and a subtraction in the logger: DP5 minus DP6 is what the filter costs. The two 40 kPa channels go on the intake because the turbo pulls harder than any aero tap.

The reference

Sealed box on the static line

The best reference for aero taps, because it's the pressure of the air the car is driving through. Seal the node's box, give it one barb, and run the probe's static line to it; all six sensors then read against static. Mount the probe where the bodywork doesn't disturb it, and leave it in the same place so runs compare.

Open box in the cabin

The easy way: leave the box open in the closed cabin and every channel reads against cabin air. The cabin's pressure moves with speed and with the ventilation, so keep the windows shut and the fan the same every run.

Two channels for a drop

For a drop across something, tap both sides and subtract: before and after the filter, in front of and behind the radiator. The reference cancels out. Each sensor is good to about 30 Pa near room temperature, so a small drop needs both zeroed together.

Bottle in the boot

A sealed box onto a bottle with a pinhole makes a slow-moving reference. It still drifts with temperature and with the cabin, so it's a last resort, zeroed before every run.

Hoses and mounting

Hose

2 mm ID silicone fits the barbs. Route it with no kinks, no low spots where water can sit, and nothing near the exhaust. Label both ends with the sensor number.

Length

A long, thin hose slows the reading down. Length matters and the bore matters far more. For a drop measured on two channels, keep their hoses about the same length, or a quick change shows up as a false difference.

Leaks

Push gently on each tap with a syringe and hold it. The reading should stay put. If it creeps back toward zero, a joint is leaking.

The node

Somewhere dry and away from heat. The cabin is easiest, with the hoses through grommets. For the static reference the box needs a sealed lid, sealed cable entries and the one reference barb; any leak lets cabin air in.

Running a test

  1. Zero first. Car parked, engine off, every tap and the reference at the same still air, then press ZERO. It's lost at power-off, so do it each time.
  2. Same road, both ways. Hold a steady speed over the same stretch in each direction and average the two. That cancels most of the wind.
  3. Divide by dynamic pressure. An aero reading divided by DP1's dynamic pressure is a pressure coefficient. It stays the same as speed changes, so a run at 110 km/h compares with one at 130.
  4. One change at a time. Baseline, change, baseline again. If the two baselines disagree, conditions moved and the change can't be judged.
  5. On track, several laps per setup. Drop the odd laps (traffic, a lift) and average the rest. Aero numbers mean most above about 100 km/h.

What the numbers look like

Speed

Dynamic pressure is about 0.5 kPa at 100 km/h, 1.9 kPa at 200 and 4.2 kPa at 300, rising with the square of speed. Floor and diffuser taps sit below the static line, so they read negative.

Intake

On road-car intakes that have been measured, the filter on its own is usually 250 to 500 Pa. On one car the airbox outlet cost more than the filter did. A turbo inlet pulls harder: my F56 intake model puts the whole tract at about 4 kPa at 355 CFM and 9 kPa at 563.

Cooling

Small. In one road test an undertray moved the difference across an intercooler from about −25 Pa to +75 Pa, and cruise intake temperature fell from 149 to 117 °F. The drop across a core rises with the square of the airflow through it, so it compares setups well but isn't a flow figure on its own.

Cowl

The base of the windscreen is usually called a high-pressure spot. On-road tests have found it often isn't the highest at the front of the car, and sometimes it's low. Measure it before cutting an inlet there.

What race teams do

The same ideas at a much bigger scale. Six channels won't map a car the way a team's scanner does, so use them a few points at a time, around the part you're changing.

Rakes in F1 testing

Every team runs aero rakes at some point in testing: rows of Kiel probes, total-pressure heads that tolerate yaw, on a frame behind the front wheels, around the rear axle and diffuser, or on the engine cover. They range from 50 to 600 mm wide and check the car on track against the wind tunnel and CFD.

Taps by the hundred

The cars carry surface taps about 0.5 mm across, plumbed to scanners that read hundreds of channels at once. The wind-tunnel model has taps in the same places, so the two compare directly.

Taped-on patches

Where drilling isn't allowed, like a historic F1 car, teams tape on aluminium patches under 20 mm wide and about 2 mm tall. One such car ran 256 of them.

Pitot-static

Higher-end race cars carry one in clean air, level and pointing forward, on a ±14 kPa differential sensor. Airspeed minus GPS speed gives the headwind around the lap, and straight-line and coast-down runs give downforce and balance estimates. Air density comes from a weather station in the pits.

Brake ducts

Taps at the inlet, along the duct and near the outlet. Air that goes into the inlet doesn't always reach the brake, and the pressure along the duct shows where it's lost.

Formula Student

Student teams check their CFD with a handful of taps, eight on the low-pressure side of an undertray in one case, and run scanners on the wings, sidepods and floor.

Sources

Where this section's practice and numbers come from. AutoSpeed was offline when this was written.

  1. How to use pitot tubesPDF, Lawrence Butcher, Racecar Engineering, 2009. Pitot-static probes on race cars: mounting, the ±14 kPa sensor, headwind, coast-down runs, weather station.
  2. Evolution of aerodynamic testing in F1: measurements, F1Technical. Surface taps about 0.5 mm across, scanners logging hundreds of them.
  3. Measuring aerodynamics at the track, F1Technical. Taps in the same places on the car and the tunnel model; constant-speed correlation runs.
  4. What is an F1 aero rake?, Motor Sport. Kiel-probe rakes in testing: where they go, how big, what they check.
  5. Aerodynamic pressure patches and Formula 1, Althen Sensors. Taped-on pressure patches: size, why not drilled, 256 channels on a historic F1 car.
  6. Unlocking brake cooling performance with embedded pressure scanning, Racetech. Brake ducts tapped at the inlet, along the duct and near the outlet.
  7. Thesis on CFD simulation and validation of a Formula Student carPDF, Michael Kirchberger, TU Wien, 2023. Eight taps on the low-pressure side of an undertray to check CFD.
  8. Chell Instruments donate latest pressure scanner technology to Széchenyi István University Formula Student Electric team, Racetech. Scanners on the front wing, rear wing, sidepods and floor.
  9. Large scale response of a vehicle wake to on-road perturbationsPDF, arXiv 2407.03948, 2024. A perforated bottle in the boot as the reference, and why it isn't constant on the road.
  10. Multi-channel high-speed pressure measurement, EcoModder forum. A pitot's static port as the reference, against a sealed vessel.
  11. Conference paper on time lag in micro-tube pressure measurementPDF, ICME 2011, paper AM-050. Hose lag grows with length and falls with the fourth power of the bore.
  12. Blog post on tubing length and transducer response, Ashcroft. Uneven hose lengths on a differential sensor give a false difference.
  13. Keep the floor honest at running ride height, PaddockLink race aerodynamics notes. Baseline, change, baseline; laps per setup; local pressure profiles rather than whole-car maps.
  14. Use dynamic pressure as your aero speed dial, PaddockLink race aerodynamics notes. Runs in both directions, averaged, to take out the wind.
  15. Article on testing intake restriction, AutoSpeed. Filter drop of 1 to 2 inches of water; the airbox exit as the bigger loss.
  16. Article on on-road pressure testing, AutoSpeed. The intercooler undertray result.
  17. Article on finding high-pressure spots at the front of a car, AutoSpeed. The base of the windscreen against other spots.
  18. Report on pressure drop through cellular radiators, NASA Technical Reports Server. Radiator pressure drop rising with the square of airflow.
  19. Pressure drop across radiator and air flow, Physics Forums. Why the drop across a core isn't a flow figure on its own.

Schematic

Drawn by function, one sheet per block, with the board's real part references and values. Click a block in the overview to jump to its sheet.

power railsignaljoinedcrosses, not joinedcontinues on another sheet
0

Overview

click a block
The car's 12 V becomes 5 V and then 3.3 V; USB can power the logic on the bench. The ESP32 reads the six pressure sensors and both ADCs over one I2C bus and sends everything out on CAN.
1

Power input and 5 V buck

J1 · F1 · D1 · D2 · C1 · C2 · C37 · U1 · L1 · R1–R5 · C3–C7
J1 12V IN 1 12V 2 GND F1 1 A D1 SS210 VIN D2 SMCJ33CA C37 22µF 100V C1 2.2µF 100V C2 100nF 100V U1 LM5164 100 V buck 2 VIN 3 EN/UVLO 4 RON 1,9 GND 7 BST 8 SW 5 FB 6 PGOOD R1 47k C3 2.2nF SW L1 33µH R4 150k C5 220pF C0G C4 2.2nF FB R2 150k R3 47k C6 22µF C7 22µF 5V BUCK R5 10k 3V3
12 V comes in on J1 through a 1 A fuse and an SS210 Schottky that blocks a reversed battery. D2 clamps load dump at about 53 V. U1 bucks it to 5 V at about 268 kHz: R1 sets the frequency, R2 and R3 set 5.03 V, and R4, C4 and C5 give the constant-on-time loop the ramp it needs (22 mV at 9 V in, 31 mV at 12 V).
2

5 V OR, 3.3 V and sensor supply

U17 · U18 · R55 · U2 · U3 · C9–C12 · R6 · R7
5V BUCK USB VBUS U17 LM66100 1 VIN 3 CE 6 VOUT 5 ST 2 GND U18 LM66100 1 VIN 3 CE 6 VOUT 2,5 GND ST 5V BUCK R55 10k OR_STATUS 5V SYS C9 10µF U2 TLV76733 1 A LDO 6 IN 4 EN 1 OUT 2 SNS 3,5,7 GND C10 22µF C11 22µF 3V3 ESP32, mux, sensors, ADC, CAN logic side U17 (buck) has priority. Its ST drives U18's CE, so one of the two is always on. 5V BUCK IO17 SENSOR_POWER_EN R6 100k U3 TPS2051B 500 mA switch 5 IN 4 EN 1 OUT 3 OC 2 GND 3V3 R7 10k IO18 SENSOR_POWER_FAULT_N 5V SENS C12 1µF 5V SENS to pin 1 of J4–J7 and J9–J12 through 0 Ω links (R32, R37 … R71)
U17 and U18 are ideal diodes, and U17, on the buck, has priority. While it conducts, R55 holds U18's CE high and USB is cut off; when the buck is missing or lower, U17's ST goes low and turns U18 on. One of them is always on, even with both supplies at the same voltage, and USB and the car can't push current into each other. U2, a 1 A LDO with room for Wi-Fi, makes 3.3 V for the logic; its SNS pin senses the output. U3 takes the buck's 5 V from ahead of the OR and switches it out to the analog-input connectors when the firmware asks, limited to 500 mA, so USB alone never powers external sensors.
3

ESP32-S3, USB and buttons

U4 · J3 · U5 · SW1–SW3 · D5 · R8–R16 · C13–C15
U4 ESP32-S3 WROOM-1-N4 2 3V3 3 EN 27 IO0 17 IO9 14 IO20 D+ 13 IO19 D− 12 IO8 IO4 IO5 IO10 IO6 IO7 IO15 IO16 IO17 IO18 1,40,41 GND I2C SDA → U6 mux, U15/U21 ADCs I2C SCL → U6 mux, U15/U21 ADCs MUX_RESET_N → U6 pin 3 CAN_TX → U16 TXD CAN_RX ← U16 RXD CAN_STANDBY → U16 STB ANALOG_EN → U13/U19 switches SENSOR_POWER_EN → U3 EN SENSOR_POWER_FAULT_N ← U3 OC 3V3 C13 100nF C14 10µF R8 10k C15 1µF SW1 RESET SW2 BOOT 3V3 R9 10k SW3 ZERO 3V3 R10 10k J3 USB-C A4 VBUS A6 D+ A7 D− A5 CC1 B5 CC2 A1 GND USB VBUS → U18 (5 V OR) R15 33Ω R16 33Ω U5 USBLC6 1,6 D+ 3,4 D− 5 VBUS 2 GND USB VBUS R13 5.1k R14 5.1k R12 1k D5 green
The WROOM module runs from 3.3 V. RESET and BOOT are the usual pull-up and button pairs; ZERO tells the firmware to zero the pressure channels. USB-C goes straight to the S3's native USB through 33 Ω resistors and a USBLC6 ESD array, and the 5.1k CC resistors get 5 V from a USB-C charger. UART0 TX/RX (IO43/IO44) go to test pads TP19/TP20 for flashing if USB won't enumerate.
4

Pressure sensor bus

U6 · U7–U12 · R17–R31 · C16–C22
IO4 SDA IO5 SCL 3V3 R17 4.7k 3V3 R18 4.7k Same bus also reaches the ADS1115s U15 (0x48) and U21 (0x49). IO10 MUX_RESET_N 3V3 R19 10k A0–A2 low: switch at 0x70 U6 TCA9548A I2C switch 23 SDA 22 SCL 3 RESET 1,2,21 A0–A2 24 VCC 12 GND 4 SD0 5 SC0 6 SD1 7 SC1 8 SD2 9 SC2 10 SD3 11 SC3 13 SD4 14 SC4 15 SD5 16 SC5 3V3 SD6/SC6, SD7/SC7 unused R20, R21 4.7k to 3V3 SDA SCL U7 DP1, 10 kPa WF100DPZ 0.1BG C17 1µF R22, R23 4.7k to 3V3 SDA SCL U8 DP2, 10 kPa WF100DPZ 0.1BG C18 1µF R24, R25 4.7k to 3V3 SDA SCL U9 DP3, 10 kPa WF100DPZ 0.1BG C19 1µF R26, R27 4.7k to 3V3 SDA SCL U10 DP4, 10 kPa WF100DPZ 0.1BG C20 1µF R28, R29 4.7k to 3V3 SDA SCL U11 DP5, 40 kPa WF100DPZ 0.4BG C21 1µF R30, R31 4.7k to 3V3 SDA SCL U12 DP6, 40 kPa WF100DPZ 0.4BG C22 1µF
All six sensors answer at the same address, so each sits on its own branch of the TCA9548A and the firmware selects one branch at a time. Each branch has its own 4.7k pull-ups and a 1 µF supply cap next to the sensor. DP1–DP4 are 10 kPa parts; DP5–DP6 read to 40 kPa for turbo-inlet vacuum.
4a

One sensor branch

U7 (U8–U12 identical)
U6 SD0 U6 SC0 3V3 R20 4.7k 3V3 R21 4.7k U7 WF100DPZ 0.1BG S6 DT 3 SDA 2 SCL 6 NC 5 VDD 4 CSB 1 GND 3V3 C17 1µF CSB high selects I2C No address pin: 0x6D or 0x6C Pin 6 open One barb in the middle of the body; reads against the air behind the die U8–U10 identical; U11–U12 are 0.4BG, 40 kPa
CSB ties to 3.3 V for I2C. The WF100DPZ has no address pin; pin 6 stays open. Pin numbers from WF's datasheet V3.6 top view.
5

Analog input (one of eight)

J4 · R32–R36 · C23 · C24 · U13 · U14 · U15 · R11
J4 AIN1 1 5V 2 GND 3 SIG R32 0Ω 5V SENS R33 200k R34 200k R35 100k C23 10nF ÷5: 0–5 V in → 0–1 V U13 TMUX1511 1 of 4 2 S1 1 SEL1 3 D1 IO16 ANALOG_EN R11 100k − + U14A TLV9004 buffer, 1 of 4 R36 1k C24 10nF U15 ADS1115 16-bit, 0x48 4 AIN0 1 ADDR 9 SDA 10 SCL 8 VDD 3 GND 3V3 SDA SCL J5–J7 repeat this into U13/U14 channels 2–4 and U15; J9–J12 into U19/U20 and U21 (ADDR high: 0x49). Pull the 0 Ω link when tapping an ECU-powered sensor.
Each input divides by five (200k + 200k over 100k), so 0–5 V reaches the switch as 0–1 V and a 12 V fault is only 2.4 V. U13 keeps the dividers disconnected until the firmware sets ANALOG_EN; R11 holds it low at power-up. U14 buffers each channel into the ADS1115 through a 1k/10 nF filter. AIN5–8 are the same circuit on U19, U20 and a second ADS1115, U21.
6

CAN

U16 · D3 · SJ1 · R52–R54 · J8
IO6 TX IO7 RX IO15 STB 3V3 R53 10k 3V3 R52 10k U16 TCAN1042 HGVDR 1 TXD 4 RXD 8 STB 3 VCC 5 VIO 2 GND 7 CANH 6 CANL 5V SYS 3V3 C34 100nF on VCC, C35 100nF on VIO CAN_H CAN_L D3 PESD1CAN 1 2 3 SJ1 open R54 120Ω Bridge SJ1 only on the board at the bus end. J8 CAN 1 H 2 L 3 G
U16 drives the bus from 5V SYS and talks to the ESP32 at 3.3 V through VIO. R52 holds STB high (standby) until the firmware pulls it low, and R53 keeps TX recessive during boot. D3 is the CAN ESD clamp. The 120 Ω terminator only connects if SJ1 is bridged.
7

ESP32 pin map

U4
GPIONetGoes to
IO4I2C SDAU6 mux, U15/U21 ADCs
IO5I2C SCLU6 mux, U15/U21 ADCs
IO10MUX_RESET_NU6 reset
IO6CAN_TXU16 TXD
IO7CAN_RXU16 RXD
IO15CAN_STANDBYU16 STB, high = standby
IO16ANALOG_ENU13/U19 switches, high = inputs connected
IO17SENSOR_POWER_ENU3, high = 5 V to the AIN connectors
IO18SENSOR_POWER_FAULT_NU3 overcurrent, low = fault
IO8STATUS_LEDD5 through R12
IO9ZERO_NSW3, low = pressed
IO0BOOT_NSW2, hold low at reset to flash
IO19 / IO20USB D− / D+J3 through R16 / R15
IO43 / IO44UART0 TX / RXTP19 / TP20 test pads
ENESP_ENSW1 reset, R8 + C15 power-on delay

These are renders from the board files, and the fitting section is the plan, not a test report. Rev E hasn't been built yet, so none of it has been on a car.