# Wing RS-485 readers, actuator link and damper timing (2026-10-07, revised)

**Decided 2026-10-07 in a changed form (Mike: "yes"): see Pending Amendments v10.2 item 386.** One Waveshare USB TO 2CH RS485 isolated reader for the tip meter and the compass, the DSD TECH SH-C30G in the hub box on its own lead to Pi A, the FTDI cable for the DO sensor. The Innomaker adapters in "The plan" below were not taken. The timing section stands.

Mike asked: research the RS-485 readers and the timing on the wing items and get it right, planned so it works, with no bench tests (Mike, 2026-10-07). Everything below comes from datasheets, seller pages and the guides; guesses are marked.

## What the guides require

- Rate damper: the wing compass's turn rate read 10 to 20 times a second, end-to-end delay under 100 ms (Wing Trim Actuator Guide, the active rate damper; Build Guide 6.4). The wing hunts at 0.5 to 2 Hz.
- Path since the slip ring: compass (HWT905, RS-485) and tip meter (Calypso CMI1017, RS-485) down through the ring to the wing box, a Taidacent RS-12V/2S on each pair, then "its own RS-485 reader"; the wing's program runs on the hull Pi; the Thomson Electrak MD on its own short J1939 CAN link (Wing Power System 8.7).
- Open in the guides: which readers, and what talks to the actuator's CAN link.

## Facts found

| Item | Fact | Source |
|---|---|---|
| HWT905-RS485 supply | 9 to 36 V, under 40 mA. **Not 5 V** | HWT905-RS485 datasheet v20-0707, sec 3.1 |
| HWT905-RS485 link | Modbus RTU, answers polls only; address 0x50; 9600 baud default; 4,800 to 230,400 (and higher) | Same datasheet; WitMotion SDK pages |
| HWT905 registers | Gyro X/Y/Z 0x37 to 0x39 (raw / 32768 x 2000 deg/s); roll/pitch/yaw 0x3D to 0x3F (raw / 32768 x 180 deg); baud at 0x04 | Same datasheet, register table and sec 5.3 |
| HWT905 reply delay | About 1 ms after a read request (MODDELAY setting) | WitMotion SDK (gitbook) |
| HWT905 heading accuracy | 1 deg on Z with good magnetic calibration | HWT905 specification |
| Calypso CMI1017 | RS-485, NMEA 0183, poll or stream, 1 to 10 Hz, set over USB | Citimarine product page. Baud not published |
| FTDI cable USB-RS485-WE | Built-in 120 ohm terminator on brown and green wires; FTDI: made for the end of a network up to about 33 ft (10 m), longer runs cut the tail to 8 in (20 cm); ESD +/-15 kV; not isolated | FTDI datasheet DS_USB_RS485_CABLES v1.4 |
| Innomaker isolated USB to RS485 | FTDI chip, power and signal isolation, ESD 15 kV, 120 ohm jumper, $24.99, sold by InnoMaker USA on Amazon (B0B2QSW67D) | Amazon listing, Oct 7 |
| DSD TECH SH-C30G USB to CAN | Galvanic isolation with its own supply, candleLight firmware (Linux sees a standard CAN port, nothing to install), classic CAN to 1 Mbit/s, 4-pin screw terminal, 120 ohm switch, TVS and fuse on each line, $19.99 (B0CDGGBZDT) | Amazon listing, Oct 7 |
| FTDI on Linux | USB latency timer 16 ms by default; set to 1 ms in sysfs or a udev rule | Granite Devices, OpenBCI, kernel notes |
| Electrak MD 025 | About 2 in/s at no load; J1939 status every 100 ms | Thomson MD catalog; System Settings 8 |
| Pi CAN ports | Both channels of each Pi's CAN HAT carry the boat's two buses | Electronics plan sec 5; Compute Bay 9.2 |

## Timing

One poll for turn rate and heading: registers 0x39 to 0x3F, 8 bytes out, 19 back.

| | 9600 baud, USB timer 16 ms | 115,200 baud, USB timer 1 ms |
|---|---|---|
| On the wire | 28 ms | 2.3 ms |
| Sensor reply delay | 1 ms | 1 ms |
| USB | up to 16 ms | about 1 ms |
| Reading's age when the command goes out | about 45 ms | about 5 ms |
| Line busy at 20 Hz | about 56% | about 5% |

Then under 1 ms to compute and under 1 ms on the actuator's CAN link. Linux can add a pause of a few ms to perhaps 20 ms on a busy Pi (a guess). At 115,200 baud the total is well under the 100 ms limit even with that pause.

**Designed so a slow reading can never do harm:** every reading carries the time it was taken; the damper uses it only if it is under 100 ms old, and otherwise leaves the tab where trim puts it for that cycle (the guides already turn the damper off when the compass goes quiet).

## Findings that change the guides

1. **The compass needs 9 to 36 V, not 5 V.** The guides feed it from the tip's 5 V converter (Wing Power System 1A, 8.5, 8.6; System Settings 8; Build Guide 6.6). It will not run on 5 V. Feed it from the tip branch's 24 V (already at the tip, on its 1 A fuse), ahead of the 5 V converter; the converter then feeds only the tip meter.
2. **No CAN port is left for the actuator.** A third one is needed: the DSD TECH SH-C30G isolated USB to CAN on Pi A's free USB port (free since the dAISy moved to the hub).
3. **Isolated readers suit these lines.** The wing pairs come off the mast through the slip ring, and the science lines keep their grounds apart (Lightning Guide 7.5). The Innomaker isolated adapter does both jobs at $24.99, cheaper than the FTDI cable ($26.99), with the same FTDI chip (the guides' rule: FTDI silicon only).
4. **With Pi B in command the wing holds its last trim.** Accepted already: Wing Power System 9 (the wing feathers in strong wind).

## The plan (Claude's recommendation, awaiting Mike's yes)

- Four Innomaker isolated USB to RS485 adapters in the hub box: DO sensor, tip meter, compass, and the SBE 45 if its label says RS-485 (about Oct 26). They fill the 7-port hub with the dAISy, the SBMS0 lead and the SDI-12 adapter.
- One DSD TECH SH-C30G on Pi A's free USB port for the actuator's J1939 link at 250 kbit/s.
- Each wing RS-485 pair runs from its Taidacent at the wing box to the hub box as shielded twisted pair.
- Compass: set once to 115,200 baud when it is set up (Wing Power System 8.5 already sets it up over RS-485); the Pi polls registers 0x39 to 0x3F at 20 Hz.
- Tip meter: streams NMEA at 1 Hz. Its baud is not published, so the Pi's program tries the standard rates at start-up (4,800, 9,600, 19,200, 38,400, 57,600, 115,200) and keeps the one that gives sentences with good checksums.
- A udev rule sets each FTDI adapter's USB latency timer to 1 ms; the wing program runs as its own high-priority process.
- Terminators off on the wing lines: at 115,200 baud a run of 40 ft or so is far too short for echoes on the line to matter (a bit lasts about 8.7 microseconds; the signal crosses 40 ft in about 0.06 microseconds). Inferred from standard RS-485 practice.
- Parts: four adapters about $100, the CAN adapter $19.99; Taidacents: the tip meter's on hand, the compass's on the Buy list, one for the DO sensor, one for the SBE 45 if RS-485.

## Option B, for the record

A wing-box node on a three-CAN Nucleo (G474 class) running the damper itself, so the wing works on Pi B too. Cost: a new node and its firmware. Not recommended while the wing on Pi B only has to hold trim.
