Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
125 changes: 113 additions & 12 deletions docs/scyrox-v8-testing.md
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,77 @@ UI strings), including a screenshot of it connected to the same mouse.
(byte 6: `01` awake, `00` asleep). While it reads `00`, every other command
goes unanswered.

## Wired V8 (0xf5f6)

- Device: "SCYROX V8", VID/PID `0x3554:0xf5f6`, release `0x0220`, mouse
firmware v2.20. Same collection layout as the dongle: `0xff02:2` with
report 8 on interface 1, plus `0xff05`, `0xff03`, `0x000c:1`, `0x0001:0x80`
and `0xff04:2`; interface 0 a keyboard, interface 2 the mouse.
- `ScyroxHidClient.readStatus` through node-hid on Windows: "Scyrox V8",
one stage at 1,600 DPI, 1,000 Hz, battery 45% / 3,829 mV charging,
profile 1, debounce 0 ms, sleep 300 s, motion sync and angle snapping on,
ripple off, performance off, 20K on, firmware "Mouse v2.20".
- Polling rate round trip: 1,000 → 2,000 Hz (read back 2,000), then
2,000 → 1,000 Hz (read back 1,000).
- Handshake (command `0x01`): a bare request fails with status 1; the
configurator's payload (four random bytes, four zeros) is answered with
payload `84 bf 7a f2 3e 01 02`: CID `0x3e`, MID `0x01` (V8), type `0x02`
(wired 1K).

## V6 identification

The configurator (S-center 3.1.0, `cfg.json`) lists only five Scyrox
product ids: wired `0xF5F6` and `0xF515`, wireless `0xF5F4`, `0xF5F7` and
`0xF517`. It tells models apart by the handshake's MID, not by product id,
and gives the V8 (MID 1, 10, 11) and the V6 (MID 2, 12, 13) the same 3950
sensor, 30,000 DPI ceiling and controls. `ScyroxHidClient` therefore sends
the handshake on the V8 entries and names the mouse "Scyrox V6" for a V6
MID. `0xF5F4` (sensor unknown) is not claimed.

- **Unanswered or refused handshake:** the client keeps the catalog's
"Scyrox V8" and asks again on a full read at least 30 s later. A mouse
asleep behind the dongle is still named once it wakes, and a silent mouse
costs one timeout per interval, not one per read. The first reply settles
it for the connection.
- **Key bytes:** the handshake sends fixed key bytes (`4f 4d 53 01`) rather
than the configurator's random ones, so captured traces are reproducible.
- **Connection type:** the reply's type byte is used too. A wired mouse
reporting type 3 (wired 8K) is offered the 8K rate list instead of the
wired 1K default.

### Writes wait for the handshake

The 3950 entries' extra writes (the lift-off stop table, angle tuning, the
30 ms Key Response Time and Restore) were tested on a V8 and a V6. The
client offers them once the handshake has named the mouse either model. A
mouse that has not answered yet gets the shared Atlantis controls only.

### V6 on hardware

The V6 was first checked read-only: identification and a full read. It was
then used and tested with the same controls as the V8 (see "Verified writes
on the V8 and V6"), which worked as they do on the V8.

The read-only check used a Scyrox V6 on the "SCYROX 8K Dongle" (`0x3554:0xf5f7`, release `0x0216`,
dongle version `02 16`), read through node-hid on Windows with a read-only
build of `ScyroxHidClient` that refuses every command but 0x01, 0x03, 0x04,
0x08, 0x0e, 0x12 and 0x1d:

- Same product id and collection layout as the V8's dongle.
- Handshake payload `e0 ec fc 6b 3e 02 05`: CID `0x3e`, MID 2 (V6), type 5
(wireless 8K). The client named it "Scyrox V6".
- Firmware `02 24`, v2.24, the version the configurator's V6 upgrade tool
ships.
- `readStatus`: three stages 500/800/1,600 with stage 2 active, 1,000 Hz,
battery 75% / 4,013 mV discharging, profile 1, debounce 0 ms, sleep 300 s,
motion sync, angle snapping, ripple control and performance on, 20K off,
firmware "Mouse v2.24". Every field carried its seal.
- The raw flash agrees field by field: rate `01 54`, three stages `03 52`,
active stage `02 53`, 20K `00 55`, legacy stages `09 09 00 43`,
`0f 0f 00 37` and `1f 1f 00 17` (500, 800, 1,600 DPI), sleep `1e 37` at 173
and 183, performance `01 54` at 181. The 3955 region at 6912 reads `ff`, as
on the V8, confirming the legacy layout.

## Protocol family

The V8 speaks the same CompX report-8 flash protocol as the V8S, and every
Expand Down Expand Up @@ -95,28 +166,58 @@ differences, which the catalog follows:

- Lift-off: the 3950 offers three heights coded 3, 1 and 2 for 0.7, 1 and
2 mm. That order fits neither the Atlantis two-stop write nor the 1-to-n
scale the V8S uses, so lift-off is not offered (`supportsLiftOffDistance:
false`, no `liftOffMillimetres`).
- Angle tuning: the configurator shows it only when 189 and 191 are sealed,
and both read `ff`, so `supportsAngleTuning` stays off.
- Key response time: the configurator allows up to 30 ms on the V8; the
Atlantis client offers 0-15 ms.
scale the V8S uses, so the catalog gives it an explicit stop table
(`liftOffStopCodes: { Low: 3, Medium: 1, High: 2 }`) and no
`liftOffMillimetres`. A byte outside the table reads as no stop
(`liftOffDistance: null`). The three stops stay on offer, so choosing one
writes a known code over it.
- Angle tuning: the configurator shows it only when 189 and 191 are sealed.
On the first V8 dump both read `ff`. A later read found them sealed (angle
0, on), and the angle round trip below was verified on that V8, so
`supportsAngleTuning` is on for an identified V8.
- Key response time: the configurator allows up to 30 ms on the V8
(`maxDebounceMs: 30`); the Atlantis client's default ceiling is 15 ms.
- Restore: the configurator's Restore button, command `0x09`, as on the
V8S (`supportsFactoryReset`).
- Sleep: 173 and 183 held the same value, as the configurator's Scyrox sleep
setter writes them, so `mirrorSleepToPerformanceTime` stays on.
- SensorMode at 185 is left alone (`supportsHyperMode: false`), as on the
V8S.
- The 20K switch at flash 8 is shown for the V8 too (`FpsScanShow: true`) and
read `00 55`.
read `00 55`. The configurator greys it out until Competitive Mode
(performance mode, flash 181) is on, so on Scyrox the two are linked:
- 20K on turns performance mode on first.
- Performance mode off turns 20K off first.
- If the second write of a pair does not take, the first is put back.
- Other Atlantis products keep the two switches separate.

The lift-off stop table, angle tuning, the 30 ms ceiling and Restore are
offered only once the handshake names the mouse a V8 or a V6 (see "Writes
wait for the handshake").

## Verified writes on the V8 and V6

On the V8 behind its 8K dongle, each was written, read back and restored:

- the polling rate
- lift-off
- the DPI stage colour
- the angle
- debounce

The V6 behind the 8K dongle was used and tested with the same controls, and
they worked as on the V8.

## Not yet verified

- DPI, lift-off, angle tuning, sleep, debounce, the 20K switch, profiles,
buttons and Restore were never written to the V8. Only the polling rate
was, as a round trip ending on the rate read first.
- Sleep, the 20K switch and its link to performance mode, profiles, buttons
and Restore have not been written to a V8 or V6.
- The sleep reading: the flash held 180 (30 minutes), while the configurator
screenshot showed "1min", likely its default before the device read.
- Whether the V6 enumerates with the same dongle id. It would read the same
fields (same sensor, same ceiling) but be named "Scyrox V8".
- The V6 on its cable (expected `0xf5f6`).
- A wired 8K Scyrox (handshake type 3); the rate list it gets comes from the
configurator's type table.
- What `0xF5F4` is. The configurator lists it as a wireless mouse id.
- The 4,000 and 8,000 Hz bytes on the V8: the configurator offers them, but
neither has been read from this mouse.

Expand Down
Loading
Loading