Verification & open questions
Status
Legend: ✅ verified · ❌ tested, no visible effect · 🟡 assumed (adopted from LA series) · ❓ unknown
| Aspect | LA series 5131:2007 | LB360-S 5131:200e |
|---|---|---|
| HID descriptor (64/64 bytes, no report IDs) | ✅ | ✅ (byte-identical, read out ourselves) |
Update packet 01 02, layout | ✅ | ✅ (2026-11-20, seglcd send/probe) |
Off packet 0F | ✅ | 🟡 effect not distinguishable from the timeout (~450 ms); the LB360-S original software does not send it |
| Update interval 250 ms | ✅ | ✅ works |
| Display turns off when the program exits | ✅ | ✅ immediately when seglcd send ends (without --off-at-end, i.e. without 0F); it stays on while running (e.g. 30 s) |
| Cause: timeout after last packet or closing the device | ❓ | ✅ Timeout – with seglcd hold --send 5 --hold 15 the display turned off when sending ended, although the device was still open |
| Timeout after last packet | ❓ (“short”) | ✅ approx. 400–500 ms (400 ms interval stable, 500 ms flashes) |
| Displayed fields | CPU temp. (integer), CPU usage + bar, fan RPM + bar | CPU and GPU temperature (integer, °C) – details |
| Full-scale value of the bar displays | ❓ (1234 RPM ≈ half fan bar) | ❓ (CPU usage has no effect) |
| °F display | ❓ | ❌ symbol stays °C; with flag 1, values ≥ 100 are only shown with two digits (mod 100) – do not use |
| Meaning of the input reports | ❓ | ❓ |
| Reconnecting after disconnect (VM redirection) | – | ✅ seglcd live reconnects by itself after the device is returned |
| String descriptors 4/5 | – | ❓ |
LB360-S: field mapping
Result of seglcd send and seglcd probe on the LB360-S on 2026-11-20:
| Offset | Field | Result |
|---|---|---|
| 3 | CPU temperature, integer | ✅ is displayed |
| 4 | CPU temperature, hundredths | ❌ no effect – the display has no decimal place |
| 6 | CPU usage | ❌ no visible effect |
| 13 | GPU temperature, integer | ✅ is displayed |
| 14 | GPU temperature, hundredths | ❌ no effect |
| 5 / 15 | CPU/GPU temperature unit | ⚠️ symbol stays °C, but the rendering changes: flag 0 caps at 99, flag 1 shows only the last two digits (122 → “22”) and “99” for value 0 – see unit flag |
| remaining (7–12, 16–35) | power, clock, voltage, speeds, time, RAM | ❓ not yet confirmed individually |
| 36–49 | extended range, incl. 37 (unknown), 48 (CPU usage hundredths) | ❌ no visible effect (seglcd probe --offsets 36-63, tested up to 49) |
| 50–63 | incl. 50 (RAM usage hundredths) | ❌ no visible effect (seglcd probe --offsets 50-63) |
LB360-S: unit flag
Test with seglcd raw (2026-11-20), with only the temperature byte and flag set in each case:
| Value (byte 3 or 13) | Flag (byte 5 or 15) | Display |
|---|---|---|
| 0 | 0 | 0 |
| ≥ 100 (100, 105, 123, 199, 255) | 0 | 99 |
| 0 | 1 | 99 |
| 31 | 1 | 31 |
| 32 | 1 | 32 |
| 50 | 1 | 50 |
| 122 | 1 | 22 (same for CPU and GPU) |
The display does not convert from °F to °C. With flag 1 the cap at 99 is dropped and only the
last two digits are shown – presumably firmware for three-digit °F values, without a hundreds digit on this
LCD. Why 0 with flag 1 appears as 99 is unclear. The symbol does not change to °F.
Conclusions for implementations:
- For the LB360-S it is sufficient to fill offset 3 (CPU) and 13 (GPU). Still send the remaining fields as specified: it costs nothing and keeps the packets compatible with the original software.
- Always send in °C with flag
0. With flag1(°F), values ≥ 100 appear with only two digits: 50 °C = 122 °F would be shown as “22”, and the symbol stays °C. A library should ignore the Fahrenheit option for this model or warn about it (seglcdwarns). - Temperatures are truncated, not rounded (byte 3 =
floor). To round instead, round the value yourself before encoding (e.g. 42.7 → 43). - Values ≥ 100 are shown by the display as 99 (tested with 99, 100, 105, 123, 199, 255). A library does not need to clamp, but should be aware that e.g. 105 °C appears as “99”.
Capture of the original software (VM)
Recorded on 2026-11-20 with tools/capture-usbmon.sh: TT SEGMENT LCD Software in the Windows VM
(qemu:///session, SPICE USB redirection, 8 vCPUs). The raw data is not in the repository; the full analysis is here.
In the VM most sensor values are 0. The software shows the tiles CPU Temp, CPU Usage, CPU Fan Speed,
CPU Frequency, CPU Power and GPU Fan Speed / Temp; only the source for CPU Usage is configurable
(Total CPU Usage, Core 0…7 Usage, Max CPU/Thread Usage, Core 0…7 Utility, Total CPU Utility).
During the capture, the source was switched once from Total CPU Usage to another source.
| Finding | Detail |
|---|---|
| Commands | exclusively 00 01 02 … (567 packets in ~117 s) – no initialization, no other commands |
Off packet 0F | is not sent – software exited normally: last update at 141.16 s, ~1 s later it closes the device (pending IN requests are canceled), no traffic afterwards. The software relies on the timeout. |
| Update interval | 200–220 ms (median 0.20 s) |
| Value refresh | approx. every 2 s; in between, the last packet is repeated unchanged |
| First packets | for approx. 1 s only time, RAM, byte 37 and byte 50 populated, all readings 0 |
| Byte 6 (CPU usage) | 0–58 %, plausible for the VM ✅ |
| Bytes 11/12 (CPU voltage) | constant 01 37 → “1.55 V” (the LA software sent floor(V) × 100 here) |
| Bytes 25–32 | local time ✅, day of week 5 = Friday ✅ |
| Byte 33 (RAM) | 28–30 % ✅ |
| Byte 37 | new, constant 5 – meaning unknown (not the vCPU count: 8) |
| Byte 48 | new, 0–99, jumps a lot, changes every 2 s → later identified as CPU usage, hundredths |
| Byte 50 | new, 11–99, changes more slowly, changes every 2 s → later identified as RAM usage, hundredths |
| Control transfers | only GET_DESCRIPTOR (device, configuration) on redirection and two SET_IDLE (HID class, bmRequestType 0x21) from Windows |
| Input reports | none – Windows keeps IN requests pending, the device never responds |
| String descriptors | are not queried by Windows |
Second capture: all usage sources
Second capture – 866 packets in ~176 s; all sources for CPU Usage (Total, Core 0–7 Usage, Max CPU/Thread Usage, Core 0–7 Utility) one after another at ~10 s intervals.
| Finding | Detail |
|---|---|
| Commands | again exclusively 01 02 – the selection does not produce a separate command |
| Byte 6 | 4–30 %, no recognizable steps in the 10 s windows – with a uniformly lightly loaded VM, total and per-core values are indistinguishable |
| Byte 37 | again constant 5 |
| Byte 48 | 1–98, jumps on every refresh (looks random), correlation with byte 6: −0.06 |
| Byte 50 | 3–98, drifts slowly (random walk), correlation with byte 6: −0.01 |
| Bytes 11/12 | again constant 01 37 |
Conclusion: with a uniformly lightly loaded VM the source cannot be identified – the third capture settles this.
Third capture: one core under full load
Third capture – while ($true) {} ran in the VM with affinity to core 5;
in the software, Core 5 Usage → Total CPU Usage → Core 7 Usage one after another.
| Section | Byte 6 | Byte 48 |
|---|---|---|
| Core 5 Usage | 100 (constant, ~12 s) | 0 in all 58 packets |
| Total CPU Usage | 23–29 (≈ 100/8 + base load) | 40–86 |
| Core 7 Usage | 8–18 | 3–85 |
Conclusions:
- The usage source selected in the software ends up in byte 6 (no separate command, no separate field).
- Byte 48 = hundredths of the CPU usage (exactly 0 at 100.00 %, otherwise uniformly distributed 0–99).
- Byte 50 = hundredths of the RAM usage: byte 33 + byte 50/100 runs smoothly in all three captures (max. jump of 0.2 percentage points between two refreshes, e.g. 27.94 → 27.97 → 27.98 → 28.03).
Open points from the captures:
- Byte 37: stayed constant
5when switching the usage source → not the core selection. - ✅ Bytes 48/50: CPU and RAM usage, hundredths (third capture).
- Test byte 37 on the display:
seglcd probe --offsets 37 --value 88. - ✅ Usage source: ends up in byte 6 (third capture).
Test plan for the LB360-S
Prerequisite: udev rule installed, seglcd list shows “Open OK”.
✅ Basic function – does the display show anything at all?
seglcd send --cpu-temp 42 --gpu-temp 57 --duration 20✅ (partially) Field mapping – set each byte individually to a conspicuous value:
seglcd probe --value 88 # interactive, Enter = next byte seglcd probe --value 88 --auto 3 # automatic, 3 s per byteFor each byte, note whether and where the display changes.
✅ Decimals / unit – no decimals, no °F (see above).
seglcd send --cpu-temp 42.75 --gpu-temp 57.25 --duration 10 seglcd send --cpu-temp 42 --gpu-temp 57 --fahrenheit --duration 10Bars/scales – determine full-scale values:
seglcd sweep 6 --from 0 --to 100 --step 5 --step-ms 700 # CPU usage seglcd sweep 21 --from 0 --to 50 --step-ms 700 # fan (×100 RPM)✅ Timeout or closing? Result: timeout (display off as soon as no more packets arrive). When the program exits, two things happen at the same time: no more packets arrive, and the process closes
/dev/hidrawN(thenusbhidalso stops polling the interrupt IN endpoint).seglcd holdseparates the two: send first, then stop, but leave the device open.seglcd hold --send 5 --hold 15- Display turns off at the message “Last packet sent” (or shortly after) → genuine timeout; note the time.
- Display only turns off at “Closing the device” → closing (end of IN polling) triggers it;
then a keep-alive in the sense of “repeating packets” may not be needed at all – to be checked with
seglcd hold --send 1 --hold 60. - Display stays on even after closing → only the off packet turns it off.
✅ Timeout narrowed down: 400 ms interval stable, 500 ms flashes. Measurement:
for i in 300 400 500 750 1000; do echo "Interval $i ms"; seglcd send --cpu-temp 42 --gpu-temp 57 --interval $i --duration 6; doneComparison with the original – USB capture of the Windows software (instructions), analyze with
seglcd decode --diffand compare with this document.
Please add results to this table. Captures belong in the local folder captures/; it is excluded via .gitignore because usbmon records the entire bus (including other devices).