TRENDnet TPE-TG PoE+: five unmanaged PoE+ switches (31/246/440/380/70 W) #1647

Merged
jaylouisw merged 1 commit from catalogue/1116-1025 into main 2026-11-16 10:32:09 +00:00
Owner

TRENDnet carried two switches and a powerline adapter before this; the vendor's unmanaged PoE+ range - the one people actually cable IP cameras and APs to - was absent. This adds five members, one per distinct port/PoE shape, all from the vendor's own product pages.

What was added

Model Copper Cages PoE budget Fabric Notes
TPE-TG50g (v4) 4x 1G PoE+ + 1x 1G - 31 W 10 Gbps smallest; budget powers one 802.3at PD, not four
TPE-TG160g (v4) 16x 1G PoE+ - 246 W 32 Gbps DIP switch: VLAN isolation, storm control, 250 m extend on ports 1-8
TPE-TG182F (v2) 16x 1G PoE+ + 2 shared RJ45/SFP 2x combo 440 W 36 Gbps combo cages modelled with comboWith/comboRole
TPE-TG262 (v1) 24x 1G PoE+ 2x 1G SFP 380 W 52 Gbps dedicated fibre uplinks
TPE-TG327 (v1) 5x 1G PoE+ + 2x 2.5G PoE+ - 70 W 20 Gbps mixed line rates on one box

All five are unmanaged: l2-forward plus poe-pse, no cli, no http-server. The 802.3af/802.3at tokens are kept as declared and reported by the importer as hardware facts with no engine behaviour, exactly as docs/DEVICE_SCHEMA.md describes.

Evidence

Per-device sources are recorded in each spec's sourceUrl and in poe.sourceUrl:

Fabric figures are self-consistent with the port counts in every spec (18 physical positions x 2 Gbps = the stated 36 Gbps on the combo box, 26 x 2 = 52 Gbps, 7 total Gbps-equivalents = 20 Gbps on the multi-gig box), which is the cross-check that caught nothing this time but is the reason the numbers were trusted.

Discrepancies and the one modelled choice

  • PoE budget vs port capability is the interesting part on every one of these. TPE-TG160g states 16 ports at up to 30 W against a 246 W budget (480 W of port capability), TPE-TG182F 16 x 30 W against 440 W, TPE-TG327 seven ports against 70 W. Each poe.note says how many devices of each class actually fit, because "16 PoE+ ports" reads as 480 W and is not.
  • TPE-TG182F port numbering: the vendor text names "2 x Shared Gigabit ports (RJ-45 or SFP)" without numbering them. They are modelled as port17/port18 (last two front-panel positions in the vendor callout) with -sfp siblings; the assumption is stated in the spec's notes so the next pass can correct it from a photo rather than guess again.
  • Not modelled: the TPE-TG160g DIP switch's VLAN port isolation is a physical toggle, not 802.1Q, so no VLAN capability was granted; extend mode's 10 Mbps/250 m behaviour is a port setting, not a separate interface. TPE-TG44g was left alone this run - its product page says 68 W PoE max while its current datasheet says a 60 W budget, and that version split deserves its own pass rather than a guess.

Gates

  • node scripts/syntax-check.js - 348 files, all OK
  • node scripts/audit-port-types.js - 2201 models, 24835 ports, 0 offending devices, 0 offending ports (no trendnet findings)
  • importer tripwire: 5 added, only the expected 802.3af/802.3at "kept as declared" warnings
  • catalogue-hygiene, capability-vocabulary, combo, poe, poe-agreement, import-devices-capabilities, doc-catalog-counts, faceplate-catalogue, link-parts, duplicate-models - all pass
  • check-control-bytes, check-canonical-urls - clean
  • combo behaviour verified by hand: cabling port17 then port17-sfp returns E_COMBO_OCCUPIED naming the occupied sibling, while port17-sfp to a TPE-TG262 SFP slot validates
  • docs/DEVICE_REGISTRY.md regenerated (2,196 -> 2,201 models, switch 370 -> 375), not hand-edited
TRENDnet carried two switches and a powerline adapter before this; the vendor's unmanaged PoE+ range - the one people actually cable IP cameras and APs to - was absent. This adds five members, one per distinct port/PoE shape, all from the vendor's own product pages. ## What was added | Model | Copper | Cages | PoE budget | Fabric | Notes | |---|---|---|---|---|---| | TPE-TG50g (v4) | 4x 1G PoE+ + 1x 1G | - | 31 W | 10 Gbps | smallest; budget powers one 802.3at PD, not four | | TPE-TG160g (v4) | 16x 1G PoE+ | - | 246 W | 32 Gbps | DIP switch: VLAN isolation, storm control, 250 m extend on ports 1-8 | | TPE-TG182F (v2) | 16x 1G PoE+ + 2 shared RJ45/SFP | 2x combo | 440 W | 36 Gbps | combo cages modelled with comboWith/comboRole | | TPE-TG262 (v1) | 24x 1G PoE+ | 2x 1G SFP | 380 W | 52 Gbps | dedicated fibre uplinks | | TPE-TG327 (v1) | 5x 1G PoE+ + 2x 2.5G PoE+ | - | 70 W | 20 Gbps | mixed line rates on one box | All five are unmanaged: `l2-forward` plus `poe-pse`, no `cli`, no `http-server`. The `802.3af`/`802.3at` tokens are kept as declared and reported by the importer as hardware facts with no engine behaviour, exactly as docs/DEVICE_SCHEMA.md describes. ## Evidence Per-device sources are recorded in each spec's `sourceUrl` and in `poe.sourceUrl`: - TPE-TG50g: https://www.trendnet.com/products/PoE/5-Port-Gigabit-PoEplus-Switch-TPE-TG50g-v4 - TPE-TG160g: https://www.trendnet.com/products/unmanaged-poe-switch/16-port-gigabit-poeplus-switch-TPE-TG160g-v4 - TPE-TG182F: https://www.trendnet.com/products/18-port-gigabit-high-power-poeplus-switch-TPE-TG182F-v2 - TPE-TG262: https://www.trendnet.com/products/poe-switch-with-sfp/26-port-gigabit-poeplus-switch-TPE-TG262 - TPE-TG327: https://www.trendnet.com/products/affordable-2-5g-poe-switch/7-port-multi-gig-poe-switch-TPE-TG327 and https://www.trendnet.com/support/support-detail.asp?prod=200_TPE-TG327 (which states which ports are 2.5G) Fabric figures are self-consistent with the port counts in every spec (18 physical positions x 2 Gbps = the stated 36 Gbps on the combo box, 26 x 2 = 52 Gbps, 7 total Gbps-equivalents = 20 Gbps on the multi-gig box), which is the cross-check that caught nothing this time but is the reason the numbers were trusted. ## Discrepancies and the one modelled choice - **PoE budget vs port capability is the interesting part on every one of these.** TPE-TG160g states 16 ports at up to 30 W against a 246 W budget (480 W of port capability), TPE-TG182F 16 x 30 W against 440 W, TPE-TG327 seven ports against 70 W. Each `poe.note` says how many devices of each class actually fit, because "16 PoE+ ports" reads as 480 W and is not. - **TPE-TG182F port numbering:** the vendor text names "2 x Shared Gigabit ports (RJ-45 or SFP)" without numbering them. They are modelled as port17/port18 (last two front-panel positions in the vendor callout) with `-sfp` siblings; the assumption is stated in the spec's `notes` so the next pass can correct it from a photo rather than guess again. - **Not modelled:** the TPE-TG160g DIP switch's VLAN port isolation is a physical toggle, not 802.1Q, so no VLAN capability was granted; extend mode's 10 Mbps/250 m behaviour is a port setting, not a separate interface. TPE-TG44g was left alone this run - its product page says 68 W PoE max while its current datasheet says a 60 W budget, and that version split deserves its own pass rather than a guess. ## Gates - `node scripts/syntax-check.js` - 348 files, all OK - `node scripts/audit-port-types.js` - 2201 models, 24835 ports, 0 offending devices, 0 offending ports (no trendnet findings) - importer tripwire: 5 added, only the expected 802.3af/802.3at "kept as declared" warnings - `catalogue-hygiene`, `capability-vocabulary`, `combo`, `poe`, `poe-agreement`, `import-devices-capabilities`, `doc-catalog-counts`, `faceplate-catalogue`, `link-parts`, `duplicate-models` - all pass - `check-control-bytes`, `check-canonical-urls` - clean - combo behaviour verified by hand: cabling port17 then port17-sfp returns `E_COMBO_OCCUPIED` naming the occupied sibling, while port17-sfp to a TPE-TG262 SFP slot validates - `docs/DEVICE_REGISTRY.md` regenerated (2,196 -> 2,201 models, switch 370 -> 375), not hand-edited
TRENDnet TPE-TG PoE+: five unmanaged PoE+ switches (31/246/440/380/70 W)
Some checks failed
CI / Classify the change (pull_request) Successful in 14s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 17s
CI / Server smoke test (pull_request) Has been skipped
CI / Docker build (pull_request) Has been skipped
CI / Test (Node 18) (pull_request) Failing after 1m30s
CI / Test (Node 20) (pull_request) Failing after 1m33s
CI / Test (Node 22) (pull_request) Failing after 1m39s
148a3a0a3c
jaylouisw scheduled this pull request to auto merge when all checks succeed 2026-11-16 10:32:01 +00:00
Author
Owner

main is red from this merge — test/property-validator.test.js, deterministic

origin/main @ 677cfb0e (this PR) and @ 1efbafd5: Test (Node 18/20/22) all fail in run 3637. Reproduced locally on a clean checkout of both commits:

✗ property-validator   18 passed, 1 failed
  not ok - 1 property violation(s) across 720 random link attempts
    [seed 65  d0:SFP56-3 -- d2:sfp1 media="console-rs232": unknown code E_BIDI_MISMATCH]

30049dc6 (the commit before this one) is green, 19/19. So the red starts here — not because this PR changed behaviour, but because five new specs change the fixture the fixed seed draws from: specs[Math.floor(rand() * specs.length)] now lands on an SFP56 port, and connecting it with a console-rs232 media override returns E_BIDI_MISMATCH from engine/optical.js:225 / engine/media.js:1339.

The assertion at test/property-validator.test.js:125 is:

if (!RULE_CODES.has(e.code) && e.code !== 'E_PORT_IN_USE') problems.push(tag + ': unknown code ' + e.code);

RULE_CODES is validation.RULES — 24 entries — but the engine declares 75 distinct code: 'E_…' values. 59 of them are outside the rules table, including E_BIDI_MISMATCH, E_BAUD_MISMATCH, E_OPTICAL_BUDGET, E_LINK_*. Any draw that reaches one of those codes fails identically; seed 65 just stopped avoiding them first.

Deleting any one of the five new specs makes the suite pass again (verified for all five) — that is the fixture draw moving, not a fix.

Needs a decision, not a guess — two candidates, both outside catalogue work:

  1. test/: check the code against the union of every code the engine declares, not just validation.RULES (then E_PORT_IN_USE special-casing goes away too).
  2. engine/validation.js: publish the media/optical/link code family into RULES.

I have not touched either — engine/ and test/ are not mine to edit. Until it is fixed every merge to main inherits a red tip and each new device spec re-rolls the draw.

**main is red from this merge — `test/property-validator.test.js`, deterministic** `origin/main` @ `677cfb0e` (this PR) and @ `1efbafd5`: `Test (Node 18/20/22)` all fail in run 3637. Reproduced locally on a clean checkout of both commits: ``` ✗ property-validator 18 passed, 1 failed not ok - 1 property violation(s) across 720 random link attempts [seed 65 d0:SFP56-3 -- d2:sfp1 media="console-rs232": unknown code E_BIDI_MISMATCH] ``` `30049dc6` (the commit before this one) is green, 19/19. So the red starts here — **not because this PR changed behaviour**, but because five new specs change the fixture the fixed seed draws from: `specs[Math.floor(rand() * specs.length)]` now lands on an SFP56 port, and connecting it with a `console-rs232` media override returns `E_BIDI_MISMATCH` from `engine/optical.js:225` / `engine/media.js:1339`. The assertion at `test/property-validator.test.js:125` is: ```js if (!RULE_CODES.has(e.code) && e.code !== 'E_PORT_IN_USE') problems.push(tag + ': unknown code ' + e.code); ``` `RULE_CODES` is `validation.RULES` — **24** entries — but the engine declares **75** distinct `code: 'E_…'` values. 59 of them are outside the rules table, including `E_BIDI_MISMATCH`, `E_BAUD_MISMATCH`, `E_OPTICAL_BUDGET`, `E_LINK_*`. Any draw that reaches one of those codes fails identically; seed 65 just stopped avoiding them first. Deleting any one of the five new specs makes the suite pass again (verified for all five) — that is the fixture draw moving, not a fix. **Needs a decision, not a guess** — two candidates, both outside catalogue work: 1. `test/`: check the code against the union of every code the engine declares, not just `validation.RULES` (then `E_PORT_IN_USE` special-casing goes away too). 2. `engine/validation.js`: publish the media/optical/link code family into `RULES`. I have not touched either — `engine/` and `test/` are not mine to edit. Until it is fixed every merge to main inherits a red tip and each new device spec re-rolls the draw.
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!1647
No description provided.