Cambium cnMatrix EX2000: five switches, a 400 W budget that binds at 30 W a port, and 60 W 802.3bt only on the multi-gig model #1580

Merged
jaylouisw merged 1 commit from catalogue/1116-0126 into main 2026-11-16 01:41:28 +00:00
Owner

The catalogue has Cambium radios (ePMP, PTP, XV2) and no Cambium switch at all. This adds the cnMatrix EX2000 series: five models off one vendor data sheet.

model ports PoE throughput
EX2010-P 8x 1G RJ45 + 2x 1G SFP 8 ports, 100 W, 30 W/port, fanless 20 Gbps
EX2016M-P 8x 1G RJ45 + 6x 2.5G RJ45 + 2x 10G SFP+ 14 ports, 240 W, 30 W on 1G and 60 W 802.3bt on 2.5G 86 Gbps
EX2028 24x 1G RJ45 + 4x 10G SFP+ none 128 Gbps
EX2028-P 24x 1G RJ45 + 4x 10G SFP+ 24 ports, 400 W, 30 W/port 128 Gbps
EX2052-P 48x 1G RJ45 + 4x 10G SFP+ 48 ports, 400 W, 30 W/port 176 Gbps

What the datasheet says that the model names do not

  • The 400 W budget is the binding constraint, not the 30 W port rating. The EX2052-P can ask its ports for 48 x 30 W = 1440 W from a 400 W supply. That is recorded in the poe.note on both 400 W models rather than left for a reader to add up.
  • Only the multi-gig model does 802.3bt. The 60 W / bt rating belongs to the six 2.5G ports of the EX2016M-P; the other four models are 802.3af/at at 30 W. The 802.3bt token is declared on that one model only.
  • Throughput is the full-duplex aggregate, and the number checks out on every model (EX2010-P 8+2 = 10 -> 20 Gbps; EX2028 24+40 = 64 -> 128; EX2016M-P 8+15+20 = 43 -> 86; EX2052-P 48+40 = 88 -> 176), so it is declared as fabric.capacityMbps.
  • Firmware note: the data sheet marks IGMP snooping v3 and AAA with "feature to be included in a future release", i.e. not in shipping firmware. Recorded in the notes rather than modelled.

Not modelled, and why

  • The out-of-band management port. The spec table lists one on every model but states its rate only for the EX2028-P (10/100 Mbps), so inventing a rate for the other four was the wrong move; it is described in the notes.
  • CLI port names. The data sheet gives counts, not names, so ports are numbered positionally (copper first, uplinks last) and the notes say exactly that.
  • Serial console is RS-232 on every model. Per docs/DEVICE_SCHEMA.md ("Serial consoles, USB, FXS go in notes, not in interfaces") it stays prose.

Validation

  • node scripts/syntax-check.js - 328 files, all OK
  • node scripts/audit-port-types.js - "No port contradicts its declared type."
  • node scripts/audit-devices.js - "No hard defects."
  • node scripts/import-devices.js - 5 added, 0 rejected, 0 refused spellings; the only warnings are the documented keep-as-declared ones (802.3bt is a real power fact with no engine token)
  • node test/importer-roundtrip.test.js - 26 passed
  • node test/catalogue-hygiene.test.js - 25 passed; node test/doc-catalog-counts.test.js - 24 passed
  • node test/spec-vs-description.test.js - 2 passed, 1 failed, and the failure is not in this branch: edgecore/AS4610-30P|30T|54P|54T each "claims 2x 20G QSFP+, 0 port(s) can do it", from the merged #1555. That description names two 20G stacking cages that the spec deliberately does not model as ports; the check reads the phrase as a port claim. It is red on main and needs a reword there (or a KNOWN entry), which is an edgecore-scope change and not made here.

Every field carries the URL it came from: the cnMatrix EX2000 data sheet (brandcentral.cambiumnetworks.com/.../cnMatrix-EX2000-Data-Sheet.pdf) plus the per-model page on cambiumnetworks.com.

The catalogue has Cambium radios (ePMP, PTP, XV2) and **no Cambium switch at all**. This adds the cnMatrix EX2000 series: five models off one vendor data sheet. | model | ports | PoE | throughput | |---|---|---|---| | EX2010-P | 8x 1G RJ45 + 2x 1G SFP | 8 ports, 100 W, 30 W/port, fanless | 20 Gbps | | EX2016M-P | 8x 1G RJ45 + 6x 2.5G RJ45 + 2x 10G SFP+ | 14 ports, 240 W, 30 W on 1G and 60 W 802.3bt on 2.5G | 86 Gbps | | EX2028 | 24x 1G RJ45 + 4x 10G SFP+ | none | 128 Gbps | | EX2028-P | 24x 1G RJ45 + 4x 10G SFP+ | 24 ports, 400 W, 30 W/port | 128 Gbps | | EX2052-P | 48x 1G RJ45 + 4x 10G SFP+ | 48 ports, 400 W, 30 W/port | 176 Gbps | ## What the datasheet says that the model names do not - **The 400 W budget is the binding constraint, not the 30 W port rating.** The EX2052-P can ask its ports for 48 x 30 W = 1440 W from a 400 W supply. That is recorded in the `poe.note` on both 400 W models rather than left for a reader to add up. - **Only the multi-gig model does 802.3bt.** The 60 W / bt rating belongs to the six 2.5G ports of the EX2016M-P; the other four models are 802.3af/at at 30 W. The 802.3bt token is declared on that one model only. - **`Throughput` is the full-duplex aggregate**, and the number checks out on every model (EX2010-P 8+2 = 10 -> 20 Gbps; EX2028 24+40 = 64 -> 128; EX2016M-P 8+15+20 = 43 -> 86; EX2052-P 48+40 = 88 -> 176), so it is declared as `fabric.capacityMbps`. - **Firmware note**: the data sheet marks IGMP snooping v3 and AAA with "feature to be included in a future release", i.e. not in shipping firmware. Recorded in the notes rather than modelled. ## Not modelled, and why - **The out-of-band management port.** The spec table lists one on every model but states its rate only for the EX2028-P (10/100 Mbps), so inventing a rate for the other four was the wrong move; it is described in the notes. - **CLI port names.** The data sheet gives counts, not names, so ports are numbered positionally (copper first, uplinks last) and the notes say exactly that. - **Serial console** is RS-232 on every model. Per `docs/DEVICE_SCHEMA.md` ("Serial consoles, USB, FXS go in notes, not in interfaces") it stays prose. ## Validation - `node scripts/syntax-check.js` - 328 files, all OK - `node scripts/audit-port-types.js` - "No port contradicts its declared type." - `node scripts/audit-devices.js` - "No hard defects." - `node scripts/import-devices.js` - 5 added, 0 rejected, 0 refused spellings; the only warnings are the documented keep-as-declared ones (`802.3bt` is a real power fact with no engine token) - `node test/importer-roundtrip.test.js` - 26 passed - `node test/catalogue-hygiene.test.js` - 25 passed; `node test/doc-catalog-counts.test.js` - 24 passed - `node test/spec-vs-description.test.js` - **2 passed, 1 failed, and the failure is not in this branch**: `edgecore/AS4610-30P|30T|54P|54T` each "claims 2x 20G QSFP+, 0 port(s) can do it", from the merged #1555. That description names two 20G stacking cages that the spec deliberately does not model as ports; the check reads the phrase as a port claim. It is red on `main` and needs a reword there (or a `KNOWN` entry), which is an edgecore-scope change and not made here. Every field carries the URL it came from: the cnMatrix EX2000 data sheet (`brandcentral.cambiumnetworks.com/.../cnMatrix-EX2000-Data-Sheet.pdf`) plus the per-model page on cambiumnetworks.com.
Cambium cnMatrix EX2000: five switches, a 400 W budget that binds at 30 W a port, and 60 W 802.3bt only on the multi-gig model
Some checks failed
CI / Classify the change (pull_request) Successful in 19s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 24s
CI / Docker build (pull_request) Failing after 12m6s
CI / Server smoke test (pull_request) Failing after 12m10s
CI / Test (Node 22) (pull_request) Failing after 12m13s
CI / Test (Node 20) (pull_request) Failing after 12m18s
CI / Test (Node 18) (pull_request) Failing after 12m22s
39ebaec775
The catalogue had Cambium radios and no Cambium switch. The cnMatrix EX2000
series is five boxes on one data sheet: EX2010-P (8 PoE + 2 SFP, 100 W, 20
Gbps, fanless), EX2016M-P (8x 1G + 6x 2.5G, 240 W, 60 W 802.3bt on the
multi-gig ports), EX2028 (24 + 4 SFP+, no PoE), EX2028-P (the same with
400 W) and EX2052-P (48 + 4 SFP+, 400 W, 176 Gbps).

Every field cites the cnMatrix EX2000 data sheet and the model page. The
400 W budget on the 48-port box is the constraint that matters: 48 ports at
30 W is 1440 W of port capability, so the supply, not the port rating, is
what binds - the note says so rather than letting the sum look compliant.

Not modelled, and said in the notes: the out-of-band management port (the
data sheet states its rate only for the EX2028-P), and the CLI port names
(numbering is positional and the data sheet prints no name).

The serial console, RS-232 on every model, is recorded in notes: a console
is a fact about the box, not a link it can carry traffic on.
jaylouisw force-pushed catalogue/1116-0126 from 39ebaec775
Some checks failed
CI / Classify the change (pull_request) Successful in 19s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 24s
CI / Docker build (pull_request) Failing after 12m6s
CI / Server smoke test (pull_request) Failing after 12m10s
CI / Test (Node 22) (pull_request) Failing after 12m13s
CI / Test (Node 20) (pull_request) Failing after 12m18s
CI / Test (Node 18) (pull_request) Failing after 12m22s
to 6bf7b73a6c
Some checks failed
CI / Classify the change (pull_request) Successful in 19s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 21s
CI / Test (Node 22) (pull_request) Successful in 3m20s
CI / Docker build (pull_request) Failing after 11m44s
CI / Server smoke test (pull_request) Failing after 11m50s
CI / Test (Node 20) (pull_request) Failing after 11m59s
CI / Test (Node 18) (pull_request) Failing after 12m2s
2026-11-16 01:41:09 +00:00
Compare
jaylouisw deleted branch catalogue/1116-0126 2026-11-16 01:41:30 +00:00
Author
Owner

This branch conflicts with main, so it was not rebased:

devices/cambium/ex2010-p.json devices/cambium/ex2016m-p.json devices/cambium/ex2028-p.json devices/cambium/ex2028.json devices/cambium/ex2052-p.json docs/DEVICE_REGISTRY.md research/incoming/devices-cambium-cnmatrix.json 

A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.

This branch conflicts with `main`, so it was not rebased: ``` devices/cambium/ex2010-p.json devices/cambium/ex2016m-p.json devices/cambium/ex2028-p.json devices/cambium/ex2028.json devices/cambium/ex2052-p.json docs/DEVICE_REGISTRY.md research/incoming/devices-cambium-cnmatrix.json ``` A conflict is a decision about behaviour rather than a mechanical fix, so this is left for a human. Rebasing by hand and force-pushing will let the automation pick it up again.
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!1580
No description provided.