Industrial Ethernet Certification Test: The Parameters That Matter
Quick answer / TL;DR: Short answer: certify every link rather than sampling, because industrial links are short and site-terminated, so a failure is not spread evenly. Record insertion loss, return loss, pair-to-pair crosstalk, propagation delay and delay skew. Use the correct adapter for the connector on each end, and calibrate the tester before the first link.
Two tiers apart: 100 MHz versus 500 MHz, gigabit versus 10G.A certifier produces a page of numbers and a pass or fail. On an industrial site the interesting question is which of those numbers will change when the machine starts running, and which are already fixed by the installation geometry.
Industrial links differ from office links in three ways: they are short, they are often terminated on site, and they are frequently disturbed. Each of those changes what a test proves.
This article covers the parameters worth recording, the adapter problem, and the sampling rule that decides how many links to test.
A certification test compares measured values against the limits for a category at a given length. It proves the channel meets the standard on the day it was tested.
What it does not prove is mechanical durability, shield effectiveness under load, or behaviour after the cable has been flexed a thousand times. Those are properties of the construction, not of the measurement.
Parameters Worth Recording
| Parameter | What it indicates | When it fails after installation |
|---|
| Insertion loss | Attenuation over the channel | Longer route than designed, or damaged conductor |
| Return loss | Impedance consistency | Tight bend, crushed cable, poor termination |
| Pair-to-pair crosstalk | Coupling between pairs | Untwisted length at a termination |
| Propagation delay | Signal travel time | Route longer than the record shows |
| Delay skew | Difference between pairs | One pair terminated differently |
Return loss is the parameter that catches mechanical damage, which is why it is the first number to read when a link that passed at commissioning now fails.
Delay skew matters more in industrial installations than in offices, because a machine may use the same link for a protocol with tight timing rather than for general traffic.
The Adapter Problem
- Test through the correct adapter for each connector, since an M12-to-RJ45 adaptor measures the adaptor as part of the link
- Account for the adapter's own contribution, because it adds insertion loss and reflections
- Verify the adapter before a campaign, because a damaged adaptor fails every link in the same way
- Record which adapter was used, so a later retest is comparable
Real-World Applications
A machine builder certifies every link on the bench before shipping, using the same adapters that will be used on site. The results travel with the machine as the reference state.
An integrator certifies at handover and finds three links with marginal return loss. All three share a termination style that the installer used on the shortest runs, where the bend was tightest.
A plant retests after a shutdown and compares against commissioning results. Two links have degraded, and both are in a tray that was walked on during the work.
A vision system installation tests every camera link at full frame rate after certification, because a link that passes electrically can still drop frames when the switch is oversubscribed.
Standards and Verification
Channel and permanent link definitions come from ISO/IEC 11801 and TIA-568, with industrial installation practice in IEC 61918. Test equipment accuracy is governed by the tester manufacturer's calibration, which should be current before a campaign.
Record the full report rather than the pass mark. The pass mark is a conclusion; the numbers are the evidence, and the evidence is what supports a comparison later.
Retest after any work that disturbs a route, and compare against the original rather than against the limit. A link that moved from comfortable margin to marginal is a finding even if it still passes.
What the Test Cannot Tell You
A certification proves the channel on the day. It says nothing about the application running on it.
A link can pass at full margin and still drop frames because the switch uplink is oversubscribed, or because a protocol expects jumbo frames that the path does not carry.
Treat certification and application testing as two separate gates. Passing the first is necessary and not sufficient.
Keeping Results Usable
Name each result by the machine and the device tag rather than by the cable number. Maintenance looks up results by device.
Keep the adapter used in the record. A retest with a different adapter produces numbers that cannot be compared.
Store the report with the as-built drawing, because the two are read together and separately they are both ambiguous.
Where a retest is commissioned after a fault, test the whole route rather than the suspect link. A cable that was walked on affects its neighbours, and the failing link is often the one that moved least.
That approach costs an hour and prevents the second visit that a single-link fix usually generates.
Related reading: Ethernet cable separation from power — blogs/news/ethernet-cable-separation-from-power
Related reading: EtherCAT diagnostics — blogs/news/ethercat-diagnostics
Related reading: Cat6A ethernet cable guide — blogs/news/cat6a-ethernet-cable-guide
A pass is not a baseline. Keep the numbers, because the next test will be compared against them and not against the standard.
Frequently Asked Questions
Should I certify every industrial link or sample?
Certify every link. Industrial runs are short and often site-terminated, so failures cluster rather than distribute evenly, which makes sampling unreliable.
Which parameter catches mechanical damage?
Return loss. A crushed or tightly bent cable changes impedance, which shows up as a return loss failure before anything else degrades.
Why does my M12 link fail certification when the cable is good?
Usually the adapter. An M12-to-RJ45 adaptor contributes its own loss and reflections, and a damaged adaptor fails every link identically.
Does a passing certification guarantee the application will work?
No. It proves the channel meets the category. Application-level issues such as bandwidth oversubscription or protocol timing are separate tests.
Sourcing cables for a project, a store, or a private-label line? Contact our B2B sales team for specifications, certification evidence, and volume terms – FOB and CIF options available. Request a quote or sample.
