Short answer: treat the ethernet port on an HMI as a view port, not a control port. The web feature most panel HMIs expose rebuilds an image of the current screen each time a browser asks for it, so a remote user sees status and presses nothing. Run the numbers: with a 5 s page reload, a 1 s load and a 2 s reaction, what the remote viewer acts on is up to 8 s old. That serves a walkthrough check of door or machine status; it cannot carry a command. Keep commands on the local touchscreen, put any real remote control on a documented supervisory system, and bring the cable in through a rated gland so the port adds no leak.
What the ethernet port actually gives you
The port is a network connection. What decides view versus control is what the panel's software exposes on it, and HMIs fall into three classes:
- Image on request. The web server rebuilds a picture of the displayed screen each time a request for it arrives. The refresh happens per download, not on a clock and not as live video, so a page nobody is asking for shows nothing new.
- Project web pages. Screens you build for a browser can read tags, counts and alarms. They carry commands only where the vendor documents a write, and that documented write is the feature to order, not something an ethernet listing implies.
- A supported remote session. Vendor tooling over VPN or VNC that mirrors the screen and passes touch into the local project. This is the class that can operate a machine, and it is a product feature with accounts and licenses to plan.
An ethernet line on the datasheet says only that the panel talks on a network. For loading a project rather than running one, a sealed service port answers a different question; see remote USB programming access at an HMI enclosure.
The screen-change test that settles it
Before you promise the customer anything, test the actual panel on an isolated network:
- Open the documented web page and confirm it shows the intended screen.
- Change a nonhazardous value on the local screen, such as a pilot-light tag on a test page.
- Fetch the image again, timing both fetches with a stopwatch.
- Compare: when the second image shows the change, the feature is view-only with refresh on request.
- Press a control in the browser image. When nothing moves at the panel, the page carries no touch.
Two results matter. The screen state at the panel is always current; the remote copy is only as new as the last fetch, so a change waits up to one full reload interval before it appears. And a healthy PLC-to-PLC link proves the network, not the web feature: master/slave exchange between controllers rides its own protocol and says nothing about what a browser can do.
The staleness math for a door-status screen with a 5 s reload
Case: door open/closed indicators on a 15.6 in panel PC in a stainless HMI housing, viewed from an office over the plant network.
| Input | Value | Where it comes from |
|---|---|---|
| Page reload interval | 5 s | example input: whoever builds the page sets it, usually with a refresh tag in the page |
| Page and image load | 1 s | example input: includes the server rebuilding the image; time it with the stopwatch |
| Viewer reaction | 2 s | example input: from the screen updating to acting on it |
| Web feature | image rebuilt on each request | the panel class this article is about |
The rule in words: a change can land just after a fetch, so it waits one full reload interval, then the page loads, then the viewer reacts. Worst-case delay = reload interval + load time + reaction time.
The arithmetic: 5 s + 1 s + 2 s = 8 s.
Result: the status the remote viewer acts on is at most about 8 s old at these inputs. Stretch the reload to 10 s and the worst case grows to about 13 s (10 s + 1 s + 2 s).
Decision: a task that tolerates 10 s of stale status passes on this path; anything that needs a command within 1-2 s of a decision fails it. Door and machine commands stay on the local touchscreen.
Where remote control belongs instead
When the customer's list includes commands, order an architecture that documents them instead of stretching the image feature:
- Engineering screens: a vendor remote-access session over VPN, with named accounts, gives an engineer the real touchscreen from a laptop. Plan the accounts like any other tool.
- Operations commands: a plant SCADA or supervisory node with its own authenticated command path, recorded commands and alarms. The HMI port stays a view port beside it.
- Network shape: put the status page on a plant VLAN separated from the control network, and give viewers routes to the status server only. A status network with no route to PLC write services cannot become a command path by accident.
The local touchscreen stays the command point, wired to the outputs. For guarded equipment such as secure-facility door controls, treat that as a hard rule: no browser page becomes an unofficial way to move a door.
Keeping the rating when the cable enters the box
The port adds no leak; the hole for the cord does. Bring the ethernet cable in through a gland sized to the cord: a typical Cat6 patch cord runs about 6 mm OD by cable-maker datasheets, and shielded or outdoor cords run nearer 8 mm, which puts most builds on an M20 gland in a 20 mm hole. The gland's datasheet lists the exact OD range it clamps, so match it before you drill.
The weakest part sets the assembly rating. A gland rated below the box drops the whole enclosure to the gland's rating, so on a manufacturer-rated Type 4X box use a gland the gland maker rates IP66 for that cord diameter, with its sealing washer seated on a flat surface. IP66 is the second digit 6 in IEC 60529: a 12.5 mm nozzle at 100 L/min. Our HMI housings are 304 stainless with a gasketed rear cover and VESA 100 mounting; the front seal is IP65-type with the display installed. The rating is the manufacturer's, not UL listed.
Cutout and gland fundamentals for a display build are in the cutouts, glands and HMI windows guide, hole sizes per gland in the M20 gland hole size article, and keeping data ports sealed in how to keep USB and ethernet ports waterproof on an enclosure.
What to put on the spec
| Spec line | What to write |
|---|---|
| Command point | local touchscreen, wired to the outputs |
| Remote status | image-on-request web page, proven with the screen-change test |
| Staleness budget | 8 s worst case at a 5 s reload; set the interval to match the task |
| Cable entry | M20 gland, 20 mm hole, IP66-rated for the cord, washer on a flat surface |
| Network | status VLAN with no route to PLC write services; VPN with named accounts for engineering |
| Housing | 15.6 in HMI housing, 304 stainless, gasketed rear cover, VESA 100 rear; delivered in 10 business days from your order, or from drawing approval for a custom build |
Then decide by case:
- Status only — door position, tank level, alarm count: order the image-on-request port, carry the 8 s budget into the handover, and seal the entry with a rated gland.
- Status plus occasional tuning — add a vendor remote-access session over VPN with named accounts; commands on guarded equipment stay local.
- True remote operation — a documented supervisory system with its own command path and records; the HMI's ethernet port stays a view port.
Housings run from 7 in to 23.8 in display sizes, so match the shell to the screen rather than the other way round. For one machine and one screen, a machine-mounted HMI enclosure keeps the cutout, the rating and the display matched from the start, and 316L housings for harsh sites cover the washdown and brine corners where 304 falls short.
Catalog dimensions to compare
Compare these catalog dimensions with your component layout. Open each enclosure for its rating, material and usable space, then check those specifications against your application.
| Enclosure | Material | Outside H x W x D / display | Rating |
|---|---|---|---|
| 15.6" HMI enclosure | 304 stainless | 11.69 x 17.64 x 4.13 in | IP65-type front seal with the display installed, manufacturer rated; not UL listed |
| 7" HMI enclosure | 304 stainless | 7.43 x 9.96 x 3.7 in | IP65-type front seal with the display installed, manufacturer rated; not UL listed |
| 23.8" HMI enclosure | 304 stainless | 15.91 x 24.96 x 4.02 in | IP65-type front seal with the display installed, manufacturer rated; not UL listed |
| 10 × 8 × 6 in | 304 stainless | 10 x 8 x 6 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
Questions
Can a browser press the touchscreen through the HMI's ethernet port?
Only where the vendor documents an interactive remote session, such as supported VNC or VPN tooling. The common web feature rebuilds a screen image on each request, so nothing gets pressed. Budget its staleness: at a 5 s reload, a 1 s load and a 2 s reaction, status is up to 8 s old — right for status, wrong for commands.
What gland keeps the rating on an ethernet cable entry?
On a manufacturer-rated Type 4X box, fit a gland the gland maker rates IP66 for that cord diameter, with the sealing washer seated on a flat surface. A gland rated below the box drops the assembly to its rating. An M20 gland fits a 20 mm hole and clamps cords near 6-8 mm OD. The rating is the manufacturer's; these housings are not UL listed.
Does a working PLC-to-PLC link mean the panel can be operated remotely?
No. Master/slave data exchange between controllers rides its own protocol and proves the network, not the web feature. A browser request rides the panel's web server, which most often returns a rebuilt screen image. Test the two separately, and set expectations from the test: at a 5 s reload, worst-case staleness on the image path is 8 s, and no press ever reaches the touchscreen.
Where do remote door and machine commands belong?
On a documented supervisory system with its own authenticated command path: named user accounts, recorded commands and alarms, and the local touchscreen still wired as the command point. For guarded equipment such as secure-facility doors, keep the browser page read-only: put it on a status VLAN with no route to PLC write services. An image-on-request page carries no touch events at all.
