Troubleshooting
Why an HMI sends email in simulation but not from the panel
Short answer
Email that leaves from PC simulation but never from the physical panel means the project logic works and the panel's own send path is broken. Work the panel's chain in order: count and send condition, gateway, DNS and mail-server reachability, then the deployed project's email options, changing one option at a time. In the case behind this pattern, one checked option in the panel's email settings was the entire fault.
On this page
Email that leaves from PC simulation but never from the physical panel tells you three things at once: the project can build a valid message, the mail service accepts it, and the panel's own send path is broken. What differs on the panel is its network route out (gateway, DNS, mail-server reachability), the mail settings as the runtime applies them, and options that act differently in runtime than in simulation. Test them in that order, one change at a time. In the installation behind this pattern — an Ethernet-only Weintek MT8102iE reached through a wireless extender while two Wi-Fi MT8103iE panels on the same mail setup kept working — the entire fault was one checked option in the panel's email settings.
Separate the count, the send request, and the delivery #
An empty inbox merges three events that have to be tested apart: the machine count changes and the panel reads it; the project's send condition fires a mail request; the SMTP transaction completes. Only the third event is "email", and the first two can fail while every mail setting is correct.
| What you observed | What it proves | What it leaves open |
|---|---|---|
| Simulation on the PC sends | The project builds a valid message over the PC's route | The panel's route and runtime behavior |
| A PC behind the same extender sends | The extender bridges a PC's traffic correctly | The panel's addressing and its project options |
| You can open the panel over the network | An inbound path to the panel exists | Any outbound traffic from the panel |
| Two Wi-Fi panels send on the same account | The mail service, account and recipients work | The settings deployed to this panel |
Inbound reachability never proves outbound access; the same one-way rule applies whenever the panel initiates traffic, as when an HMI sends SMS through an HTTP GET modem. And a feature that works on the PC but dies on the panel is a known HMI failure class — WAV event sounds that play in simulation but not on the HMI panel cross the same boundary, because the PC supplies resources the panel must provide itself.
Replacing the extender with a new unit and seeing no change already did useful work here: the extender hardware left the suspect list, leaving the panel's addressing and the deployed project.
Prove the panel itself reaches the send request #
- Back up the project from the panel and write down its IP address, subnet mask, gateway and DNS values before changing anything.
- Confirm the project running on the panel is the one you simulate. If there is any doubt, pull the panel's current project and test that copy; a simulation of an edited file proves nothing about what is deployed.
- Compare the part count shown on the panel with the controller's value. If they disagree, fix the driver, station address or tag before touching mail settings.
- Establish the trigger type. An edge-triggered send needs the condition to drop and rise again; a level-triggered send needs the maintained state; a scheduled send needs the panel clock. A send armed once and never re-armed looks exactly like a dead mail path.
- Force one controlled send and watch the send-status indication or diagnostic the runtime provides. If the email settings include a test function, run it from the panel, never from the PC.
Decision at this point: a wrong count points at the controller link; a correct count with no send request points at trigger logic or object binding; a send request followed by an error points downstream — continue. And before trusting any simulated result at all, spend the effort on testing the simulated project against the live PLC, because a simulation feeding itself test data proves nothing about the panel's tags.
Test the panel's route to the mail server #
Illustrative addresses, real structure of the check: panel 192.168.1.50, mask 255.255.255.0, workstation 192.168.1.100, mail server 203.0.113.10.
- 192.168.1.50 AND 255.255.255.0 = 192.168.1.0
- 192.168.1.100 AND 255.255.255.0 = 192.168.1.0
- 203.0.113.10 AND 255.255.255.0 = 203.0.113.0
The workstation and the panel share subnet 192.168.1.0, so reaching the panel from the PC needs no gateway: upload, remote access and browsing all work with the gateway field empty. The mail server sits on 203.0.113.0, a different subnet, so the panel must hand that connection to its default gateway. In this example the gateway field must read 192.168.1.1, the extender's LAN address; a blank or 0.0.0.0 gateway strands the panel on its own subnet and produces exactly this symptom split — inbound fine, outbound dead.
Name resolution is likewise the panel's job: the mail hostname must resolve with the DNS the panel holds, not the DNS your PC uses. On ports, RFC 5321 defines SMTP and RFC 8314 requires TLS for mail submission — port 587 with STARTTLS or port 465 with implicit TLS — leaving port 25 to server-to-server relay, which client networks commonly block outright. Set the panel to the port and encryption the mail service documents, and confirm that port is permitted outbound from the plant network.
Compare the mail settings that exist only at runtime #
The deployed project, not your simulation copy, is what sends. Check, in this order:
- Credentials: a project copied to the panel without username and password cannot authenticate, while the PC you simulate on may hold working credentials the panel never had.
- Sender policy: many services accept mail only from a registered sender address, and the deployed project must carry it.
- Encryption: none, SSL/TLS and STARTTLS are separate settings, and the choice must match what the server offers on that exact port.
- Clock: TLS presents an X.509 certificate valid only between its not-before and not-after times, so a panel date off by a single day fails the handshake before authentication starts.
- Runtime-only options: settings that act differently once the project runs are a recognized HMI trap — HMI user management that works in simulation but fails on the panel crosses the same boundary.
In Weintek's EasyBuilder Pro the mail account and send-event configuration belong to the project's system settings; confirm the exact page and option names in the manual for your installed version rather than from memory.
Find the one option that breaks sending #
In the reported installation the fix was unchecking a single option in the email settings. Nobody recorded its exact label, so treat the search method — not a remembered checkbox name — as the deliverable:
- Save a baseline copy of the project and capture the current email settings screens.
- List every option whose documented function can affect whether or how a message is sent.
- Change exactly one option, download to the panel, and repeat the same controlled send test.
- Record the option label, its old state, its new state and the result; revert anything that changes nothing.
- Stop at the first option whose change makes the panel send — that option, in its old state, was the fault.
Each cycle eliminates exactly one suspect, so the search converges, and one deploy-and-test cycle per option is the whole cost. Do not disable authentication or encryption to make a test pass; if the panel cannot speak the security the service requires, put an approved internal relay in front of it instead.
Decide by what you observe #
| Observation | Fault location | Action |
|---|---|---|
| Panel count disagrees with the controller | Controller link, driver or tag | Fix addressing before any mail work |
| Count correct, no send request fires | Trigger logic or object binding | Correct edge versus level, rebind the event |
| Send request, then a connection error | Gateway, DNS, port or clock | Set the gateway, resolve the host, use 587 or 465 with TLS, set the clock |
| Connection opens, then rejection | Authentication or sender policy | Deploy real credentials and the registered sender |
| Everything verified, still silent | Runtime-only email option | Run the one-option-at-a-time search |
Start at the row that matches what you observed and stop when the observation changes. In the installation that produced this article, that search ended on one checked option, and clearing it restored delivery with the network untouched.
Questions asked
Does a successful send from the engineering PC prove the mail server accepts the panel?
It proves the mail server, the account and the credentials are good, and nothing about the panel. The PC uses its own gateway, DNS resolver and TLS stack, so its success narrows the fault to the panel side: addressing, deployed mail settings, or a runtime-only option. Repeat the same test from the panel before touching anything on the server.
Why do my Wi-Fi panels send email while the Ethernet panel behind an extender does not?
A Wi-Fi panel associates directly with the access network and receives its addressing there. An Ethernet panel behind an extender depends on the extender bridging correctly, on the gateway and DNS entered in the panel, and on any upstream rule that treats wired clients differently. Swapping the extender for a new unit with no change, as happened here, removes the hardware from the suspect list and points at the panel's settings.
Can the panel's clock stop email from sending?
Yes, when the connection uses TLS. A certificate is valid only between its not-before and not-after timestamps, so a panel clock wrong by a day or more fails validation and the connection drops before authentication. Set the panel date and time, or feed it from a controller clock, and retest before changing any mail setting.
Do I need DNS on the panel if I configure the mail server by IP address?
No. A numeric address bypasses name resolution entirely, which makes it a legitimate isolation step: if email leaves with the IP address configured and fails with the hostname, DNS on the panel is the fault. Prefer the hostname in production, because providers move their server addresses.