PLC restarts when solenoid valves switch: suppress and ground first

What each symptom points to

Read the controller first, before you touch anything. Its status LEDs and its outputs after an event tell you whether it lost power or only lost the conversation.

• Full restart, machine re-initializes. The controller's 24 V feed sagged or its ground reference shifted hard enough to reset it. Note which valve fired at that instant, then check the supply that feeds the controller and how its negative is bonded.
• Communication timeout with the PLC still running. The controller never stopped. The tell is outputs frozen at their last state while the panel alarms, which is the case the fix for an HMI communication timeout when outputs hold state walks through. The disturbance hit the communication path or the shared reference.
• Blank HMI screen. The HMI's own feed dipped, or it is finding its return through the communication cable because nothing bonds the DC negatives. Start with power at the HMI, as checking an HMI that goes blank and freezes lays out, then do the bonding below.

One valve switching can produce any of the three, depending on which path the spike finds and how hard it hits. Suppress the coil and bond the references, and all three usually stop together.

Why relay outputs do not protect you

A relay output interrupts the coil current at the contact, exactly where a transistor output would. The field in the coil collapses, and because current through an inductance cannot change instantaneously, the voltage across the opening contact climbs until something clamps it. Wiring guidance for small relay-output PLCs puts a number on the unclamped case: the spike can run well past 1000 V. That is enough to erode the contacts and to couple into every parallel wire in the same wire trough. PLC installation manuals carry the same warning in practical terms: high-frequency noise from poorly suppressed inductive loads disrupts controller operation. Relay or transistor, the coil behaves the same.

Suppress each coil at its own terminals

Fit a TVS diode directly across each 24 VDC valve and contactor coil, at the coil's terminals. Not at the PLC terminal strip and not at the supply: a suppressor mounted away from the coil leaves the whole cable run exposed to the spike. A bidirectional part suits a 24 V rail with no polarity decision to get wrong; with a polarized part, the cathode lands on the positive side. Size the standoff voltage for the rail and check the clamp voltage against what the contact and wiring withstand. Both numbers are on the suppressor datasheet, the coil datasheet prints the coil's own rating, and nothing about the supply's wattage selects the suppressor.

A plain diode across the coil also protects, but it clamps at a single diode drop and lets the coil current crawl to zero, so the valve releases slowly. A TVS absorbs the energy quickly and drops the valve out promptly. Pick one method that suits the release time the machine needs, from the coil and suppressor datasheets. For the clamp physics worked through with real numbers, this guide to TVS voltage-spike suppression is worth the read.

Bond every DC negative and the controller jumper to one stud

Isolate and lock off the panel before you open anything; the 120 V branch in that trough is lethal. Many DIN-rail supplies float their 24 V output, so treat each negative as floating until you bond it. With power off, run a conductor from each supply's negative output terminal to one panel ground stud, and land the controller's ground there as well. Check the installation manual for the controller in the panel: this class of small PLC documents a metal jumper between its two ground terminals, and the wiring chapter sends that point to the panel ground. One stud, star arrangement, short conductors.

The bond gives transient current a defined local return. With floating negatives, the communication cable becomes the return path between equipment sitting at different potentials, and the disturbance announces itself as timeouts on the panel. If the timeouts come and go as valves fire, that is the pattern behind why an HMI shows an intermittent PLC communication timeout. If any loop's own documentation requires a floating reference, follow that manual for that loop and bond the rest.

A controller bolted to a grounded backplate is not grounded by the bolts; the conductor does that work. And if this job comes with a new screen to fit in the door, the flush and panel mount guide gives the real cut-out sizes before you cut.

Worked example: the 7.5 A branch carrying 180 W of 24 VDC supplies

Take the branch as built: 120 V on its own breaker, fused down to 7.5 A, through a 5 A power-line filter, feeding a 60 W supply for the controller and analog sensors and a 120 W supply for the valves and contactors, all at 24 VDC. Two supply swaps and a move of the valve supply to its own 120 V circuit have already left the fault in place. Assume 85% efficiency, typical for small DIN-rail switchers; the datasheets print the real figures.

The rule: primary-side current equals total output power divided by branch voltage times efficiency.

• Total output: 60 W + 120 W = 180 W.
• Primary current: 180 W ÷ (120 V × 0.85) = 180 ÷ 102 = 1.76 A, before the HMI's adapter adds its share.
• Against the filter: 1.76 A ÷ 5 A = 35% of its rating.
• Against the fuse: 1.76 A ÷ 7.5 A = 23% of its rating.
• On the DC side, the valve supply holds 120 W ÷ 24 V = 5 A for the coils.

Result: the branch runs at about 1.8 A steady, a third of the filter and a quarter of the fuse. Nothing on the 120 V side is heat-stressed, and a fuse clears on accumulated heating, i²t, which a sub-millisecond spike does not deliver. So the arithmetic rules out steady overload as the weak link and says another supply swap attacks the wrong problem. The mechanism that survives this math is the unsuppressed coil and the floating negatives. The branch as drawn is otherwise sound: the valves already run on their own supply, which is the arrangement sizing a separate 24 VDC supply for HMI inrush current applies when a panel needs one.

How to check the fix held

1. Power down and isolate. Meter continuity from each supply's negative terminal to the ground stud, and from the controller's jumpered ground point to the same stud. Both should read near zero ohms.
2. Fire one valve at a time, by hand, and watch the HMI and the controller's status at each firing.
3. Run the panel's normal sequence for a full cycle, then for a full shift.
4. If a glitch still lands, capture it with a scope or a transient recorder across the controller's supply terminals; a handheld meter will not see an event this fast.
5. After any panel rebuild, confirm every suppressor still sits at its coil terminals. Devices migrate to terminal strips during maintenance.

Recommendation: In a panel where valve switching produces restarts, blank screens or timeouts, fit a bidirectional TVS across every 24 VDC valve and contactor coil at the coil terminals, and bond each supply's negative output plus the controller's jumpered ground point to one panel stud, before you swap or upsize any supply. Do the bonding with the panel isolated and lockout applied, then re-test one valve at a time. If your valves share the controller's supply instead of running on their own, the same suppression and bonding apply with more urgency; the shared version of this fault is why a shared DC supply with solenoid coils freezes an HMI.

More shop-floor screen faults, and the screens built for those walls, sit in the shop-floor screens and andon boards section.

Questions