The turbine is spinning at full load and the control room calls: the overspeed alarm just tripped, but the tachometer on the panel says everything is normal. You walk to the SpeedSys T20 and see the RPM digits jumping ±15% every few seconds. Is it a bad speed sensor, a loose connector, or something in the relay logic? The ISTEC SpeedSys T20 is a dual-channel speed monitor designed for SIL-capable overspeed protection on rotating machinery — but when the display won't settle, the entire safety loop is in question.
The T20 accepts pulse inputs from two independent speed sensors (proximity probes, Hall-effect pickups, or optical encoders) and compares them against programmable overspeed and underspeed setpoints. Each channel drives a dedicated relay output, and both channels report over Modbus RTU to the DCS or PLC. When the reading is unstable, the problem almost always lives in one of four places: the sensor gap, the wiring path, the power supply, or the alarm latch logic.
Sensor Gap and Alignment Check
The most common cause of RPM fluctuation on the T20 is an incorrect proximity probe gap. If the gap between the probe tip and the gear tooth is too wide, the pulse amplitude drops below the T20's trigger threshold and the channel counts intermittently — producing a reading that jumps between the true speed and half-speed or zero. If the gap is too tight, thermal expansion at operating temperature can close the gap to zero, saturating the sensor and producing a flatline.
- Target gap
- Set per the sensor manufacturer's datasheet. For a typical M8 inductive probe on a 60-tooth gear, the gap is 1.0–1.5 mm. Use a non-metallic feeler gauge — a steel gauge on a magnetic pickup will give you a false reading.
- Runout check
- Rotate the shaft by hand and measure gap variation across all teeth. Total indicated runout exceeding 0.25 mm can cause channel-to-channel mismatch alarms on the T20.
- Sensor mounting rigidity
- A bracket that resonates at the machine's operating speed will amplitude-modulate the pulse train. If the RPM fluctuation frequency matches the running speed or a multiple of it, the bracket is suspect.
Wiring and Noise Rejection
Speed sensor signals are low-voltage pulse trains traveling from the machine frame to the control cabinet — often through the same cable tray as VFD motor leads. The T20's inputs are differential and include a configurable low-pass filter, but a poorly grounded shield or a damaged twisted pair can still introduce enough common-mode noise to produce false pulse counts.
- Verify the sensor cable shield is grounded at one end only — the T20 terminal side. Grounding at both ends creates a ground loop that injects 50/60 Hz noise directly onto the signal pair.
- Check the cable type: the T20 pulse input expects twisted-pair, shielded instrumentation cable (0.25–1.5 mm²). Substituting unshielded tray cable is a common field mistake that produces intermittent counting.
- If the T20 is installed near a VFD, enable the onboard low-pass filter via the Modbus configuration register. Start at a 1 kHz cutoff for a 60-tooth gear at 3000 RPM (generating a 3 kHz pulse frequency) and decrease if noise persists.
- Measure DC resistance across the sensor terminals with the sensor disconnected. A reading that drifts while you wiggle the cable indicates an intermittent conductor break inside the jacket.
Power Supply Stability
The T20 is powered by 24 VDC. Supply ripple exceeding 200 mV peak-to-peak can cause the internal threshold comparator to false-trigger — especially when combined with a marginal sensor gap. Check the 24 V rail with an oscilloscope, not a multimeter. A switching power supply that shares a bus with solenoid valves or motor contactors may droop below the T20's minimum 18 VDC during load switching, producing a momentary RPM dropout that self-recovers faster than the alarm delay, leaving no record except an unexplained Modbus diagnostic counter increment.
On one ethylene compressor train, an intermittent T20 overspeed trip was traced to a 24 V power supply that sagged to 16 V for 80 ms every time the lube oil pump contactor pulled in. The T20 brownout reset took 200 ms — longer than the 100 ms alarm delay — so every pump start generated a spurious trip. The fix was a dedicated 24 V supply for the overspeed loop, isolated from the MCC bus.
What do the Modbus diagnostic registers tell me about the RPM fault?
The T20 exposes a set of diagnostic registers over Modbus RTU that are more useful for root-cause analysis than the front-panel display. Register 0x0100 holds the pulse interval measurement for Channel A in microseconds; register 0x0101 holds Channel B. If one channel shows a stable interval while the other jumps, the problem is isolated to that channel's sensor or wiring. Register 0x0200 counts the number of "pulse dropouts" — intervals where no pulse was detected within the timeout window. A non-zero count that increments during operation confirms an intermittent signal loss before it triggers an alarm. Register 0x0202 stores the last overspeed trip RPM value, so even if the trip was transient and self-clearing, you can read the captured speed and compare it against the setpoint to rule out nuisance trips caused by setpoint drift.
How does the T20 handle a sensor failure — will it trip or will it silently stop monitoring?
The T20's response to a sensor failure depends on the configured fail-safe mode. In de-energize-to-trip mode (default), a loss of pulse input for longer than the timeout period forces the relay to the tripped (de-energized) state — the machine shuts down. In energize-to-trip mode, a sensor failure holds the relay energized (non-tripped), which means the machine keeps running without overspeed protection. The second mode is sometimes chosen for machines where a spurious trip carries unacceptable cost, but the trade-off is explicit: the T20's Channel Status LED will flash the fault code, Modbus will report the dropout, but no automatic shutdown occurs. The decision between these modes should be documented in the safety requirement specification, not left to the commissioning technician's preference.
Can I replace a single-channel speed relay with the T20 without changing the sensor setup?
In most cases, yes — the T20 accepts the same pulse input types as legacy single-channel relays (NAMUR, NPN, PNP, and passive magnetic pickup). The input impedance is configurable via DIP switch on the backplane. However, the dual-channel architecture means you must either connect two physical sensors or configure the second channel as "unused" in the Modbus setup to avoid a permanent Channel B fault. If you are upgrading from a single-channel relay and the machine has only one speed sensor, order the T20 with the single-channel firmware option (field-configurable) and wire the sensor to Channel A. Adding a second sensor later requires only mechanical mounting and a cable run — no hardware replacement of the T20 itself. For a broader look at comparable overspeed monitors, the SpeedSys T30 adds a third channel and a 4–20 mA analog output that tracks RPM continuously, useful for trending in a historian alongside the Modbus data.
