The spec sheet rarely tells the full story
Accuracy and repeatability get top billing on every flow meter datasheet — and rightly so, since they define how close a reading is to the true value. But neither number says anything about when that true value is captured. A meter can be perfectly accurate at steady flow and still miss what's actually happening in a dynamic process.
That gap is response time, and it's the spec most clamp-on ultrasonic flow meters leave off the page entirely.
What response time actually measures
Response time is expressed as a time constant, τ: the time it takes a sensor's output to reach 63.2% of its new value after a step change in flow. This follows standard first-order system behavior:
V(t) = V_final × (1 − e^(−t/τ))
Settling to 99% of the final value takes roughly 5τ. So a sensor specified at τ = 1 second doesn't fully stabilize in one second — it takes closer to five.
This matters because almost no industrial process runs at constant flow. Pumps start and stop. Valves open and close. Recipes change mid-batch. Each of these is a step change in flow, and the meter's response time determines whether that change shows up in the data — or gets quietly averaged away.
What happens when response time is too slow
A slow-responding meter doesn't just delay the signal — it can actively distort the picture of the process:
Transients disappear from the data. A 300-millisecond pump surge is invisible to a sensor with τ = 5s. The event happened in the pipe; it never reaches the log.
Control loops degrade. PID controllers depend on timely feedback. A meter reporting flow from several seconds ago introduces phase lag that can destabilize an otherwise well-tuned loop.
Trends look stable when the process isn't. A slow meter acts like a built-in low-pass filter — it smooths the real fluctuations along with the noise. The chart looks calm; the process may not be.
Batch errors accumulate. When dosing or totalization depends on flow integration, every missed transient adds a small volumetric error. Over thousands of cycles, those errors stop being negligible.
Why meters aren't simply built to respond as fast as possible
Response time and measurement noise are linked by a direct trade-off. A faster response means less signal averaging — the output tracks real flow changes closely, but it also passes through more raw electrical and acoustic noise. A slower response averages more, suppressing noise — but it suppresses genuine transients along with it.
There's no universally "correct" response time. The right value depends entirely on the application: how fast the flow actually changes, and how much noise the application can tolerate in exchange for speed.
A meter with one fixed response time is built around someone else's process, not yours.
How the FM800H handles this trade-off
The FM800H clamp-on ultrasonic flow meter offers 7 user-selectable response time settings, from 0.25 seconds to 30 seconds.
The setting is changed directly from the meter's menu or remotely over Modbus RTU — no additional software or tools required. Engineers can select the appropriate setting during commissioning and adjust it again if the process changes.
Why this matters specifically for semiconductor applications
Semiconductor front-end processes present a particular challenge for flow measurement: small pipe diameters (12.7–25.4 mm) mean flow velocity changes faster for a given volumetric change, and bellows or diaphragm pumps introduce frequent, rapid pulsation. A meter built around a single fixed response time — designed, for example, for steady municipal water flow — isn't equipped to track that kind of process.
The FM800H was designed with response time as a primary parameter from the outset, not an afterthought tuned in firmware.
All seven response time settings are published on the datasheet — the same standard applied to accuracy and repeatability. Engineers selecting instrumentation for a process should have the complete picture; an unpublished spec is a risk the engineer doesn't know they're taking on.