Sensors in the field rarely report on a tidy schedule. One transmitter updates every second, another every five, a third only when its value changes, and each carries a timestamp that lands wherever the scan happened to fall. Edge time alignment and resampling is the step where a gateway takes those ragged, mismatched streams and reworks them onto a single, evenly spaced time grid before publishing, so that a calculation combining several tags is actually comparing values from the same instant. This page explains what alignment and resampling do to the data itself, why it matters for downstream math, and how it differs from keeping clocks in sync.
Edge Time Alignment/Resampling in one line: Edge time alignment and resampling is the gateway-side process of snapping readings from sensors with different scan rates and irregular timestamps onto a common, uniformly spaced time grid before the data is published. By interpolating or holding values to shared grid points, it produces aligned samples so that calculations combining multiple tags operate on values from the same moment, rather than on readings that were actually taken seconds apart.
In a real installation the tags that feed a single calculation almost never arrive on the same rhythm. A flow meter might scan at one rate, a pressure transmitter at another, and a temperature reading only on change, so at any given moment the newest value for each tag was captured at a slightly different time. Add normal scan jitter, where even a nominally one-second scan wanders by tens of milliseconds, and the timestamps become a scatter rather than a grid. Individually each reading is fine; together they do not line up.
This becomes a real error the instant you combine them. Compute a mass flow from a volumetric flow, a pressure, and a temperature, and if those three came from moments seconds apart during a transient, the result is a blend of three different plant states that never coexisted. During steady operation the mismatch hides, but exactly when it matters most, a startup, a swing, an upset, the values are moving and the misalignment injects an error that looks like real process noise but is actually a timing artifact.
Resampling fixes this by choosing a common grid, for example one point every second, and deriving each tag's value at every grid point. For a fast tag that means picking or averaging the samples nearest each point; for a slow or change-based tag it means interpolating between the last two known values or holding the last value forward until a new one arrives. The output is a set of streams that all share the same timestamps, so any two of them can be added, multiplied, or differenced with confidence that they describe the same instant.
It is easy to confuse this with clock synchronization, but they solve different problems. Time synchronization, using protocols like NTP or PTP, makes sure every device agrees on what time it is, so a timestamp of ten o'clock means the same thing everywhere. That is a prerequisite, but a perfectly synchronized clock still leaves you with tags sampled at different rates and different offsets. Synchronization aligns the clocks; it does nothing to align the samples themselves onto a shared cadence.
Time alignment and resampling operate one layer up, on the values after the clocks already agree. They take correctly timestamped but irregularly spaced readings and reshape them into a uniform series, deciding how to represent each tag at each grid point through interpolation, averaging, or last-value hold. The choice of method matters: interpolating a signal that steps in discrete increments can invent values that never existed, while holding a fast analog signal flat across a wide interval can miss real movement. Matching the method to the nature of each signal is part of doing alignment well.
Where the resampling happens is also a design decision. Doing it at the edge means the gateway publishes clean, grid-aligned data that every downstream consumer can trust as-is, which keeps calculations and historians consistent and reduces the bandwidth of change-based streams. The cost is that the raw, native-rate detail is reshaped before it leaves the site, so systems that want the original samples must be given a separate raw path. Many designs publish both: an aligned series for calculations and dashboards, and a raw series preserved for analysis.
For a cloud SCADA platform such as Merobix pulling data from many remote sites, edge alignment is what lets values from different sensors and even different devices be combined without timing artifacts. When a gateway resamples a wellpad's flow, pressure, and temperature onto one grid before publishing, the platform receives a coherent snapshot it can total, allocate, or run calculations against directly, instead of having to reconcile three streams that each marched to their own scan clock.
This matters more across a distributed operation than in a single control room, because the tags feeding a network-wide figure may originate at sites hundreds of miles apart, each with its own devices and scan rates. If a mass balance or an allocation sums flows that were each captured at their own irregular cadence, small timing mismatches accumulate into a balance that never quite closes. Aligning each site's data to a shared grid at the edge gives the platform comparable, contemporaneous samples to work with, which makes those network calculations both cleaner and easier to defend.
Alignment also smooths the human experience of the data. An operator overlaying trends from several tags on one chart expects the lines to share an x-axis; if each tag carries its own ragged timestamps, the overlay looks jittery and features appear shifted from one another even when nothing is wrong. Resampled, grid-aligned data plots cleanly, so a pressure spike and the flow change it caused appear at the same point on the timeline, and an operator reading a remote site can trust that what lines up on the screen actually happened together.
Time synchronization, using NTP or PTP, makes every device agree on the current time so that timestamps mean the same thing everywhere. Time alignment and resampling assume the clocks already agree and instead reshape the readings themselves, snapping tags sampled at different rates onto a common grid. Synchronization aligns clocks; resampling aligns the data, and both are needed for reliable multi-tag calculations.
Because the latest value of each tag was captured at a different moment, so combining them mixes readings from times that were seconds apart. During steady operation that is invisible, but during a transient the values are all moving and the mismatch injects error into any calculation that combines them. Resampling to a shared grid ensures the values you combine describe the same instant.
It reshapes the data, and if done carelessly it can distort it: interpolating a signal that changes in discrete steps can invent intermediate values, and holding a fast signal flat across a wide interval can hide real movement. Matching the method to each signal avoids this, and many gateways publish both an aligned series for calculations and a preserved raw series so the original native-rate detail is never lost.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.