- November 20, 2026
- Posted by: editor
- Category: RECD & Emission Control Devices
The alarm history makes a clear pattern once someone lines it up with the building’s event log. The retrofit emission control device (RECD) is quiet through hours of steady running. Then a lift motor starts, a chiller kicks in, an automatic transfer switch (ATS) moves load onto the generator, or a large block of load drops off, and an alarm appears within seconds. A few minutes later it clears. The rest of the time everything is fine.
This is almost always a timing mismatch. The alarm limit was set for a world where the set runs at steady load, and the event is a world where flow, temperature and pressure are all moving at different speeds. Pressure follows exhaust flow almost immediately. Temperature follows slowly, because metal and ceramic have thermal mass. Sensors respond at their own speed and controllers sample at their own rate. For a few seconds, the numbers on the screen describe a system that is not in equilibrium, and a limit tuned for equilibrium trips.
That does not mean the alarm is wrong. Sometimes a load step does push the exhaust system somewhere it should not go. The task is to tell a nuisance from a warning. This guide builds the picture in four parts: the anatomy of a load step, a four-by-two matrix of alarm types against step direction, a staged step test that produces evidence, and tuning options that do not hide faults.
The anatomy of a load step
A load step is a sudden change in electrical load on the generator. The engine responds by changing fuel delivery within a fraction of a second, and the exhaust system feels the consequences almost immediately. Four things change, at four speeds.
| Quantity | How fast it responds | Why |
|---|---|---|
| Exhaust mass flow | Seconds | The engine’s air and fuel flow follows load, limited by turbocharger response |
| Pressure in the exhaust (back-pressure and ΔP) | Almost immediately with flow | Pressure drop across a restriction rises sharply with flow |
| Exhaust temperature at the filter inlet | Tens of seconds to minutes | Thermal mass of the manifold, pipework and canister delays the change |
| Bed or filter temperature | Minutes | Ceramic and mat hold heat |
Pressure follows flow quickly, temperature lags and thermal mass in the filter smooths out the slowest of all. In the first seconds after a step up, the system has more flow but not yet more heat. After a step down, it has less flow but still a hot filter.
Step up
When load rises suddenly, a diesel engine has to deliver more fuel, but the turbocharger needs time to build boost. Combustion runs richer for a moment, and soot output rises briefly. Exhaust flow rises as the engine breathes more. Pressure across the filter rises with it, because the pressure drop across the filter and pipework climbs steeply with flow. Meanwhile the exhaust temperature at the filter inlet has not yet caught up.
Result: ΔP and back-pressure spike for a few seconds, inlet temperature is still low, and a brief soot puff passes into the filter.
Step down (load rejection)
When load falls suddenly, fuel is cut back, flow falls quickly and exhaust temperature starts to fall, though the filter stays hot. If a regeneration was in progress, the flow through the filter drops and the heat that was being carried away by the gas is no longer removed. Soot that is still burning may burn faster in a lower flow of gas that is rich in oxygen but carrying less cooling. That is one of the conditions in which a filter can overheat. A patent filing on exhaust gas purification describes how a sudden drop in exhaust flow during DPF regeneration makes the inlet temperature hard to control, and why control logic must detect it, which is the same physical problem a stationary set can face.
Result: pressure falls with flow, temperature drops in the pipework but the bed may stay hot or briefly climb, and a regeneration may be interrupted. See what happens when the generator stops during RECD regeneration.
What the standards say about steps
Generator makers describe load acceptance in terms of the ISO 8528-5 standard. A Kohler white paper explains that ISO 8528-5 defines transient performance classes and specifies frequency and voltage limits for different load steps, and that the load steps are defined in terms of engine mean effective pressure, so a given kW step represents a different fraction of capability on different engines. A Cat white paper on transient performance notes that a generator set will typically reach full load in three to six steps under the standard’s definition. These documents concern electrical performance: frequency and voltage dips and recovery. They do not address the exhaust, but they matter because they tell you how large a step the set is designed to take and how fast it recovers. A building that applies a large step will produce a large exhaust transient.
Four alarm types, two directions
Not every RECD alarm responds the same way to a load step. The matrix below sets out the main types and what a step up or a step down typically does to each. It is a thinking tool and will not match every controller.
| Alarm type | Step up (load applied) | Step down (load rejected) |
|---|---|---|
| High ΔP (filter pressure) | Likely: ΔP rises with flow before any compensation | Unlikely: ΔP falls |
| High absolute back-pressure | Likely, short spike | Unlikely |
| High temperature (inlet or bed) | Possible if load is applied during a hot regeneration | Possible: bed stays hot as flow falls; a burn can run away |
| Low temperature (inlet below working range) | Possible: flow rises before heat does | Possible: flow falls, pipework cools, inlet temperature drops below a limit |
Two additional alarm types often accompany steps even though they are not pressure or temperature alarms.
- Regeneration interrupted or aborted. A step in either direction can end a cycle.
- Sensor or communication fault. A step can trigger electrical noise or a supply dip, producing a spurious reading. See RECD alarm versus generator trip control response.
Why the high-ΔP alarm is the usual culprit
ΔP across a filter does not rise linearly with flow, and in the pipework and silencer the pressure drop rises roughly with the square of flow. A load step that doubles flow can more than double the pressure drop in turbulent sections. If the alarm limit was set from a steady reading at moderate load, a step to a much higher load can cross it while the filter itself is perfectly clean. This is not a filter problem. It is an operating-envelope problem.
The check: compare ΔP at the new, higher load once steady with the supplier’s figure at that load. If the steady value is within limits, the spike is a transient of flow. If the steady value at the higher load is also above the supplier’s figure, the filter or the pipework has a real problem; see high backpressure after RECD installation.
Why the temperature alarms can be real
Temperature alarms deserve more respect, because temperature is what damages filters. A step down during a hot regeneration is the classic risk. A step up during a hot regeneration can also change the burn, since flow rises while the bed is already hot. If your alarm history shows temperature alarms on load steps during regeneration, do not treat them as a nuisance until the supplier has reviewed the data. Our article on repeated failed regeneration and filter damage describes what can go wrong.
What the controller sees: sampling, response and delay
Several instrument and controller properties turn a real transient into an alarm.
Sensor response time. A pressure transmitter responds quickly, but a sensing line with condensate or a long length of tubing slows and distorts the signal. A thermocouple in a thick sheath responds slowly, and a bare junction responds quickly. Different sensors in the same system can therefore disagree for a few seconds even when the exhaust is behaving normally. See RECD temperature sensor placement.
Controller sampling rate and filtering. A controller that samples once per second and has no smoothing will see every pulse and spike. One that averages over thirty seconds may miss a short real event.
Alarm logic. Some controllers trip on a single sample above the limit, some require the value to persist for a set time (a debounce or delay) and some use rate-of-change rules. The same event can alarm on one controller and pass on another.
Compensation. Better controllers adjust ΔP for flow, using the generator’s load signal or a calculated flow, so that the alarm compares like with like. Others compare the raw number to a fixed limit.
Hysteresis. An alarm that clears only when the value falls well below the limit can appear to “hang on” after the event.
Our article on reducing nuisance RECD alerts without hiding faults explains how these settings interact with safety. The summary: any tuning must keep the protective function, and any change should be agreed with the supplier and documented.
The staged step test: producing the evidence
Before changing anything, run a test that shows what the system does during steps. A load bank makes this safest, but a real building load can serve if the plan is agreed with operations.
Preparation.
- Agree the test with the facility manager, the electrical lead and the RECD supplier. Include the building’s critical load considerations.
- Install or check time-synchronised logging for load (kW), ΔP, absolute back-pressure if available, inlet and bed temperature, controller alarms and ATS state. See aligning RECD sensor timestamps with generator runtime logs.
- Ensure that no regeneration is running unless that is part of the test.
- Record ambient conditions, fuel level and the state of the filter (time since last regeneration).
Procedure. The steps below are a template. Adapt step sizes to what the engine and the building can accept, and ask the supplier whether any step should be avoided on a loaded filter.
- Warm up at a moderate steady load (for example about 30 to 40%) until temperatures are stable. Log a baseline for several minutes.
- Step up to about 50% and hold until steady. Note the peak ΔP, the time to settle and any alarm.
- Step up to about 75% and hold. Repeat the observations.
- Step up to the highest load the plan allows, hold and log.
- Step down in the same stages, noting temperature behaviour at each.
- Repeat one large single step representative of the real event (for example the chiller start) if it is safe to do so.
- Repeat the sequence at least twice, since one trace can mislead.
- If the supplier approves, repeat a step during an assisted regeneration on a separate occasion, with the supplier watching.
What to extract.
| Measure | Why it matters |
|---|---|
| Steady ΔP at each load | Gives the flow-ΔP curve of the installation |
| Peak ΔP during each step and time to settle | Shows how large and how long the transient is |
| Peak back-pressure during steps | Compares with the engine limit |
| Inlet temperature at the step and 60 seconds later | Shows the lag of heat behind flow |
| Alarm limits crossed, with times | Links alarms to specific steps |
| Controller state changes | Shows which alarms are logic-driven |
Plot ΔP against load, and ΔP against time for each step. If the steady ΔP curve is within the supplier’s figures and only the transient crosses the limit, the limit is the issue. If steady values already approach the limit at the higher load, the device or layout is the issue.
Reading the pattern: a lookup table
| What the step test shows | Likely explanation | Direction of response |
|---|---|---|
| Steady ΔP within spec; peak spike crosses limit for a few seconds on step up | Alarm logic too tight for transients | Discuss a short delay or flow compensation with the supplier |
| Steady ΔP at high load already near or above limit | Real restriction or undersized device | See the back-pressure fault-finding guide; review the sizing |
| Spike much larger than steady rise, noisy trace | Sensing line or response artefact | Check lines, condensate, damping; compare with an independent instrument |
| High-temperature alarm on step down during regeneration | Real thermal risk | Supplier review; consider regeneration inhibit during load changes |
| Low-temperature alarm after step down | Normal cooling of pipework, or a low limit | Check limit and delay; consider whether regeneration is interrupted |
| Alarm appears only when the ATS transfers, not on gradual load changes | Speed of the step is the trigger | Staged load pickup or sequence delay; electrical authority involved |
| Alarm accompanied by a supply dip | Control power | See battery-supported controls |
| Alarm only after the filter has been loaded for weeks | Real soot load narrowing the margin | Check trend; see why an RECD keeps clogging |
Tuning options, from safest to riskiest
If the evidence shows that the alarm is a nuisance, several adjustments are possible. They are listed here from the least to the most risky. Each requires the supplier’s agreement and a written record.
- Compensate the reading for flow. Use load or calculated flow to normalise ΔP, so that the alarm compares like with like. This addresses the root of the problem rather than hiding it.
- Add a short persistence delay (for example a few seconds, set from your step test) to the nuisance alarm only, while keeping a hard high limit with no delay.
- Use a rate-of-change filter that ignores very fast, short spikes but still catches slower, sustained rises.
- Add hysteresis so the alarm clears cleanly and does not chatter.
- Split the alarm into two levels: a warning that logs and notifies, and a protective alarm that acts. Keep the protective level unchanged.
- Reduce the step size by staging load pickup or adding a short delay on the ATS transfer, which is a building-side change. This is the cleanest fix when the real problem is the speed of the step, but it involves electrical and operational decisions.
- Raise a limit. The riskiest option. Only do this if the supplier and the engine maker confirm that the higher limit is safe, and record why.
Never do the following:
- Disable an alarm without a replacement.
- Raise a temperature limit to avoid a nuisance without the supplier’s written agreement.
- Accept “it is just a transient” without a step test.
Our article on RECD controller settings access control explains why changes should be controlled and recorded.
Protecting regeneration during steps
Many step-related problems occur during regeneration. Operational measures can help.
- Avoid deliberate steps during a cycle where the building allows it, such as scheduling non-critical load transfers after the cycle completes. See RECD regeneration while supplying building loads.
- Use inhibit signals that prevent a regeneration from starting when load is about to change. See RECD regeneration inhibit signals.
- Make heater power robust. A transfer that drops heater supply for a few seconds can abort a cycle.
- Brief operators on what to do if a step occurs mid-cycle.
- Consider generator choice for rapidly changing loads. See RECDs for generators with rapidly changing loads.
A worked illustration: a chiller, a spike and a limit
The following is an illustrative scenario, not a real customer. A commercial complex’s 630 kVA set starts a large chiller within a minute of an ATS transfer on most grid failures. The RECD controller raises a high-ΔP alarm each time, clearing about ninety seconds later.
The log alignment. The ATS log and the RECD log are aligned. The alarm occurs four to six seconds after the chiller contactor closes, and the set’s load jumps from about 35% to about 70% in that interval.
The step test. During a planned load-bank test, the team steps the set from 35% to 50%, 70% and 85% with holds of ten minutes each. Steady ΔP is 21, 29, 41 and 54 mbar at those loads, all inside the supplier’s figures. A single step from 35% to 70% produces a peak of 63 mbar for about four seconds before settling at 41 mbar. The controller’s alarm limit is 60 mbar with no delay.
Reading it. The steady curve is healthy. The transient overshoots because of the speed of the flow change, together with the response of the sensing line. The alarm limit is correct for steady running but too tight for a transient.
The agreed changes. The supplier agrees to compensate ΔP for load, add a five-second persistence delay to the 60 mbar warning and keep an unchanged hard limit that is well above the 63 mbar peak and below the engine back-pressure limit, with no delay. Facilities also agree to move the chiller start to a sequenced delay of ten seconds after the ATS transfer, which reduces the step size. The next grid failure shows no alarm, and the log records a smooth rise to 41 mbar.
The team did not raise the limit. They proved the steady curve was healthy, used compensation and delay for the nuisance, kept the hard limit and reduced the step at its source.
Frequently asked questions
Does a load-step alarm mean the filter is blocked?
Not by itself. A clean filter can show a transient ΔP spike because flow rises faster than the system settles. Check the steady ΔP at the new load against the supplier’s figure.
Is it safe to ignore a load-step alarm that always clears?
Do not ignore it until you know why it occurs. A nuisance alarm is acceptable only after a test shows steady values are healthy and the transient is well below any protective limit.
Can load steps damage the filter?
Mostly the risk is thermal, and it is highest when a step happens during a hot regeneration or when a heavily loaded filter is suddenly exposed to high flow. Keep soot loads within limits and protect regeneration cycles. See soot versus ash in RECD regeneration.
Should I change the ATS settings?
Only if the electrical authority and the building’s operations agree. Staged pickup or a short delay can reduce steps, but the priority is keeping critical loads supplied.
What standard covers generator load steps?
ISO 8528-5 defines transient performance classes and load-step testing for generator sets. It covers frequency and voltage, not exhaust pressure, so for the exhaust side you rely on the engine maker’s and the RECD supplier’s data.
Bringing it together
An RECD alarm that fires only during generator load steps usually reflects a fast change in flow meeting a limit that was set for steady running. Understand the anatomy of the step, separate the four alarm types and two directions, run a staged step test to get the steady curve and the transient peaks, and tune in order from compensation to delay to sequencing, keeping protective limits intact and documenting each change. Temperature alarms during regeneration deserve extra caution.
If you would like help designing a step test for your site or reviewing alarm settings with an engineer, see our retrofit emission control devices, browse the FAQ or contact the Aceget team.
Related Reading
- RECDs for Generators with Rapidly Changing Loads
- RECD Alarm vs Generator Trip: Control Response
- RECD Regeneration While Supplying Building Loads
- RECD Fitment and ATS Testing Coordination
- RECD Pressure Spikes Only at Startup: What Could Explain Them?