RECD Pressure Spikes Only at Startup: What Could Explain Them?

The pattern is oddly specific. The DG set starts, the retrofit emission control device (RECD) reports a pressure spike within the first minute or two, and then the number settles to something normal and stays there for the rest of the run. Nothing else is wrong. Later in the day the same set at the same load reads fine. A different shift sees the spike and calls it an alarm, and a third shift dismisses it as “just the startup”.

A pressure spike that appears only at startup is worth understanding because it can mean three very different things. It might be a real pressure event in the exhaust, caused by something that happens in the first moments of running. It might be a sensor or line artefact, where the instrument is reporting something that is not happening in the exhaust. Or it might be a control event, where a power dip or a start sequence is causing the reading to jump. The remedies are completely different, so the first job is to find out which of the three you have.

This guide follows the first 120 seconds of a start, slice by slice, and looks at what can plausibly happen in each. It then gives six suspects with a discriminating test for each, a capture method that separates real pressure from artefact, and a decision table for what to do. The scope is differential pressure (ΔP) across the filter and, where relevant, absolute back-pressure. If your alarm appears only during load changes rather than at start, see our companion article on alarms during generator load steps.

Two cautions before we begin

Temperature does not usually explain a cold-start ΔP spike. It is tempting to think that cold, dense gas is harder to push through a filter. For a ceramic wall-flow filter at a given mass flow, the effect of temperature usually works the other way, because gas flows through the wall in a viscous way and the volume of gas and its viscosity both rise with temperature, so ΔP tends to be higher when the filter is hot than when it is cold. Forum discussions of vehicle filters describe the same trend, noting that pressure drop depends on temperature as well as flow rate. Treat this as a general engineering expectation and confirm with your supplier. The practical point: if ΔP spikes when the system is coldest, look for something other than gas properties.

Do not compare readings taken at different loads. ΔP depends strongly on flow, so a start at light load and a later reading at 60% load cannot be set side by side. See comparing RECD pressure readings at different generator loads.

What happens in the first 120 seconds

Think of the start as a series of short episodes. Each one has its own physics and its own typical artefacts.

Slice 1: seconds minus ten to zero (before the engine fires)

The controller and its sensors may power up with the set’s control supply, and in some arrangements they power up at the same moment as the starter engages. The RECD’s pressure transmitter may be doing its own start-up routine, with an output that is not yet valid for a second or two. If the controller supply is shared with the engine’s starting battery, the cranking dip can reset or glitch the controller. Our article on RECD regeneration and battery-supported controls explains how.

Typical artefacts: zero offsets, brownout resets, a first sample that is not a measurement at all.

Slice 2: zero to about ten seconds (cranking, firing and run-up)

The engine turns over, fires and accelerates to rated speed. In the first seconds exhaust flow is irregular, with very rich or irregular combustion on a cold engine, and the flow settles as speed stabilises. The set may also accept load almost immediately if an automatic transfer switch (ATS) is waiting to transfer. Exhaust pulses are strong and uneven.

Typical artefacts: pressure pulses that are real but brief, a pressure sensing line that is excited by pulsation, an alarm comparator that sees one high sample.

Slice 3: ten to about thirty seconds (speed stable, load possibly applied)

The engine is at rated speed. If an ATS transfers load at this point, exhaust flow rises abruptly, which raises ΔP and back-pressure quickly. Meanwhile the exhaust system is still cold, and any condensation that formed while the set was idle is still liquid.

Typical events: a real flow-driven rise as load is applied, with the system still cold and possibly wet.

Slice 4: thirty to ninety seconds (warm-up and moisture)

The exhaust begins to heat the pipework and filter. Water that condensed in the system during the idle period starts to evaporate or be blown out. Condensate in the pressure sensing lines may be pushed or boiled, which can make the reading jump or hunt. The filter and mat may release moisture, and any soot cake that absorbed moisture may behave differently for a few minutes.

Typical events: moisture-related transients, water in lines, gradual settling of ΔP.

Slice 5: ninety to 120 seconds and beyond (settling)

Temperatures have begun to stabilise. Moisture has largely gone. The signal settles to its steady value for the load.

Typical outcome: a stable ΔP at the load, which is the number to compare with the baseline.

This timeline is a framework, not a promise. A set’s controls and loading arrangements decide when each slice begins and ends.

Six suspects and the test for each

Each suspect below has a mechanism, evidence you may already have in the logs, and a test that separates it from the others.

Suspect 1: Condensate in the pressure sensing lines or at the sensor

Mechanism. The sensing tubes run from taps on the hot exhaust to a transmitter that is kept cool. Between starts, moisture condenses in the cooler parts of these lines. On the next start, hot gas and pulses push or boil the water, causing the reading to jump, wander or stick for a short time. A fleet-maintenance article notes that the tube from the filter port to the transducer can leak or crimp, and that these are problems a diagnostic tool cannot solve, which highlights how often the cause lies in the line and not the filter.

Evidence. The spike is larger after long idle periods, damp weather or cold nights. It is smaller on a hot restart. The signal looks noisy or saw-toothed rather than a clean rise.

Test. Compare a cold start with a hot restart shortly after a run. Drain or blow out the sensing lines with the system cold and isolated and repeat the cold start. If the spike shrinks, condensate is the cause. Check that lines slope so that water drains back to the tap and not toward the transmitter, and that there are no low points where water can sit. See protecting an RECD against rainwater and condensation.

Suspect 2: A real transient at light-load or no-load start

Mechanism. At cold start the engine may run rich, producing unburnt fuel and wet soot that packs into the filter wall. In the first seconds the exhaust is also at its most pulsating. These produce a real but short rise in pressure.

Evidence. The ΔP and absolute back-pressure both move, the spike is repeatable, smoke or smell is present for a few seconds, and the spike is larger when the engine had been idling for a long time before the stop.

Test. Measure ΔP and absolute back-pressure together with a second, independent instrument with a fast response. If both rise at the same time and by comparable amounts, the event is physical. Check whether the engine’s starting behaviour is normal: a cold, rich start with heavy smoke suggests fuelling or glow-system issues for the engine specialist.

Suspect 3: Load applied immediately by the ATS

Mechanism. If the ATS transfers load within seconds of the set reaching rated speed, exhaust mass flow rises from almost nothing to a large value in a short time. Pressure follows flow, so ΔP and back-pressure rise quickly. Because the system is cold, the temperature has not caught up, so any controller logic that compensates ΔP for temperature or load may be confused for a short time.

Evidence. The spike coincides with the ATS transfer in the building’s event log, and disappears if the set is started with no load transfer (for example in a test). The spike’s size scales with the load transferred.

Test. Align time stamps between the ATS or DG controller and the RECD controller. See aligning RECD sensor timestamps with generator runtime logs. Then run a start with a delayed transfer or a staged load. A spike that tracks the transfer rather than the start is a load-step effect. Our article on RECD fitment and ATS testing coordination discusses the relationship.

Suspect 4: Control power dip or controller start-up behaviour

Mechanism. The cranking dip on the controller supply can cause a reset or a short invalid reading, while a controller that boots slowly may output a default or last-held value until its inputs stabilise. Sensor power-up can also produce a spike for a second or two.

Evidence. The “spike” is a single sample or a very short event right at the start. The log shows a power-up or reset entry at the same time. There is no corresponding change in a second, independent instrument.

Test. Log controller supply voltage during several starts and compare with the controller’s minimum voltage. Check whether the controller’s log shows resets. See RECD monitoring during power failure and RECD regeneration and battery-supported controls.

Suspect 5: Moisture and wet soot inside the filter

Mechanism. After a long stop, especially in humid or wet conditions, the filter and its insulation can hold moisture. The first heating drives it off, and the soot cake, if oily or damp, may briefly resist flow. This effect is real and temporary. It is more likely after storage or a long idle. See restarting a DG set with an RECD after long inactivity.

Evidence. The effect fades as the system warms, is larger after idle periods and in the monsoon, and is accompanied by white vapour at the outlet or a slight steam smell at start.

Test. Compare the spike after short stops (hot) with that after a long stop (cold and damp). A bigger spike after long stops points to moisture. A loaded run of sufficient length should clear it. If it persists after several hot starts, suspect wet soot from light-load running. See RECD regeneration after wet stacking.

Suspect 6: A damper, bypass or valve sequence

Mechanism. If the system has a bypass, a diverter or an isolation damper, the start sequence may move it. A damper that takes several seconds to reach position can briefly restrict the flow path, and an actuator with a sticky stroke can produce a spike. The controller may also be coded to run the start in bypass and switch to the filter after warm-up, creating a step in flow through the filter.

Evidence. The spike’s timing matches the damper’s stroke time. There is a position signal that changes at the same moment.

Test. Compare the damper position log with the pressure log. Watch the damper during a start. Check the actuator and limit switches. A change in the sequence logic is a question for the supplier.

A summary table

SuspectIs the pressure real?Typical size and shapeCold versus hot startBest discriminating test
1 Condensate in linesReading, not exhaustNoisy, wandering, can be largeWorse after long idle and dampCompare cold with hot start; blow out lines
2 Real start transientYesShort, matches a second instrumentWorse after long idleTwo independent fast instruments
3 ATS load applied earlyYes, flow-drivenScales with load transferredIndependent of temperatureAlign logs; delay transfer
4 Control power dipReading, not exhaustSingle sample or very shortIndependentSupply voltage logger; reset log
5 Moisture or wet soot in filterYesSlow rise that fades during warm-upWorse after long idle and dampCompare long stop with short stop
6 Damper or bypass sequenceYesStep matching damper strokeIndependentCompare damper position log

Notice that two suspects (1 and 4) produce readings that do not reflect real pressure in the exhaust, and four (2, 3, 5 and 6) produce real pressure. If you have a single instrument, you cannot tell the two groups apart. That is why the capture method below matters.

The capture method: one afternoon, two instruments

A good capture separates real events from artefacts and takes about an afternoon.

What you need.

  • The controller’s own ΔP log.
  • An independent differential pressure instrument with a fast response and data logging, connected at spare taps or in parallel, according to the supplier’s advice. See pressure sensors versus pressure switches in RECD monitoring.
  • A data logger on the controller supply voltage.
  • Time-synchronised records from the ATS or DG controller.
  • A notebook, a stopwatch and a phone for photographs.

Procedure.

  1. Synchronise clocks across the instruments.
  2. Plan four starts: a cold start with ATS transfer, a cold start with no load for two minutes then load, a hot restart with transfer and a hot restart with no load. This matrix separates temperature effects from load effects.
  3. For each start, note the time of crank, fire, rated speed and transfer, the ambient temperature, the time since last run and the weather.
  4. Log at a rate faster than the event. If the spike lasts a few seconds, one sample per minute misses it. Aim for a sample every second or faster.
  5. Mark events in the notes: smoke, steam, noises, damper moves.
  6. Repeat each start type at least twice, to avoid reading too much into a single odd event.
  7. Compare the controller’s trace with the independent instrument. If they agree on a spike, it is real. If the controller spikes and the independent instrument does not, it is an artefact of the controller or its lines.

Reading the results.

ResultInterpretation
Both instruments spike, only on cold startsMoisture or wet soot (5), or a cold-start transient (2)
Both spike, only with transferATS load step (3)
Both spike, timing matches damperDamper sequence (6)
Controller spikes, independent instrument does notSensing line or controller artefact (1 or 4)
Controller spike aligns with a supply dipControl power (4)
Neither spikesThe alarm may come from a threshold or filter setting; check the controller configuration

How big a spike is acceptable?

There is no general figure. A spike matters if it exceeds an alarm limit, if it grows over time or if it is accompanied by other signs. Questions to settle with the supplier:

  1. What is the alarm limit, and is it intended to apply during the first minutes after start?
  2. Does the controller use a start-up inhibit or a delay, and what are the values?
  3. How is ΔP compensated for flow or temperature, if at all?
  4. What is the maximum ΔP the device and the engine can tolerate for a short transient?
  5. What is the sensor’s response time and the controller’s sampling rate?

A US EPA technical bulletin on diesel particulate filter installation stresses that back-pressure monitoring should alert maintenance staff to service needs and should itself be periodically inspected to confirm it works. A monitor that false-alarms every start trains operators to ignore it, which undermines its purpose. See reducing nuisance RECD alerts without hiding faults for how to tune responsibly.

When a start-up spike is a warning, not a nuisance

Most start-up spikes fall into the harmless categories. A few patterns deserve prompt attention.

  • The spike is growing. If each month’s spike is higher than the last, it may reflect rising baseline ΔP, wet soot or an increasingly blocked filter.
  • The settled value is also rising. A spike followed by a higher settled ΔP at the same load suggests real loading. See why an RECD keeps clogging.
  • The spike approaches the engine’s back-pressure limit. A transient that touches the limit is a risk to the engine and to the filter. See how exhaust back-pressure affects a DG set.
  • The spike comes with smoke, smell or a bang. These suggest oil or fuel in the exhaust or a physical event.
  • The spike triggers a trip or derate. An engine that trips on start-up back-pressure is an availability risk, particularly for a standby set.
  • The spike appears after a repair, a cleaning or a change of settings. Something about the change may be involved.

Fixes that follow from each suspect

SuspectTypical fixWho does it
Condensate in linesRe-route lines with continuous slope; add drain or trap if the design allows; keep lines short and insulated where advisedSupplier or installer
Real start transientEngine check for cold-start fuelling; consider loaded warm-up before transfer where operations allowEngine specialist
ATS load applied earlyAgree with the building team whether a short delay or staged load is acceptable; consider ATS settings only with electrical authorityFacilities and electrical lead
Control power dipDedicated or conditioned supply for the controller; DC-DC converter; battery maintenanceElectrical contractor and supplier
Moisture and wet sootLoaded warm-up runs, better insulation, moisture protection, change in test routineOperations and supplier
Damper sequenceAdjust actuator or logic after supplier reviewSupplier
Alarm settings too tightAdd a start-up inhibit or delay agreed with the supplier, with a clear maximum timeSupplier

Do not shorten the problem by raising alarm limits yourself. Limits exist to protect the engine and the filter, and changes should be agreed with the supplier and recorded. See RECD controller settings access control.

A worked illustration: the spike that was in the tubes

The following is an illustrative scenario, not a real customer. A data centre’s 800 kVA set has an RECD whose controller raises a high-ΔP alarm at about 40 seconds after each start, clearing within two minutes. Operators have been acknowledging it for weeks.

First look. The alarm history shows it occurs on about 80% of starts, mostly on cold mornings and after rain, and rarely on tests done shortly after a previous run. Settled ΔP at 60% load is stable.

The capture. An independent logging manometer is connected in parallel. Four start types are tested over two days. The controller spikes to roughly double its settled value in the first minute on cold starts. The independent instrument shows a modest, smooth rise of about 10% that tracks the load transfer, with no spike.

Reading it. The controller trace is noisy and spiky, the independent trace is not. The sensing line from the high-pressure tap runs horizontally for 1.5 metres with a low sag. A cold morning inspection finds a small amount of water in the sag. The spike is mostly a line artefact (suspect 1) with a smaller real component from the ATS transfer (suspect 3).

The fix. The installer re-routes the lines with a continuous slope back toward the tap, shortens them and insulates the run. The supplier also agrees a 90-second start-up inhibit on the high-ΔP alarm only, with the absolute back-pressure alarm left active. After the changes, cold starts show a smooth rise of the same size as the independent instrument, and the alarm stops. The team notes the new baseline in the event log.

The sensible order matters: they proved the spike was an artefact before suppressing the alarm, and they left the independent protection in place.

Frequently asked questions

Is a brief pressure spike at startup harmful to the filter?

A short, modest transient is usually harmless, but you need evidence. Check that the spike does not approach the alarm or engine limits, and that the settled ΔP at the same load is steady. A growing spike deserves investigation.

Why does it happen only on cold starts?

Cold starts involve condensed moisture, richer combustion, cold lines and a cold filter. Moisture in sensing lines and damp soot are the most common reasons for a cold-only effect.

Can I just add a start-up delay to the alarm?

A delay is a legitimate tool once you know the cause and agree it with the supplier. Applied blindly, it can hide a real event. Keep a hard limit active, keep the delay short and record the change.

Does a spike affect emission test results?

Tests are normally done at stable, warmed-up conditions, so start-up transients are usually outside the test window. See warm-up time in RECD testing.

Should I worry if the spike happens only after a long idle?

It fits moisture or wet soot. Run a loaded session of sufficient length and see whether the effect shrinks. If it does not, ask the supplier to review it. See choosing an RECD for a generator running at low load.

Bringing it together

An RECD pressure spike at startup is not one problem but a family. Slice the first 120 seconds, separate real pressure from sensor, line and control artefacts with an independent instrument, and use a four-start matrix to split temperature from load. Most spikes turn out to be moisture, a sensing-line sag, an early load transfer or a control dip. Fix the cause, record the new baseline and tune alarms only after you understand what they are reporting.

If you would like help designing a start-up capture or checking sensing-line routing for your installation, explore our retrofit emission control devices, read the FAQ or contact the Aceget team.

Related Reading

Sources



Leave a Reply