- October 15, 2026
- Posted by: editor
- Category: RECD & Emission Control Devices
A facility manager in Gurugram receives two numbers for the same Retrofit Emission Control Device. The commissioning report says the device cut particulate matter by 78%. Six months later, a periodic check says 63%. Before anyone blames the device, someone pulls the load logs. The commissioning baseline was taken while the plant’s chillers were running in May, at about 82 kW. The post-installation sample was taken on a quiet Sunday at about 54 kW. The periodic check was taken the opposite way round.
The device didn’t drift by fifteen points. The load did.
Of all the variables in an RECD before-and-after test, generator load has the most direct effect on the result, and it is the easiest one to get wrong without noticing. This article explains why load changes what an RECD sees, how a mismatch inflates or deflates the reduction figure, and how to set up a load-matching protocol that holds up when a supplier, auditor or State Pollution Control Board officer reads your report.
The Short Answer
An RECD’s measured particulate reduction is a ratio of two readings: emissions without the device and emissions with it. Diesel engine emissions change sharply with load, and so does the efficiency of the RECD itself. If the two readings are taken at different loads, the ratio mixes the engine’s change with the device’s change, and you can’t separate them afterwards. Match load in kilowatts (kW), not kVA or “approximate percentage”, hold it within a tight tolerance, and log it continuously during every sample.
What “Load” Actually Means in an Emission Test
People use “load” loosely, and that causes problems. In a DG set there are at least four different numbers people call load:
| What people say | What it really is | Is it what the engine “feels”? |
|---|---|---|
| “100 kVA load” | Apparent power at the alternator terminals | No, it includes reactive power |
| “80% load” | Usually kVA as a share of nameplate kVA | No, same problem |
| “80 kW load” | Real electrical power delivered | Close (plus alternator losses) |
| Engine shaft power (kWm) | Mechanical power the engine produces | Yes, this sets the emissions |
The engine responds to mechanical power. That is the electrical kW delivered plus the alternator’s losses, which are a few percent depending on size and loading. Reactive power, the part of kVA that isn’t kW, barely loads the engine at all, though it does heat the alternator.
That is why two tests that both show “100 kVA” on the controller can put very different loads on the engine. At a power factor of 0.8, 100 kVA is about 80 kW. At a power factor of 1.0, typical of a resistive load bank, 100 kVA is 100 kW. That is a 25% difference in engine load, and the display looks the same in both cases.
If kVA, kW and power factor still feel unfamiliar, our guide on how to read a generator’s kVA and kW rating explains them simply.
The rule: record and match load in kW, and record power factor alongside it so anyone reading the report can see the engine conditions were the same.
Five Ways Load Changes What the RECD Sees
An RECD doesn’t treat exhaust as a neutral quantity of “pollution” to be cut by a fixed percentage. It is a physical and chemical system, and its performance depends on the gas passing through it. Load changes that gas in five important ways.
1. Air-fuel ratio and soot formation
At light load, a diesel engine burns a small amount of fuel in a full cylinder of air, so the mixture is very lean. As load rises, more fuel is injected into roughly the same air charge, and local rich zones in the combustion chamber form more soot. Near rated output, especially on older or poorly maintained engines, soot output can rise steeply. That is why black smoke usually shows up under heavy load.
A baseline taken at high load therefore starts with more soot to remove. Whether that helps or hurts the reduction percentage depends on how the device handles extra carbon. The point is that it is a different starting condition.
2. Exhaust temperature
Load has the strongest effect on exhaust temperature. Typical DG exhaust can range from under 200 °C at light load to 450 °C or more near full load, depending on the engine and where you measure.
Temperature matters to an RECD in two ways:
- Catalyst activity. Oxidation catalysts that convert CO, HC and the organic part of PM need a minimum temperature to work well. Below it, conversion drops sharply.
- Passive soot burning. Catalysed filters use NOâ‚‚ produced on the catalyst to burn trapped soot continuously, but only above a certain temperature. At light load the filter may keep loading without regenerating, and its behaviour changes as it does. We look at this in RECD testing during regeneration.
A device that performs well on hot exhaust can look mediocre on cool exhaust, and that’s true of the same physical unit.
3. Particulate composition
Diesel particulate isn’t a single substance. It is a mix of solid carbon (soot), organic material from unburnt fuel and lubricating oil, sulfates, and small amounts of ash. At light load, the organic share is often larger. At high load, carbon dominates.
Filter-type RECDs trap solid carbon very efficiently. Some organic material passes through the filter as vapour and only becomes particle mass when it cools further downstream. If the baseline was carbon-heavy and the post-RECD test ran at a lighter load with more organics, the device can look worse than it is. The reverse can make it look better.
4. Exhaust flow and residence time
More load means more exhaust flow through the same device. Higher flow means:
- Less time for gas to spend in contact with the catalyst, which can lower conversion at very high flow
- Higher face velocity through the filter walls
- Different particle deposition behaviour
Most RECDs are sized for their rated flow, so this effect is moderate within the normal operating range. It is still another reason the device’s behaviour at 30 kW and at 90 kW isn’t identical.
5. Back-pressure
Back-pressure rises with exhaust flow, and the RECD adds to it. At high load the total back-pressure on the engine is at its highest, which slightly changes how the engine breathes and how much fuel it burns for a given output. If back-pressure is excessive, it can itself increase smoke. Our article on how exhaust back-pressure affects a DG set explains the mechanism.
How a Load Mismatch Flatters or Punishes the Device
To see the size of the effect, take a 125 kVA DG set rated about 100 kW electrical. Assume that at matched loads the RECD performs like this. These are representative figures for illustration, not from a specific test:
| Load (kW) | PM without RECD (g/h) | Device efficiency | PM with RECD (g/h) |
|---|---|---|---|
| 80 | 18.0 | 75% | 4.5 |
| 55 | 12.0 | 67% | 4.0 |
| 30 | 8.5 | 62% | 3.2 |
Now look at three ways a site test might be run:
Scenario A: matched at 55 kW. Before 12.0 g/h, after 4.0 g/h. Reduction = 66.7%. This is the device’s true performance at that load, and it fails a 70% requirement.
Scenario B: baseline at 80 kW, post-RECD at 55 kW. Before 18.0 g/h, after 4.0 g/h. Reduction = (18.0 − 4.0) ÷ 18.0 = 77.8%. The same device now passes comfortably, because the engine was dirtier during the baseline.
Scenario C: baseline at 55 kW, post-RECD at 80 kW. Before 12.0 g/h, after 4.5 g/h. Reduction = 62.5%. A device that actually performs better at 80 kW is penalised because the baseline was cleaner.
The swing between B and C is more than 15 percentage points from the same hardware.
Doesn’t g/kWh fix this?
Partly. Expressing results in grams per kilowatt-hour divides by the work done, so it removes the simple “more load, more gas” effect. In Scenario B, the before result becomes 18.0 ÷ 80 = 0.225 g/kWh and the after result becomes 4.0 ÷ 55 = 0.073 g/kWh. That gives a reduction of about 67.7%, much closer to the true value.
It doesn’t remove the error completely, for two reasons:
- Specific emissions vary with load. Diesel engines are often dirtier per kWh at very light load than at mid-load, so the “before” g/kWh itself depends on the operating point.
- Device efficiency varies with load. The RECD’s own percentage changes from 62% to 75% across this range, and no unit conversion can correct for that.
Using g/kWh is good practice. Our explainer on mg/Nm³ vs g/kWh for RECD data covers when to use each. But a unit conversion doesn’t replace matching the load.
Matching the Load Isn’t Only About the Number
Two tests can both hold exactly 55 kW and still not be comparable, because of how the engine got there.
Approach direction and thermal history
If the baseline sample at 55 kW was taken after the set had been running at 90 kW, the exhaust system and RECD are cooling down towards their 55 kW equilibrium. If the post-RECD sample at 55 kW was taken after warming up at 30 kW, everything is heating up. Temperature-dependent processes in the device, such as catalyst activity, soot burning and condensation, will differ even though the kW reading is identical.
The fix is to approach each load point the same way in both tests (for example, always stepping down from the higher mode) and to wait for real thermal stability before sampling. We explain stability criteria in detail in how to account for warm-up time when testing an RECD.
Load stability during the sample
A gravimetric PM sample typically runs for tens of minutes per load point. If building load drifts during that time, for example when a lift cycles, a compressor kicks in or a chiller unloads, the sample averages across a range of conditions. Two samples that average 55 kW can contain very different load profiles.
Log load every minute during sampling, not once at the start. A report showing “Load: 55 kW” with no trend gives no assurance that it stayed there.
Building Load or Load Bank?
Most site tests use one of two load sources. Each has trade-offs:
| Factor | Building / plant load | Resistive load bank |
|---|---|---|
| Represents real operation | Yes | Partly (unity power factor) |
| Stability during sampling | Often poor | Excellent |
| Ability to hit exact kW | Limited | Precise |
| Can test high loads anytime | Only when the plant demands it | Yes |
| Power factor | Usually 0.8–0.95 | ~1.0 |
| Extra cost and logistics | None | Hire, cabling, safe placement |
| Effect on operations | Needs a planned changeover | Plant can stay on grid |
The most important rule is don’t mix them. A baseline on building load and a post-RECD test on a load bank (or the reverse) introduces a power factor difference, a stability difference, and usually a kW difference as well.
If the set normally runs on building load and you want realism, use building load for both tests, schedule both at similar times of day and week, and accept a slightly wider tolerance. If you need precision, which is usually the case for contractual acceptance, use a load bank for both.
Some testers use reactive (resistive-inductive) load banks that can reproduce a lagging power factor like 0.8. That gets closer to real operation while keeping control, and it is worth asking about for larger sets where alternator heating matters.
Choosing Which Load Points to Test
The type-approval reference
For constant-speed genset engines, type approval is based on the ISO 8178 D2 five-mode cycle: 100%, 75%, 50%, 25% and 10% load, each with a weighting factor. DieselNet’s summary of ISO 8178 cycles gives the modes and weights. RECD type approval in India is carried out under the CPCB system and procedure by recognised agencies, and the certificate tells you which engine range and conditions the device was proven on. Our guide on how to read an RECD test certificate explains what to check.
The site-acceptance reality
You don’t need to repeat type approval on site. You need to show that this device on this engine performs as promised. That calls for:
- At least one load point where the set really operates. Many standby DG sets in commercial buildings spend most of their running hours at 25–50% load. Your DG load logs and a look at load factor and duty cycle will tell you where.
- One higher load point, typically 75%, where exhaust is hot and the device should be at its best.
- A light-load point if the set often runs light. This is where weaker devices show their limits. If that describes your set, read choosing an RECD for a generator running at low load.
Whatever points you choose, write them into the test plan in kW, before testing day, and use the same points for both tests.
What about fluctuating loads?
Steady-state tests at fixed points don’t capture what happens when load jumps around, which is common in factories with cyclic machines, cold stores and construction sites. If your set behaves like that, read RECD selection for generators with rapidly changing loads and treat transient behaviour as a separate question from the steady-state comparison.
A Load-Matching Protocol You Can Write Into the Test Plan
Here is a protocol that can be pasted into a test plan or purchase order and adjusted for site conditions:
- Define load points in kW (electrical, at the alternator terminals), with the rated kW of the set stated alongside.
- Set a tolerance. About ±2–3% of rated kW for load-bank tests. A wider band may be agreed for building load, but it must be stated.
- Use the same load source (building or bank) for the before and after tests.
- Use the same approach sequence, for example 75% → 50% → 25%, stepping down each time.
- Log kW, kVA, power factor, voltage, current and frequency every minute from the controller or an independent power analyser.
- Start sampling only after thermal stability at the load point, using the same criterion in both tests.
- Void and repeat any sample where load leaves the tolerance band for more than an agreed time (for example, two consecutive minutes).
- Report average and range of kW for each sample, alongside the emission results.
- Record engine-side indicators: exhaust temperature before the RECD, back-pressure, and fuel consumption if available. These confirm the engine was in the same state.
- Both parties sign the load log on the day.
Our guide on writing measurable performance requirements into RECD tenders shows how to word clauses like these, and RECD performance guarantees tied to operating conditions explains why the guarantee should name the load points it applies to.
The Load Log: What Good Looks Like
A useful load record for each sample looks something like this (abbreviated):
| Time | kW | kVA | PF | Exhaust temp before RECD (°C) | Back-pressure (mbar) | Note |
|---|---|---|---|---|---|---|
| 10:41 | 55.2 | 56.0 | 0.99 | 362 | 48 | Sampling start |
| 10:42 | 55.0 | 55.8 | 0.99 | 363 | 48 | |
| 10:43 | 54.7 | 55.5 | 0.99 | 363 | 48 | |
| … | … | … | … | … | … | |
| 11:10 | 55.4 | 56.2 | 0.99 | 364 | 49 | Sampling end |
| Avg / range | 55.1 / 54.6–55.6 | 363 | 48 | Within ±2% |
With this log alongside the emission result, anyone can confirm the load was matched. Without it, the result rests on trust.
Common Excuses, and Why They Don’t Hold
“We tested at roughly half load both times.” “Roughly” can mean 45% one day and 60% the next. With emission and device performance both varying with load, that gap can move the reduction figure by several points.
“The DG controller showed the same kVA.” Same kVA with different power factor means different engine load. Check kW.
“We used g/kWh, so load doesn’t matter.” Normalising by kWh helps but doesn’t correct for device efficiency changing with load, as the worked example above shows.
“The plant load is whatever it is, we can’t control it.” Then use a load bank for both tests, or schedule both tests for the same operating pattern and log continuously, so the comparison is honest even if it isn’t perfect.
“The type-approval certificate already proves 70%.” It proves the device met the requirement on the test engine and cycle. Site acceptance proves the installation on your engine. Both matter, and they answer different questions. See factory acceptance vs site acceptance tests for RECD projects.
Where Load Matching Fits With the Other Test Variables
Load is the first of several variables that must be controlled for an RECD comparison to mean anything. The others are covered in companion articles:
- How to account for warm-up time when testing an RECD
- RECD testing during regeneration: what should be recorded
- How exhaust sampling location affects RECD test results
- How dilution can mislead RECD before-and-after measurements
For the overall plan and the reduction calculation, see planning an RECD emissions test and how to calculate RECD particulate reduction.
Frequently Asked Questions
What load tolerance should an RECD comparison test use?
There isn’t a single legally fixed figure for site tests, so agree it in writing beforehand. With a load bank, holding within about ±2–3% of rated kW is practical. With building load a wider band may be unavoidable, but then continuous logging becomes even more important.
Should I test at full load to get the best RECD result?
Full load often gives the best device efficiency because the exhaust is hottest, but it may not reflect how your set actually runs. A test only at full load can make a device look better than it will perform day to day. Include the load where the set spends most of its hours.
My DG set never runs above 40% load. Do I still need a high-load test?
For site acceptance, the most important point is where the set operates. A higher-load point is still useful as a reference, and it confirms the installation can handle peak flow and back-pressure if the set is ever loaded heavily, for example during a grid outage when all the building load transfers.
Can I compare a baseline taken last year with a post-RECD test today?
Only with caution. Besides load, engine condition, fuel and ambient conditions may all have changed. If the original load log is detailed and you reproduce the same kW points, the comparison is more defensible. Otherwise, consider an upstream/downstream test on the same day.
Does alternator size affect the comparison?
Not if the same alternator is used in both tests and load is matched in kW. The alternator’s efficiency slightly affects how much mechanical power the engine delivers for a given electrical kW, but it is the same in both tests.
Get Load Points Right Before the Test Starts
Most disputed RECD results can be traced to something that could have been settled on paper beforehand, and load is usually first on that list. Aceget supplies CPCB type-approved Retrofit Emission Control Devices for DG sets from 25 kVA to 2500 kVA, and we help clients define kW load points, tolerances and load sources before installation. You can review our certifications, read about DG set emission regulations, or contact our team with your DG nameplate photos. Here’s a guide to which nameplate photos we need.
For the regulatory background on RECDs in India, see the Central Pollution Control Board website and our explainer on why RECD is mandatory in India.