An acceptance test exists to turn a supplier’s promise into four numbers that can be signed: airflow at the duty point, static pressure across the fan, capture at the hood perimeter, and noise with the background subtracted. Almost every dispute about a kitchen exhaust system that “does not work” traces back to one of three procedural errors, a measurement taken at the wrong plane, a measurement taken under favourable conditions, or no measurement at all. This is the field procedure, the instrument list, the arithmetic and the sheet to hand to the commissioning engineer.
1. What the test is for, and what it is not
A commissioning visit is not a demonstration. A demonstration shows the fan turning and the lights on; an acceptance test produces values that can be compared with a design document and a tolerance. The purpose is narrow and commercial: it defines the moment the system is treated as delivered, it separates a design failure from an installation failure, and it gives the buyer a defensible reason to hold back the last payment.
The trap is that a system can pass on the day and fail in month two. A test run with clean filters, closed doors, make-up air running at full rate and a single hood in operation measures the best case the equipment will ever see. The numbers that matter are taken at the worst credible configuration, because that is the configuration the kitchen will be in at 20:00 on a Saturday.
2. The instrument kit
Six instruments cover the whole procedure, and the calibration date matters as much as the instrument. Anything with a certificate older than twelve months will be challenged the moment a result is inconvenient.
| Instrument | What it measures | Specification to insist on |
|---|---|---|
| Pitot-static tube plus micro-manometer | Velocity pressure and static pressure in the duct | Resolution 1 Pa, range to 2,000 Pa, tube long enough for the duct depth |
| Vane anemometer | Face velocity at the hood opening, room air movement | 100 mm head, averaging over 30 seconds |
| Smoke pencil or smoke generator | Capture at the hood perimeter, air movement direction | Non-toxic, low velocity, no draught of its own |
| Sound level meter | Noise at 1 m and at the boundary | Class 2 minimum, A-weighting, fast response |
| Acoustic calibrator | Verification before and after each measurement set | 94 dB at 1 kHz |
| Differential pressure gauge | Room pressure relative to adjacent spaces | Range to 50 Pa, readable to 1 Pa |
Add a tape measure, a ladder that reaches the access panel, and a printed sheet. A test recorded on a phone note and transcribed three weeks later is not evidence.
3. Airflow: the pitot traverse, done properly
Airflow is measured as velocity across a plane and multiplied by the area of that plane. The instrument is a pitot-static tube facing the airstream, connected to a manometer that reads velocity pressure. Mean velocity comes from mean velocity pressure, not from averaging the individual velocities, because the relationship is a square root:
v = square root of (2 x velocity pressure / air density), with air density around 1.2 kg/m3 at 20 C and standard pressure. A duct of 800 x 600 mm has an area of 0.48 m2; a mean velocity of 9.9 m/s gives 0.48 x 9.9 x 3,600 = 17,100 m3/h.
Three procedural rules separate a real reading from a decorative one. First, the traverse plane must sit at least six to eight equivalent duct diameters downstream of any disturbance and two to three diameters upstream of the next one: an elbow, a damper, a transition or a fan outlet all distort the profile, and a single-point reading in a distorted profile can be wrong by 30 percent. Second, the plane is divided into a grid, 25 points in a five by five pattern for a large duct and 16 points in a four by four pattern for a small one, with holes on two axes at right angles so the horizontal and vertical profiles can be checked against each other. Third, the tube is pushed to each point depth measured from the same side, and the reading is taken only when the value settles.
Five places where the measurement is meaningless, and all five appear in real reports: directly at the hood face with a vane held by hand, because the flow is turbulent and recirculating there; at the fan outlet in free air, because the system effect disappears and the fan looks better than it is; within one diameter of a damper, because the damper is the thing being tested and it is throttling the profile as well as the flow; in the grease-laden airstream with a hot-wire probe, because the wire fouls within minutes and the reading drifts downward; and with the kitchen doors open, because the measurement then includes whatever the building is doing to the kitchen rather than what the system is doing to the kitchen.
The acceptance value is airflow within plus or minus 10 percent of the design figure, at the duty point with the system configured as it will operate. Anything tighter than that is not achievable in a field measurement; anything looser than that will not capture a genuine shortfall.
4. Static pressure: the number that explains a low airflow
Static pressure taps at the fan inlet and outlet, read together with airflow, tell you whether the system is under-delivering because the fan is wrong or because the resistance is wrong. The two are fixed in completely different ways, and the readings separate them in about ten minutes.
Compare the measured total pressure rise with the fan curve at the measured airflow. If the fan is delivering its curve but the airflow is short, the system resistance is higher than the design assumed: a closed or partly closed damper, an unopened fire damper, a filter or cell section loaded beyond its clean state, a flexible connection collapsed under negative pressure, an undersized duct run added after the design, or an elbow used where the drawing showed a radius. If the fan is not delivering its curve at all, look at rotation direction, belt tension, sheave setting, variable-speed drive parameters, and whether the impeller is the one that was ordered.
One number closes most arguments: an airflow shortfall of more than 20 percent with a static pressure higher than design is an installation or ductwork problem, and it is not solved by telling the operator to clean the filters more often.
5. Capture: the smoke test that is not optional
Airflow in the duct does not prove capture at the hood. The two are linked by geometry, and geometry is what fails in retrofit kitchens where the canopy is 100 mm short of the appliance line, the overhang is less than the 150 to 300 mm that keeps the plume inside the canopy, or a make-up air grille is blowing across the cooking surface.
The test is simple and hard to fake. With the kitchen in its worst credible state, extract running, make-up air at its normal setting, the most remote appliance lit or simulated with a smoke source, pass a smoke pencil slowly along the perimeter of the hood opening: along the front edge, both ends, and the rear. Capture passes when the smoke is drawn into the canopy within one to two seconds anywhere on the perimeter, with no spill at the corners and no reversal at the ends. It fails when smoke lingers, drifts sideways along the ceiling, or is pulled back out of the canopy by a draught.
Then repeat it with the back door open and the make-up air off, which is what happens during a delivery. That second run is the one that predicts complaints. A system that captures only when the building is sealed is a system with a make-up air problem, not a hood problem, and the fix belongs in the air balance rather than in a bigger fan.
Recirculating hoods need one extra step, because there is no duct to traverse: verify that the unit is actually moving design airflow through the filtration path, that the ductless range hood range is running with its cell energised and its recirculation path unobstructed, and that no site modification has opened a bypass around the filters. A recirculating canopy that captures but leaks unfiltered air past the cell passes the smoke test and fails the reason it was bought. Mobile units add a third case: a cart tested indoors passes and then fails at a market with a cross-draught, so the mobile food cart range should be tested in the open with a side wind, which is the condition it was designed for.
6. Make-up air and room pressure
Measure the pressure difference between the kitchen and the adjacent space with everything running. A kitchen held between 0 and minus 10 Pa relative to the corridor is doing what it should: slightly negative, so odour does not migrate, but not so negative that doors are hard to open and the extract cannot get its air. A kitchen at minus 25 Pa is starved, and the symptom appears as poor capture at the far end of the cookline, which is often misdiagnosed as a fan problem.
Make-up air is normally designed at 80 to 90 percent of extract volume, with the balance drawn through transfer paths. Check three things on site: that the make-up unit is actually running at its design speed, that the grille it discharges through is not pointing at the hood, and that the transfer path exists. A kitchen with a closed fire door and no transfer grille has no make-up air at all, whatever the drawings say.
7. Noise: measure it with the background subtracted
Noise complaints are the most common reason an otherwise compliant system gets switched off, and the measurement is the most commonly botched. Procedure: sound level meter on A-weighting and fast response, at 1 m from the fan or hood, 1.5 m above floor, at least 1 m from any reflecting surface, and at the property boundary if that is the contractual limit.
Background must be measured first, with the system off, at the same positions. The correction is mechanical and non-negotiable: if the measured level is 10 dB or more above background, no correction is applied; a difference of 6 to 9 dB takes 1 dB off; a difference of 4 to 5 dB takes 2 dB off; and a difference of 3 dB or less is not a valid measurement at all, because the result is dominated by the background and must be repeated when the site is quieter.
Two contract figures usually apply together: 65 dB(A) at 1 m from the equipment, and a boundary limit that is set locally and is often 55 dB(A) by day and lower at night. A rooftop fan that meets the first and fails the second is a real and common result, and it is fixed with a silencer, a slower fan, or a different discharge direction rather than with a promise.
8. Purification performance: what can and cannot be verified on site
Collection efficiency cannot be measured in a working kitchen. It is a laboratory figure established by a type test on a representative unit, and the acceptance document should be that report, referenced by number, not a site claim. What can be verified on site is whether the equipment is operating in the condition the report assumed.
For an electrostatic stage that means: cell energised, current draw at the rated value, no audible or visible arcing, insulators dry and clean, collector plates straight and correctly spaced, and the access door interlock functioning. A cell drawing half its rated current is not collecting half the grease, it is collecting almost none, because the field collapses unevenly. The ducted ESP purifier range is therefore commissioned on current and voltage rather than on appearance, and those two readings belong on the test sheet beside the airflow figure.
Odour is verified at the boundary, not at the hood: three visits at different operating times, on the prevailing downwind side, assessed by people who did not install the system. If the contract contains an odour limit expressed in odour units, the sampling is a laboratory job and the acceptance schedule must name who pays for it.
9. The test sheet
One page, signed by the commissioning engineer and countersigned by the client representative. Every row carries a value, a location and an acceptance band, and a row left blank is a defect in the handover, not a minor omission.
| Parameter | Where measured | Acceptance |
|---|---|---|
| Airflow at duty point | Traverse plane, 6 to 8 diameters downstream of the last fitting | Design plus or minus 10 percent |
| Static pressure, fan inlet and outlet | Taps at the fan connections | Matches the fan curve at the measured airflow |
| Face velocity at the hood opening | Average over the open area, 30 second samples | 0.4 to 0.5 m/s heavy duty, 0.25 to 0.3 m/s light duty |
| Capture at the perimeter | Front edge, both ends, rear edge, worst-case configuration | Drawn in within 1 to 2 seconds, no spill |
| Capture with the back door open | Same points, make-up air at normal setting | No reversal at the ends |
| Room pressure | Kitchen to corridor, all systems running | 0 to minus 10 Pa |
| Make-up air volume | Measured at the supply unit or grille | 80 to 90 percent of extract |
| Noise at 1 m | Fan and hood, background subtracted | 65 dB(A) or the contractual figure |
| Noise at the boundary | Nearest receptor, background subtracted | Local limit, commonly 55 dB(A) by day |
| Cell current and voltage | Power supply display or meter, per cell | Rated value, no arcing |
| Grease duct velocity | Traverse in the main riser | 5 to 10 m/s transport velocity |
| Access and cleaning provisions | Panels, hatches, drain points, clearances | Openable, reachable, documented |
10. When it fails: the diagnosis order
Work in this order, because each step rules out a cheaper cause than the one after it. First, re-check the measurement: plane position, tube alignment, damper position, and whether the reading was taken before the system stabilised. Second, check the air path: fire damper open, balancing damper open, filters and cells in the clean state the design assumed, no collapsed flexible connection. Third, check the fan: rotation, belt, sheave, drive parameters, impeller identity. Fourth, check the make-up air, including the transfer path. Only then question the design, because in field data the design is at fault far less often than the installation.
Two commercial points belong in the contract before the test happens: who pays for a repeat test, and what happens to the retention. A five to ten percent holdback released on a passing test sheet, rather than on delivery, converts a commissioning visit from a courtesy into an event with consequences.
11. What the test protects
A signed test sheet does three things that no warranty clause does. It proves the system met its numbers at handover, so a later shortfall is a maintenance question rather than an installation argument. It gives the operator a baseline for the clean state, which is the only way to know when a cell is loaded or a filter is spent. And it moves the conversation from opinion to arithmetic, which is where a dispute about kitchen exhaust usually belongs.
None of it compensates for the wrong architecture. A ductless canopy fed by a multi-stage electrostatic cell and a carbon stage is commissioned on capture, airflow through the filtration path and cell current; a ducted system is commissioned on traverse and static pressure; a cart is commissioned in the open air it will work in. The N-series and ND-series canopies, the ESP Purifiers cell sections and the carts are all handed over against the same sheet, because the sheet tests the installation rather than the equipment.
