RTU Filter and Belt Program for a Property Manager

Purpose

A filter and belt program sold to a property manager does not fail on technique. It fails because nobody wrote down what is on the roof, so the truck arrives with the wrong filter sizes, and because the visit interval was a number somebody picked rather than one the equipment earned. The manager is buying two things: units that keep running, and a paper trail they can hand their owner. A shop that delivers only the first loses the account to whoever delivers both.

The measurement this program runs on is filter pressure drop against that unit's own recorded as-new drop, and its one counterintuitive rule is that a low reading proves nothing on its own. A filter that has been pulled out of its track by the airflow reads exactly like a clean one.

Scope

Covers a recurring filter and belt program across multiple rooftop units at one managed property: the asset register, the as-new baseline, the change trigger, per-unit cadence correction, belt condition tracking and the per-visit report.

Does not cover the annual service pass, owned by the light commercial RTU service SOP, or belt replacement mechanics and alignment, owned by the belt and sheave SOP, which this program calls out to. Does not cover roof access or fall protection, owned by their own SOPs.

Roles and responsibilities

Procedure

1. Build the asset register on the first visit, on the roof. For each unit: tag, position on a hand sketch of the roof, make, model and serial, filter size, count and type, belt cross-section and length, sheave part numbers, motor nameplate, and how you got up there. Acceptance: every field filled or marked not present, verified at the equipment rather than copied from the manager's spreadsheet. What wrong looks like: a truck arriving with five of the right size and one of the wrong, which means one unit runs a quarter unfiltered until the next trip. Stop rule: a unit you cannot access on the register day is recorded as inaccessible with the reason, not estimated. Hazard: this is roof work, so access is governed by the ladder SOP and everything on the roof surface by the fall protection SOP.

2. Record an as-new differential for each unit. With new filters in place and the coil in known condition, read static across the filter bank with a manometer through ports upstream and downstream of the rack, and record where those ports are. Acceptance: a baseline in inches of water column per unit, with port locations noted so the next tech reads the same two points. What wrong looks like: substituting a published initial resistance off the filter carton, which was measured at a face velocity this unit may not run. Stop rule: the baseline belongs to that rack at that coil condition, so it is re-recorded after any change to the rack, the filter specification or the coil. Hazard: drilling test ports means opening a cabinet, so the unit is isolated and proved dead per the disconnect SOP and you confirm what sits behind the panel before the bit goes in.

3. Set the change trigger before you set a calendar. Filters are changed when the measured drop reaches twice that unit's recorded as-new drop, or the filter manufacturer's published final resistance, whichever comes first. Acceptance: both numbers written on the register for every unit, so any tech can apply the trigger without a phone call. What wrong looks like: a single trigger applied across six units with different racks, face velocities and loading rates. Stop rule: a unit with no recorded baseline has no trigger, so it goes back to step 2 before it enters the cadence. Hazard: none physical, a decision made from the register.

4. Measure the differential at every visit and confirm the media is actually in the airstream. Read the drop with the unit running, then open up and confirm each filter is seated in its track with bypass gaps blocked. Acceptance: a differential value and a visual confirmation, both recorded, for every unit. What wrong looks like: a low reading accepted as a clean filter, when a filter sucked out of its track, or a 1 in filter in a 2 in rack, produces the same low number while the coil loads. Stop rule: any reading at or below that unit's as-new baseline is treated as a bypass finding until proven otherwise, because a used filter cannot restrict less than a new one. Hazard: the differential is read with the blower running through existing ports, and the visual check happens only after the unit is isolated and proved dead.

5. Fix the filter specification in writing with the manager, not on the truck. Agree the type, thickness and efficiency rating for every unit and record it once. Acceptance: a signed specification on file and the same specification on the kit list. What wrong looks like: a higher-efficiency filter fitted as an upgrade on a unit whose blower cannot carry the extra resistance, which drops airflow and hands you a comfort complaint and a frozen coil. Stop rule: efficiency ratings come from ASHRAE Standard 52.2, a consensus standard that binds only where adopted or specified, and whether a given unit tolerates a higher-rated filter comes from that unit's manufacturer data, not from the filter carton, so a change of rating is quoted and signed rather than substituted. Hazard: none physical, an office agreement.

6. Correct the cadence per unit from the trend, one step at a time. After each visit, compare the measured drop to the baseline. A unit that arrived past its trigger moves to a shorter interval; a unit comfortably under moves to a longer one. Acceptance: cadence changes made in 30-day steps, never more than one step per cycle, with the ratio that justified it recorded beside the change. What wrong looks like: a cadence halved and then doubled again on consecutive visits, which is a program oscillating instead of settling. Stop rule: a unit that hits its trigger two cycles running at the shortest interval you offer is not a cadence problem; it is a filtration or an infiltration problem, and it gets investigated rather than visited more often. Hazard: none physical, arithmetic done from the record.

7. Track belts on condition and on life, not on the calendar. At each visit inspect for glazing, sidewall wear, cracking and tension, and record the belt's installed date. Acceptance: a condition note and the running age per belt, with any replacement done to the belt and sheave alignment SOP, including its alignment measurement and run-in re-tension. What wrong looks like: belts changed on a fixed schedule, which replaces good belts and still misses the one on a misaligned drive that will not last the interval. Stop rule: a second belt failure at the same unit inside a year stops the replace-and-move-on cycle and triggers a drive geometry investigation. Hazard: the drive is inside a cabinet with a live control section, so isolation and the before-and-after instrument check happen before the guard comes off.

8. Produce the report the manager can forward, every visit. One page: units serviced, each measured differential against its baseline, filters changed, belt condition, exceptions with photos, cadence changes with their ratios, and anything needing approval. Acceptance: the report sent within one business day and filed against the register. What wrong looks like: a stack of individual work orders, which is data the manager has to assemble themselves before they can show their owner anything. Stop rule: an exception needing spend goes with a price and a consequence, not just a photo. Hazard: none physical, an office task, and the one that renews the agreement.

The record this produces

The register itself, plus a per-visit line for each unit: measured differential, the baseline it is compared against, the computed ratio, seated or bypassed, filters changed and to what specification, belt condition and age, cadence before and after, and every exception with a photo and a price. The signed filter specification and any approvals sit alongside it.

The ratio column is what makes the program defensible. A raw differential means nothing to a property manager and nothing to a tech who has not been to that roof, but "0.51 against a 0.22 baseline, 2.3 times, interval shortened to 60 days" explains itself to both, and it is the sentence that justifies the invoice without an argument.

Worked pass: six rooftop units at a managed strip center, first 90-day visit

Register and baselines were recorded in spring. Baselines: RTU-3 0.22 in w.c., RTU-4 0.24 in w.c., RTU-5 0.20 in w.c. All six on a 90-day interval to start.

RTU-3 measured 0.51 in w.c. running. 0.51 divided by 0.22 is 2.3, past the 2x trigger, so the unit had been running restricted for an unknown part of the interval. Filters changed, media confirmed seated, and under step 6 the interval moved one step from 90 days to 60.

RTU-5 measured 0.28 in w.c. 0.28 divided by 0.20 is 1.4, comfortably under the trigger. Filters changed as scheduled, and the interval moved one step the other way, from 90 days to 120. Same gate, opposite outcome, and the difference is the tenant: RTU-3 serves a bay with a roll-up door open most of the day, RTU-5 an office suite.

RTU-4 FAILED step 4. Its measured drop read 0.09 in w.c. against a 0.24 baseline. A used filter cannot restrict less than a new one, so under the stop rule this was treated as a bypass finding rather than a clean unit. With the unit isolated and proved dead, two of the four filters were found lying flat in the bottom of the rack: 1 in filters had been fitted into a 2 in rack by a previous vendor, and airflow had pulled them out of their tracks. The coil face behind them was matted.

The corrections were separated in the report: the filter specification for RTU-4 was corrected on the register and clips were fitted the same day, which the manager had already signed for; the coil cleaning went out as a priced exception with the photo, because it is spend the manager has to approve. Under step 2, RTU-4's baseline was marked void, to be re-recorded after the coil is cleaned, since the old figure belongs to a coil condition that no longer exists. Its cadence was not adjusted at all, because a ratio against a void baseline is not a measurement.

Belts on all six were inspected: five serviceable, one on RTU-2 glazed on both sidewalls and replaced to the belt and sheave SOP with its alignment measured and a run-in re-tension logged. The one-page report went to the property manager the next morning with the three ratios, the RTU-4 photos, the priced coil cleaning and the two cadence changes.

What the failure teaches: RTU-4 had the best-looking number on the roof. On a program that reads differentials and does not open the cabinet, it would have been reported as the healthiest unit at the site while running most of its air unfiltered across a loading coil, and the first symptom anyone would have seen is a high head pressure call in July.

References

• ASHRAE Standard 52.2 for filter efficiency reporting, a consensus standard binding only where adopted or specified; the tolerance of a specific unit for a higher-rated filter comes from that unit's manufacturer data
• The filter manufacturer's published initial and final resistance at a stated face velocity, used as the second half of the change trigger
• See related: RTU Belt and Sheave Alignment; Light Commercial Rooftop Unit Service; Rooftop Access and Ladder Safety on a Service Call