The Airlock That Looked Like a Dead Pump

Why this matters

Two circulators went into this building in five weeks, both replaced on a "pump is dead" call, and neither one was dead when it was pulled. The fault was reconstructed afterwards almost entirely from the paperwork, because what the file did not contain turned out to be the finding. No visit recorded a pump differential. No visit recorded system fill pressure except one. That combination is what allows a spinning pump to be condemned twice, and it is common enough that a shop can fix it with two extra fields on a form.

The safety calls this job forced first

• Venting a high point on a hot loop discharges hot water. The tech shut the boiler and let the loop fall below 120 F before opening any manual vent. Where waiting is not an option, pipe the vent through a hose into a floor drain, stand to the side of the discharge, and wear a face shield and heat-rated gloves. Water held above 212 F under pressure flashes to steam when the vent opens, so read the loop temperature before touching a vent, and do not open one above that.
• Before pulling a circulator, lock the disconnect open and verify the motor is dead, proving the meter on a known live source before and after (NFPA 70E-2021, 120.5). The mechanical and stored-pressure isolation is 29 CFR 1910.147; work inside an energized panel is 29 CFR 1910.333(b)(2).
• Relieve loop pressure and confirm zero on a gauge before breaking the circulator flanges. A closed loop at fill pressure will empty itself through an open flange at whatever temperature it is currently holding.
• Do not gag, plug or adjust a weeping relief valve to stop it. That instruction matters here specifically, because a version of it is what caused this whole failure.

What the file said

Three visits, one building, a three-story loop off a boiler in the basement.

• Visit 1. "No heat, second and third floor. Circulator hums, no flow. Replaced circulator. Reset boiler high limit."
• Visit 2, eleven days later. "Same complaint. Circulator replaced under warranty. Reset boiler high limit. Advised customer may need larger pump."
• Visit 3. "Same complaint."

Three months before visit 1, a fourth entry that nobody connected to anything: "Relief valve weeping on heat-up. Backed down fill pressure. Monitor."

Both replaced circulators were bench-tested at the supply house. Both passed. That is the fact that broke the case open, because it removed the only explanation anyone had offered.

What the file did not say

Reconstructing backwards, the gaps line up in one direction:

• No pump differential on any visit. Nobody put a gauge on both sides of the circulator. This is the reading that separates a pump that cannot make head from a pump that has nothing to make head in, and it is the only reading that separates them.
• No system pressure on visits 1 or 2. It appears exactly once, at the bottom of visit 3.
• No expansion tank precharge, ever. No entry in the file mentions the tank at all.
• No pipe temperatures by floor. "No heat on two and three" was recorded as a customer complaint, not as a measurement.

Two things that were recorded turned out to matter more than anyone realized. The boiler high limit tripped on both visits. A boiler that is making heat and tripping its own limit is telling you the heat is not being carried away, which is a flow statement, not a heat-source statement. And the pumps kept passing on the bench.

The one number somebody wrote down

Visit 3: system pressure 6 psi, cold.

Convert it. One psi supports 2.31 ft of water column, so 6 psi supports 6 x 2.31 = 13.9 ft of water above the gauge.

The highest point of this loop is roughly 30 ft above that gauge, at the third-floor ceiling.

So the water in the risers stood about 13.9 ft up, and the top 16 ft of the loop had no water in it at all. Not an air pocket. A void the height of a story and a half, sitting exactly where the complaint was.

What the building actually needed

The fill pressure has to lift water to the highest point and then leave it positively pressurized, so that air vents up there have something to push against and so the loop cannot pull air in through a vent or a packing.

minimum fill = (height above the gauge / 2.31) + a cushion
minimum fill = (30 / 2.31) + 4 = 13.0 + 4 = about 17 psi

The 4 psi cushion is common hydronic practice rather than a physical constant; take the actual figure from the system design or the air-vent manufacturer's requirement. The reducing valve was set to 18 psi, a little above the 17 psi floor.

Against that, 6 psi was not slightly low. It was about a third of what the building required, and no pump made was going to fix it.

Now the three-month-old entry reads differently. The relief valve was weeping on heat-up because the expansion tank had lost its precharge and could no longer absorb the water's expansion, so pressure spiked every cycle and lifted the relief. Backing down the fill made the weeping stop, exactly as intended, and drained the top of the building in the process.

Why two working pumps read as dead

A centrifugal pump produces head in feet of the fluid it is actually pumping. Thirty feet of head in water is 13 psi. Thirty feet of head in air is essentially nothing, because air is roughly 830 times less dense. So a circulator with a void on its suction side spins at full speed, sounds completely normal, and produces a pressure differential a gauge can barely see.

Measured on visit 3, before anything was touched: 0.4 psi across the circulator, against a published shutoff head for that model in the range of 14 ft, which is about 6 psi. Not a weak pump. A pump with nothing to work on.

Motor current supported it without proving it. Nameplate 0.9 A, measured 0.55 A. On a centrifugal machine, low current means the operating point moved toward the left of the curve, not that the motor is failing. Be honest about the weight of that clue: on a small wet-rotor circulator the current swing between full flow and no flow is modest, so treat it as corroboration for the differential reading rather than as evidence on its own.

The second pump is the interesting one, because it genuinely was dead by the time it came out, and that is why the wrong theory kept getting confirmed. A wet-rotor circulator uses the pumped water as its bearing lubricant and its motor coolant. Run one in a void and the bearings run dry and hot. The first pump was pulled before that finished; the second one ran eleven days and did not survive. The shop's own repair created the evidence for its own misdiagnosis.

The repair, in the order it had to happen

Order matters here, and getting it wrong wastes the visit.

1. Expansion tank first. With the system side isolated and drained, check the tank's air precharge and compare it against the intended cold fill pressure. They should match. A tank that reads no pressure at the valve, or that spits water out of the valve, is finished. Replace or recharge it. Doing this after the fill is raised means doing it twice, because the tank has to be set against the pressure you are about to establish.
2. Set the reducing valve to 18 psi cold.
3. Purge from the bottom up, floor by floor, isolating branches so the fill water is forced through one path at a time rather than taking the easy loop. Manual vents at each high point, opened one at a time, after the loop is below 120 F.
4. Fit or verify the air removal that should have been there: an air separator near the boiler outlet and an automatic vent at each high point. A loop that cannot shed air will re-accumulate it every time it is refilled.
5. Then, and only then, run it and take readings.

How they proved it

Three measurements, none of which had ever been taken in this building:

• Circulator differential: 5.6 psi, which is 5.6 x 2.31 = 12.9 ft, and lands where the published curve puts that circulator in this loop's flow range. Against the 0.4 psi measured before the repair, this is the reading that settles it.
• Supply and return temperatures at the boiler: a 22 F spread, against a design spread of 20 F. Report that honestly rather than rounding it into agreement: 22 F against 20 F implies flow running roughly 9 percent below design, which is small enough to be reading tolerance on two surface probes and large enough to be worth re-taking next visit with a well-mounted probe. It is not a failure, and it is not a match either.
• Third-floor riser hot to the hand along its full height. Qualitative, and the fastest confirmation available.

Cold fill pressure was rechecked at the following visit and still read 18 psi, which confirms the tank is holding and the loop is not feeding.

What goes in the record so this cannot repeat

The gaps were the fault. Close them with fields, not with training.

• System pressure, cold, on every hydronic visit. One number, five seconds, and it would have ended this on visit 1.
• Circulator differential on any no-flow or low-heat complaint. A pump is never condemned on sound. It is condemned on a differential compared against its published curve.
• Expansion tank precharge at every service interval, with the reading written down rather than "checked."
• Any adjustment to a fill or relief setting gets a reason and a follow-up flagged. The entry that caused this said what was done and not why the symptom existed. An entry reading "relief weeping on heat-up, suspect expansion tank, reduced fill as temporary measure, return within two weeks" would have carried the diagnosis forward instead of burying it.

References

• ASHRAE Handbook, HVAC Systems and Equipment, for hydronic fill pressure, expansion tank sizing and air control
• Circulator manufacturer published curve and shutoff head data for the installed model
• OSHA 29 CFR 1910.147 for mechanical isolation and stored pressure; 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5 for verifying the motor circuit is dead
• See related: Airlock vs Blockage (decision tree); How an Operating Point Moves; What a Closed Loop Does That an Open One Does Not