The New Standby That Could Not Start the Heat Pump

The call

A 22 kW air-cooled standby on natural gas, installed in the autumn. The first winter outage came in January. The unit started, transferred and carried the house for about forty minutes. Then the thermostat called for heat, the lights dimmed hard, the engine note dropped, and a few seconds later the generator shut down. The controller restarted it after its delay, it transferred again, and the same thing happened the next time the heat pump called. The customer gave up, set the thermostat off, and called the shop from a cold house the next morning.

The remote monitoring log showed three shutdowns, each logged as underfrequency, each within about two seconds of a load step the controller recorded at the start of a heat call. That timing is the first clue: a running overload builds over minutes as loads stack up, while a motor start that the engine cannot carry pulls frequency down in the first second or two.

Every test below meant running the unit under load and putting meters on energized conductors. The unit was locked out under The Generator Auto-Start Lockout Before Service SOP whenever hands went into the enclosure or the transfer switch, and the inrush reading was taken as How to Measure AC Compressor Inrush with a Clamp Meter lays out: jaws placed dead, read live, by a qualified person in the PPE the shop's assessment calls for.

What the site-walk sheet said

The sizing sheet listed the heat pump's compressor at its rated-load amps, 18 A at 240 V, about 4.3 kVA running, plus the air handler. It listed the rest of the house loads at running values and arrived at a peak running load comfortably under the unit's output. It did not record the compressor's locked-rotor amps, did not record a measured inrush, and took the unit's rated output from the propane column of the spec sheet.

Two things were wrong with that before anyone opened a meter. A heat pump's compressor starts on several times its running current, so the running figure says nothing about whether the engine survives the start. And an air-cooled unit on natural gas is rated for less output, and less starting capability, than the same unit on propane, because natural gas carries less energy per cubic foot of mixture the engine can draw. The spec sheet for this unit listed, say, a starting capability of 80 A on natural gas and 95 A on propane at the sheet's stated voltage dip.

Ruling out the fuel supply

A natural gas unit that bogs under a big load step is very often a unit starving for gas: an undersized line or meter lets pressure fall under load, and the engine loses power. That had to be eliminated first, because if it were true, no amount of electrical work would fix the heat pump problem.

The tech read gas pressure at the unit's test port, static and at full load, as in How to Measure Gas Pressure at the Generator, Static and Under Load, using a load bank to hold the unit at its natural gas rated output. Both readings sat inside the maker's spec, and the full-load reading held steady for the full run with no sag. The supply was not the cause.

Ruling out a running overload

The second candidate was the house simply asking for more than the unit could carry once heating started. Heat pumps carry auxiliary electric heat strips, and if the strips energize with the compressor during a defrost or a large temperature recovery, the running load jumps by several kilowatts at once.

Two facts eliminated it. The installer had put the strips on a load-shed relay that the transfer switch's load management module held off whenever the generator was the source, and the tech confirmed the relay stayed open on generator power. And the shutdown timing did not fit: with the strips held off, the running house load with the heat pump going was about 9 kW of other loads plus about 4.3 kVA of compressor and the air handler, well inside the unit's natural gas output. An overload trip would not have arrived two seconds after the contactor closed.

Ruling out the engine and governor

The third candidate was the generator itself: a governor, ignition or valve problem that let speed fall under a large step. On the load bank the unit took resistive steps up to its natural gas rated output and held frequency within the controller's normal recovery each time. An engine that recovers from a full resistive step is not the weak link. A resistive load bank cannot reproduce a motor start, which is the point: the engine handled any steady load the house could present, so what remained was the start itself.

The number that settled it

With the gas, the running load and the engine eliminated, the tech measured the compressor's start. The condenser plate read 112 A locked-rotor. The clamp meter in inrush mode, on one leg at the disconnect, read 88 A on a normal start after the system had equalized. Starting kVA is the inrush current times the voltage: 88 A x 240 V = 21,120 VA, about 21.1 kVA. The unit's natural gas starting capability was 80 A, which at 240 V is 19,200 VA, about 19.2 kVA. The compressor asked for about 1.9 kVA more than the unit could deliver at its rated dip, about 10 percent over, before counting the 9 kW of the house already on the unit. How starting kVA and the NEC code letter fit together is owned by Motor Starting kVA and the NEC Code Letter; this case only needed the measured figure.

On propane the same unit's 95 A capability would have covered the 88 A start. Read from the propane column, the sizing looked fine. On the fuel the house actually had, it was short.

The tech reproduced the failure on purpose with a simulated outage: utility opened to the transfer switch, the house on the generator at its normal load, then the thermostat turned up to call for heat. The lights dipped, frequency fell, and the controller shut the unit down on underfrequency two seconds after the contactor pulled in, exactly as logged. Utility was restored at once.

The fix and the retest

The interim, first. The HVAC contractor could not fit anything for a week, and a second outage in that week would repeat the cycle: start, transfer, heat call, shutdown, restart, every few minutes, which is hard on the engine and the transfer switch and leaves the house with neither heat nor reliable lights. So before leaving that day the tech set the load management module to hold the heat pump off entirely whenever the generator was the source, confirmed it on a simulated outage, and told the customer in writing: the generator would carry the house but not the heat pump until the fix, and a gas fireplace or a listed electric space heater on a generator-backed circuit was the fallback. Not a charcoal grill, not a gas oven, not anything run indoors that is not built to be.

There were three options for the real fix: a larger generator, keeping the heat pump off the generator, or making the start smaller.

• A larger unit would have worked and was the costliest by a wide margin, and it would have left a generator oversized for every hour of the outage except the few seconds of each compressor start.
• Keeping the heat pump off the generator would have left the house without heat in a winter outage, which defeats the purpose of the install.
• A compressor soft starter reduces inrush by ramping the compressor up rather than slamming it across the line. Installed by the HVAC contractor with the equipment maker's approval, so the condenser warranty stood, it was the smallest change that fixed the actual problem.

After the soft starter went in, the inrush reading on the same leg, same meter, same equalized start, was 44 A: 44 A x 240 V = 10,560 VA, about 10.6 kVA, half the 21.1 kVA before and well inside the 19.2 kVA capability on natural gas.

The tech also changed one setting in the load management module, per Load Shed Module Programming: the heat pump was assigned to a managed input that lets the compressor start only when the house load is below a set threshold, and holds it off for a few seconds after any other large load starts. That keeps the compressor from starting at the same moment as a well pump or a dryer, which could push even the softened start over the edge.

The house as it ran on generator power after the fix, with the two loads the module controls:

 [22 kW standby, natural gas]
            |
    [transfer switch]-----[load management module]
            |               |                 |
      [house panel]         | holds open on   | start only
            |               | generator       | below the
            |               |                 | threshold
            +--[load-shed relay]--[strips]    |
            |                                 |
            +--[soft starter]--[heat pump compressor]
            |   start 88 A before, 44 A after
            +--about 9 kW of other house loads

Retest. Simulated outage again, house on the generator at about 9 kW, thermostat turned up. Frequency dipped briefly at the compressor start and recovered, the lights barely flickered, the unit stayed online. Three more heat calls over the next hour, all carried. Utility restored, retransfer and cool-down normal, controller back in auto. The remote log showed the next real outage, three weeks later, carried through the night with eleven compressor starts and no shutdowns.

Where the sizing went wrong

The sheet was built on running load, which is the right way to find the size a unit has to carry for hours and the wrong way to find whether it can start the biggest motor. The site walk for any house with a central air conditioner or a heat pump needs the compressor's measured inrush, or at least its plate locked-rotor amps, compared against the starting capability for the fuel the house actually has. How to Size a Standby Generator From a Site Walk owns that procedure, and its starting check would have flagged this unit before it was sold: 88 A measured, or the plate's 112 A locked-rotor, against 80 A is a fail on paper. The code points the same way: NEC Article 702, as adopted by your jurisdiction, requires an optional standby system with automatic transfer to have capacity for the load it serves, or a load management system that limits that load to what the unit can carry, in substance; check the current edition.

Two things would have changed the answer. A compressor already fitted with a soft starter or a variable-speed drive starts on far less, and the same unit would have carried it from day one. And a site on propane, with the same compressor, would have started it on the 95 A capability without any change; the fuel decided this case as much as the motor did.

References

• See related: Motor Starting kVA and the NEC Code Letter
• See related: How to Measure AC Compressor Inrush with a Clamp Meter
• See related: How to Size a Standby Generator From a Site Walk
• See related: Load Shed Module Programming
• NEC Article 702, as adopted by your jurisdiction (optional standby systems sized to the load they serve)