Dayton City Schools HVAC Installation 2025

Why one plant got a buffer tank and another did not

All three Dayton buildings lost equipment that held a great deal of water. Two of them had thermal storage tanks, and the third had a cooling tower and a heat recovery chiller. When the new air-cooled chillers went in, Residence Park received a 600-gallon buffer tank, Ponitz received a 400-gallon tank, and Charity Adams Earley received none. The difference is a calculation, not a preference.

A chiller needs water to push against

A chiller cools the water passing through it and watches the temperature of the water coming back. In a loop with very little water, the chilled water makes the round trip in seconds, the return temperature drops almost at once, and the machine shuts off, then restarts moments later. Compressors are rated for a limited number of starts per hour, and short cycling is one of the fastest ways to wear one out. Manufacturers therefore set a minimum system volume, usually expressed as a loop time of two to three minutes. At a typical 2.4 gallons per minute per ton, that works out to roughly 5 to 7 gallons of water in the loop for every ton of chiller.

Removing storage removed the volume

The old plants never had this problem, because a thermal storage tank is thousands of gallons of buffer by design. Demolishing the tanks took that volume out of the loop, and what remained was the water in the pipes and coils. Each building was checked against its new chillers. At Charity Adams the piping volume and flow were enough, and the contract records that no buffer tank was required. At Residence Park and Ponitz they were not, and a tank sized to make up the difference went into the chilled water specialties along with the new air separators and expansion tanks.

Converting LON to BACnet: which wire can stay

The control conversion produced a rule worth copying. Existing temperature control wiring was reused only where a run was whole from the sensor through the electrical trough to the terminal block, with no wire nuts and no junctions. Anything spliced was replaced. The network wiring was demolished and replaced everywhere, with no exceptions, and the reason is topology.

The LON networks in these buildings ran on free-topology wiring, which allows stars, loops and T-taps, so installers branched the cable wherever it was convenient. BACnet MS/TP runs on RS-485, which does not allow that. It requires a single daisy chain from device to device, with no stubs, terminated at both ends, on low-capacitance shielded twisted pair of 22 or 24 gauge, up to about 4,000 feet per segment. A branch on an MS/TP trunk reflects the signal and drops controllers off the network at random. The copper in the old LON cable was adequate. The layout was wrong, and it could not be corrected without pulling new cable.

The reuse rule for sensor wiring follows the same reasoning. A splice hidden in a ceiling is the usual cause of an intermittent fault, and those faults cost far more to find than the cable costs to replace.