How Is HVAC Noise Control Engineering Done in Buildings?

HVAC noise control engineering designs mechanical systems to meet room-by-room noise criteria (NC or RC ratings). It selects equipment for sound power, sizes ducts for low velocity, places silencers based on acoustic analysis, and details diffusers, returns, and equipment isolation so the system conditions the air without being heard. The fundamental relationship is velocity: noise in ducts rises steeply with air velocity, so quiet systems move a lot of air slowly through big ducts. Every shortcut — smaller ducts, faster fans, cheaper equipment — shows up as noise in the occupied space.

The acoustic analysis is the core deliverable: an octave-band calculation from each noise source (fans, VAV boxes, diffusers, duct fittings) through the duct system to each room, compared against the room's NC target. This tells the designer exactly where silencers are needed and how much insertion loss each must provide — no more, no less. Equipment selection carries the rest. Fans, air handlers, and rooftop units all publish sound-power data, and the engineer selects for it the same way they'd select for efficiency. A unit that's 5 dB louder at the source needs that much more silencing downstream — or it simply can't serve a quiet room. Vibration isolation (springs, hangers, inertia bases) keeps structure-borne energy out of the building frame.

Mechanical acoustics fail when equipment is value-engineered after the acoustic design. Swap the selected air handler for a louder one and the silencer schedule no longer works — but nobody re-runs the analysis. The acoustic selections have to be protected like structural ones. Here's what I lock in on mechanical projects.

• Set NC targets room by room in the design criteria: the analysis needs targets to design against
• Select equipment for sound power, not just capacity: the data sheets exist — use them
• Size ducts for low velocity in quiet zones: velocity is the master variable
• Protect acoustic selections from value engineering: a louder substitution voids the analysis
• Verify with field NC measurements: testing after startup catches what the math missed