Cooling Coil Bypass Factor and Apparatus Dew Point

Air does not pass through a cooling coil in a single uniform stream. Some of it grazes a fin face and leaves saturated at the metal temperature; some of it shoots through the middle of a gap and leaves barely touched; most of it does something in between. Modelling that properly means solving a heat and mass transfer problem across every row. The trade solves it with one number instead, and the number works: imagine the coil as a PERFECT coil in parallel with a HOLE. A fraction BF of the air goes through the hole and comes out exactly as it went in; the remaining (1 − BF) goes through the perfect coil and comes out saturated at a single temperature called the apparatus dew point. Mix the two streams back together and you get the leaving condition you actually measured.

Because the mixing is linear, the bypass factor is just a ratio of temperature differences: (mathrm{BF} = (t_{la} - t_{adp})/(t_{ea} - t_{adp})). A four-row chilled-water coil with 27 °C on, 13.4 °C off and an 11 °C apparatus dew point has BF = 2.4/16 = 0.15. Its complement, the CONTACT FACTOR, is 1 − BF = 0.85, and some references and most European texts work in that instead — same model, same arithmetic, opposite labelling, so read which one a table is giving you before you use it. Rows and fin spacing set the value: roughly 0.30 for two rows, 0.15 for four, 0.08 for six, 0.05 for eight at ordinary face velocities.

The bypass factor is not a property of the coil. That last phrase — at ordinary face velocities — is where selections go wrong. Push the face velocity up and BF climbs with it, because each parcel of air spends less time against the metal. A coil that dehumidified properly at 2.5 m/s stops doing so at 3.5 m/s, which is exactly what happens when somebody speeds a fan up to chase an airflow figure and then cannot work out why the space has gone humid: the sensible capacity barely moved, the latent capacity fell off a cliff, and the room is now cold and clammy. Dirty fins do the same thing from the other direction — they raise the pressure drop, the airflow falls, BF improves, and the coil starts overcooling.

The apparatus dew point is worth its own paragraph because of what it is not. It is NOT the chilled-water supply temperature; it sits several degrees above it, because the metal is warmer than the water inside it and the fin tips are warmer still. It is not a temperature you can go and measure with a probe. It is the point where the coil process line, drawn on a psychrometric chart from the entering condition through the leaving condition, hits the saturation curve if extended — an equivalent surface temperature, defined by the model rather than found by instrument. It is genuinely useful all the same: because it lies on the saturation curve, its dry bulb, wet bulb and dew point all coincide, so reading the humidity ratio at saturation for that temperature tells you the moisture content the contacted air leaves with, which is what sets the coil's latent capacity. That is also how a selection runs backwards — pick the ADP that gives the required leaving humidity, then pick the rows that give the BF that produces the required leaving dry bulb.

Two honest limits. The straight line the model draws is really a curve, gently, so the ADP is an equivalent rather than an exact state; and if the process line drawn from your two measured points misses the saturation curve entirely, the coil is running dry — no condensation at all — and there is no apparatus dew point to find, only a sensible-cooling process that the model was never built to describe.