Engineering Calculators
● Effectiveness-NTU Guide

Heat Exchanger Sizing: The Effectiveness-NTU Method

A plain-language guide to sizing a liquid-to-air heat exchanger with the effectiveness-NTU method: what UA, NTU, Cmin and Cr mean, why crossflow effectiveness has its own formula, and what limits a core's heat rejection.

Open the Heat Exchanger Sizing Calculator →

Why effectiveness-NTU beats "just use LMTD"

The classic way to analyse a heat exchanger is the log-mean temperature difference (LMTD), but it has an awkward property for design work: it needs the outlet temperatures you're usually trying to find, forcing an iterative guess-and-check. The effectiveness-NTU method flips the problem around. It asks: of the maximum heat this exchanger could possibly transfer, what fraction does it actually achieve? That fraction — the effectiveness ε — comes from a closed-form expression, so a radiator or oil cooler can be sized in a single forward pass with no iteration.

The three numbers that define an exchanger

Each fluid stream carries heat at a capacity rate C = ṁ·cp (W/K) — how many watts it absorbs or gives up per degree of temperature change. The smaller of the two streams, Cmin, sets the ceiling: the most heat that could ever be transferred is Q_max = Cmin·(T_hot,in − T_cold,in), because the weaker stream would then leave at the other stream's inlet temperature — thermodynamics allows no more. The ratio Cr = Cmin/Cmax describes how lopsided the two streams are. Finally, NTU = UA/Cmin (Number of Transfer Units) measures how much heat-transfer hardware the exchanger has relative to the load its weakest stream can carry — UA being the overall conductance, the product of the effective heat-transfer coefficient and area.

Why crossflow gets its own formula

Effectiveness depends not just on NTU and Cr but on the flow arrangement. Counterflow is the best possible; parallel flow the worst; and a radiator — where air crosses the tubes at right angles and neither stream mixes sideways — sits in between, with its own well-known approximation: ε = 1 − exp[(1/Cr)·NTU^0.22·(exp(−Cr·NTU^0.78) − 1)]. Two limits are worth knowing. When Cr → 0 (one stream overwhelmingly larger, or changing phase), every arrangement collapses to the same curve, ε = 1 − e^(−NTU). And as NTU grows, ε creeps toward 1 with brutally diminishing returns — doubling the core area of an already-effective exchanger buys very little extra heat.

Where UA comes from

UA is a series resistance network, exactly like an electrical circuit: the air-side film (usually the bottleneck, which is why cores are covered in fins), the tube wall (usually negligible for thin metal), and the coolant-side film. The fins don't count at full value — a fin's tip runs cooler than its base, so its area is discounted by a fin efficiency factor before it enters UA. Once UA is known, the whole solve is arithmetic: NTU, then ε, then Q = ε·Q_max, then each outlet temperature from a plain energy balance.

What actually limits a radiator, and a sizing checklist

Which side of the core is the real bottleneck, why more fin isn't always the answer, what louvers buy you, and a step-by-step checklist for a first-pass core sizing.

The air side is almost always the bottleneck

Liquid-side heat transfer coefficients run one to two orders of magnitude higher than air-side ones — water in a small tube manages hundreds to thousands of W/m²K where air over a fin manages tens to low hundreds. That imbalance is the entire reason finned surfaces exist: the air side compensates with area what it lacks in coefficient. It also means improvements should be aimed where the resistance is: a better coolant, higher coolant flow, or thicker tube walls barely move the needle once the air side dominates 1/UA.

The levers, in rough order of power

  • Frontal area — more face means more air mass flow and more fin area at the same time; the strongest single lever.
  • Face velocity — more air helps twice (higher h and higher C_air), but with diminishing returns and a steep fan-power/pressure-drop penalty.
  • Fin density — more area, but tighter channels raise pressure drop and eventually choke the airflow that feeds them.
  • Louvers — restarting the boundary layer at each louver edge buys roughly 2–3× on the air-side coefficient, at a comparable pressure-drop penalty; this is why virtually every automotive core is louvered.
  • Core depth / multiple rows — added rows keep helping, but each deeper row sees air that's already been heated, so the second row transfers less than the first.

First-pass sizing checklist

  1. Fix the duty: heat to reject, coolant inlet temperature and flow, worst-case air temperature and available face velocity.
  2. Compute both capacity rates; identify Cmin and Q_max = Cmin·ΔT_inlet. If Q_max is below the duty, no core geometry can save you — change the flows or temperatures first.
  3. Estimate air-side and coolant-side coefficients, discount the fin area by fin efficiency, and build UA as a series network.
  4. NTU → ε → Q. Compare against the duty with margin for fouling, non-uniform airflow, and correlation scatter.
  5. If short: grow frontal area first, then face velocity, then fin density/louvers — and re-check the fan operating point after any air-side change.
  6. Validate a final design against surface-specific correlation data (Kays & London-type) or test — a first-principles screen gets the trends and rough magnitude, not the last 20%.

← All guides · Open the Heat Exchanger Sizing Calculator