Thermal Comfort and ASHRAE 55

Every thermal comfort problem is, at root, an energy balance problem. The human body is a heat engine that never fully idles: a seated adult releases heat at roughly the rate of a small incandescent lamp, and that rate multiplies several fold with activity. Thermal comfort is the subjective report of a body that is disposing of its heat without calling on emergency measures. When heat production and heat loss are matched while skin temperature and sweat rate rest near their neutral values, occupants describe themselves as neutral. When the balance tips, the autonomic nervous system intervenes, first quietly by redirecting blood flow and nudging skin temperature, then overtly through shivering or sweating, and comfort deteriorates. HVAC engineers therefore treat comfort as a physics problem first: balance the body's heat flows and the sensation usually follows.

The production side of the ledger is the metabolic rate, expressed in met units. One met equals 58.2 W/m², the heat release of a seated person at rest, normalized by body surface area (about 1.8 m² for an average adult, the DuBois area). In whole-person terms, 1 met is therefore roughly 105 W. Representative values used in design are 1.0 met for seated quiet rest, 1.2 met for office work and typing, about 1.6 met for standing light activity, and about 2.0 met for slow walking. The per-area normalization matters because the balance that governs sensation is per unit area, whatever the body size. There are two practical consequences for the engineer. First, metabolic rate sets the absolute cooling load the space must remove. Second, it appears inside the comfort equations themselves, so a wrong met assumption corrupts both the load calculation and the comfort prediction at once.

The resistance side of the ledger is clothing, expressed in clo units. One clo equals 0.155 m²K/W, the insulation of a typical business suit. Garments in an ensemble add approximately in series: light summer clothing totals about 0.5 clo, while winter indoor dress totals about 1.0 clo. Two subtleties routinely escape designers. First, clothing resists moisture transport as well as dry heat flow, so heavy clothing traps humidity at the skin and raises discomfort faster than its dry insulation alone would suggest. Second, real ensembles vary with posture, fit, and furniture: a padded office chair adds measurable insulation beneath a seated occupant. The design consequence is direct. The clo value assumed in design selects the required room temperature, so a space tuned for 1.0 clo will feel cool to occupants dressed at 0.6 clo. That is why competent designs carry separate summer and winter clothing assumptions instead of one number for the whole year.

Fanger's experiments showed that six factors move the average thermal sensation vote in a meaningful way. Two are personal: metabolic rate and clothing insulation, discussed above. Four are environmental: air temperature, which drives convection; mean radiant temperature, which drives radiation; air speed, which accelerates both convection and evaporation; and humidity, which sets the ceiling on evaporative loss. A wall thermostat measures exactly one of the six, which is the original shortcoming of naive comfort control: holding air temperature constant while radiation, air movement, clothing, or activity drift is no guarantee of comfort, and most chronic complaints trace back to one of the other five factors.

Convection is the exchange between skin or clothing and the moving air around it. Even in a nominally still room the body drives its own gentle plume of rising warm air, so natural convection never fully stops. Forced air movement from diffusers, fans, or open windows raises the convective heat transfer coefficient, which is why a breeze cools. The same mechanism turns hostile when it cools one body part much faster than the rest: the neck and ankles are particularly sensitive, and that local overcooling is called draft, treated in Session 4.

Radiation is frequently the largest sensible path indoors, and it follows the fourth power of absolute temperature, which makes it brutally sensitive to hot or cold surfaces. Because real rooms are non-uniform, engineers use the mean radiant temperature (MRT): the uniform temperature of an imaginary enclosure that would produce the same radiant exchange with the body as the actual room. A cold window dominates perimeter comfort because the body faces it across a large view factor; occupants near glazing complain of chill in rooms whose air temperature is perfect. The net radiant exchange between a surface and its surroundings takes the familiar Stefan-Boltzmann form:

Evaporation is the body's pressure relief valve. Moisture diffuses through the skin continuously, and sweat glands add regulated evaporation under heat stress; each kilogram evaporated carries away roughly 2400 kJ. It is the only heat path that still works when the air is warmer than the skin, which is why humidity dominates comfort in hot conditions and barely registers near neutral. High humidity does not make a temperate room feel much warmer, but it makes a hot room feel oppressive, because it throttles the one path the body was counting on.

Conduction and respiration close the account. Conduction through contact with chairs, desks, and floors involves small areas, so it is minor for the whole body but decisive locally: a cold slab underfoot ruins an otherwise balanced room. Respiration warms and humidifies inhaled air; a few percent of total loss at rest, it grows with activity as breathing deepens. Neither path usually drives a design decision alone, but both sit inside the full heat balance, and conduction explains several classic local-discomfort complaints.

Taken together, the five paths must sum to metabolic heat production for the body to sit in balance. Fanger's achievement was to write that balance as a single equation, estimate the skin temperature and sweat rate a comfortable person would exhibit, and link the residual imbalance to the average vote of building occupants. That equation, and the dissatisfaction curve derived from it, is the subject of Session 2.

Key takeaways

• Thermal comfort is the sensation of a body whose heat production is balanced by heat loss with minimal regulatory effort (no shivering, no sweating).
• Metabolic rate (1 met = 58.2 W/m²) sets both the cooling load and the comfort sensitivity; clothing (1 clo = 0.155 m²K/W) sets the insulation between skin and room.
• Six factors govern comfort: metabolic rate and clothing (personal), plus air temperature, mean radiant temperature, air speed, and humidity (environmental).
• Heat leaves the body by convection, radiation, evaporation, conduction, and respiration; radiation usually dominates sensible exchange indoors, while evaporation dominates in hot conditions.
• A thermostat senses one factor of six, so air-temperature control alone cannot guarantee comfort.

In the late 1960s, P. O. Fanger posed a question that still organizes comfort engineering: can the average thermal sensation vote of a large group be predicted from physics alone? Working with subjects in climate chambers, where every one of the six factors from Session 1 could be set independently, he collected thousands of votes on a seven-point sensation scale and regressed them against the body's heat balance. The result was the comfort equation and, from it, the Predicted Mean Vote (PMV): a single number that estimates how a large group will vote, on average, in a given uniform, steady environment.

The core of the model is the thermal load, L. Imagine a person whose skin temperature and sweat rate are fixed at the values a comfortable person would show for their activity level. L is the difference between the body's internal heat production and the heat loss that would occur under those neutral skin conditions, in W/m² of body surface. When L equals zero, production and loss are matched at neutral skin conditions, and the predicted mean vote is zero. When L is positive, the body retains more heat than a comfortable person would, and the vote moves to the warm side; when negative, to the cool side. L therefore compresses all six factors into one signed imbalance, which is what makes the model so useful for design iteration: change a diffuser layout, a glazing choice, or a setpoint, recompute L, and read off the comfort consequence.

PMV maps that load onto the familiar seven-point scale through an activity-dependent sensitivity factor:

Two things about this equation deserve attention. First, PMV is a mean, not a promise to any individual. A PMV of +1 means the group's average vote lands on "slightly warm"; some occupants vote neutral, a few vote warm, and the distribution around the mean is what the PPD curve quantifies. Second, the sensitivity factor in parentheses shrinks as metabolic rate rises (about 0.065 at 1 met, falling toward 0.03 at high activity). Physically, an active body with elevated blood flow and sweating capacity absorbs a given watt-per-square-meter imbalance with a smaller shift in sensation. Designers sometimes miss this: the same 10 W/m² load swings PMV further in a quiet office than on a factory floor.

Dissatisfaction follows from the vote through the PPD equation, the Predicted Percentage of Dissatisfied:

The curve's most important feature is its floor: at PMV = 0, PPD equals 5%, not zero. Even in a thermally perfect room, about one person in twenty votes dissatisfied, because people differ in basal metabolism, body composition, fitness, expectation, and plain preference. Unanimous comfort is physiologically impossible, and any design brief demanding it is incoherent. The working target derived from this curve, and the one embedded in the ASHRAE 55 analytical method, is PPD below 10%, which corresponds to PMV between -0.5 and +0.5. That band, not the single point of neutrality, is what engineers design to.

The model needs all six factors as inputs: the four environmental quantities (air temperature, mean radiant temperature, air velocity, and humidity as vapor pressure) plus the two personal estimates (metabolic rate and clothing insulation from tables or observation). In practice the environmental four are measured or simulated and the personal two are estimated, which makes met and clo the dominant uncertainty in most PMV calculations. A clo estimate wrong by 0.2 shifts PMV by a noticeable fraction of a scale unit, so serious comfort work always includes a sensitivity pass on the assumed clothing and activity.

Finally, know the model's boundaries; Fanger derived it for steady, uniform conditions, and it is published with application intervals (roughly 0.8 to 4 met, 0 to 2 clo, air temperature 10 to 30 C, mean radiant temperature 10 to 40 C, air speed 0 to 1 m/s, vapor pressure 0 to 2700 Pa). Do not apply PMV where its premises break: transient exposures such as lobbies where people arrive from outdoors, strongly non-uniform spaces (use the local-discomfort criteria of Session 4 instead), or occupant-controlled naturally ventilated buildings, where field studies show the adaptive model of Session 3 predicts better. PMV also says nothing about air quality, and it cannot capture the psychological boost occupants get from personal control over their environment. Used inside its limits, it remains the sharpest whole-body comfort instrument the profession has.

Key takeaways

• PMV predicts the mean thermal sensation vote (-3 cold to +3 hot) of a large group from the thermal load L, the signed imbalance between heat production and neutral-condition heat loss.
• The sensitivity factor falls with rising metabolic rate, so active occupants respond less per W/m² of imbalance than sedentary ones.
• PPD bottoms out at 5% at neutrality; the practical design target is PPD < 10%, i.e., PMV within -0.5 to +0.5.
• All six factors are required inputs; met and clo estimates are usually the largest source of uncertainty.
• PMV is valid for steady, uniform environments inside its published intervals; use local-discomfort criteria for non-uniform spaces and the adaptive model for naturally ventilated buildings.

For decades, comfort standards were built almost entirely on climate-chamber data, and then field researchers started measuring real buildings and found something the chambers had missed. In the ASHRAE RP-884 project, Richard de Dear and Gail Brager compiled thousands of paired occupant votes and indoor measurements from offices around the world. In air-conditioned buildings the votes tracked PMV reasonably well. But in naturally ventilated buildings, occupants reported comfort across a far wider range of indoor temperatures than the chamber model predicted, and their preferred indoor temperature drifted with the seasons. People were not passive sensors; they were adapting, and the standard had to account for it.

The adaptive hypothesis is disarmingly simple: if a change produces discomfort, people react in ways that tend to restore their comfort. De Dear and Brager organized those reactions into three modes. Behavioral adaptation is the most visible: opening windows, switching on fans, adjusting blinds, changing posture, altering clothing, moving to a cooler spot, or simply slowing down. Physiological adaptation is acclimatization over days and weeks: sweat glands respond earlier and more copiously, blood flow redistributes, and the body tolerates heat it would have protested a week earlier. Psychological adaptation is the subtlest and, in mild conditions, the most powerful: expectations shift with the season, and perceived control transforms tolerance. A slightly warm room bothers far less when the occupant chose the window position than when a sealed facade imposed it. Chambers suppress all three modes by design, which is exactly why chamber data under-predicts comfort in free-running buildings.

The bridge between outdoor climate and indoor expectation is the prevailing mean outdoor temperature: an exponentially weighted running mean of recent daily mean outdoor temperatures, with a memory of several days. It is not the instantaneous outdoor reading, because the body acclimatizes over days rather than minutes, and it is not a monthly average, because a heat wave last week matters more than one two months ago. This running mean captures both seasonal drift and recent weather swings, which is why it predicts the indoor temperature people will find comfortable better than any single outdoor snapshot.

The RP-884 regression distills the field data into one equation for the optimum operative temperature in naturally conditioned spaces:

Around that optimum, the standard defines a comfort band rather than a single temperature: a narrower band for stricter acceptability and a wider band for standard acceptability, both expressed in operative temperature. The band idea is the adaptive model's great practical gift. In a mild week with a prevailing mean of 20 C, the optimum sits near 24 C, and the acceptable band stretches several degrees on either side, which no PMV-based design would permit at constant clothing and activity. Note what the equation implies at the extremes: it is a creature of moderate climates and free-running operation, and its database thins out near the edges of the 10 to 33.5 C range.

Applicability is where engineers get into trouble, so be precise. The adaptive method applies to occupant-controlled, naturally conditioned spaces: operable windows (or equivalent occupant control of ventilation), no mechanical cooling in operation, occupants engaged in near-sedentary activity, and dress codes flexible enough to allow behavioral adaptation. It does not apply to sealed buildings, to spaces where mechanical cooling is running (use the PMV method there), or to high-activity occupancies. Mixed-mode buildings live in both worlds: when the cooling plant runs, evaluate with PMV; when the building free-runs on windows and fans, evaluate with the adaptive model. The changeover logic between modes should be an explicit design decision, not an accident of controls programming.

The design payoff is substantial. A wider acceptable indoor range in mild weather means more hours when natural ventilation alone satisfies comfort, which means smaller plants, lower energy use, and facades designed for operability rather than sealed perfection. Ceiling fans, operable windows, and night ventilation stop being amenities and become comfort equipment with a legitimate engineering basis. But the same field data carries a warning: adaptation is not infinite. Push indoor conditions far beyond the band and behavioral options saturate, windows stay open because nothing else helps, and the votes collapse toward the chamber predictions. The adaptive model extends the comfort envelope; it does not abolish physics.

Key takeaways

• Field studies (de Dear and Brager, ASHRAE RP-884) showed occupants of naturally ventilated buildings comfortable across wider indoor ranges than chamber-based PMV predicts.
• Adaptation works through behavior (windows, fans, clothing), physiology (acclimatization), and psychology (expectation, perceived control).
• Prevailing mean outdoor temperature, an exponentially weighted running mean of recent daily means, is the predictor in T_comf = 0.31 × t_pma + 17.8 (C).
• Comfort is a band around the optimum, not a point; the band widens or narrows with the acceptability class chosen.
• The adaptive method applies only to occupant-controlled, naturally conditioned, near-sedentary spaces with no mechanical cooling running; mixed-mode buildings must switch methods with operating mode.

A room can average "neutral" while one occupant's neck sits in a cold supply jet and another's feet rest on a cold slab. Whole-body indices like PMV integrate heat flows over the entire body surface, but thermal receptors report locally, and the skin of the neck, head, and ankles is disproportionately sensitive to local cooling rates. ASHRAE 55 therefore treats local thermal discomfort as a separate compliance topic with its own criteria: draft, radiant temperature asymmetry, vertical air temperature difference, and floor surface temperature. A design that passes PMV and still generates complaints has usually failed one of these four.

Draft is unwanted local convective cooling, and its mechanism explains why mean air speed alone is a poor predictor. Moving air strips the thin thermal boundary layer clinging to the skin, raising local heat loss; turbulence intensity matters because fluctuating velocity produces fluctuating heat transfer, and receptors read intermittent gusts as stronger cooling than a steady flow at the same mean speed. That is why a diffuser with a ragged, highly turbulent throw feels drafty at an air speed that would be unnoticeable from a smooth displacement flow. The standard's draft model combines the three drivers into a predicted percentage of people bothered by draft:

Read the equation as engineering guidance: draft risk climbs as air temperature falls, as air speed rises above about 0.05 m/s, and as turbulence intensity grows. The head and neck region is the most draft-sensitive, followed by the ankles, which is why supply jets that wash across seated occupants and floor-level leakage paths cause the classic complaints. In design, the remedies are geometric before they are thermal: keep diffuser throws out of the occupied zone's breathing and neck region, select diffusers with low-turbulence throw characteristics where occupants sit close, and remember that raising supply air temperature a degree or two trims the (34 - t_a) term directly.

Radiant temperature asymmetry attacks from a different direction. The body exchanges radiation with every surrounding surface through view factors, so a cold window chills the window-facing side of the body while the room-facing side stays neutral, and an overheated ceiling panel roasts the head while the feet stay cool. The sensation is one-sided discomfort in a room whose air temperature is textbook perfect. The standard limits how asymmetric the radiant field may be, with the strictest attention on warm ceilings (the head is both sensitive and richly supplied with blood) and correspondingly defined limits for cool walls, cool ceilings, and warm walls. The design lesson is that glazing performance is a comfort device, not just an energy device: high-performance glass, perimeter heating that raises the cold surface temperature, and interior shades all act directly on the radiant asymmetry term, which no thermostat will ever see.

Vertical air temperature difference is stratification made personal. Warm air rises, so tall spaces, displacement ventilation systems, and perimeter heating can all leave a seated occupant with a warm head and cold ankles. The standard caps the temperature difference between head level (1.1 m) and ankle level (0.1 m) for seated occupants at 3 C, a widely cited limit worth memorizing. Exceeding it produces the characteristic complaint of "stuffy head, cold feet" that no setpoint adjustment fixes, because raising the thermostat warms the head further while the ankles lag. Displacement ventilation designers live with this constraint constantly: the supply-to-exhaust temperature gradient that makes the system efficient is the same gradient the standard polices.

Floor surface temperature works through conduction to the feet, a small area with an outsized vote. Cold slabs on grade, stone or tile finishes, and overheated radiant floors all drive complaints that occupants describe as general chill or warmth but that trace to the soles of the shoes. The standard defines an acceptable floor surface temperature range for normally shod occupants engaged in light, mainly sedentary activity; outside that range, even a perfect air temperature will not save the space. Radiant floor designers should note the asymmetry of the problem: floors can be only mildly warm before discomfort starts, while they can be surprisingly cool, which bounds how much heating a floor system can usefully deliver in comfort-critical spaces.

When a complaint lands on your desk, diagnose locally. Measure at the complainant's seat, not at the thermostat: air temperature and speed at ankle, waist, and head height, globe temperature for radiant asymmetry near glazing, floor surface temperature underfoot. Most local-discomfort problems yield to local fixes: adjusting a diffuser vane, adding perimeter heat under a cold window, changing a shade schedule, or rebalancing a floor system. Reserve system-level changes for problems that appear in every seat, and always re-measure after the fix; local discomfort is one of the few comfort topics where a handheld traverse settles the argument in an afternoon.

Key takeaways

• Whole-body neutrality does not guarantee comfort; ASHRAE 55 judges draft, radiant asymmetry, vertical temperature difference, and floor temperature as separate local criteria.
• Draft risk rises with falling air temperature, rising air speed, and rising turbulence intensity; the neck and ankles are the most sensitive regions.
• Radiant asymmetry causes one-sided discomfort through view-factor geometry; glazing performance and perimeter heat act directly on it.
• The head-to-ankle air temperature difference for seated occupants is capped at 3 C; floor surface temperature has its own acceptable range for shod, sedentary occupants.
• Diagnose complaints at the occupant's seat with multi-height measurements; most local problems yield to local fixes.

Comfort quantities are mostly derived rather than directly sensed, so a measurement strategy has to mirror the six factors: sense the four environmental quantities with instruments, estimate the two personal factors by observation and tables, and ask the occupants themselves, because the vote is the ground truth every model is trying to predict. A measurement campaign that skips any of these three legs ends up arguing with complaints it cannot explain.

Air temperature seems trivial until you watch how often it is measured badly. The sensor must be shielded from radiation: a bare sensor in sunlight or facing a hot ceiling reads a fiction compounded of air and radiant effects. It must sit away from heat sources, supply jets, and exterior walls, and it must represent the occupied zone, which is why the standard defines measurement heights by posture: 0.1, 0.6, and 1.1 m for seated occupants, and 0.1, 1.1, and 1.7 m for standing occupants, with the results averaged appropriately. A single wall thermostat at 1.5 m, often mounted above a heat-producing device or in a dead corner, satisfies none of these conditions; treat its reading as a control input, not as evidence about comfort.

Mean radiant temperature is measured with a globe thermometer, most classically the 150 mm matte-black copper globe. The globe settles at an equilibrium where its convective exchange with the air balances its radiant exchange with the room, so its temperature encodes both; knowing the air temperature and air speed lets you solve for the radiant component. The standard relation is:

Two practical notes on globes. First, the 150 mm globe takes on the order of twenty minutes to equilibrate, so it is a survey instrument, not a control sensor; smaller 40 mm globes respond in a few minutes and are popular for walk-throughs, at the cost of greater sensitivity to air speed that the diameter term in the equation corrects for. Second, a globe in direct sun measures the sun, which is legitimate if solar-loaded comfort is the question and misleading if it is not; note the exposure on every reading.

Humidity is the easiest factor to sense and the easiest to let drift out of calibration. Capacitive relative-humidity sensors are standard; they age, they foul in dusty or chemical environments, and a two-year-old uncalibrated sensor can read several percent off, which matters when you are diagnosing edge cases. Derive humidity ratio or dew point from the reading for engineering use, since relative humidity alone conflates temperature and moisture content. Co-locate the humidity sensor with the air-temperature sensor so the pair describes the same air.

Air speed demands the most care. Thermal (hot-sphere) anemometers with omnidirectional probes cover the 0.05 to 1 m/s range that matters for comfort; vane anemometers stall at low speeds and ultrasonic instruments, while excellent, are overkill for most surveys. Average over minutes rather than quoting spot gusts, because comfort responds to sustained exposure, and map the space: near diffusers, at seated head height in the throw path, at ankle level near entrances. If you need turbulence intensity for the draft equation, you need fast sampling and a proper statistical reduction, which is a level beyond the walk-through traverse; most practitioners estimate it from diffuser and flow-regime knowledge instead.

Occupant surveys supply the ground truth. Use the same seven-point thermal sensation scale the models were built on (-3 cold through +3 hot), ask for "right now" sensation so the vote pairs with concurrent measurements, and add two companion questions: thermal preference (want warmer, no change, want cooler) and acceptability. Keep the survey short; long questionnaires bias the sample toward the aggrieved. Sample across zones and seasons rather than polling once in October, and be honest about response bias: chronic complainers answer surveys, contented occupants often do not, so weight the silent majority by pairing votes with the measured environment.

Long-term monitoring turns snapshots into understanding. Trend operative-temperature proxies on the BMS, sample perimeter versus core zones separately, and keep the logging running across seasons; a space that passes in October can fail in February when the glazing runs cold or in August when solar loads peak. The highest-value use of monitoring data is validating design assumptions: compare measured conditions against the met, clo, and occupancy patterns the design assumed, then tune setpoints, schedules, and zoning to the building as operated rather than the building as imagined. Before-and-after logging around a retrofit is similarly the only honest way to prove a comfort intervention worked.

Key takeaways

• Measure the four environmental factors with instruments, estimate met and clo by observation, and survey occupants; the vote is the ground truth.
• Air-temperature sensors must be shielded, representative of the occupied zone, and read at posture-specific heights (0.1/0.6/1.1 m seated; 0.1/1.1/1.7 m standing).
• The globe thermometer yields mean radiant temperature through the standard convection-radiation balance relation; mind equilibration time and sun exposure.
• Average air speed over minutes, map it spatially, and treat turbulence intensity as a separate, harder measurement needed only for formal draft evaluation.
• Run short "right now" surveys on the seven-point scale across zones and seasons, and use long-term monitoring to validate design assumptions and prove retrofits.

Everything so far has described comfort as the body experiences it; this session turns it into control decisions. The central insight is a change of controlled variable. Thermostats sense air temperature, but occupants feel operative temperature: the combined effect of air temperature and mean radiant temperature, weighted by the convective and radiative heat transfer coefficients. In a well-insulated interior zone the two track closely and the distinction barely matters. In a glass perimeter on a winter night, with 22 C air and a 16 C mean radiant temperature, the occupant feels something near 19 C, files a complaint, and the "correct" setpoint takes the blame for a radiation problem.

Operative temperature formalizes the combination:

In ordinary indoor conditions the two coefficients are close in magnitude, so the weighted average collapses to the familiar approximation t_o ≈ (t_a + t_r) / 2. The design use is immediate: estimate the mean radiant temperature a zone will actually produce (cold glazing in winter, warm ceiling under a sun-baked roof in summer), then back out the air-temperature setpoint needed to hold the target operative temperature. A perimeter zone with a 3 C radiant deficit needs its air setpoint raised, or better, its radiant deficit repaired with perimeter heat or better glazing; raising air temperature to compensate for radiation is legitimate but energetically expensive, which is exactly the trade-off the equation lets you quantify.

Setpoints should sit inside the comfort band, not on a remembered number. Take the acceptable operative-temperature band for the season's clothing and activity (from the PMV ±0.5 criterion or the adaptive band), place the heating setpoint near its lower edge and the cooling setpoint near its upper edge, and separate the two with a deadband in which neither heating nor cooling operates. The deadband is pure profit: every degree of widened separation trims simultaneous heating-and-cooling waste and lets the building float on internal gains and thermal mass. The failure mode is the too-clever narrow deadband that hunts, with reheat fighting cooling across the same afternoon.

Setbacks trade comfort risk for energy savings and must be engineered, not just scheduled. Dropping the heating setpoint overnight saves energy, but the building must recover to the comfort band before occupants arrive, which is what optimum-start controls calculate from the building's thermal response. Two limits bite in practice. First, heavyweight buildings and radiant systems recover slowly; a deep setback that a light VAV system shrugs off can leave a radiant slab building uncomfortable until mid-morning. Second, setback during unoccupied hours is only free if the recovery completes in time; trend the first occupied hour, not the 3 a.m. low, when you verify the strategy.

Zoning is comfort's floor plan. Zone by solar exposure first: east, south, and west perimeters peak at different hours and should never share a thermostat with each other or with the core, whose load is dominated by people and equipment rather than envelope. Zone by use second: conference rooms spike in occupancy, equipment, and metabolic rate, and they need their own control (plus ventilation) response. Zone by schedule third: spaces with different occupied hours should not be dragged along by the longest schedule on the floor. The classic pathology is one thermostat near the core return grille governing a floor with three glazed exposures; the core is satisfied while the perimeters swing between complaints, and no setpoint pleases everyone because the zone itself is the error.

Comfort and energy couple most elegantly through elevated air speed. When occupants control the air movement (a ceiling fan, a personal desk fan), the standard permits higher operative temperatures at the same comfort level, because the extra convection and evaporation offset the warmth. This is not a loophole; it is the physics of Session 1 applied deliberately, and it lets cooling setpoints rise several degrees in suitable spaces. Adaptive operation in mixed-mode buildings is the seasonal version of the same idea: let the building free-run inside the adaptive band during mild weather instead of mechanically holding a PMV setpoint. One caution on humidity: near neutral conditions, humidity's leverage on comfort is small, so over-dehumidifying for comfort alone wastes energy; control humidity for indoor air quality, condensation, and mold, and let temperature and air movement carry the comfort load.

Close the loop by commissioning the comfort intent, not just the equipment. Trend operative-temperature proxies by zone through real occupied hours, run a short occupant survey after the first season, and compare the results against the met, clo, and occupancy assumptions the design was based on. Designs rarely fail because the equations were wrong; they fail because the assumed clothing, the assumed occupancy schedule, or the assumed glazing performance never matched the operated building. Measurement from Session 5 is how you find out, and this session's setpoints, deadbands, and zones are what you adjust once you know.

Key takeaways

• Control to operative temperature, the air-plus-radiation combination occupants actually feel; back out the required air setpoint from the zone's expected mean radiant temperature.
• Place heating and cooling setpoints at the edges of the comfort band with a real deadband between them; the deadband is energy saved, not comfort lost.
• Zone by solar exposure, by use (conference rooms earn their own zone), and by schedule; one thermostat cannot serve a core and three perimeters.
• Elevated, occupant-controlled air speed legitimately raises the acceptable cooling setpoint; adaptive free-running extends the same principle across seasons.
• Engineer setbacks with recovery in mind, control humidity for IAQ and moisture rather than comfort alone, and commission the comfort intent with trends and surveys.

This session works one realistic case end to end, by hand, so the Fanger formulation becomes a tool rather than a black box. The scenario is a perimeter-adjacent open office in summer: seated occupants doing computer work. Every number below is an illustrative assumption for teaching the method, labeled as such; in a real project each input would come from measurement, simulation, or a documented table value.

Inputs (all assumed/illustrative): metabolic rate 1.2 met for typing and office work, so M = 1.2 × 58.2 = 69.84 W/m², with external work W = 0 (no useful mechanical work is done). Clothing 0.8 clo for trousers and a long-sleeve shirt, so I_cl = 0.8 × 0.155 = 0.124 m²K/W. Air temperature t_a = 24 C, mean radiant temperature t_r = 24 C (an interior zone with no strong radiant asymmetry), relative air velocity v_ar = 0.1 m/s (gentle background air movement), relative humidity 40%, giving a vapor pressure p_a of about 1194 Pa at 24 C.

The clothing area factor, which accounts for a clothed body presenting more surface than a nude one, follows from the insulation value. Since I_cl = 0.124 exceeds 0.078 m²K/W, f_cl = 1.05 + 0.645 × 0.124 = 1.130. The clothing surface temperature t_cl is implicit: it appears on both sides of its own equation because the heat conducted through the clothing must equal the heat convected and radiated away from the clothing surface. The equation is:

Solve by successive substitution with light damping, a standard treatment for this stiff equation. Start with a guess of t_cl = 30.0 C. The convective coefficient is the larger of 2.38 × 6^0.25 = 3.72 and 12.1 × √(0.1) = 3.83, so h_c = 3.83 W/m²K (forced convection from the background air movement governs). Radiant loss from the clothing surface is 3.96 × 10^-8 × 1.130 × (303⁴ - 297⁴) = 29.0 W/m²; convective loss is 1.130 × 3.83 × (30.0 - 24.0) = 25.9 W/m². Feeding these into the right-hand side gives a raw new value of 35.7 - 0.028 × 69.84 - 0.124 × (29.0 + 25.9) = 26.93 C, and damping toward it lands the next iterate near 28.5 C. One more pass at 28.5 C: radiant loss 21.4 W/m², convective loss 19.3 W/m², raw new value 28.7 C. The iteration converges quickly to t_cl = 28.57 C, with h_c = 3.83 W/m²K, radiant loss 21.9 W/m², and convective loss 19.8 W/m². Notice the physics the numbers reveal: the clothing surface sits about 4.5 C above room temperature, and radiation carries slightly more than convection at this low air speed, exactly the balance Session 1 predicted for indoor conditions.

With t_cl known, the thermal load L assembles term by term. Internal production minus external work is 69.84 W/m². Against it stand the losses: moisture diffusion through the skin 12.36, regulatory sweating 4.91, latent respiration 5.55, sensible respiration 0.98, radiation from the clothing surface 21.94, and convection from the clothing surface 19.77, all in W/m². Total loss is 65.51 W/m², so L = 69.84 - 65.51 = +4.34 W/m². The sign is positive but small: the body retains slightly more heat than a neutral person would shed, a gentle warm-side imbalance. Two observations for the designer: dry heat loss (radiation plus convection, 41.7 W/m²) dominates evaporative loss (about 23 W/m²) at these mild conditions, and the single largest adjustable term in the whole balance is the clothing insulation sitting in front of every surface loss.

The last two steps are arithmetic. PMV = (0.303 × exp(-0.036 × 69.84) + 0.028) × 4.34. The sensitivity factor evaluates to 0.0525, giving PMV = +0.23. Then PPD = 100 - 95 × exp(-0.03353 × 0.23⁴ - 0.2179 × 0.23²) = 6.1%. Interpretation: the mean vote sits just to the warm side of neutral, comfortably inside the -0.5 to +0.5 band, and about 6% of occupants would be dissatisfied, which passes the PPD < 10% acceptance criterion. This is a space performing as designed: the average occupant reports near-neutral, and the dissatisfied fraction sits just above the theoretical 5% floor that no design can beat.

Step back and note what the hand calculation teaches that software hides. First, the iteration converges in a handful of passes because the equation, though implicit, is well-behaved with damping; anyone can check a program's t_cl against a two-pass hand estimate. Second, the term-by-term breakdown shows where design leverage lives: clothing insulation scales every surface loss, air speed acts on convection and evaporation together, and humidity barely moves the total at 24 C (the evaporative terms sum to roughly a third of dry losses). Third, the sensitivity test at the end is the real engineering deliverable: comfort work is not about one PMV number but about the slope of PPD against the variables you control. A space at PMV +0.23 with a steep slope toward warm is a space that needs its setpoint defended; the same PMV with margin on every side is a space you can stop worrying about.

Key takeaways

• A complete PMV calculation needs six labeled inputs; in this illustrative office case (1.2 met, 0.8 clo, 24 C air and radiant, 0.1 m/s, 40% RH) the result is PMV +0.23 and PPD 6.1%, which passes.
• The clothing surface temperature is implicit and solved by iteration; damping makes successive substitution converge in a few passes to t_cl = 28.57 C here.
• The term-by-term load breakdown reveals design leverage: clothing scales every surface loss, air speed acts on convection and evaporation together, humidity matters little near neutral.
• Comfort work is about slopes, not points: a 2 C rise in this space drives PPD from 6% to 36%, so setpoint discipline and margin matter more than hitting neutrality exactly.
• Always sanity-check software with a hand case, watching for unit errors (M in W/m², p_a in Pa) and implausible intermediate values.