Demand control and constant flow ventilation compared in an exhaust ventilated bedroom in a cold-climate single-family house

ABSTRACT

A convertible, zoned ventilation system was field-tested in a modern, airtight Swedish home when occupied either by an experimental team or by a family. Indoor air quality in the master bedroom was monitored under four ventilation strategies. Relative to constant air volume strategies (CAV), demand-controlled ventilation (DCV) that was responding to CO₂ concentration extracted more air when people were present, but less in total over 24 h. This elevated the indoor air humidity, beneficial in climates with dry winter air. Multiple monitors within the bedroom indicated that vertical CO₂ stratification occurred routinely, presumably due to low mixing of supply air from a wall-mounted diffuse vent, spreading the air radially over the wall. This seemingly improved air quality in the breathing zone under local (ceiling) extract ventilation but worsened it during more typical, centralised extract ventilation, where air escapes the room via an inner doorway. The local extract arrangement thus seemed to yield both improved ventilation efficiency and reduced contaminant spread to other rooms. The noted air quality variations within the room highlight the importance of sensor placement in demand-control ventilated spaces, even in small rooms such as bedrooms.

KEYWORDS:

• Indoor environmental quality
• ventilation
• occupant comfort
• intelligent building
• digital homes

Introduction

Energy losses by ventilation during the heating season are responsible for approximately half the energy use of a cold-climate home if the building envelope itself is well insulated (Laverge et al. 2011). Even where energy sources are relatively low carbon – grid electricity production in Sweden had an annual average footprint of 8.8 gCO₂(e)/kWh in 2020 (EEA 2021) – reductions in heat demand during peak periods can limit carbon emissions, local pollution and stress upon heat and power grid infrastructure. Demand reduction in electrical space heating eases the transition away from fossil fuels of nascent sectors such as transport. Ventilation strategies for managing levels of indoor environmental quality (IEQ) – criteria for health and freedom from disturbance – alongside energy use are therefore relevant to all homes of sufficient airtightness. Systems that alleviate summer indoor overheating will have increasing relevance under climate change.

Ventilation in Swedish homes is governed by the Boverket Byggregler statutory building code (Boverket 2020), which requires a baseline airflow proportional to floor area of 0.35 l·s⁻¹m⁻², with the option to run at 0.1 l·s⁻¹m⁻² when a space is unoccupied. This allowed reduction creates opportunity for ventilation energy savings. The norm is for properties to be reasonably airtight, with fresh air admitted or mechanically supplied into the cleanest and most often occupied areas (bedrooms, lounge) then internally transferred and vented out from areas of lowest air quality (bathroom, kitchen). In this way, undesirable humidity, odour or other contaminants can be removed with minimal spread. This approach does not, however, lend itself to containing pathogens associated with people breathing. The need from early 2020 onwards to control airborne vectors of coronavirus highlighted the health role played by ventilation in public environments (Morawska et al. 2020). Many residential studies focussed on leakage between dwellings in apartment blocks (Lin et al. 2021). Guyot et al. (2022) modelled air transfer within homes and found that manual window opening, to maximise fresh air, can increase the risk of infecting others in the same flat or house.

Field studies in domestic settings have been limited in number since 2003 (Guyot, Sherman, and Walker 2018). There is a potential risk, therefore, alongside advances in smart systems that real-world experiences are sidelined. ‘Smart’ in this context refers to ventilation that reacts to conditions inside the building: events such as people cooking, or outcomes such as odours. It is complex to balance excluding outdoor pollution and extracting indoor sources of air contamination while also considering comfort, energy, cost, environmental impact and operational disturbances such as draughts and noise (Hesaraki and Holmberg 2015). Modelling that balance, without knowing exactly how the space is used, is to make significant assumptions. Misalignments may arise because of the choice of IEQ indicators monitored, or alternatively where, when, or how the readings are taken. Test chamber experiments by Pei et al. (2019) showed that the location of sensors in a room, coupled with the extent of air mixing the ventilation system achieves, are relevant to how well readings reflect conditions experienced by the occupants. In a bedroom context, Laverge et al. (2013) investigated the distribution of tracer gas released from specific locations around a manikin ‘sleeping’ in an experimental test chamber with otherwise well-mixed air. Contaminants – relating both to exhaled air and chemicals that might arise from bed materials – inhaled by the manikin demonstrated that exposure in the breathing zone is strongly affected by a person’s thermal plume.

A key factor is whether to evaluate indoor air quality (IAQ, the characteristics of the air itself) via direct or proxy indicators. Such proxies are often convenient signals of occupation, allowing systems to infer that indoor air is contaminated or at least not fresh. High CO₂ is associated with accumulation of bio-effluents and other contaminants, and a limit of 1000 ppm is routinely recommended for acceptable air (Mainka and Zajusz-Zubek 2015). That value was previously advised for Swedish workplaces, although since 2020 no guide figure has been given (Arbetsmiljöverket 2020). The gas alone can lower alertness, even at moderately elevated levels – Snow et al. (2019) observed effects at 2700 ppm.

People sense some direct indicators readily (temperature and draught), others more variably (odour, humidity, allergens). In addition, we seldom notice odourless chemicals, particulates and pathogens. Among those inconspicuous contaminants are substances that raise health concerns, and that may be present regardless of occupancy by people. A smart system designed to satisfy parameters that are not apparent to occupants can judge its performance acceptable, whether or not the householders are comfortable. If outdoor air is polluted by wood smoke or traffic emissions, increasing the air supply may worsen indoor conditions (Langer and Bekö 2013), yet for some systems is the only response to detecting contamination.

The use of relative humidity (RH) as a direct indicator for control of domestic ventilation rates is widespread in France (Guyot 2019). In a ten-year field study there, improved IAQ was reported under humidity demand-controlled ventilation (DCV) relative to constant air volume (CAV), alongside DCV flow rates during the heating season reduced by 30%, despite what is termed ‘over-occupation’ of some of the 31 apartments. Indoor CO₂ concentrations exceeded 2000 ppm. When carbon dioxide is a governing factor, lower levels are maintained. The MONICAIR project (van Holsteijn and Li 2014) monitored IAQ in 62 Netherlands apartments and small houses with natural or mechanical ventilation, including some under demand control by carbon dioxide. One measure was the cumulative time and exposure level above 1200 ppm CO₂. Relative to CAV, that exposure during CO₂ DCV varied by between –70 and +11%. Buildings more prone to air leakage did not offer good IAQ, because infiltration was seldom into areas where occupants spent time. People in homes with mechanical ventilation were observed to manually supplement or adjust its operation, but by habit more than in response to poor IAQ, even CO₂ readings as high as 3500 ppm. This highlights a risk of entirely occupant-controlled ventilation, since many crucial indoor contaminants are difficult or impossible for humans to sense.

Judging the probability of a space being occupied using a wide range of direct and proxy IAQ indicators – body heat (passive infrared detection), air temperature and humidity, sound level, and volatile organic compound or carbon dioxide concentration – was tested by Pedersen, Ubbe Nielsen, and Petersen (2017), who noted that detecting arrival and departure was best achieved by different measures, for instance pairing infrared with VOC. An earlier Danish field study (Nielsen and Drivsholm 2010) saw the mechanical ventilation in a 4-person, 140 m² house modified such that it ran at either normal rate (building code) or, when assumed to be unoccupied, at 40% of code. The mode-switching was governed by contrasts in the absolute humidity or CO₂ between the centralised exhaust air handling unit and outside air, aiming to rapidly detect occupation and restore full ventilation. The relative humidity target was a maximum of 50% (aimed at avoiding conditions suitable for mites), an approach different to that expected in colder climates, where insufficient moisture is problematic during the heating season. Excess humidity was successfully avoided but the system occasionally failed to prevent ‘too-high CO₂’, defined in this case as over 1000 ppm.

The Swedish context of indoor climate and wellbeing has been discussed holistically (Wierzbicka et al. 2018). People’s reactions to disturbance may prioritise obvious effects such as fan noise over factors to which they are less sensitive, leading to underventilation. How people interact with their indoor and outdoor space differs widely, and many choose simple responses like opening windows, rather than learning to manage complex systems. A separate study (E. Pedersen, Gao, and Wierzbicka 2021) interviewed apartment residents and found they did not strongly connect their indoor behaviours with energy consumption.

Our group aims to investigate IEQ in a single-family house in as practical a manner as possible. This study monitored DCV approaches led by presence/absence of people (specifically, by carbon dioxide as a proxy indicator), and assessed air quality and other outcomes. This has relevance in times of coronavirus, if increased home working means a dwelling is seldom ever empty. Conversely, low air flows during unoccupied periods may fail to dissipate contaminants. Although this case study is limited to one family, and indeed focussed on one room, the value of having real-life occupation patterns with which to monitor the effectiveness of ventilation that responds to those patterns is significant. Relatively weak interactions between occupation, CO₂ and demand-controlled airflow elsewhere in the house, justify our primary focus on the bedroom.

Objectives

• With a focus on human-sourced contaminants and comfortable humidity, compare IAQ under different exhaust ventilation modes – constant air volume or demand control – during habitation of a bedroom by an experimental team and a real family.

• Study possible local air quality variations in those settings, and evaluate implications for demand-controlled ventilation.

• Suggest IEQ control refinements such that exhaust ventilation systems for single-family homes may better respond to conditions to achieve good outcomes for residents.

• Shed light on how existing ‘smart’ approaches to attaining IAQ may be affected by lifestyle changes due to coronavirus behaviour recommendations.

Method

This study measured the indoor air quality in a test house with a zoned mechanical ventilation system (see Figure 1). Two occupation setups were followed:

1. controlled conditions, when researchers themselves used the house and could observe directly

2. real-life conditions, during which a family lived in the villa as their permanent home, and observations were made by automated monitors and written feedback.

Figure 1. Exhaust ventilation system schematic layout for whole villa, showing seven zones with sensors and dampers. During CAV Indirect mode, the master bedroom exhaust duct was sealed off.

Schematic diagram of villa's ventilation, across upper and lower floors. Single fan at central AHU extracts air from seven zones, with dampers on each zone adjusted according to sensors (dry rooms CO2; wet rooms humidity/VOC).

Four ventilation modes were used (see Table 1). The key room air characteristics logged were temperature, relative humidity and carbon dioxide concentration.

Table 1. Ventilation modes used in the bedroom during the two occupation setups.

Ventilation mode	Label	Flow rates (l s^–1·m^–2)	Extract path	Occupation conditions
Demand control by CO2	DCV Normal	0.13–0.45	Ceiling duct	Real-life only
Demand control by CO2	DCV Intense	0.13–0.54	Ceiling duct	Both
Constant	CAV Direct	0.35	Ceiling duct	Both
Constant (Swedish standard)	CAV Indirect	0.35	Via internal doorway	Controlled only

The test house

Dalarnas Villa is a newly built, single-family house in Dalarna, inland central Sweden (60°N) with high airtightness, measured as 0.6 ACH (hourly air changes by volume) at 50 Pa underpressure. It is a two-storey wooden building on a concrete foundation. The internal footprint is 88 m², with floor area 150 m² and volume approximately 500 m³. A ground-source heat pump provides warm water circulation to underfloor heating throughout. Floors are wood or ceramic tile. Unusually, the thermal and sound insulation is of natural materials: cellulose and wood fibre.

The focus in this paper is the master bedroom – due to the time people spent there as well as the IAQ effects identified. Unless stated otherwise, descriptions, results and discussion refer there. The bedroom was one individual ventilation zone, normally with two adult occupants overnight. Its layout and the monitoring and ventilation installations are shown at Figure 2 and Figure 3. The bedroom floor area was 14.6 m² including a built-in wardrobe that occupied approximately 2 m² and influenced air exchange within the room. During the controlled conditions study, the bedroom door was always kept closed, with a 10 mm air gap below. The family generally did not shut doors either day or night during the real-life conditions.

Figure 2. (a) Plan of Villa lower floor with air inlets and ceiling exhaust ducts. Floor plans of master bedroom under (b) controlled and (c) real-life conditions, showing also LAQ monitor positions.

Plans show the entire lower floor layout (4 room areas plus hallway), then more detail of just master bedroom. Bedroom air inlet shown on outside wall, with ceiling air extract duct and experimental IAQ monitors (two different heights) positioned 3–4m away.

Figure 3. Bedroom details, left to right: monitors during controlled conditions; Luvian LAQ at 1.1 m on shelf beside door during real-life conditions; inlet wall vent.

Three photographs show (1) room with bed in position, two LAQ monitors on tripods and ceiling exhaust vent labelled (2) LAQ on shelf beside door (3) air inlet fully open with diffuser plate at an angle, on wall towards corner with ceiling above.

Ventilation hardware and modes

Throughout the test period, a single central Renson Healthbox 3 air handling unit (AHU) provided mechanical exhaust ventilation (Figure 1). It monitored IAQ in seven zones and extracted a constant or variable air volume per second through separate exhaust ducts, each equipped with an independent automated damper. Replacement supply air entered each bedroom and the living room through window or wall vents (Figure 2). The master bedroom had an exhaust duct at ceiling level while an unpressurised diffused inlet (Fresh TL100D) admitted air through an outside wall at a height of 2 m. Extracting air directly from bedrooms is unusual in Swedish homes (the norm is from kitchen and bathrooms) but is a feature of demand control systems that ventilate where and when is most appropriate. Such an installation provides a route to extract stale or polluted air near its source, limiting the transfer to other rooms. The open-plan kitchen/living room had a ceiling exhaust duct to the AHU and additional, manually activated extraction above the stove. Relative underpressure in the bathrooms prevented outflow to adjoining areas.

The ventilation strategies trialled included CAV in accordance with the Swedish building standard (0.35 l·s^–1·m^–2, approximately 0.5 ACH) as well as demand-controlled approaches with independent zones each having variable air extraction between 0.13 and 0.54 l·s^–1·m^–2. The permutations are listed in Table 1, including the label used subsequently in this report. When under DCV, dampers for dry zones (bedrooms, living room, kitchen) were governed by CO₂ concentration; wet zones (toilet/bathrooms and laundry room) responded to high humidity and/or increased volatile organic compounds (VOCs). The CO₂ concentration threshold in the exhaust air from dry rooms was 950 ppm, above which level the airflow would sharply rise from the background rate up to the maximum for the mode. Under CAV Indirect, the extract path to the AHU from the bedroom was not via the dedicated ceiling duct. Instead that outlet was physically sealed and the room air was extracted, principally under the door, by underpressure generated in other rooms. The physical disruption involved prohibited a real-life conditions test of CAV Indirect.

Monitoring, calculations and third-party data

Air monitoring was conducted by two separate systems. The AHU control sensors at its exhaust duct terminals generated a log of temperature, absolute/relative humidity (with declared uncertainty ±2%) and carbon dioxide concentration (uncertainty ±5%). The data was recorded at 5-minute intervals. Freestanding IAQ sensors (Luvian LAQ) were also deployed, which logged temperature, relative humidity and carbon dioxide at intervals of 10 min. Each LAQ actively drew in sample air and obtained its CO₂ readings via a Winsen MH-Z19 sensor, with temperature compensation and a reported uncertainty below 2000ppm of ±(50 ppm and 5%). The LAQs were tripod-mounted in the bedroom during the controlled conditions, at heights of 0.6 m (level with the bed mattress) and 1.1 m, and placed at least 0.5 m from walls and furniture. When the house was in real-life use, the one bedroom LAQ was fixed to a shelf at 1.1 m height, with its intake 15 cm from the wall (see Figure 3). The AHU and a further LAQ (at 1.1 m) monitored the adjacent kitchen/living room area. External calibration of the non-dispersive infrared CO₂ sensors in the Renson and Luvian devices was not possible, although all were less than one year old, and performed their own background self-calibration at intervals. The independent monitors rapidly arrived at consistent readings when the bedroom was unoccupied.

Air flow from the bedroom under CAV Direct and DCV Intense was derived from the logged proportion of airflow – the damper setting – relative to the benchmark. The total air extraction from the whole house was measured independently by a differential pressure flow monitor (SPI-160 C Iris damper, Systemair AB, Sweden) located downstream of the AHU, with readings at 1-minute intervals. Under controlled conditions and the two constant air volume modes, the rate of decline of the LAQs’ excess CO₂ concentration (indoor reading less outdoor background value) in the unoccupied bedroom during daytime was used to calculate the effective ACH value based on an exponential decay rate, the internal door being closed throughout. There were no obvious nearby outdoor sources or sinks of CO₂ – plant growth is limited until May – and based on indicated minima an ambient level of 405 ppm was assumed. In this study it is not essential that the absolute values of CO₂, humidity and temperature are correct, since the interest lies in their short-term variations and localised differences within the building.

Outdoor air characteristics were sourced from an SMHI meteorological station at Borlänge Airport, approximately 15 km distant.

The adult occupants’ experience of the indoor environment was evaluated through structured written feedback, covering their recent prior use of the building – e.g. thermostat settings, open/closed internal doors – as well as their recollection of the IEQ in the previous seven days – e.g. extent to which they noticed disturbance or discomfort due to air composition/temperature, or noise.

Results

The two complementary occupation stages of this study (controlled vs real-life) were designed to help interpret realistic situations. The initial period under controlled, closely monitored experimental conditions was time-limited, only 24 h for each ventilation mode.

Controlled conditions

Air characteristics established for the three modes tested in the bedroom under controlled conditions are summarised in Table 2 and presented graphically (with times of occupation) as Figure 4. In all ventilation modes, there were indications of CO₂ stratification when the room was occupied, with higher values recorded by the AHU, extracting air at the ceiling height of 2.5 m, than by the LAQs at 1.1 or 0.6 m height. The weather at the time (early May) reflected a cold spring period, with occasional snowfall, sunshine and outdoor temperatures ranging between –4 and +14 °C.

Figure 4. Timelines of bedroom CO₂ concentration from three sensors with occupant count during controlled conditions overnight periods, 2200–0900.

Three lineplots show CO2 levels and times of bedroom occupation from 2200 and through the night under each ventilation mode. Start concentrations roughly 400–600 ppm. Rapid rise once people enter. Stable level reached by 0400 under CAV modes at over 1200 ppm, but before midnight under DCV Intense at 800–1100 ppm.

Table 2. Key values from 24-hour observations in bedroom under controlled conditions, in each of three ventilation modes.

	Ventilation mode
Reading/value description, with sampling height and source, and timing where relevant	CAV Direct	CAV Indirect	DCV Intense
CO2 concentration (ppm)
Ceiling (AHU) before overnight occupation	485	–	475
Ceiling (AHU) overnight peak	1562	–	1112
0.6 m LAQ before overnight occupation	468	568	445
0.6 m LAQ overnight peak	1357	1674	903
1.1 m LAQ before overnight occupation	398	610	–
1.1 m LAQ overnight peak	1297	1923	–
Excess CO2 hours above 1000 ppm in 24 h (ppm.h)
Ceiling (AHU)	949	–	306
1.1 m LAQ	655	6393	0
Air change rate (ACH, 24-hour average, h^−1)
Ceiling, logged by AHU	0.46	–	0.37
Calculated from 0.6 m LAQ CO2 decay	0.50	0.45	–
Calculated from 1.1 m LAQ CO2 decay	–	0.42	–

Carbon dioxide overnight concentration trend and peak

As is evident in Figure 4, the room CO₂ level responded rapidly to the presence of people. A clear differential between the 0.6 m LAQ and AHU also arose during room occupation, typically 150–200 ppm higher at the ceiling extract point, the two having been in close agreement (within 30 ppm) beforehand. Under DCV Intense, the indicated CO₂ ppm rose sharply initially, then stabilised once the AHU’s demand control threshold of 950 ppm was reached. Because constant air volume does not vary its flow rate, the initial CO₂ rise under CAV was slower but it then reached a higher equilibrium level. The CAV Indirect situation peaked considerably higher than the two other modes.

Carbon dioxide excess hours

Excess CO₂ hours, as a measure of the time extent of elevated CO₂ periods, were quantified based upon a threshold of 1000 ppm. For each 10-minute interval under each ventilation mode, the concentration level as ppm above threshold was multiplied by the duration. CAV Indirect recorded by far the greatest excess hours value, as is clear from Figure 4.

Average airflows

Under CAV Direct the logged and calculated air extraction rates were close to the 0.5 ACH required by the standard. The alternative CAV Indirect, in which the whole house extraction met the required standard, saw the bedroom ventilation, as calculated from CO₂ decay at the higher LAQ, fall to 0.42 ACH. Under DCV operation the recorded average 24-hour ACH was lower than either CAV mode, despite apparent IAQ being as good or better under DCV.

Real-life conditions

The period of family observation took place between February and April 2020 and was therefore partly affected by lifestyle changes (home working and school disruption) prompted by coronavirus. The datasets for each ventilation mode were 28–33 days in duration. 14-day timelines of CO₂ and trends in relative/absolute humidity (RH/AH), alongside bedroom airflow under each mode are shown at Figure 5. Where AH was not obtained directly, it was calculated via the saturated vapour pressure of water according to Bolton (1980). To allow comparison with nearby areas, Figure 6 plots airflow and CO₂ concentration in the kitchen/living room and the bedroom, for two weeks when those zones ran under DCV Normal.

Figure 5. Timelines for bedroom CO₂ concentration, humidities, and room extraction airflow rate over 14-day period in each of the three ventilation modes during real-life conditions.

Figure 6. Timelines for CO₂ concentration and room extraction airflow rate for DCV Normal in both master bedroom and adjoining kitchen/living room. Monitor locations labelled as in Figure 5.

Similar plot to Fig 5 repeats the DCV normal bedroom CO2 and airflow timelines and overlays comparable data (same 14 days) from the adjacent kitchen/living room. Close CO2 alignment during daytime periods; clear bedroom excess overnight. Kitchen airflow at minimum level for vast majority of hours in the two-week period.

Carbon dioxide time profiles, overnight peak/equilibrium

Figure 5 indicates for the CAV Direct case that CO₂ levels peaked significantly lower during family use than during the controlled conditions (Figure 4). This is likely to be caused by the occupants’ tendency to keep the bedroom door open. The distinction – when people were present – between ceiling exhaust air (AHU reading) and occupation zone (LAQ reading) is clear. The ceiling air routinely read 300 ppm higher under DCV Intense; 200 ppm higher under DCV Normal; 150 ppm higher under CAV Direct.

The LAQ values showed more consistency within and between overnight periods, with maxima below 950 ppm at all times. With regard to the overnight exhaust (AHU) readings, under DCV Intense, levels also routinely fell below the 950 ppm threshold, such that reduced (or varying) ventilation occurred after 0200 and sometimes from midnight. This was not as frequently the case for DCV Normal, which tended to reach an equilibrium in the 950–1000 ppm range, above the DCV activation threshold. Under CAV Direct, the readings routinely stabilised between 700–900 ppm, despite lower air flow rate via the zone extraction point. This appears related to the steady CAV flow yielding lower CO₂ levels during early evenings than did DCV.

Carbon dioxide excess hours

Excess levels throughout the entire period for each ventilation mode were averaged per complete 24 h. The daily excess ppm hours at the LAQ were zero throughout. The daily averages in exhaust air (ceiling level), as reported by the Healthbox AHU, are shown in Table 3. Although the bedroom was occupied during those times of excess CO₂, none of these later values suggest any extended period of high concentration, rather they reflect that the DCV modes held low ventilation until their 950 ppm trigger level was reached. Comparing with values in Table 2, the DCV Intense result was lower here than under the controlled occupation conditions, and the CAV Direct value considerably so. This pattern is significant within the assessed measurement uncertainty, and likely to be caused by the family keeping inner doors open.

Table 3. Excess CO₂ values, in daily ppm.h above 1000 ppm CO₂, during the three real-life ventilation modes, as measured at two locations in the bedroom.

	Ventilation mode
Sampling height (monitor)	CAV Direct	DCV Normal	DCV Intense
Excess CO2 hours above 1000 ppm, average per 24 h (ppm.h)
Ceiling (AHU)	46	309	158
1.1 m (LAQ)	0	0	0

Daytime CO₂ removal

As indicated by the CO₂ decay periods in Figure 5, the AHU and LAQ readings were frequently in very close alignment during daytime, i.e. when the bedroom was unoccupied. There was also reasonable agreement between the two different sensors through the early evening period. There was evidence – the lowest indoor CO₂ level being higher than 500 ppm, and therefore noticeably above outdoor background – that under DCV modes the room may not always have achieved complete air refresh during the daytime.

Also relevant is whether people were elsewhere in the home. It was observed that when the family’s two children returned from school, their presence alone in the house raised CO₂ but did not trigger any DCV response. Figure 6 shows how the kitchen/living room was typically ventilated at the AHU’s minimum level, because the 950 ppm CO₂ threshold was seldom reached in that zone. There was also little evidence of CO₂ stratification there. During daytime and evenings, uniform concentration was routine throughout the two adjoining zones. Overnight CO₂ levels in the bedroom were significantly higher than occurred at any time in the kitchen/living room.

Temperature and humidity

The bedroom temperature was fairly stable and noticeably warm, typically 23.5 ± 1 °C recorded at the LAQ and 24.5 ± 0.5 °C at the AHU. This suggests a slight temperature stratification in the room. The general trends of indoor relative humidity were also stable, as indicated in Figure 5, with some influence of outside conditions, specifically that extremes of RH were associated with periods of high or low outdoor absolute humidity (AH), with a few hours’ delay.

Unlike the CO₂ readings, the raw absolute humidity levels reported by the AHU and LAQ did not coincide during periods of non-occupation of the bedroom, as was expected. The timeline graphs at Figure 5 incorporate a constant offset adjustment of 0.5 g·m⁻³ to achieve such alignment. It is notable that during bedroom use both the rise and stratification of AH were smaller than that of CO₂, thus additionally confirming AH stability. The increased indoor AH – attributable to generation from occupants – is presented in Table 4 for each ventilation mode. Those with higher ACH rate (see Table 2) demonstrate lower AH uplift, as expected. All modes compare favourably with surveyed values of Swedish single-family homes, where average uplift is 1.8 g·m⁻³ (Boverket 2010).

Table 4. Average bedroom AH uplift between indoor (at 1.1 m LAQ) and outdoor (SMHI meteorological station), under three ventilation modes, real-life conditions.

	Ventilation mode
Value	CAV Direct	DCV Normal	DCV Intense
Absolute humidity, additional indoor component (g·m^–3)
Mean	2.39	2.81	2.70
Standard deviation	0.94	0.98	0.76

Occupants’ subjective experience

The family commented on their satisfaction with various aspects of the ventilation system, towards the end of running each mode. They raised cold draughts at times, and overnight noise disturbance, particularly under DCV Intense. Measurements of ventilation noise level in the bedroom, using a basic consumer-grade meter, showed the sound pressure to be around 34 dB(A), 44 dB(C) when DCV Intense extraction was at its maximum. That was comparable to when both fridge-freezers in the adjacent kitchen ran, with the bedroom door open. The Boverket regulations generally require ventilation noise not to exceed 30 dB(A) in sleeping areas (Boverket 2020).

Discussion

Three of the four ventilation modes tested in this study provided acceptable air, as judged by avoiding significant accumulation of carbon dioxide: CAV Direct, DCV Normal and DCV Intense. The CO₂ levels recorded under these modes were significantly lower than those by van Holsteijn and Li (2014) whose real-life monitoring reported average excess dose above 1200 ppm of 190–258 ppm.h per day for a comparable exhaust ventilation system. Under real-life conditions the analogous daily excess dose at 1.1 m height in the bedroom of Dalarnas Villa was zero, and indeed concentration never exceeded 1000 ppm. The night-time CO₂ concentration being significantly higher in the bedroom – especially at ceiling level – than in the adjoining kitchen/living room indicates that direct exhaust from the bedroom achieved localised contaminant removal, despite the bedroom door being open. This prevented much of the bedroom air spreading into the building.

The mode that performed poorly in CO₂ terms was CAV Indirect, during which exhaust air passed under the closed bedroom door. Despite the setup complying with building regulations, and representing the most common exhaust ventilation solution in Swedish dwellings, air quality in the occupants’ breathing zone was noticeably worse. In the studied room, rather cold outdoor air entered through a wall vent (Figure 3) and spread over the inner wall surface, entailing some risk for negatively buoyant air to drop to the floor and exit through the door slot, without much benefit to the occupied zone. It appears likely that this arrangement results in weak air mixing, and in heat-associated contaminants such as human CO₂ and bio-effluents accumulating by buoyancy in the upper part of the room.

With local ventilation exhaust at the ceiling of the bedroom, vertical stratification of CO₂ levels occurred in the room during occupancy under all ventilation modes tested, with higher concentrations at the ceiling (exhaust) than nearer breathing height (0.6 and 1.1 m). The present study indicates CO₂ concentrations at breathing height around 200 ppm lower than in the exhaust, indicating a beneficial air flow pattern in the room. As reasoned above, it appears the supply air had too low inlet speed to induce strong air mixing in the room, the cool air instead presumably dropping to the floor. In combination with air exhaust at ceiling level, the situation resembles a displacement ventilation system, where contaminated air travels upwards keeping the lower, occupied zone cleaner. This kind of ventilation is likely to be particularly effective in bedrooms, where the breathing zone is low down and the occupants are fairly static. Peoples’ movements otherwise tend to mix room air and impair the displacement function (Mattsson 1999). Well functioning displacement ventilation may be utilised to reduce ventilation rate while keeping inhaled air quality good.

Under both controlled and real-life conditions, it was apparent that CO₂ stratification preceded any demand-controlled ventilation air flow increase. This suggests that the location of any IAQ sensors is significant if they are usefully to regulate DCV. In Dalarnas Villa, the distinction between the bedroom being vacant or occupied was not clear-cut during real-life conditions, as the family routinely kept internal doors open. (While complicating interpretation, this was crucial to the value of the field study over simulated or managed experiments.) Other researchers have considered ways to reliably spot occupation of zones, drawing on strategies such as internal and external air comparison (Nielsen and Drivsholm 2010) and complementing IAQ with infrared or physical presence detection (Wierzbicka et al. 2018). Our experience in Dalarnas Villa suggests that the CO₂ contrast between paired sensors at different heights within the same ventilation zone may be an alternative solution, free from arbitrary concentration levels and able to register an occupant (or other source of respiration/combustion) that would not trigger a motion detector. More studies seem however to be needed on air quality in realistic bedroom breathing zones, and how to measure it in a practical way for DCV. The experimental chamber used by Laverge et al. (2013) had supply air around 24 °C rather than the unheated, sometimes subzero intake of our study.

The humidity in Dalarnas Villa was low throughout the test period, as is in fact typical for Swedish homes (Boverket 2010). Relative humidity seldom reached the 40% minimum associated with optimal human comfort and health. Indoor vapour sources such as people, plants and cooking made a broadly similar absolute humidity contribution as did the outdoor air. Despite strong temporal variations in these sources, indoor air humidity was fairly stable, suggesting significant buffering by interaction with surfaces, plausibly due to extensive use of wood-based materials. In comparison, CO₂ fell more sharply when people left the house. Thus, reduced ventilation during non-occupancy will not only save heating energy but also limit escape of accumulated moisture, a positive outcome in climate zones with dry winter air. The lower daytime ventilation rates offered by DCV during non-occupation indeed had some RH benefit at background level across long timescales. It is of note that the more aggressive demand-controlled exhaust ventilation (DCV Intense) was still able to keep bedroom RH above 30% for a two-week period in February (classically a dry air period), due to long periods of reduced ventilation.

Demand control was evaluated with two maximum flow rates, 128% of nominal for DCV Normal and 154% for DCV Intense. Given that there was little evidence of a lack of ventilation (brief CO₂ peaks notwithstanding), it appears there was no advantage in the highest flow rate being available in the bedroom. Indeed the residents’ self-reported dislike of overnight noise – especially its fluctuation – and of draughts, suggests drawbacks to the DCV Intense mode. It would however be relevant to limit spreading of airborne pathogens between family members. The more mixed air environment of the large open-plan kitchen/living room was ventilated at minimum DCV airflow for lengthy periods – including during house occupation – because the activity of those present did not raise CO₂ levels to the zone threshold. Intermittent supplementary high extraction via the cooker hood must be allowed for, but the potential for contaminant accumulation remains. Lowering a zone threshold for CO₂ could reduce that risk, but would lessen the usefulness of running DCV. A possibly ‘smarter’ DCV control might instead prompt high ventilation upon the detection of a CO₂ increase (someone entering), followed (after a purging period) by airflow variation relative to a threshold. Separately, were ventilation modes that aim to steer air change rates described to users with reference to circumstances they easily recognise (examples: disease control or outdoor pollution rather than boost or eco) residents might better appreciate how they themselves could improve the air they breathe.

The impact of coronavirus precautions and changes to home behaviours was not designed into this study. Even so, some patterns with relevance to homeworking and other possible future trends were apparent. A period of continuous occupation of the house during April 2020 (the DCV Normal period) saw CO₂ never falling to outdoor background level and may have interfered with automated sensor calibration systems that assume the lowest steady concentration in a timeframe that extends over several days represents background. Hence zero calibration might occur at a CO₂ concentration that is erroneously high, in turn raising the true threshold for a DCV response. A potential feedback loop exists where inadequate ventilation reinforces itself via false calibration. Care needs to be taken to ensure auto calibration algorithms avoid this effect.

Conclusions

The key function of any demand-controlled ventilation (DCV) system is to respond to actual indoor air quality, including changes related to the presence of people and their activity. Monitoring in Dalarnas Villa showed that both DCV modes achieved acceptable IAQ by the measure of indoor air CO₂ concentration, as did the constant air volume mode, CAV Direct. However CAV Indirect (much more typical in Swedish homes) did not seem effectively to supply air to the breathing zone of sleeping occupants, resulting in a high overnight CO₂ level.

With mechanical exhaust ventilation at ceiling level and cool outdoor air entering the bedroom directly through a wall vent that spread the air radially over the wall, a room air flow pattern resembling displacement seemed to appear – yielding stratification in CO₂, humidity and temperature, with greater values nearer the ceiling. This is likely to provide for better air quality in the breathing zone of sleeping occupants and may possibly be utilised to reduce ventilation rate, with energy and noise benefits. The potential for this is particularly high in bedrooms, motivating more detailed studies on actual air quality in the breathing zone, and how to measure this in a practical way for ventilation control. More generally, the possible occurrence of spatial IAQ variations, as observed in this study, is important when placing sensors for demand control ventilation. Measuring in the room’s exhaust means – in the present case – ‘playing safe’ by sensing worse IAQ than the occupants are exposed to, thereby potentially leading to excessive ventilation. Further studies in this regard may include additional influential factors like furniture locations, and how a room is used (e.g. by adults or children, and routinely standing, sitting or lying down).

The DCV modes performed better than CAV with regard to mitigating low indoor humidity that is problematic in cold inland climate zones. On wider IEQ criteria, the residents’ dislike of ventilation noise appeared to relate to changing noise levels as much as the absolute intensity, suggesting DCV that continually adjusts its flow rate may be disfavoured. Response times for variable ventilation systems, as well as physical silencing, should be carefully considered by designers and installers.

It was apparent that in the Villa’s larger, open-plan area, people could be present for very long periods of time with no DCV response, raising the prospect of accumulated pollutants. In winter, low ventilation during occupation avoids draughts and retains beneficial humidity, so seasonal adjustment need not be a problem unless moisture, odours, particles or other contaminants are excessive. Devices that sense a range of IEQ indicators, and reassure users that important but less obvious matters are being considered, may improve acceptance of intelligent systems in residential settings and thereby uphold potential health and environmental benefits.

Acknowledgements

Dalarnas Villa is a collaboration between Dalarnas Försäkringsbolag and Högskolan Dalarna, with assistance from Luvian AB and Renson BV, and with financial support from Region Dalarna and the European Regional Development Fund, Norra Mellansverige, through the project Energiinnovation 2.0.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the European Regional Development Fund (administered by Tillväxtverket) and by Region Dalarna.

Notes on contributors

Ian Garman

Ian Garman is a PhD candidate with a background in energy efficiency and demand-side management.

Magnus Mattsson

Dr. Magnus Mattsson is a lecturer in building energy and indoor environments.

John Are Myhren

Dr. Jonn Are Myhren and Dr. Tomas Persson are associate professors in building construction and energy technology.