Lawrence Berkeley National Laboratory
1 Cyclotron Rd.
Berkeley CA 94720, USA
Lawrence Berkeley National Laboratory
1 Cyclotron Rd.
Berkeley CA 94720, USA
Lawrence Berkeley National Laboratory
1 Cyclotron Rd.
Berkeley CA 94720, USA
This study used laboratory testing to evaluate low-cost (about $200 US) IAQ monitors that measured PM2.5 to determine if they are suitable for controlling IAQ for ventilation or air cleaning systems.
Keywords: Sensors, Controls, PM2.5, IAQ, laboratory testing
Current Indoor Air Quality (IAQ) standards and approaches to minimizing pollutants depend almost exclusively on using dilution with outdoor air for some generic, continuously generated contaminant (e.g., ASHRAE Standard 62.2-2016). Some standards include measurement of CO2 (e.g., EN 13779 standard (CEN, 2007) and NEN 8088 (NEN, 2011)) – however this is not because CO2 itself is a pollutant of concern, but rather because it can be used as an occupancy indicator or as something that correlates with bioeffluents. Ideally, we would like to measure contaminant concentrations directly and ventilate to control their concentration within acceptable limits. This would ensure that concentrations do not got too high (as they can if emission rates exceed our assumptions) and also allow for ventilation reductions, and resulting energy savings, if concentrations are low. Until recently, it was impractical to consider direct contaminant control in residential (and many commercial) spaces due to the high cost and maintenance requirements for monitoring equipment. In the past couple of years low-cost sensors have been developed for some contaminants of concern – the greatest example of which is for particles. These sensors have been incorporated into low-cost (< $250 US) IAQ monitors. This has opened up the possibility of direct control of ventilation (and filtration systems) by sensing particles. However, it is important to evaluate these monitors to determine if their results are sufficiently good to control a ventilation system.
In this study we performed laboratory experiments to compare the output of seven low-cost monitors to laboratory grade and reference particle measurement methods. The laboratory tests used a range of particle sources to determine if the monitors can reliably detect common household particle emission events. It should be noted that an important aspect of this work is consider particle size. The current state of the art is that particles less than 2.5 microns in diameter (PM2.5) are of the most concern for health. There are some concerns that smaller submicron particles (less than 0.1 microns in diameter) may also be a health hazard, although the evidence for these smaller particles being a significant health hazard is not as strong. Most of the monitors evaluated for this study do not directly have a way of disaggregating results by particle size. For those that do we used the total and did not size disaggregate the results. Instead they employ calibrations or signal conditioning to take the raw output of a particle sensor and translate it into a reported particle concentration. Therefore, our results include both the response of the particle sensor and whatever calibration has been used by the manufacturers of the IAQ monitors. Not all particle emission events emit particles in these ranges (as will be shown in our results). The results can be used to assess the ability of these devices to provide a reasonable control signal for a ventilation system. The results of this study are discussed in more detail in Singer and Delp (2018) together with more information regarding individual sensor performance and discussion of potential indoor particle sources.
The experiments were conducted in a 120 m³ laboratory with a 5.8 m by 7.1 m floor plan, as illustrated in (Figure 1). The room was continuously mixed with fixed direction and oscillating fans. Consumer and research particle monitors were placed on a wire shelving unit co-located with reference instruments in the central area, sufficiently far from source activities that measurements should reflect average room conditions rather than concentrated plumes. In a subset of experiments, filter samples were collected for time-integrated, gravimetric mass determination. During source activities, outdoor air exchange was provided solely by natural infiltration with no mechanical ventilation systems operating. After the source activity ended, particles were allowed to naturally decay over a period of variable duration. In most experiments, after about one hour of decay exterior doors at opposite ends of the laboratory were opened to rapidly ventilate the room and remove residual particles and co-pollutants. During the experiments, the outdoor air exchange rate was measured by a tracer decay method and varied from 0.5 to 1.1 h-1 with a median of 0.7 h-1. Several mixing fans were placed in the test chamber to create uniform particle concentrations. Baseline measurements were taken prior to each source activity. The baseline values were subtracted from the measurements taken during the source activity and integrated over time to calculate mass integration measurements for the source.
Figure 1. Plan view of test laboratory.
The purpose of this study was to compare consumer and research monitor response to reference instrumentation for typical indoor-generated aerosols, therefore, we did not try to precisely control the source emissions. Instead, we used sources that might commonly occur in a home. There were 16 distinct sources: recreational combustion included candles, cigarettes, and incense; mineral sources included an ultrasonic humidifier without a filter, Arizona test dust, and shaking of a workshop dust mop; cooking sources included heating oil in a steel wok on gas or electric burners, frying bacon and toasting four slices of bread in a toaster oven, and stir-frying green beans in oil on a gas burner. Cooking sources that produced large numbers of particles with low to moderate mass concentrations and almost all below 0.3 µm included heating water in a covered pot on a gas stove, heating a gas oven, cooking a pizza in the gas oven, cooking pancakes on a lightly oiled pan over medium heat, and toasting bread in a well-used electric toaster oven. Each source was active for about 10–15 minutes.
The highest 5-min baseline-subtracted mass concentrations (adjusted Mini-WRAS data) varied from 21 µm m-3 for the pancakes to 721 µm m-3 for one of the green-bean stir-fry experiments. The highest number concentrations varied from the AZ test dust and dust mop experiments (at 2–5 x 103 cm-3) to the Bacon + Toast experiment at 2 x 105 cm-3. The five experiments with the lowest peak mass concentration (Pancakes, Pots on gas burners, Oven, Pizza, Burnt Toast) were in the middle of the distribution of peak number concentrations (6th to 18th).
Table 1. Particle sources
Ultrasonic humidifier, cleaning cartridge removed
Incense stick (Shanthimalai Red Ng Champa)
AZ test dust (0–3 micron) manually puffed from bag
150 g frozen green beans, 15 g canola oil stir fried in steel wok on gas stove
Single piece of bread, medium-toasted in used electric coil toaster oven
280 g bacon fried on gas stove; 4 slices bread med-toasted in toaster oven
15 g of canola oil brought to bubble in steel wok on gas stove
Slice of bread, dark-toasted in used electric coil toaster oven
Aggressive shaking of a 90 cm wide workshop dust mop
5 unscented dinner candles, lit with butane lighter
Two covered 5 L pots, half-filled w/H2O, heated on gas stove
Gas oven heated to 400F over 12 min after ~4 y of no use
Two batches pancakes cooked on a lightly oiled fry pan on gas burner
Gas oven heated to 400F 14 min; frozen pizza cooked
3 cigarettes lit with butane lighter, smoldered until self-extinguished
15 g canola oil brought to bubble in fry pan on an electric coil burner
Particle Monitoring Devices
A Grimm Mini Wide Range Aerosol Spectrometer Model 1.371 (Mini-WRAS) was used as a reference for 1-minute resolved data and also to provide distributions of particle number and mass concentrations. The Mini-WRAS combines an electrical mobility analyzer that counts particles in 10 size bins from 10 to 200 nm with a laser-based optical particle counter that provides particle counts in 15 size bins from 0.2 µm (200 nm) to 2.5 µm plus 16 bins between 2.5 and 35 µm. The Grimm estimates volume concentration from the size-resolved number concentrations by assuming the particles are spheres, then calculates mass assuming a density of 1.68 g cm-3. The density is a user-definable parameter that enables the Mini-WRAS to measure aerosols with varying composition.
Seven consumer grade monitors were selected for testing. All devices were available for retail purchase in the US in early 2017 and had either been tested previously by the US EPA or AQ-SPEC for outdoor use, or the research team had learned of their use in one or more citizen science projects. We also used two research-grade monitors that are often used in field and laboratory studies. Summary information for the monitors is listed in Table 2. All of the monitors have a data reporting interval that is reasonable for use as a ventilation controller where we are unlikely to want to make changes on a time scale of less than 5 or 10 minutes.
Table 2. Consumer and research grade monitors evaluated in this study.
Full schematics and program available on github
Web site mentions PM2.5 several times, but does not list the specs
Air Quality Egg [AQE]
Talks about PM2.5, but lists the operating range 0.5-10 mm
Sensor developed by AirVisual. Nominally reports PM2.5 for particles 0.3–2.5 mm.
Product lit describes measurement as “PM”. Range of 0–500 µm/m³. (This corresponds to linear range for voltage output as specified on Sharp sensor sheet.)
Proprietary ‘learning’ algorithm applied to the signal.
Product literature describes as PM2.5 covering range of 0.3–2.5 mm
0–1,300 µm/m³ ±4 mg or ±20%
PurpleAir PA-II [PA]
Reports # in 6 size bins; PM1, PM2.5, PM10. Calibrated to ambient PM in Beijing.
Counting efficiency: 50% @0.3mm 98% >= 0.5 mm
Consistency error: ±10 mg/m³@0–100 mg/m³, ±10% 100–500 mg/m³
Calibrated with AZ dust. Machine learning algorithms applied to sensor signal.
Product literature notes range of 0.5–3 mm.
Thermo pDR-1500 [PDR]
Calibrated with SAE Fine AZ dust. Precision: ±0.5% of reading or ±0.0015 mg/m³, whichever is larger, for 10 sec averaging time.
Accuracy: ±5% of reading ±precision
MetOne BT-645 [BT]
Calibrated with 0.54 mm diameter polystyrene latex spheres.
Figure 2 presents some example results illustrating the different time and magnitude responses of the monitors. The plot shows two very large sources – candles and oil heated on a gas burner – that produced clear and substantial, though not fully quantitative responses by all analyzers. The Egg, Awair and Speck all reported only a small fraction of the actual mass concentration. Emissions from the dust mop, which were concentrated in the largest particles, produced responses of some but not all devices; though for this source the Speck response was substantially higher than the estimate of actual mass concentration. Using the gas cooktop burners to heat water in covered pots and heating the empty oven produced large numbers of particles below 100 nm and modest mass concentrations as indicated by the Mini-WRAS signal (GRM), but no perceptible response of the other monitors. It is also interesting to note that the Speck baseline appears to have shifted following the candle source event.
Figure 2. Example results from five source experiments. For devices that sampled more frequently than each minute, data have been averaged for 1-min resolution. The top portion of plot shows the distribution of mass by particle size.
Figures 3 through 7 summarize the test results for the monitors, broken down by the type of particle source. The inset plot in each figure shows the particle size distribution. This is important because these results show that the monitors are sensitive to particle size – in particular for the < 0.3 micron particles from combustion & cooking (in Figure 7).
Figure 3. Response to dust and humidifier sources (generally larger particles).
Figure 4. Response to recreational combustion sources.
Figure 5. Response to frying and toasting.
Figure 6. Response to heating oils on gas and electric cooktops.
Figure 7. Response to cooking events that mostly emit small particles
The results show that the research-grade devices worked well, with the exception of events dominated by small submicron particles. This is expected due to the inability of the light-scattering sensors in these devices to detect such small particles.
Among the consumer grade monitors, the Egg and Speck showed the most problems. The Egg had very low responses across a wide range of events. The Speck had inconsistent correlation. The Awair had decent correlation but also had responses that were consistently low in magnitude. The AirBeam, AirVisual, Foobot, and Purple Air were better at detecting a wide range of sources, with the Foobot and Purple Air generally showing the most accurate responses. The exception for the Purple air was low responses to events dominated by large particles, such as the Arizona dust and the duct mop. For both the Speck and Awair the manufacturer claims to have upgraded the sensor since these experiments were performed.
As a group, the monitors generally missed sources that had the vast majority of their mass below 0.3 microns, including use of gas oven or cooktop cooking that did not include frying with oil. The exception was the Purple Air, which had low responses but high correlations, suggesting that it saw the events but not quantitatively. Note that PurpleAir is particle counter, which is more sensitive to low-levels.
Overall, these results suggest that the low-cost monitors will miss important sources of ultrafine particles and sources that may contribute to PM2.5 mass exposures in homes, and therefore we cannot generally say that these devices are suitable for controlling ventilation systems. However, four monitors were acceptable for identifying sources that emitted a lot of PM2.5: AirBeam, AirVisual, Foobot, Purple Air, and may have some value not as a primary control for a ventilation system, but perhaps to control and auxiliary boost function to increase ventilation during high particle events. They could also be used to operate stand alone filter units.
Funding was provided by the U.S. Dept. of Energy Building Technologies Program, Office of Energy Efficiency and Renewable Energy under DOE Contract No. DE-AC02-05CH11231; by the U.S. Dept. of Housing and Urban Development Office of Healthy Homes and Lead Hazard Control through Interagency Agreement I-PHI-01070, by the U.S. Environmental Protection Agency through Interagency Agreement DW-89-9232201-7. The authors thank Bill Roe of USA Operations Grimm Technologies for providing the Mini-WRAS for the duration of testing. Simon Walker assisted by executing many of the experiments and assisting with data capture.
CEN, EN 13779:2007. Ventilation for non-residential buildings — Performance requirements for ventilation and room-conditioning systems. ( withdrawn and replaced by EN 16798-3:2017)
NEN, 2011. NEN 8088-1:2011 nl Ventilatie en luchtdoorlatendheid van Gebouwen - Bepalingsmethode voor de toevoerluchttemperatuur gecorrigeerde ventilatie- en infiltratieluchtvolumestromen voor energieprestatieberekeningen - Deel 1: Rekenmethode (Ventilation and infiltration for buildings - Calculation method for the supply air temperature corrected Ventilation and infiltration air volume rates for calculating energy performance - Part 1: Calculation method).
Singer, B. C., & Delp, W. W. (2018). Response of consumer and research grade indoor air quality monitors to residential sources of fine particles. Indoor Air, 44, 1539–16. .