Condensation Particle Counters, Ultrafine Number and Efficiency

A condensation particle counter, or CPC, measures airborne particle number by enlarging tiny aerosol particles into droplets that can be detected optically. The number printed in particles per cubic centimetre is not PM2.5 mass, and it is not necessarily a count of particles below 100 nanometres. The instrument measures particles within its own size-dependent detection range, sample path and operating conditions.

To read CPC data correctly, begin with the minimum detectable size, the counting-efficiency curve, sample flow, working fluid or water-growth design, coincidence correction, time interval and quality flags. A concentration that rises sharply beside a road may reflect many small fresh particles while a gravimetric mass monitor changes little. That is possible because particle number and particle mass emphasise different parts of an aerosol size distribution.

This article walks through first principles, dilution and count arithmetic, the meaning of a fifty-per-cent cut point, false zeros, saturation, losses, colocation, uncertainty, worked exercises and answer checks. Hypothetical examples are clearly labelled rather than presented as observations. For operational measurements, the relevant instrument model’s documented inlet, counting range and calibration requirements override simplified teaching equations.

Why invisible particles can become countable droplets

An individual aerosol particle with a diameter of twenty nanometres scatters very little light compared with an optical droplet many times larger. A CPC does not usually try to see that original nanoscale particle directly. Instead it creates controlled supersaturation of a condensable vapour around particles, allowing some of them to grow into droplets large enough for an optical detection system to register.

Many conventional designs use an alcohol-based working fluid, while water-based designs use controlled temperature and saturation paths to achieve similar enlargement. The precise growth mechanism differs with instrument architecture. The key intellectual distinction is between the original particle diameter entering the CPC and the much larger diameter of the detected droplet. The optical pulse reports an event associated with a particle that successfully activated; it does not directly measure the original particle’s size.

Successful activation is size dependent. Tiny particles may fail to grow under a given supersaturation, while somewhat larger particles are detected more efficiently. The result is a counting-efficiency curve, not a perfect on-off boundary. A commonly reported d50 is the particle diameter at which a defined test aerosol has a fifty-per-cent probability of detection under specified conditions. It does not mean that every particle above d50 is counted and all below d50 disappear.

A device advertised as a ’10-nanometre CPC’ may mean its nominal fifty-per-cent size under a particular calibration aerosol and setup. That label must not be rewritten as ‘counts every particle from 10 nm upward with perfect accuracy’. Particle composition, working-fluid conditions, flow and internal design can influence activation. Source studies of water-based CPCs show that size response can vary with test aerosol, even when the same device remains mechanically healthy.

From optical pulses to particle number concentration

The simplest count calculation is number concentration equals detected particle counts divided by sampled air volume. If a fully efficient hypothetical instrument counts 12,000 valid events while exactly 2 cubic centimetres of representative sample air pass through the detector, the resulting concentration is 6,000 particles per cubic centimetre. The numerator is a number of countable events and the denominator is an air volume. No mass conversion has occurred.

A real CPC often uses internal flows, split streams, optical counting modes or algorithms that complicate direct reconstruction from raw pulses. The nominal sample flow at the inlet may not be the volume that passes through a particular optical detection region. When checking data, use the flow and correction definitions provided for the actual instrument model rather than guessing which pump specification belongs in the denominator.

Suppose a valid volumetric counting path is 0.05 litres per minute, and the interval lasts exactly two minutes. Its sampled volume is 0.10 litres or 100 cubic centimetres. If the corrected and valid event count were 250,000 over those two minutes, the concentration would be 2,500 particles per cubic centimetre. The exercise illustrates litres-to-cubic-centimetres conversion; it is not a claim that every commercial CPC exposes or uses that raw path.

The time axis can mislead. ‘Particles per cubic centimetre’ is a concentration, not a number per second. A time series with readings of 5,000 particles/cm³ every second does not mean the instrument detected exactly 5,000 individual events per second. The actual event rate depends on the sample volume per second and detection efficiency, while each displayed concentration may also include algorithmic smoothing or averaging.

If an instrument reports number concentration as particles per millilitre, that is numerically identical to particles per cubic centimetre because one millilitre equals one cubic centimetre. A value in particles per litre is different by a factor of one thousand. For instance 4,000 particles/cm³ equals 4,000,000 particles/L. Record both the figure and its unit before comparing two monitoring networks.

What a detection threshold actually means

The phrase ‘lower cut point’ is easily misunderstood. A CPC has no universal cliff-edge below which its probability falls instantly to zero. For a specified test aerosol the detector may approach full counting efficiency over a range of sizes, pass through a fifty-per-cent response point at a smaller size, and decline further below it. A small shift in the distribution toward its lower threshold can change reported counts even when the real total aerosol number stays constant.

Consider two hypothetical atmospheres with the same total number of particles but different sizes. Air A contains 10,000 particles/cm³ mostly well above a particular CPC’s effective lower response transition. Air B also contains 10,000 particles/cm³, but most are very small clusters near or below the detection transition. The CPC can report a lower count for Air B, not because particles ceased to exist, but because fewer of its particles were activated and detected.

The counting-efficiency curve depends on what ‘size’ means in calibration. Aerosol scientists use mobility diameter, aerodynamic diameter, optical diameter and other definitions for different instruments. These are not interchangeable labels. A CPC size-response experiment commonly uses particles selected by an electrical-mobility method, and a comparison to aerodynamic size from another instrument requires a physical conversion with assumptions.

The calibrated counting response can also depend on composition, particularly at sizes near activation limits. A water-based CPC may treat certain organic particles differently from some inorganic particles at the same nominal selected size. This does not mean one device should never be trusted; it means that a manufacturer specification reflects a described test population and controlled procedure, not every ambient aerosol mixture.

The right question is whether the apparent environmental change falls in a size region where counting efficiency is stable. If two instruments with different d50 values diverge during a burst of extremely small particles, that difference may reveal their detection windows rather than one unit failing. A size-resolved instrument can help test this explanation, but it brings its own charging, inversion and flow assumptions.

Why number and mass tell different stories

Particle mass scales approximately with volume and therefore with the cube of particle diameter for equal-density spheres. Doubling diameter multiplies volume and mass by eight under those simplified assumptions. Ten small particles do not necessarily weigh as much as one larger particle. A concentration expressed as particles/cm³ is thus not convertible to micrograms/m³ without substantial information about size, shape, density and composition.

This explains a familiar monitoring pattern. A fresh traffic-related particle-number plume can produce tens of thousands of very small particles/cm³ yet add comparatively little to PM2.5 mass. A dust event with fewer large particles may produce a pronounced mass signal instead. The CPC and gravimetric monitor are addressing different physical questions, and agreement cannot be evaluated by checking whether both numerical values rose by the same percentage.

Suppose two toy spherical particles share density and one is 20 nm in diameter while the other is 200 nm. Their diameter ratio is ten, so their individual mass ratio is 10 cubed, or one thousand. In this idealised model, one 200-nm particle has the same mass as one thousand 20-nm particles. Real atmospheric particles vary in density and shape, but the cube relationship explains why number and mass may behave very differently.

A CPC also cannot directly tell whether the counted particles were black carbon, organic droplets, salts or nucleation-mode clusters. It detects qualifying particles after a growth and optical counting process. Source attribution needs temporal evidence, meteorology and often composition or size-resolved measurements. A sudden CPC peak near a road is consistent with traffic influence, but not proof of a unique vehicle type or chemical substance.

When reporting to non-specialists, keep the measured metric inside the main sentence. Say ‘particle number concentration increased’, not merely ‘pollution increased’, if the only evidence is CPC number data. Higher number concentrations may be environmentally important, but they do not supply a directly equivalent PM2.5 mass concentration or a universal health-risk conversion.

Coincidence, dead time and concentration limits

At low event rates a detector may resolve separated droplet pulses. At high rates several droplets can pass close together, pulses may overlap, or detector electronics may be temporarily unable to register the next event. These mechanisms can cause undercounting. Instrument software may correct this nonlinearity within a validated concentration range; outside that range the correction can become unreliable.

In a simplified nonparalysable dead-time model, the observed rate r relates to a true event rate R by r = R/(1 + R times tau), where tau is dead time in seconds. Rearranging gives R = r/(1 – r times tau) when the assumptions hold. The expression illustrates why the correction grows rapidly as r times tau approaches one. It is not a substitute for the calibrated coincidence model of an actual CPC.

Take an invented detector with 0.0001 seconds of nonparalysable dead time and an observed count rate of 500 events/s. Multiplication gives 500 times 0.0001, or 0.05. The corrected rate under this toy model is 500/(1 – 0.05), approximately 526.3 events/s. The correction is about 5.3 per cent relative to the observed rate. A reader who simply adds 0.05 events/s has confused a dimensionless factor with a count rate.

If the observed rate were 9,000 events/s at the same assumed dead time, the denominator becomes 0.1 and the arithmetic produces a huge correction. That does not prove the instrument can truthfully resolve the environment at that rate. Close to model singularities, uncertainty is greatly amplified and detector assumptions may fail. Practical instruments specify maximum validated number concentrations, modes, dilution and status flags precisely to avoid unjustified extrapolation.

Dilution may make a high-concentration aerosol measurable by moving its count rate into a validated range. Yet multiplying the diluted concentration by a nominal factor is justified only if the dilution ratio is measured and aerosol losses are characterised. Small particles can deposit on surfaces during dilution and tubing transport. The dilution itself therefore introduces uncertainty and can modify the population being counted.

Transport losses from the outdoor inlet

Ultrafine particles can be lost in sampling tubing by diffusion to internal surfaces. The smaller the particles, the more strongly Brownian diffusion can affect penetration under many conditions. Losses also depend on tube length, diameter, flow and bends. A CPC with excellent internal counting efficiency may still underestimate particles at the outdoor inlet if the sampling line transmits only part of the relevant aerosol.

Long inlet tubing is therefore not a harmless convenience. A monitor on a rooftop connected through an extended indoor tube may receive a size-distorted particle distribution relative to a short direct-inlet arrangement. The record should describe inlet height, tubing and residence time. Where the scientific question concerns the smallest particles, transport-loss estimates or checks become especially consequential.

A flow fault can distort both transport efficiency and computed concentration. If volume flow falls and software uses a stale nominal flow, the same count rate could be normalised by too large or too small an assumed volume. A pressure-dependent instrument may apply mass-flow or volumetric corrections. The analyst must identify which form the instrument reports and at what reference conditions, rather than correcting pressure a second time by intuition.

Leaks can create unexpected consequences. A leak upstream of a counting path may dilute the outdoor aerosol with indoor air or admit unconditioned air. A leak downstream of the detection region may affect flow stability differently. The location of the leak matters. The appropriate response is to perform the prescribed flow and leak checks and document maintenance, not apply an arbitrary numerical adjustment to the final time series.

Reading a plausible urban episode

Imagine that a station normally records 3,000 to 6,000 particles/cm³, and a sharp rise to 28,000 appears during a morning interval. This is a hypothetical signal only. Before interpreting it as a local emission event, verify whether the concentration is within the instrument’s validated counting range, whether a dilution mode changed, and whether sample flow and working-fluid or growth-tube temperatures remained acceptable.

Then align the CPC series with other measurements. A nearby traffic count, wind direction, black-carbon monitor or size-distribution instrument may help discriminate competing causes. If the number spike coincides with an unusual wind shift but not with traffic variation, a source farther away may be plausible. A consistent temporal association still does not substitute for full source apportionment.

A third step is to inspect timestamps and smoothing. Some output channels display a moving average, while others deliver short internal concentration estimates at regular intervals. A smoothed value can lag a sudden plume, and two differently averaged CPCs can disagree during the event even if they would agree under steady concentrations. For comparison, harmonise the analysis window rather than shifting one series until it looks aligned.

Suppose the station’s one-minute concentrations are 5,000, 10,000 and 30,000 particles/cm³ and all three minutes are fully valid. Their time-weighted three-minute mean is 15,000 particles/cm³. If the third minute is missing rather than 30,000, the two-minute observed mean is 7,500. It is not defensible to label 7,500 as the complete three-minute mean without a stated missing-data rule.

A field note should separate detected concentration from ambient particle formation mechanisms. Elevated particle number can follow direct emission, atmospheric nucleation and growth, or transport of a particle-rich air mass. CPC data alone generally cannot tell these processes apart. Combining observations across a size spectrum, wind and chemical context may help, but one number is a starting clue, not the whole mechanism.

Evaluating two counters side by side

A proper side-by-side study begins by checking whether both devices have comparable lower size responses. One CPC might detect a larger fraction of freshly formed nanometre-scale particles than another. If the ambient distribution resides well above both lower transitions, agreement may improve. A disagreement that depends on time of day or aerosol source can therefore be more informative than a single overall correlation coefficient.

Compare like-for-like flow conditions, sample lengths, dilution settings and instrument ranges. If one monitor is connected to a long sampling manifold, the two devices may receive genuinely different size distributions. The comparison then tests the entire sampling systems rather than just their internal counting chambers. That distinction belongs in the method description.

Plot paired number concentration on appropriate scales, including a one-to-one reference. A ratio of 1.2 tells a different story at 1,000 versus 100,000 particles/cm³ when counting statistics and saturation limits vary. Examine absolute differences, proportional differences, detection thresholds and outliers. A strong correlation can coexist with a systematic undercount due to a different minimum detectable size.

At low counts, Poisson-like counting variation can be noticeable when relatively few detection events occur during an interval. At very high counts, coincidence and dilution uncertainties may dominate. The error mechanism changes across the range. A single constant percentage uncertainty for all concentrations should not be asserted unless supported by a model-specific uncertainty assessment.

Practical questions and transferable calculations

A school group hears that a CPC counted 20,000 particles/cm³ and asks whether this means 20,000 micrograms/m³. No: one result is particle number divided by volume, the other is particle mass divided by volume. They cannot be equated because particle sizes and masses have not been measured. A worked conversion would require a measured size distribution, density assumptions and a careful integration, not a replacement of the unit label.

An analyst receives two number results: 12,000 particles/cm³ and 15 million particles/m³. The second equals 15 particles/cm³ because one cubic metre contains one million cubic centimetres. The first is eight hundred times the second. If the analyst treats those values as nearly equal based on the zeros in the display, the error is purely volumetric unit conversion.

A student asks whether d50 of 10 nm implies that a CPC measures PM10. It does not. PM10 denotes a mass-based aerosol size fraction around ten micrometres aerodynamic diameter, whereas a CPC’s 10-nm d50 concerns detection efficiency for nanometre-scale particles. The similar numeral disguises a thousand-fold difference in nominal length unit and a different measurement principle.

A monitoring chart shows a brief CPC reading of zero while optical PM2.5 remains elevated. Possible causes include very low detectable number, sampling disruption, detector or fluid faults, telemetry suppression, or a particle population not captured efficiently. The zero does not demonstrate particle-free air. Inspect detection status and count-rate channels before claiming an environmental absence.

Derive a correction without mistaking a model for data

Counting efficiency introduces another mathematical layer. If a controlled calibration aerosol at one selected size sends a known reference rate of 1,000 particles per second into a comparable detector path and the CPC detects 650 events per second after appropriate background and flow corrections, the estimated counting efficiency at that test size is 0.65, or 65 per cent. The numerator and denominator must describe matching aerosol streams and comparable conditions; otherwise the ratio reflects transport or flow differences as well as detection.

Now repeat the calibration with several sizes and plot efficiency against diameter. A response that rises from twenty per cent to eighty per cent over a narrow size range is not proof that the physical particle concentration changed during the experiment. It shows size-dependent detection probability. Fitting or interpolating a d50 from that curve requires enough calibration points near the transition and attention to the reference aerosol’s sizing uncertainty.

Imagine an idealised ambient aerosol in two narrow size categories. It contains 2,000 particles/cm³ near one size with 40-per-cent CPC efficiency and 8,000 particles/cm³ at another with 95-per-cent efficiency. Expected detected number, before other losses, is 2,000 times 0.4 plus 8,000 times 0.95, giving 800 plus 7,600, or 8,400 particles/cm³. Actual total is 10,000. This simplified calculation demonstrates why dividing every CPC value by one fixed efficiency is unreliable for changing size distributions.

If the first category increases to 6,000 while the second stays at 8,000, the true total becomes 14,000 but expected detected number is 6,000 times 0.4 plus 8,000 times 0.95, or 10,000. The observed count rises by only 1,600 even though the real total rises by 4,000. A second CPC with better smallest-particle efficiency might show a larger increase. Instrument disagreement can therefore provide a clue about shifting particle sizes rather than just sensor malfunction.

The efficiency curve should never be applied outside its validated conditions without qualification. Operating temperatures, working fluid, pressure, aerosol chemistry and aging of internal surfaces can influence activation. Laboratory standards allow controlled calibration, but an ambient distribution can be broad and compositionally mixed. The appropriate interpretation is often ‘number concentration above an instrument-dependent lower response window’, accompanied by the documented response if quantitative comparison demands it.

Time response, zero checks and environmental drift

A CPC may respond quickly to a step in incoming concentration, but its displayed signal is not necessarily an infinitely fast reflection of outdoor air. Residence time inside the tubing and growth region introduces delay, while software may smooth count-rate fluctuations. A ten-second time lag between devices could result from sampling lines and processing, not distinct source locations. Align physical sampling intervals before extracting the arrival time of a plume.

Testing with clean filtered air can expose a background count attributable to leaks, contamination or spurious optical events. Yet a near-zero reading on such a test does not validate the high end of the counting range. Background, concentration linearity and cut-point response are separate properties. An instrument can pass one and fail another, which is why a comprehensive quality-control record identifies the purpose and outcome of each procedure.

Water-based and alcohol-based CPCs have different operating details. A device may depend on wick condition, reservoir level, controlled temperature gradients, stable liquid supply or humidity management. These are not interchangeable maintenance steps. A technician should consult the manufacturer’s manual, not a generic CPC checklist, because an intervention appropriate for one growth system can damage or destabilise another.

A sudden step coinciding with refill or servicing should be treated cautiously. It may reflect real external air variability, but the service record offers a competing explanation. Where instruments automatically suppress data during warm-up or fluid stabilisation, preserve those flags. Filling a plot with interpolated values can conceal the very conditions necessary to assess whether the post-service readings are valid.

Pressure and temperature can alter air volume, and the concentration unit depends on the reference basis for that volume. Instruments may report concentrations corrected internally to a stated condition. Do not apply a second ambient-to-standard conversion unless the documented data product requires it. Confirm whether the quantity is ambient number per unit volume or concentration at reference conditions before combining observations from different networks.

Sampling location changes the question being answered. A CPC at a busy curb may detect fresh exhaust-related bursts that disappear a short distance away as the plume dilutes and particles evolve. A station at rooftop height may represent a more mixed air volume. Neither location is intrinsically wrong; each represents a distinct exposure or atmospheric process. State the inlet placement before generalising one result to a community.

Build a defensible uncertainty statement

Counting statistics matter when the number of detected events in a measurement interval is small. For an ideal count process with independent events, a Poisson approximation gives standard deviation near the square root of count N. At N equal to 100, the standard deviation is approximately ten events, or ten per cent; at N equal to 10,000, it is one hundred events, or one per cent. This only represents counting scatter under the assumed model, not total CPC uncertainty.

A very busy aerosol may violate simple independence assumptions because of overlapping events and concentration changes within the interval. At high loads, coincidence correction, saturation and dilution can dominate. Likewise, a concentration averaged over one second during an exhaust plume may have genuine environmental variation much larger than Poisson counting noise. The uncertainty budget must reflect both instrument response and atmospheric variability.

An analyst could separate uncertainty contributions into calibration, sizing and counting efficiency; sampling flow; particle losses; background; coincidence correction; and time representativeness. Some are systematic, such as a persistent loss in a long inlet tube. Others vary in time, such as unstable sample flow or rapid plumes. Averaging many hours will not automatically erase a fixed low-size transmission bias.

When comparing two instruments, one may have a broad validated range and the other may apply strong correction at high concentration. A regression slope near one is useful but cannot resolve all differences. Examine paired data near the low threshold, at typical urban background and during peaks. State which range was verified rather than describing all number concentrations as interchangeable.

A confidence interval should be anchored to the actual test design. If measurements are serially correlated, treating every one-second reading as an independent replicate overstates precision. Blocks of time, instrument maintenance periods, weather changes and source episodes may be more relevant units of analysis. A transparent report says how uncertainty or variability was estimated rather than printing a percentage without its provenance.

Worked investigations readers can reproduce

A clean-air challenge produces two counts over ten-second windows: seven events in one and eleven in another. The approximate raw background rates are 0.7 and 1.1 events per second. They do not by themselves reveal ambient particles/cm³ because the effective sample volume and possible detector background corrections have not been applied. The right next step is to consult the instrument’s zero acceptance criteria and counting path.

A field technician sees 60,000 particles/cm³ while a device manual specifies a validated operating range that ends below this number in its current mode. The display does not extend the manufacturer’s validation. The appropriate response is to flag the exceedance, consider a suitable dilution or higher-range method, and preserve the raw trace as contextual information rather than treating the number as fully quantitative.

A student compares four sequential one-minute values: 2,000, 2,000, 8,000 and 8,000 particles/cm³. The four-minute mean is 5,000. If the last minute failed quality control and is removed, the accepted three-minute mean becomes 4,000. Neither mean represents the complete four-minute ambient history once the last minute is missing; describe the accepted coverage and any required completeness rule.

A laboratory has a measured dilution ratio of ten to one, and its diluted CPC stream reads 2,500 particles/cm³ within valid range. A first-pass estimate of undiluted concentration is 25,000 particles/cm³. This estimate assumes a well-characterised ratio and negligible losses or a validated loss correction. If small-particle penetration is only 80 per cent through the diluter, the corrected original number would be 25,000/0.8, or 31,250, under that narrow model. It is not acceptable to assume 80 per cent without evidence.

A reader sees ‘count median diameter’ in a size-distribution report and believes the CPC measured it. A stand-alone CPC generally supplies integrated number concentration, not a full size distribution. Median diameter requires size-resolved information or an additional method. Pairing a CPC with a differential mobility analyser can support a particle-size experiment, but interpretation must then include charge selection, transfer functions and inversion assumptions.

A final arithmetic challenge asks whether 1,200,000 particles per litre is greater than 2,000 particles/cm³. Convert one litre to one thousand cubic centimetres: 1,200,000 per litre equals 1,200 per cubic centimetre. Therefore 2,000 particles/cm³ is higher. This is a unit exercise, not a statement about which air sample is more hazardous or chemically concerning.

Translate results without overclaiming

A well-formed statement reads: ‘The CPC reported an hourly average particle-number concentration of 8,400 particles/cm³ for the instrument’s stated detection range, with 58 of 60 minutes valid.’ That sentence preserves the particle metric, time basis, unit, approximate detection window and data coverage. It is substantially more informative than the phrase ‘8,400 pollution particles’ without a definition.

In a scientific figure, place the size-response information in the caption or method notes. If the device has a nominal lower d50 of 10 nm, say that this is a calibration response point, not an absolute ambient threshold. In a comparison study, state both instruments’ response curves where known. Otherwise an apparently precise difference between counters may conceal the fact that they are not responding to the same particle population.

For a public-facing interpretation, do not convert a number peak to a health effect using a universal multiplier. Health research may consider ultrafine particle exposures, but measurements, populations, mixtures and endpoints differ. A CPC number series is meaningful environmental evidence; its limits are part of that meaning. Point readers to appropriate health authorities rather than presenting an unsupported dose-response equation.

The transferable lesson is to reason backward through the measurement: distinguish an original nanoscale particle from its grown detection droplet; confirm which particles have a reasonable chance of activating; check count-rate corrections, flow, tubing and time aggregation; and only then attach environmental significance. Once these definitions are clear, a CPC trace becomes an interpretable record rather than an impressive but ambiguous string of digits.

One deceptively simple question is whether a reported particle number is an arithmetic mean of individual counts or a mean of already-normalised concentrations. If sampled volume changes during the interval, those calculations need not agree. Consider two equal-duration periods, the first sampling one cubic centimetre with 2,000 events and the second sampling two cubic centimetres with 6,000 events. The respective concentrations are 2,000 and 3,000 particles/cm³. Their equal-time mean is 2,500, whereas combining all events and all volume produces 8,000/3, or about 2,667. The right quantity depends on whether the purpose is time-average ambient concentration or pooled sampled-volume concentration.

The distinction matters when instrument flow varies or downtime interrupts an interval. For a properly controlled CPC the flow should be stable within validated tolerances, but a field data analyst must still know the aggregation convention. Software can store instantaneous estimates, count-weighted values or arithmetic means after quality screening. Before recalculating a published number, locate the product definition and reproduce a short known period. A numerical result that seems close is not proof that the procedure used the same denominator.

Environmental gradients also make calibration transfer difficult. A CPC tested with laboratory particles of one composition may perform predictably with those test particles while undercounting a different ambient distribution concentrated near its activation limit. The appropriate response is not to invent a simple site correction. Instead, describe the calibration aerosol, test sizes, aerosol line and operating settings. An independent size-distribution instrument can help test whether a shift in lower-size particle abundance explains a systematic difference.

Particle-number data can be especially misleading when plotted on a linear scale alongside another variable with a much wider numerical span. A large-looking visual spike may simply reflect axis scaling. Show units, transparent baseline, data completeness and chosen averaging interval. If using a logarithmic axis to display a range from hundreds to hundreds of thousands of particles/cm³, label it plainly and do not calculate arithmetic differences from equal-looking vertical distances.

A credible interlaboratory exercise gives the participants a shared set of raw intervals, flow records, dilution ratios and validity flags. Each team independently calculates accepted concentration, records excluded periods and explains whether the objective is time-weighted or volume-weighted averaging. Agreement in both the calculation and the stated limitations is more persuasive than agreement in a plotted trend alone. Any unresolved differences should be traced to units, thresholds, corrections or sampling losses rather than silently harmonised.

For a personal understanding check, cover the numerical display and reconstruct the result in words. Identify what qualified as a detected particle, which fraction of the original aerosol reached the counter, how a raw pulse became an event rate, and what sampled volume supplied the concentration denominator. If you cannot answer one of those questions from the documentation, the correct conclusion is that interpretation remains conditional, not that the missing information can be guessed from the number of digits on the screen.