The manuscript presents a comparison of four impactors with regards to their sampling performance of UFP. Laboratory tests and ambient intercomparisons were performed and results on measured cut-off diameters, the sharpness of the cut-off curve and internal particle losses are presented, while further potential sampling artefacts including evaporative losses and particle bounce are discussed in more speculative ways. Generally, the manuscript is reasonably well written and structured and the topic is important and interesting. There are quite a number of issues, however, that should be addressed before the paper can be considered for publication.

Major issues

• The authors selected 2 commercial impactors that have since long been used for ambient sampling (MOUDI, ELPI) and two more compact instruments that do not seem to be commercially available (ultraMOUDI, PENS). Unfortunately, however, the commercial and established impactors were used in a rather unconventional way, by removing one or more of the lower stages and replacing them with a quartz-fibre after-filter. These modifications likely change the performance characteristics of these impactors compared to their original configuration and the results of the present study are therefore difficult to transfer to other applications of MOUDI and ELPI. Quartz filters have different sampling characteristics as the typically applied flat impaction substrates (in terms of adsorptive and evaporative and bounce-off artifacts, for example), the pressure drop is different in the modified version and possibly also the internal air flow characteristics have changed. What was the motivation to apply these modifications instead of using the impactors as intended by their manufacturers? In any case, the modifications should be made very clear already in the abstract and potential impacts on the UFP sampling performance should be discussed in the manuscript.
• In some places in the manuscript, the authors seem to imply that UFP sampling is only “correct” with a cut-off diameter exactly at 100 nm (e.g. L125 or L481-485). Given that the value of 100 nm is rather a convention than a physical law, I would suggest to relax such statements. Properties and impacts of particles change rather gradually than abruptly at 100 nm, so slight deviations from a nominal 100 nm cut-point are likely not very relevant, especially with regards to mass-based sampling and analyses as done in this study.
• More important are potential artefacts from evaporation and bounce-off. These are discussed in several places of the manuscript, but often in rather qualitative and speculative ways only. I give some suggestions below, but would like to encourage the authors to try and use their data to improve the quantitative understanding of these critical issues in UFP sampling.  
• The Introduction suffers from lengthy parts describing text book knowledge or irrelevant historical developments, while lacking a proper appreciation of the literature of b) impactor comparisons, including for nanoparticle characterisation, and b) applications of impactors for chemical UFP studies.
• In many places, the manuscript is written in slightly imprecise language, which is inappropriate for a scientific manuscript. Some examples follow below, but more can be found throughout the manuscript and a careful check by the more experience co-authors should be done to eliminate any ambiguity and impreciseness.

Further issues

• The abstract could be improved by shortening the more descriptive parts and including more quantitative results, especially for the field intercomparisons.
• L15-16 “techniques developed for larger particles”: Unclear, which techniques are referred to here. Impactors have since long been applied for sampling in the UFP size range as well.
• L22-24: Unclear sentence structure, please rephrase
• L24 “successfully”: Did the authors expect otherwise? Is it really worth noting in the abstract that the impactors work as designed and intended?
• L29: evaporative and bounce-off artifacts have long been known as critical issues in impactor sampling. It might be misleading to argue they were identified in this study.
• L31: Based on the results that follow later, I would argue they were partially mitigated at best.
• L35: What does “semi-industrial air” mean?
• L38: The inclusion of 6-PPD as a typical UFP marker is interesting. Is there evidence that tyre wear is a primary source of UFP? Given the mechanical abrasion process, I would expect these particles in the coarse fraction only.
• L40: …were about 29% higher of 30% lower…?
• L45: …with decreasing pressure drop…?
• L46: Isn’t it trivial that marker concentrations increase with higher air concentrations? What do you mean here?
• L94: These might not be the best references for the statement, as they do not seem to deal with UFP composition.
• L124-129: Misleading and confusing paragraph, please rephrase. The cut-off diameter is a property of any impactor stage, it does not “determine a threshold” to distinguish fine from ultrafine particles. The aerodynamic diameter is just one of several diameter definitions, it is not “an abstract measure describing a sphere…”. And the cut-off diameter does not have a “sharpness”, which varies with particle properties.
• L131-132: What do you mean by “larger” and “smaller” particles here? I doubt sedimentation plays a role inside an impactor.
• L133-134: Particles can bounce without breaking apart as well! What does “falsely adding material” mean?
• L136-138: Such evaporative losses are likely different between quartz-fibre filters, where particles are exposed more individually and directly to ventilated air as compared to compact deposited material on flat impaction substrates.
• L138: How would evaporative losses lead to mass gain? What do you mean by “hampered conclusion”, conclusion on what?
• L148: Why have exactly these 4 impactors been selected? Have other models been considered?
• L150: I guess the pressure was regulated between impactor and pump, but the air flow rate was measured at the impactor inlet? The current phrasing might be slightly misleading.
• L158: While the MOUDI does use 47mm substrates, the area the particles are deposited on is actually smaller than 47 mm.
• L162: What do you mean by “clean flow path”? How is it different (cleaner?) than with the MOUDI I version?
• L165 and 195: If the lowest stages of the MOUDI and ELPI are removed and replace with an after-filter holder, how does this change pressure drops and air flow rates through the impactors? Could this lead to different sampling characteristics of the upper stages and/or influence particle bounce?
• L167-168: The given cut-points sum up to 8 stages only, while nine stages are mentioned in the text.
• L168+175+199+209 and others: Were the applied after-filter holders custom-made or original parts of all four impactor models?
• L171: Who developed the ultraMOUDI, the authors or msp/TSI? Was it indeed developed specifically for this specific study?
• L173: The given cut-points sum up to 2 stages (0.1 – 1 and 1 - 2.5 µm), while three stages are mentioned in the text.
• L188: Same here, the number of given cut-points does not fit the number of stages.
• L207: How was the original PENS modified and by whom?
• L212 and elsewhere: I’d suggest to always use “cut-off diameter” or “cut-off size” or maybe “cut-points” instead of just “cut-off”.
• Section 2.2.1: The detailed descriptions of the operating principles of MPSS and DMS are not really needed and could be replaced by proper references.
• L253-255: Confusing. Why exactly was a 5 m long line needed in the first place? The line was not heated, as it seems, but in L253 it says “heated line”. Which one is correct?
• L270: Figure S1 is helpful in understanding the experimental setup and should be placed in the main manuscript.
• L285: Eq. 2 is textbook knowledge and not needed here. If so, it needs to be referred to and all variables explained.
• L295: Was the particle filter included or removed for this application?
• L303: Flushing the chamber with ambient air seems unusual. What are ambient particles and gases needed for in the chamber runs? Is the resulting SimSOA actually a mixture of chamber SOA and ambient aerosol? If so, what fraction of ambient aerosol particles does SimSOA contain?
• L323: I can’t follow how step 3 was actually performed, please explain more. What is left, if all impactor stages are removed? Or do you mean the entire impactors were removed and the inlet line directly detected to the outlet line to pump and MPSS?
• L325: The motivation for step 3 is not entirely clear to me. Did these measurements show different concentrations than step 1, i.e. PNSD upstream the impactor? If so, why?
• L339, Figs. 2&3: Check your definitions and explanations of dp10 and dp90. Do you mean dp90 refers to the diameter at which 90% of particles are deposited at the pre-UFP stage, i.e. 10% of particles at this diameter are collected as UFP? Similarly for dp10, the phrasing is unclear. The diameters are defined in different ways in Figure 2 and 3, please harmonize.
• L343: What is the “original” transmission curve? Are there different versions of it?
• L356: What is meant by “semi-industrial environment”? Is there heavy industry in the area? Do you expect strong UFP emissions nearby?
• L357: 30 L/min through 3/8” tubing should lead to turbulent air flow, I suspect. To avoid in-tube losses, laminar flow is typically preferred. Did the tubing add significant back-pressure to the setup? Did you check the sampling flow rate at the tubing inlet or at the impactor inlet (without the tubing)?
• L362: Please report the O3 scrubbing efficiency.
• L373: What temperature was used for filter baking?
• L391: Custom-made glass frits or commercial ones? Why not use conventional syringe filters instead?
• L397: Can you estimate the volume of the residual droplet?
• L398 and elsewhere: “Millipore water” does not really fit, as your system was not a Millipore system.
• L410: Which reference material?
• L428: Which NIST standard? Where and how was this used, it has not been mentioned before.
• L431: Please give the IC LODs in the Supplement Table as well, not only the air-equivalent ones.
• L436ff: There might be an error here. The given sampling volumes refer to the entire filter, while only half of the filter was actually extracted (L415). This needs to be included in the air-equivalent LOD calculation and the given LODs might be different by a factor of 2. Please check.
• L441: Better explain the different considered uncertainties here (briefly) and give a Table of their values in the Supplement (rather than text).
• L456 and Table 1: Aerodynamic cut-off diameters should be calculated not only for NaCl, but also for SimSOA and soot, using appropriate values for density and shape factor. Only these can be compared to the nominal impactor cut-points, which are defined in aerodynamic diameter as well.
• L466: See comment above on dp10 and dp90.
• L469: Unclear and possibly misleading statement, please check. What do you mean by “lower efficiency for capturing small particles”? Did you consider the differences between aerodynamic and electromobility diameter here?
• L472: The stickiness of NaCl depends on RH and their water shell. Not sure they can generally be considered sticker than SOA, which often is semi-liquid and thus rather sticky.
• L481ff: Unclear paragraph. What is a “suitable” cut-off? Are slight deviations from 100 nm really relevant, especially when given in a different diameter (electromobility) and for a mass-based analysis? Unclear, how MOUDIs would “shift” their cut-off diameter in “more realistic air mixtures”, as the cut-pint is actually fixed by design. Please rephrase entire paragraph.
• Table 1: The figures of the impactors are misleading, as the Table characterizes their last stage only (or close-to-last, if modified). They could be removed to make the Table more compact.
• Figure 3, caption: What does “original impactors” mean? Were these experiments performed in the original configuration of MOUDI and ELPI, i.e. without removing last stages?
• L520: Transmission of the ELPI is quite visibly increasing at largest diameters. I doubt, the statement is justified. It is also unclear, which “uncertainty” the authors refer to here. A transmission of 6% is likely significantly larger than 0, depending on the total transmission concentration.
• L521: Similarly, the MOUDI line in Figure 3 does not seem to exactly reach 0, i.e. “point of no transmission”.
• L534: Losses for the MOUDI were basically the same as for the ultraMOUDI. For ELPI, they were just slightly higher (L505). The statement of “increased losses” is not exactly correct. Also, slightly higher losses likely have much less impact on final mass-based concentrations as bounce-off of large particles.
• L592ff and Fig. 5: These data should be used to estimate the potential mass bias from bounce-off. Stating that 98.5% of particle mass was captured might be misleading, if the actual UFP mass was much smaller than 1.5%. From Fig. 3, it seems the UFP mass was ~0.7 µg/m3. If transmission to the UFP stage was indeed 3.3 µg/m3, this would mean that only ~20% of UFP mass concentration is actually from the ultrafine particle size range, while 80% of mass (and chemical composition) are bias from the bounce-off sampling artefact. Marker concentrations from such biased sampling would be difficult to trust. If such data is available for the other impactors as well, similar estimates should be done and reported. Bounce-off is a well-known and critical artefact of impactor sampling and should be characterized as quantitatively as possible from these lab experiments. It might also be worth mentioning that SOA is likely not a worst-case scenario for bounce-off, given the comparably sticky nature of SOA particles.
• Section 3.4: Page 25 largely repeats what has been given in the experimental section already. Please make sure to remove redundancies throughout the entire manuscript.
• L641: Better “deviation” instead of “over- or underestimation” as there is no ground truth available.
• L648: 37% maximum difference seems low (from Fig. 6), please double-check.
• L648: How do you know about >90% particle phase fraction of PAHs?
• L689-690: Given that with the impactors applied, the potential of evaporative losses is inversely related to the potential of bounce-off artifacts, I am not convinced such a statement can be made. From Fig. 7A it can be seen that between MOUDI and ELPI, marker concentrations often increase, even though the ELPI has the largest pressure drop. This could point to bounce-off having a stronger impact on marker concentrations than evaporation. Potential evaporative losses should better be discussed based on the volatilities of the marker compounds, possibly in combination with the pressure drops experienced.
• L698: Not sure if I follow here. How would the PENS attribute twice as much mass to UFP as there is total ambient mass available? And for the MOUDI it would be 18% of the total ambient PM mass? There is quite large scatter in Fig. 7B, and on logarithmic axes. Please double-check if these statements are regarded robust.
• L740: transmission of large particles is not entirely eliminated
• L750: The loss is not 69%.
• L752: Considering the confounding with potential bounce-off artefacts, the concentration differences between ELPI and ultraMOUDI are not high enough to dismiss its suitability for semi-volatile UFP components, in my opinion. The same applies to the MOUDI. Not considered at all in this study are possible differences of evaporative losses from quartz fibre filters (particles deposited individually on fibres) as compared to the more typical flat substrates (particles deposited in dense spots or layers).
• L752: What does “stable” mean in this context?
• L760: If 690 hPa (ELPI) is “high” and 420 hPa (ultraMOUDI) is “moderate”, I would consider 600 hPa as rather high as well.

This study inter-compares different cascade impactors able to sample UFPs. As the earlier devices based on particle collection by inertial impaction did not have a good size-resolution below 0.1 um, different modifications/extensions of the original impactors were performed. These modified versions implying different underlying principles and design brought a variety of advantages and disadvantages, strengths and weaknesses. Therefore, it is an original idea and a useful task to compare them. It is one of the strengths of this manuscript that the selected devices were compared both under laboratory and field conditions. It is also remarkable that the authors used different test aerosol systems with different properties allowing for the study of evaporation and bouncing, two confounding factors of precise aerosol size distribution measurement. Another strength of the work is the exigent planning, completion and analysis of the related experiments, as well as a precise description of their work and presentation of the results both in the manuscript and in the supplementary material.

In the light of the use of modified impactors instead of the original ones an imminent question is the relevance of current results regarding the original impactors. The authors should include a section or paragraph commenting on this important issue.

In addition, the authors did not evaluate losses for particles with diameter below 30 nm. They mentioned that „Due to the relatively larger uncertainties in the reference instruments for very small diameters, i.e. d_m <20 nm, we decided to evaluate the particle number concentration at 30 nm for determining the losses in the ultrafine fraction”. As deposition by diffusion increases steeply below this size, it would be important to reflect on this issue, at least by expert judgement or/and using data from the open literature.

Finally, the manuscript is quite long. If the authors can find a way to compact it without loss of pertinent information, it would be nice. For instance, the introduction could be shortened, but I would let the authors to decide on what to shorten.