The manuscript by Stavros Amanatidis et al. entitled “Efficacy of a portable, moderate-resolution, fast-scanning DMA for ambient aerosol size distribution measurements” reports an intercomparison of a novel SEMS or MPSS system consisting of the “SPIDER” DMA and a “MAGIC” CPC. The main question raised in the paper is whether this system operated a relatively low DMA resolution is able to catch the key characteristics of ambient aerosol size distributions.

The paper comprehensively evaluates the transfer function of the “Spider DMA“, an also reports field measurement data showing very good agreement between the novel system and a more traditional scanning DMA system.

From this reviewer’s point of view, the paper is very welcome to be published in “Aerosol Measurement Techniques” although it should not be accepted for publications until following remarks have been considers in a revised version of the manuscript:

1)    P1/L18ff 

The authors state that traditional mobility analyzers are large and most often not suitable for UAV, but that the “Spider DMA” would be appropriate to be used on moving platforms. 

It would be helpful if the authors could also elaborate on how the sheath air flow supply of the Spider DMA would look like when used on n UAV or other moving platform – especially when compared to “traditional” sheath flow supplies.

2)    P3/L74 ff. 

The authors report the use of a soft X-ray charge conditioner. It would beneficial to specify brand / make of the specific instrument. It is not the objective of this paper, but nowadays soft x- ray ionizers are often used for size distribution measurement, without knowing the actual charging probabilities or knowing if the Fuchs charging theory is applicable without any adaptions. Therefore, at least the type of the used instrument should be stated.

3)    P5/L107: 

It should be named “Fuchs charge distribution” 

4)    P5/L114 ff 

The fact that down-scan peaks have a higher maximum number ratio and are also narrower than the up-scan peaks confuses me. Typically, one would expect the opposite, where the down-scan transfer function also often exhibits a distortion or tail. Therefore, the – as far as this reviewer can say – the most common way would be to use the up-scan data for scanning DMA data rather than the down-scan data.

It would be extremely important for a clear understanding – especially for non-DMA expert readers - to elaborate in more detail on this topic and explain the differences between the down-scan/up-scan transfer functions of the Spider DMA vs. traditional DMAs.

The authors present the application of the “Spider”-DMA to ambient measurements, demonstrating that despite the lower resolution compared to typical SMPS/DMPS systems, the characteristics of the size-distribution can be retrieved with high accuracy making the “Spider”-DMA a suitable instrument for lightweight particle size-distribution measurements. The article is written in a concise style, the presented Figures are of high quality and the conclusions are scientifically sound. The topic is certainly of interest for the community, as the interest in e.g. unmanned aerial vehicle measurements is increasing due to the usage of drones, where this device could come in handy. I can recommend publication in Atmos. Meas. Techn. subject to some minor revisions:

• 2, l.26-28: Doesn’t this also come with the benefit of requiring smaller sheath-flow supplies?
• 3, l.80: Standard SMPS systems often use particle counters with 1 lpm flow rate. This would increase the counting statistics compared to the spider DMA. Does the reduced resolution still outweigh the benefit of an increased sample flow in terms of counting statistics?
• 3, l.84: Why is a different sampling rate used for the LDMA system and the Spider system?
• 4, l.90: Please be more specific on the inversion algorithm. Tikhonov regularization? What was the choice for regularization parameter? Was the same method applied to the LDMA?
• 5, l.117: Please be more specific where this difference comes from. What’s the difference between downscan and upscan voltage operating mode which explains these discrepancies? Figure 4 also shows that there are slight changes in the inverted distribution. Is this happening repeatedly? And if one scan gives more precise results, why using an up- and downscan procedure instead of one direction only?
• 6, l.129: It would be very helpful to see the same for the LDMA in order to demonstrate the higher counting statistics provided by the spider DMA. What is the advantage we gain in “counts” compared to the other device?
• Figure 5: Compared to the LDMA there seem to be some spikes in the reconstructed total number in the Spider DMA which do not appear in the LDMA? On the contrary, the contour plot clearly shows more scatter for the inverted LDMA size-distribution. Is this caused by the lower counting statistics (see previous comment) or by a different inversion algorithm (requiring less smoothing)?
• Figure 6: Similar to the above comment, the spider size-distribution looks extremely “smooth” here. What is causing this?
• 11, l.185: Are the GMD and total number the only key characteristics of a size-distribution? I would be a bit more defensive with that statement.
• 11, l.185: Related to the fact that lower resolution is perfectly suitable in reconstructing aerosol formation and growth rates there is a recent paper in Atmos. Chem. Phys. Discuss. by Ozon et al. (https://doi.org/10.5194/acp-2021-99), showing that a wider resolution can indeed help to reconstruct aerosol dynamics process rates. Even if it is at preprint stage, it could be mentioned here as it is making a similar case that size-distribution reconstruction in ambient or chamber experiments does not necessarily require high resolution.

The authors provide a concise and well reasoned analysis of a relatively low resolution spider DMA system as a means to balance classifier size and robustness of data capture. The analysis of the transfer functions were performed by steady state laminar CFD analysis, which is appropriate for this system (done in separate study). The modeled results are in line with findings from other studies and give a quantified potential for weight and resolution tradeoffs. As a whole the article is worth of publication and should be accepted once addressing the following comments.

My largest disappointment in the article is that it did not engage more broadly with the larger question of what are the theoretical and practical tradeoffs that allow classifiers (in this case DMAs) to be optimized for specific applications. This work gives one data point as to a classifier that works well for the desired application, but does not give the reader insights into whether the system could be further optimized by tradeoffs. A particularly useful inclusion would be general scaling laws that might guide others in the field who are seeking to develop custom classifiers for small (either volume or mass limited) applications.

Minor points for the article are as follows:

Line 55: Despite the work being based on a previous publication (Amanatidis, 2020) it would be useful to have a schematic in Sec 2.2 to depict basic model parameters and methodology, such as labeled boundary conditions and a few rudimentary results.

Line 68: It was unclear why the intervals for injection were chosen and what impact it had on the simulation results.

Line 107: Why was Wiedensohler approximation chosen for an x-ray charger when the ion properties of soft x-ray have been shown to be different than Wiedensohler’s results which are calibrated to radioactive neutralizers?

Line 116: It would be nice to have the maximum number ration and “narrower” transfer functions discussed in quantified terms.

Fig. 3b: I like the visual representation of the scans in Fig 3b, but a quantitative figure would also be useful depicting higth and width of the transfer function.

Line 130: how are the smooth lines fit to the data? What are the smoothing parameters?

Fig. 6: The smaller modes of the distributions appear to be less distinct in the right plots making them indistinguishable. I would appreciate the authors view of these findings.

Typesetting – the authors provide correct spacing between the number and engineering unit, e.g. L/min, in most cases, but fail to provide spaces between numbers and seconds or volts in several places.