This paper presents a thorough and quantitative comparison of two commercial instruments for measuring droplet number-size distributions, based on shadowgraph and phase Doppler interferometry. The study is particularly useful because the two techniques have not been commonly used in cloud measurements, so this helps define the strengths and limitations of the methods. The paper is well written and the presentation is generally clear. I recommend that the paper is accepted, subject to minor revisions. The following suggestions should help improve the work: 

Figures 3 and 4 and the associated text leave some confusion about the orientation of the instruments relative to the wind direction. On line 225 I understand that the instruments are facing in a direction that allow optimal sampling of wind coming from 320 degrees. This could be indicated in the last panel of Figure 4, or at least mentioned in the caption. If that understanding is correct, what does it imply about the measurements on 13-Jul, 14-Jul, and 13-Aug when the wind direction is close to 90 degrees? Is it expected that the measurements are heaviliy influenced by the instrument housings? That may cause problems with the sample volume estimation for the phase Doppler interferometer. This is relevant to the presented results, e.g., see line 263 where average wind direction is stated as 225 degrees. Please address this aspect of the study, and clarify the directions as needed. 

In the presentation of results on droplet number concentration (e.g., Figure 9) please provide some information on the range of number of droplets sampled for the data points that are shown. That would allow Poisson uncertainty to be evaluated. I expect the numbers are large enough that Poisson sampling uncertainty is not a significant contribution, but please show or discuss this directly. 

In several of the plots I believe more insight could be obtained by showing results on logarithmic axes. For example, log-log coordinates might help in the interpretation of the droplet number concentration results. But for sure it would be helpful in the presentation and interpretation of the LWC results in Figure 13. This would allow a broader range of LWC to be observed, rather than results being compressed near the origin because of a relatively small number of outlier points. I expect the least-square fits will be significantly different in log-log coordinates as well. 

Another comment on data presentation: in Figure 10 it would make sense to show the y-axis in log scale so that the full variability can be observed. The polynomial fit seems rather arbitrary. Is there any justification for this form? Perhaps log-scale will suggest something else. 

Another comment related to Figure 10: At large mean diameter there was a strong undercounting of the shadowgraph method or overcounting of the PDI method. If you use the polynomial fit to 'adjust' for this bias, how much improvement would result in the LWC results in Figure 13? This would be one way to further disentangle the number concentration and D^3 contributions. It seems possible that the PDI instrument is overestimating number concentration when the mean size is large, and that this is leading to the unreasonably high values of LWC. 

The approach of the study is an instrument intercomparison, but without a known reference or standard. Generally, the authors point out differences without suggesting that any specific technique is correct or incorrect, and I agree with that cautious approach. (Note that one exception is on line 311, which should be reworded, or evidence shown that one method is indeed overestimating size.) In some places, however, it may be possible to argue that one instrument is more closer to reality, such as in Figure 13 where the PDI is showing significant numbers of data points with liquid water contents above 1 g/m^3. Based on other measurements at the UFS and similar mountain stations, I expect these values are much too high. More importantly, the paper is missing a discussion in the Conclusions section about how the investigation could be taken to the next level. Is there a way to perform an absolute calibration or provide additional instruments in future studies, so that the discrepancies can be understood and resolved? In the end, the community needs to know which of these instruments can be considered reliable, in what range of conditions. 

The manuscript “Cloud microphysical measurements at a mountain observatory: comparison between shadowgraph imaging and phase Doppler interferometry” shows a study comparing two methods to size cloud droplets and explores the settings and methods of how these instruments should be used to yield matching results between the two methods. The authors used the VisiSize D30 as well as a Phase Doppler Interferometer during a mountaintop field campaign. These in-situ methods outside the laboratory can help to improve cloud measurements, as the reliability of various measurement techniques are uncertain and a thorough examination and evaluation of techniques and how to improve them is important for interpretation of collected data as well as the conducting of future measurements. In general, the manuscript is well written, and I recommend publication with minor revisions.

In the introduction of the coincidence filter, it would be good to explain why this effect happens. If I understand correctly, you assume that there was only one droplet, but it got somehow recorded twice by the instrument and you only keep the average size and velocity. Could actual coincidence also be a contributing factor, if two droplets (or fragments) pass the sample volume? Since other instruments have issues with recording two drops at a time, it would be worth noting.

For Figure 5, a size distribution of dNdlogd would be more useful. This way, it would be easy to see at which sizes the two instruments deviate. Since they seem to match in panel d, a dNdlogd size distribution would also show agreement there and show a deviation for smaller sizes. This would also show a comparison in the number concentration. Right now, it is not easy to see from panel a if the PDFs should be shifted or if one instrument detects less particles than the other; that only becomes obvious when considering panel d. This way, you could also include both methods for the PDI to show if those two have the same size distribution but shifted up/down (due to the different sample volume) or if they are skewed. It would help to also understand table 3. Probably the size distributions in Figure 5 only match for very large cutoff because D_min should not be a sharp cutoff of the PDI, but it might miss smaller droplets gradually?

In Figure 10 especially in panel A the fit does not seem to be very good for large sizes. Why did you choose polynomial, not a different shape, and how many terms did you keep for the polynomial fit?

For general presentation I would suggest increasing the size of some figures for better readability and decrease the amount of white space between them, such as in figure 13. It looks very cramped, but there is plenty of white space to be used between the panels.

Equation 2: tau_Tot was not introduced.

Line 121-122: grammar (missing an “and” or something similar)

Line 316: typo: “because”