The instrument described in the paper is based on a good idea and it was built with care. The paper presents a thorough description in clear language appropriate for an AMT publication.

The main novelty of this instrument is to separate the droplet production microfluidic device from the testing section where the cooling of the sample and the observation of freezing events takes place. The advantage derived is a better control of the sample temperature, minimizing internal temperature gradients that are the limiting factor to accuracy in some other droplet freezing devices.

On the production section, the choice of materials is crucial. This is well described in the paper but would find it helpful to clarify two things: Why is a surfactant (line 216) needed for a water in oil suspension? How are air bubble introduced (line 221) and why? In the end, are the water droplets in contact with the tubing and air, or also with some oil? How particle-free is the air? Is the surfactant likely to be covering the droplets in the test section?

The precision of droplet sizes is indicated in Tables 1 and 2 in terms of the estimated variation in droplet diameters. The ±5 μm amounts to about 6.5%. This translates into a volume variation of about 20% which is not negligible in the evaluation of the results. This is a greater limitation to the overall performance of the instrument than is acknowledged in the paper. The authors' comment on this would he helpful.

The small droplet size and the immersion of the tubing in a liquid are the main features regarding temperature accuracy. However, mention is made of a stack of glass slips (line 163) being placed below the tube. How does this limit the flow of the cooling liquid around the tube and to what extent does it introduce further temperature gradients. Could this be clarified?

The spatial uniformity of temperature is demonstrated in Figs. B1 and B2. This display in terms of x and y coordinates` is somewhat unclear. Do both the x and y coordinates of all droplets in a sample are included? Probably yes. Also, is the x and y coordinate system given with respect to the internal dimension of the test chamber? A simple change to using the distance from the walls would be easier to understand.

The results and comparisons to other works are presented as the fraction frozen versus temperature. This is a straightforward manner of showing the results. However, it is specific to the volumes of the sample in the experiment. For even slightly polydisperse populations of drops the function looses generality and makes the calculation of the nucleation rate J for homogeneous freezing contain an error. It also influences the comparison of the three runs with microcline, as, according to Table 2, the drop volumes were about 20% larger for run 1 than for runs 2 and 3. The volume-dependence makes the FF(T) functions inadequate for comparisons with other experiments. It is not clear if any adjustments were made in Fig. 5 to overcome the problem.

In any case, this problem with the volume-dependence is not critical for this AMT paper. It would be more important for a science paper. To fully account for the volume variations in the samples is not a trivial matter. For the comparisons with literature results an appropriate caveat regarding the constant-volume assumption is probably sufficient. A more thorough step to bring results of different experiment on a comparable basis is conversion of the FF data into spectra (eq. 4 in Vali, G.: Revisiting the differential freezing nucleus spectra derived from drop-freezing experiments: methods of calculation, applications, and confidence limits, Atmos. Meas. Tech., 12, 1219-1231, doi: 10.5194/amt-12-1219-2019, 2019.).

Regarding the image analysis process. As described it is a demanding process. Is there some future improvement possible so that the apparatus would really become as user-friendly as it aims to be (line 134)?

This paper describes the development of an improved microfluidic device that has lower temperature gradients and less water and gas permeability. The paper is suitable for AMT and should be publishable after some revisions.

This paper does need a thorough editing for content and especially length. The paper should be considerably shortened before final publication. There are also grammatical errors and statements that are not quantified. Some of these are called out specifically below but I encourage the authors to edit and shorten the paper before resubmission.

Abstract:

1. The Abstract is too long and reads like an introduction to a paper. Please remove extraneous details and cut the text could by 50% which can be done without loss of important content. Please concentrate on the instrument being presented.
2. “requirements: (i) high accuracy and precision in measuring droplet temperatures within 0.2 K (ii) ability to reach the homogeneous freezing point of pure water, with a median freezing temperature of 237.3±0.1 K…” These appear to be capabilities, not requirements? Also, here and throughout the paper : the uncertainties here seem contradictory. Is the latter the uncertainty or spread in the freezing temperature? The numerous uncertainties, specifically in temperature, used throughout the paper need to be explicitly stated for clarity.
3. The Title and Abstract state MINCZ is going to provide homogeneous freezing data but the above quote and paper then says this is restricted to only pure water – something that doesn’t exist in the atmosphere. This means a qualification – “…a platform for homogenous water and …” needs to be added.
4. “to detect mediocre and poor ice-nucleating particles” please define what mediocre and poor means (in temperature and saturation)? Please note that throughout the text words that are qualitative are often used when quantitation is necessary.

Materials and Methods

1. Many parts of Figure 1 seem extraneous and the figure overall confusing. Inclusions of things such as syringes, computer, cooling bath complicate the figure without adding any detail not in the text. Parts c and d are the most important and could be combined with a very simple composite of a and b to improve the figure. Please consider what really needs to be in the figure and can’t be in the text and eliminate for clarity.
2. Can the authors explain in the text why a ~75 μm droplet size was chosen? Is this a fabrication limit? While it is understood microfluidic devices can’t attain the small size of most atmospheric droplets it is important to detail why this size was chose and what implication this volume has in relation to the atmosphere.
3. What information is detailed in Figure 2 that is not given in the text? It appears this figure simply repeats what the text says in a flow chart format.
4. The caption of Figure 3 is a method description, which belongs in the text, not a figure caption.

Results

1. See also abstract. Can the authors explain the temperature uncertainties in the paper, specifically this section? For example, 2 significant figures in “237.41 ± 0.04” is in excess of the earlier statements on the equipment capabilities of .2. Figure 4 then appears to show a yet larger range in data which seems to be the most important uncertainty. A comprehensive explanation of the uncertainties and data ranges would greatly improve this paper.
2. The captions in Figures 4 and 5 are too long and include detail that is not a description of the figure. This text needs to be move to the main text.

Conclusions

1. Consider not using “homogeneous” to describe the droplet size here as it is then used for the freezing mechanism; this is confusing.