Comments on PRF-18-001, Andreas Meyer



* abstract
  The first sentence emphasises the methodology. Propose to rather emphasise the results presented.
  For shortening, some of the details, e.g. about the CMS calo geometry, could be removed from the abstract.

* paragraphs in line 16 and/or 26:
  Propose to introduce Particle Flow already here, as it is a crucial technique in CMS and 
  it sets the stage for the role of the calorimeters, esp. Hcal,
  In the current version, PF is mentioned for the first time on p5. 

* introduction (around lines 33/34):
  Suggest to rephrase here avoiding the term "complications", which sounds bad, and address the 
  "complications" in detail later. 

* introduction (around lines 33/34):
  We would propose to explain and quantify the non-linearity 
  such that the reader understands the source and the size of it.

* introduction (around lines 33/34):
  Why is large shower size w.r.t granularity a problem? Would small shower size w.r.t granularity
  not be much worse ? Is it _longitudinal depth in interaction length we are referring to here ?

* line 50: That HF is "only sensitive to the EM portion of the shower" requires some explanation.

* section 3: Is this section specific to collision events ? If not, it may be useful here, to specify possible
  differences in the reconstruction between collision events, cosmics, test beam data and possible 
  other types of data.

* line 159: what is the (typical) zero-suppression threshold ?

* line 177: is this out-of-time PU or also in-time PU ?

* line 185 and following: it may be useful to explain that PF is only really relevant in the rapidity 
  covered by the tracker where - for  collision events - the main functionality of the Hcal is 
  the measurement of neutral hadrons, neutrons and KL.

  In the context of particle-type-dependent detector response, it may be useful to explain whether 
  any attempt is made (in the Ecal or Hcal) to distinguish between EM and Had energy deposits and to scale them 
  accordingly (software compensation). 

* section 5. For the relative calibration in phi, are there any backgrounds from the beams
  which would lead to a phi-dependency ?

* lines 258 and following: we assume that Elow and Ehigh refer to the energy in a single channel.

* figure 2 caption: suggest to replace "an" by "one" for clarity that these are single channels.

* figure 3: would be interesting to understand better what the dataset is that is shown here,
  and what "before the corrections" exactly means. Is this after basic calibration using test beam information ?

* section 5.2: 
  Off-hand a noise of 200-300 MeV per channel seems high and the signal height seems small. 
  Maybe a remark is in order here. As such it is surprising that a calibration can be derived
  reliably and reproducably if the signal only makes percent level deviations to the noise distribution.
 
* In figures 4-6, it is not quite clear what the scale factor exactly is.
  What is the starting point before calibration ?

* section 5.3:
  Combination of the above results appears possible only if the aforementioned non-linearity (how big is it?, see above) 
  is the same across the channels, at least to a good approximation. Does the agreement within 5% 
  come from such discrepancies in linearity, or other effects leading to limited precision of (one of) the methods ?

* line 378 and following: This explanation would be good to have right after line 351.

* One wonders whether one could not also get a calibration using particles _with_ sizable 
  energy deposits in the ECAL. How is the ECAL veto energy threshold chosen ?

* figure 8 right: one would expect that possible systematic effects due to PU are eta-symmetric.
  This does not seem to be the case here.

* Figure 11: A word about correcting the resolution of the MC to the data ?
  If resolution corrections are possible, can/should we produce a figure in which data and 
  resolution-corrected MC are overlayed ?

* line 527: "suboptimal" seems not a good term. Suggest to replace by short explanation what the difficulty is.

* Summary, line 577: "variety of collision data" seems in contradiction to the fact that also cosmics, test beam and 
  no-beam data are used.

* Summary, line 590: the removal of non-linearities is never really explained 
  (except around line 45, as coming from simulation (?)).



==== type A ====

* line 158: selections -> selects