Main comments on chapter 6 :

The improvement of the trigger is one of the main motivations for GE 1/1, but it seems that we are ignoring the phase-1 improvements described in the L1 TDR (esp. improved PT assignment via larger LUTs and tail-clipping), which were said to reduce the rates considerably (although this came with a non-negligible efficiency loss), by a factor of 1.5 to 2 for a L1 threshold of 15-20 GeV :
As this issue has been discussed in detail between the reviewer and the TDR proponents, we will summarize here the main points of that discussion as a way of documenting it. We indeed are using the run-i configuration of the muon trigger in our studies as the simulation of the trigger with the tail clipping technique was unavailable. It is true that the tail-clipping technique allows reducing trigger rates and it efficiently works in the barrel region yielding a factor of several reduction in the trigger rate at the cost of a couple percent inefficiency. Fir the endcap (see Fig. 2.2 of the L1 Upgrade TDR) the reduction in rate requires an additional 12-13% efficiency loss in the entire endcap region relative to the run-1 version of the L1 muon trigger. While this is already a very large loss, the true degradation is actually more severe because the denominator used for estimated in Fig.2.2 for the endcap requires muons with multiple reconstructed stubs. Our studies using generator level muons (of which 99.5% are reconstructable in the offline) were showing that at high luminosity the run-1 L1 muon trigger in the endocarp has efficiency of less than 90%, so even for a perfectly working system, Level-1 muon trigger efficiency would be below 80% in the endcap. With the GE1/1 upgrade, integration of the GEM data into the trigger will allow building a trigger that will have a significantly lower overall rate and efficiency of 96%. Combing the strengths of different techniques can only improve these estimates yielding low rate and high efficiency muon trigger. (A.S.).

In the L1TDR, we wrote (Tab 2.4) that, at 2.2e34, one can keep a single muon trigger with an offline threshold of 22 GeV; it would have a rate of 14 kHz. Here, we say (l 1818 - 1823) that one can not maintain a L1 threshold of 15 GeV (i.e. offline threshold of ~ 20 GeV) and O(15 GeV) is the L1 threshold that we seem to target. -> the trigger motivation related to the threshold should be made more clear in view of the L1TDR (and a link with the L1TDR should be made - currently, the L1TDR is only refered to in the chapter on electronics). Are we saying that we want to lower the L1 threshold of the inclusive muon trigger by ~ 2 GeV compared to what was shown in the L1TDR ? or do we actually target an even lower threshold (see l 193-194) ? and/or are we trying to reduce the rates but without the non-negligible price in efficiency that was shown in the L1TDR ?

If the bandwidth allocation cited for the inclusive muon trigger is not reduced, CMS should be able to reduce thresholds by noticeably more than 2 GeV. There is a number of improvements that will become possible with a better use of redundancy as the new track finders will be able to use data from all detectors. If those improvements are implemented (which should bring at the very least a factor of 2 in the trigger rate based on ongoing studies in the trigger group) and the tail-clipping is deployed in the barrel, the total rate of the Level-1 muon trigger will be heavily dominated by the region of eta>1.6. With GE1/1 deployment that rate can drop by a factor of 6-10. Because of many uncertainties, it is difficult to quantify the final shift in pt, but a simple example calculation could look like this: - with run-1 muon trigger, barrel and endcap carry equal contribution to the full rate. - About 70% of the rate in the endcap comes from eta>1.6 - if rates in the region below eta=1.6 are reduced by a factor of 2 due to redundancy and the rate in the barrel is reduced by a further factor of 3 from tail clipping (from Fig. 2.2), the distribution of contributions will look as follows: barrel: 0.15 x 0.5 of the current total rate, 0.3 x 0.5 x 0.5 of the current rate in the endcap part below 1.6, and 0.7 x 0.5 x 0.15 in eta>1.6. - the sum is 0.075+0.075+0.05=0.20 of the current rate, which is a factor of 5 reduction. The L1 upgrade TDR target reduction was about a factor of 3, so the thresholds of the new trigger can be moved to the left until the total rate increases by 5/3. Looking at Figure 2.1 in the upgrade TDR, that ought to be at least 5 GeV. Given that these estimates are very conservative (I completely dropped tail-clipping in the endcap and assumed no reduction in rate due to being angles in ME1/2 which are usable as the Bdl in there is much larger than for eta>1.6), the actual rate improvement should be larger and the shift would be more than 5 GeV. In addition, the efficiency in the barrel can stay in high 90s, which is at the very least 15% higher than what went into the L1 Upgrade tables. (A.S.).

In any case, the numbers for the single muon rates do not look very consistent: - L1TDR, Tab 2.4 : 14 kHz at 2.2e34, offline threshold of 22 GeV (i.e. L1 threshold of about 17 GeV) - Fig 6.9 left, upper curve : about 8 kHz at 2e34 for a L1 threshold of 17 GeV, in the region 1.6 - 2.2 alone (which, without GE 1/1, accounts for roughly 30% of the rate in the range eta < 2.2 - by the way, I think that one should give this information). Hence 27 kHz in the full range. Since the legend of Fig 6.9 says "phase-1", one expects this curve to be consistent with the numbers given in the L1TDR. I guess that the "phase-1" curve in Fig 6.9 actually does not account for the phase-1 improvements of the PT assignment (since I think they are not yet implemented in the software anyway) ? Some additional explanation should then be added.
You are right, we compare things everywhere to the run-1 trigger configuration, which, as you correctly stated, is the only option available now. (A.S.).

With GE 1-1, the rate in 1.6 - 2.2 will be a very small fraction of the rate in the full range, and the limitation for a low threshold inclusive single muon trigger will come from eta < 1.6. To make the statement on l 2142-2143, one should give the rate in eta < 1.6 for a O(15 GeV) threshold. I think that the phase-1 upgrades should be good enough for 15 GeV, but if one would like to lower further the threshold, what would be the additional handles to reduce further the rates in eta < 1.6 ?
We have intentionally stayed vague as there are many uncertainties about the trigger rate below 1.6. However, the reduction factors are significant and the improvements will come from the use of redundancy (main problem of the current trigger is mismeasurements and the rate is dominated by tracks that are misimeasured because they have only a few points. it should bring a large improvement in the entire eta<1.6. Then, for the endcap, one can use the bending angle measured in station ME1/2 - as the BdL there is larger, the lever arm does not need to be as large as we need in eta>1.6, plus one can also use bending angle in station ME2/2. That should give a factor of at least 2. In the barrel, DT experts should be the source of information, I can only speculate here: one could also use the bending angle in stations 2 and more aggressively use bending angle in both stations 1 and 2 for muons that are partially going through the holes between the wheels which give a large contribution to the rate, which must have a good potential. One other simple consideration is that everything being equal, one would expect that the barrel would yield a noticeably smaller total rate compared to the endcaps where we have lots of problems due to the falling B-field and insufficient thickness of the chambers. (A.S.).

Section 6.2.3 and non-pointing or displaced muons : The efficiencies shown in Fig 6.10 are just the efficiencies to reconstruct a muon object, and not trigger efficiencies as written in the caption. To get the trigger efficiency, the PT of the muon object must be above the trigger threshold. -> did you check the PT resolution for non-pointing / displaced muons ? It may not be as good as for prompt muons. (by the way, we are still investigating why the efficiency for L1TkMuons is so low for d_xy of a just a few mm).
That is correct, no pt is applied. We added a comment to make that explicit. We know that for muons with small dxy and relatively large Lxy pt assignment seems to be working, for modest dxy it also works (a good figure of merit would be how well one can measure the impact parameter by extrapolating a muon from the muon detector). For large deviations, we will need to develop new "roading rules" for the track finders where one would require consistency between the bending angle and the position of a hit in the next station in trajectory building and adjust pt assignment. This work is only starting now, but it should be doable at least for "about straight" muon tracks. (A.S.).

Other comments :

- The phase-2 TP and the GEM TDR should be harmonized regarding the installation of GE1/1. The phase-2 TP says that one considers advancing this installation to LS2 (e.g. section 1.3.2, l 1139 in the version of mid-september), while in the GEM TDR, the baseline plan is to install in LS2.

This is intentional and is dictated by CMS upgrade management, which does not want consideration of "early" installation of a 4 MChF project to derail discussions with funding agencies of the full 275 MChF project (JH).

- suggest to use one eta range consistently, for example the average between the acceptance of the long and short chambers. Currently, many eta ranges are given (1.5-2.4 and 1.55-2.18 in the abstract, 1.6-2.2 in l 161, 1.64-2.14 in Fig 1.2, 1.5-2.2 in l 218, etc)

Yes. The introduction has been made more clear, and in general 1.6-2.2 will be quoted except in parts where the long and short chambers are specifically discussed (JH).

- Fig 1.2 : the caption is not correct. Since the plot is for PU = 50, this must correspond to about 2e34, and not 4e34 as written - cf Fig 6.9 left, here the caption is correct.

KH: Good catch. Fixed.

- l 191: unclear if this refers to phase-2 (with Track Trigger in), or to before LS3. With the Track Trigger, a threshold of 20 GeV on the single electron trigger should be easy to accommodate, with a L1 bandwidth of 750 kHz.

KH, JH: The reference to electron trigger is removed, and the sections have been reworded to make the distinction of pre- and post-LS3 situations more clear.

- l 268: the GEB and the optohybrid have not been introduced yet

KH: Moved GEB definition to the first mentioning (GEB = GEM electronic board). Same for optohybrid (OH). tried to harmonize the spelling of OptoHybrid but only in the introduction.

- l 330 : what does that mean ? is it the data volume reduction that is mentionned in l 1400-1401 and 1408-1409 ? or is it in view of processing in the HLT ? anyway, l 330 sounds strange.

Reworded to make it clearer. (JH).

CHAPTER 2 starts here

- l 395 : chapter 7: no

Good catch. Now use proper latex ref. to Ch.6 here. (MH)

- l 406: maximum hit rate of 5 kHz / cm2 at 5e34. From Fig 6.6, I read that this maximum hit rate is lower, 1.5 kHz / cm2 or so ?

Fig. 6.6 left only describes background hit rates. The placeholder plot on the right also has points for muons, but it is not clear to me if those are from hard primary scatters or from background muons from decays in flight. The quoted 5 kHz rate includes hit rates from both signal muons and background based on scaling up the measured muon rates in run 1. (MH)

- Fig 2.16: add numerical values on the r and z axes

Fixed. (MH)

- Table 4.3 : why are the data rates the same for both formats, for the tracking data ? do you mean that, for the tracking lossless, the zero-suppression is made in the optohybrid, instead of being made in the VFAT for the SPZS scheme ? but how is this rate of 0.17 Gb/s for the lossless data (for L1A of 100 kHz) consistent with l 1396 ? the tracking output data coming out from the optohybrid are not the ones that arrive on the MP7 ??

- l 1400 : you could give more explanation on how this data reduction in the MP7 is made, and what is the data reduction factor.

- l 1827: "by an order of magnitude" : that's right at high thresholds, but for the thresholds of interest, of about 15 GeV, the reduction is less, O( 6 - 7), cf Fig 6.9.
You are correct and we will adjust the statement, but the reduction factor can be increased by pushing the bending angle cuts that are set incredibly lose now. Even better to combine the muon track finder tools, including tail clipping, with the bending angle to maximize the discrimination. (A.S.).

- l 1837: one does not need an inclusive muon trigger for these two processes, especially for the second one with two photons.
You are right, two photons have showed up there by mistake. For H->hh->tau tau bb the inclusive muon trigger can be helpful, but the upgrade will also allow reduce thresholds for the mu+jet trigger, yielding higher acceptance and better efficiency (A.S.).

- Fig 6.6 right: did you say how you determine the prompt background ?
RH: The prompt particle rates are evaluated using sample with minimum bias events simulated with CMSSW. The integrated number of hits in a given eta partition is normalized to the sensitive area of the partition and the full simulated time. The time of flight of the particles has been counted with respect to the primary interaction and the time cut-of has been set to 500 ns. However it is important to note that about 70% of the particles cross the GE1/1 detection planes within the first 50 ns.

- l 2064-2065 : the 1st part of the sentence is clumsy
We re-phrased it. (JH).

- Fig 6.11 : what is the PT threshold used here ?
This is plateau efficiency, but the trigger configurations were optimized to keep about the trigger rate within eta=1.6-2.2 of the two options acceptable and of the same order. The pt threshold we were looking at was ~12 GeV. (A.S.). - l 2305 : should be 450 microrad
corrected (Anna).

a few misprints: l 1259, 2049 (comparies), caption Fig 6.10 (funcion)