Continuing my investigation of the mass dependence of raw dijet yields as a function of time. Emphasis of this page is on the effect of dead RDO boards ...

The previous investigation (blog page here) showed that there was a correlation between the change in jet and dijet averages and a drop in the NHitsFit for tracks in sector 18. After some investigation, it was confirmed that these features were related to the loss of RDO board 18-1 (board 1 is closest to the beam pipe). Gene has a blog page in which he lists RDO boards which have died, but I believe I can get better resolution and accuracy by looking for anomolies in the NHitsFit and NHitsPossible spectra. I have two pdfs which show the average number of hit points used in track fitting and the number of possible hit points for each sector of the TPC, I also record some observations below.

PDF showing the average number of hit points per track in each sector.
PDF showing the average number of possible hit points per track in each sector.

General observations:

•  According to Gene's blog, sectors 4, 5, 6, 11, and 20 have RDO boards which were dead for the entire relevant run period. A clear drop in the average NHitsFit values can be seen for these sectors with respect to the other sectors. A less obvious effect can be seen in the average NHitsPossible plots.

•  There appears to be a small dip in the average NHitsFit value for all sectors around a run index of ~600. Not sure what is causing this, maybe luminosity effect?

List of dead sectors:

Sector 2: RDO 3 dead starting on day 161

Sector 4: RDO 1 dead entire run

Sector 5: In addition to RDO 5 being dead the entire time, RDO 6 failed on day 144 and was fixed on day 147

Sector 6: In addition to RDO 5 being dead the entire time, RDO 6 failed on day 142 and was fixed on day 147. Also, RDO 4 failed on day 178.

Sector 8: RDO 6 failed on day ??? andw was recovered on day 155

Sector 11: RDO 1 dead entire run

Sector 12: RDO 3 had DAQ problem on and was fixed on day 158

Sector 16: RDO 6 died on day ??? and was fixed on day 162

Sector 18: RDO 1 died on day 155 (note this was listed as day 158 in Gene's blog). RDO 6 died on day 173

Sector 19: RDO 6 died on day 180

Sector 20: Part of RDO 6 dead entire run

Sector 21: Small section of RDO 6 dead entire run

The figures below show the per run averages summed over all TPC sectors for several quantities such as dijet mass, jet pt, number of tracks/towers in a jet, etc. In many cases, features can be seen which correspond to the failure of RDO board 1 in sector 18 on day 155. The features can be seen in the full jet reconstruction and the TPC only reconstruction, but not in the EMC only reconstruction indicating that the feature is due to the TPC.

In addition to the summed plots below, I have also made pdfs showing the average track and tower pt and the average number of tracks and towers for each TPC sector. Note that the 'by-sector' track and tower average plots are incremented even if there is no track or tower in that sector, so the averages are much lower than for the summed sector plots.

PDF of the average number of tracks per sector
PDF of the average number of towers per sector
PDF of the average track pt per sector
PDF of the average tower pt per sector

Figure 1: This figure shows the average dijet mass as a function of run index. The upper left panel is for full jet reconstruction, the upper right panel is for emc only jet reconstruction, and the bottom left panel is for TPC only jet reconstruction. The fits run from index 0-600 which covers the period before RDO 18-1 died, 600-900 which covers the periods after RDO 18-1 died but before 18-6 died, and 900+ which covers the period after RDO 18-6 died.

Figure 2: Same as figure 1, but now showing the average jet pt.

Figure 3: The average number of tracks in a jet. Top panel shows full jet reconstruction and bottom figure shows TPC only jet reconstruction.

Figure 4: Same as figure 3, but now showing the average number of towers in a jet.

Figure 5: Same as figure 3, but now showing the average track pt in a jet.

Figure 6: Same as figure 3, but now showing the average tower pt in a jet.

Figure 7: Same as figure 3, but now showing the average scalar sum of tracks in a jet.

Figure 8: Same as figure 3, but now showing the average scalar sum of towers in a jet.

It appears there may be two effects going on. The first, and most obvious, is a drop in the average number of tracks in a jet which seems to be driven by the failure of RDO board 18-1 around day 155. Because of the 100% subtraction scheme, this leads to an increase in the average tower ET. The second effect seems to be a slight increase in the average track pt, although this does not seem to be localized to any one sector. These two effects lead to an increase in the average jet pt and therefore to an increase to the average dijet mass which causes the 'bow-tie' effect shown in previous posts.

Looking at the loss of RDO board 18-1 in the embedding database via the NHitsPossible spectrum. I will compare the NHitsPossible spectra in data and embedding, note that the mapping between run index and run number is different for data and embedding.

Figure 9: This is the average NHitsPossible spectrum for sector 6 in data (top) and embedding (bottom) for reference. The feature seen in the data starting around index 420 and ending around 520 (525 is the field change, several RDOs were recovered during this access) corresponds to runs 10143008 - 10146087. In the embedding, this feature can be seen extending from roughly index 300 to index 400. The feature present in the last several runs can also be seen in data and simulation.

Figure 10: Same as figure 9, but now showing sector 18. The slight dip seen in the data between index 420 and 520 is again visible (barely) in the embedding from roughly index 300 to 400. The feature at index 900 in the data corresponds to run 10173051 which translates to index 786 in the embedding. There is no evidence of RDO board 18-1 dying on day 155 in the data (should be at index 600) or in the embedding (should be at index 483).

Figure 11: The average NHitsFit in the embedding sample for sectors 6 (top panel) and 18 (bottom panel). The effect of RDO board 18-1 dying can clearly be seen in the bottom panel around index 480.

We see in figure 11 that the average number of hits used to fit a track decreases in the simulation at the place where board 18-1 dies on day 155. This is confusing as the only way the simulation would know that board 18-1 had died would be if it was in the database, however, we see that the number of hits possible does not change in the simulation (same behaviour as the data) indicating that from the NHits possible standpoint, the simulation believes that sector 18-1 is fine. I would expect that it the simulation thinks NHitsPossible is fine, that it would also think the number of hits used in the fit would be fine as well. The only thing I can think of is that the number of possible hits is obtained from one database, and the status of the various RDO boards is obtained from some other database and NHitsPossible database didn't get updated but the RDO status database did. Just a wild guess. This is probably a question for Gene and embedding experts.

Regardless of why the simulation shows a drop in NHitsFit but not NHitsPossible, it certainly seems to be the case and the simulation matches the behaviour of the data. Then Renee's question at the jet meeting on 2/25/14 is pertinant: why does the fully unfolded cross section still show a 'bow-tie' effect when I plot ratios of the RFF and FF cross sections against the full cross section?

Figure 12: The average number of tracks per jet in sector 18 for the simulation sample.

Figure 13: The average tower Et in jets summed over all sectors for the simulation sample.

Figure 14: Same as figure 13, but for average track pt in jets.

Figure 15: The average jet pt for the simulation sample.

Figure 16: The average dijet mass for the simulation sample.

From figures 12 - 16, it appears as though the reduced number of hits in sector 18 does not affect the average number of tracks, average track or tower pt, or the average jet pt / dijet mass as it does in the data. It is unclear why, but it seems like the embedding does not simulate this aspect very well. Maybe it isn't too suprising as the changes in average track/tower pt, number of tracks etc. seen in the data are small, on the order of a few percent.

In addition to the questions about the effect of board 18-1 dying in the simulation, Carl wanted to see how the track losses due to the death of board 18-1 were distributed kinematically. To do this, I compared the eta and pt distributions of tracks in sectors 18 and 24 (sector 24 had no problems) for days 130-139 (before board 18-1 died) and days 160-169 (after board 18-1 died).

Figure 17: Comparison of track eta spectra. The top four panels compare sector 18 to sector 24 with the left column showing sectors 18 and 24 for days 130-139 (before board 18-1 died) and the right column showing sectors 18 and 24 for days 160-169 (after board 18-1 died). The bottom four panels compare sector 24 before and after board 18-1 dies (left column) and the same for sector 18 (right column)

Figure 18: Same as figure 17, but now showing the track pt spectra.

Figure 19: Simultaneous comparison of track pt and eta between sectors 18 and 24. The left panel shows the ratio of sector 18 to sector 24 for days 130-139 (before board 18-1 died) and the right panel shows the ratio of sector 18 to sector 24 for days 160-169 (after board 18-1 died).