2011 Spin QA: Transverse Running

This page summarizes my fill by fill spin QA for the transverse running period.

For an explanation of the methods and scripts I use to do the QA, see my blog page describing the 2009 200GeV QA here. Note: I have changed the loopMainAll.tcl to reject runs shorter than 3 min and exclude some different test runs.

Run 11 consisted of two proton-proton configurations; transverse collisions followed by longitudinal collisions. This page deals only with the transverse period. I take the beginning of the transverse period to be Tuesday, February 1st 2011 and the end of the transverse period to be Friday, April 8th (run 12098045).

The original runlist created from the loopMainAll.tcl script which begins with fill 15003 (2011-02-01 run 12032004) and ends with fill 15419 (2011-04-08 run 12098036) can be found here.

The list of fills that the scripts determined had consistent spin patterns can be found here.

The list of fills that the scripts determined had inconsistent spin patterns can be found here.

Table 1: This table shows the 6 fills that were flaged as bad along with the runs in the fill, the run type, and the completions status.

Below is a brief description of the problems with each of the above fills and how I address them.

F15003: My initial analysis showed this fill starting with run 12032004 on 2/1/11 (00:47:15 BNL) and ending with run 12033084 on 2/2/11 (12:36:57 BNL). However, the cdev record has run 12032004 in fill F15029 and run 12033084 in fill F15033. In addition, a look at the RHIC Scalars over this period shows a number of apparent fills. Looking over the run log browser for this period, it appears that the 'RHIC Summary' section is frozen, showing the same values for over 2 days. The fill info I use in my runlists is grabbed from the 'beamInfo' table in the STAR database, I have looked at a number of entries from this table over the period of interest and all the entries are the same. It appears that some link in the chain which fills this table broke for a period of time. Given these observation I feel confident that the runs listed in the table above as belonging to fill F15003 in fact belong to a number of different fills (none of them F15003). Because this occured so early in the run and because all the affected runs appear to have been for FMS calibration, I have decided not to try to assign the runs to the actual fills and will just exclude this period from the spin database.

F15102: This fill contains only one run: 12037092 which started on 2/6/2011 (20:19:12) and ended on 20:54:12. The run was a pp500_production_2011 marked as bad. The fill lasted only ~40 min. I plan on just excluding this period from the spin database.

F15263: This fill contains only one run: 12062039 which started on 3/3/2011 (15:08:54 BNL) and ended on 15:55:29. The run was a pp500_production_2011 marked as bad. I plan on excluding this period from the spin database.

F15312: My analysis indicates that this fill contains 3 runs (12079003 - 12079005) which meet my criteria. Two runs were successful production runs. The fill lasted ~1h 20min. This fill was marked bad due to a discrepency between the nominal start of the fill and cdev timestamps. I define the start of the fill as 20min before the start of the first run, in this case run 12079003 starts at 08:01:40 BNL so the nominal start of the fill is 07:41:40. The cdev database is then queried at this time stamp and if the fill number returned is not what is expected, the fill is flagged. Because the cdev database is filled every 5min, there is also a 310 second limit on the allowed difference between the fill start and the corresponding cdev entry. Looking at the cdev record during this period, it appears that there is a gap between 07:14:21 and 07:54:55. This means that when we query the database at 07:41:40 we get the record from 07:54:55, which is 795s from the nominal start of the fill. This discrepency is greater than the 310s allowed, so the fill was marked as bad. Appart from this discrepency, everything else about the fill seems fine. The spin patterns spanning F15312 all have the same spin pattern. I believe that this fill is fine as is and that the indicated spin pattern is valid for the times listed in the summary files.

F15324: My analysis indicates that this fill contains 20 runs (12081049 - 12082014) which meet my criteria. The nominal start of the fill was on 3/22/2011 (19:12:02 BNL) and the nominal end was on 3/23/2011 (04:26:48 BNL). This fill was marked bad due to a spin pattern discrepency at the end of the fill. Similar to the way I define the beginning of a fill, the end of a fill is defined as the end of the last run plus 10 min. If the cdev information from this time stamp does not match the information from the beginning of the fill, the fill is marked bad. If we look at the cdev records from the beginning or middle of the fill we see a spin pattern of P3 (B+-+--+-+ Y--++--++). The last run of this fill ended at 4:16:48 BNL, if cdev is queried at 4:26:48 (the nominal end of fill), a spin pattern of P4 is found. Note that the cdev fill number is still 15324 for this querey. We can use the RICH Scalars to see when the fill was dumped. This plot shows that the fill was dumped around 4:11:??. It appears that run 12082014 was running during the beam dump. This run should not be included in the spin db. The second to last run is 12082012 which ends at 4:02:54 BNL, which is before the beam dump as indicated by the RICH Scalars. If I add 10 min to the end of the run and query cdev, I get an entry at 4:17:00 which is after the beam dump and which has a spin pattern of P4. If I query cdev at the actual time of the end of the run (4:02:54) I get an entry from 4:06:48. This is before the beam dump and is consistent with the cdev entries from the beginning and middle of the fill. So the spin information for F15324 is valid up to a time of 4:06:48 BNL time or 1300867608 unix time.

F15384: My analysis indicates that this fill contains 9 runs (12092036 - 12092046) which meet my criteria. The nominal start of the fill was on 4/02/2011 (20:10:35 BNL) and the nominal end was on 4/02/2011 (23:45:23 BNL). The run log browser says that run 12092032 (start 13:35:19 BNL) is the first run in fill F15384 and the cdev record from this time stamp has the same fill number. However, the next run, 12092033 starts at 19:56:14 BNL, a 6+ hour gap. The cdev record from this time stamp indicates that run 12092033 is in fill F15386. In addition, according to cdev, all 9 runs which my analysis indicated where in fill F15384 are actually in F15386. We can look at the RICH Scalars to get an idea of the fill structure during this period. This plot shows the RICH Scalars for the period of interest, we see a small amount of activity around the time of run 12092032 and then a long gap until around 4/02/2011 20:00:00 BNL. This corresponds roughly to run 12092033 which is the first run in F15386 according to cdev. The original run list has fill F15386 ending with run 12093014 which ends on 4/03/2011 (04:30:32 BNL) which matches well with the end of the fill as shown in the RICH Scalar plot. In addition, the cdev record shows that fill F15386 ends between 04:54:56 and 04:59:51 BNL. Finally, the difference between the end of run 12092046 (run log fill F15384) and run 12092047 (run log fill F15386) is only 4 min, not enough time to start a new fill. As with fill F15003, it appears that the RHIC Summary section of the run log browser is 'frozen' between runs 12092033 and 12092046, the information displayed here is the same for the runs indicated. I dumped the beamInfo table of the database for these runs and the db entry time for all of them is set at 04/04/2011 17:48:42. Starting with run 12092047, the RHIC Summary seems to come back to life and the db entry time matches within a couple of min to the run begin time as expected. Given the above observations, I believe that runs 12092036 - 12092046 actually belong in fill F15386 and fill F15384 does not include any runs which meet my criteria for inclusion in the spin QA.

Using the information about the 6 bad fills I have outlined above, I have made the following changes to the runlist:

• F15003 Removed
• F15102 Removed
• F15263 Removed
• Run 12082014 from F15324 Removed
• Runs 12092036 - 12092046 from F15384 have been moved to F15386 and F15384 has been removed

The final runlist can be seen here. A file which shows the differences between the original runlist and the modified runlist can be seen here.

Using the final runlist, I have redone the analysis and generated new lists of good and bad fills. The new bad list still contains two fills (F15312 and F15324) but this is expected. F15312 is fine as is and can be moved to the good list after the error codes are deleted. F15324 can be included in the good list after the error codes are removed and the end fill time stamp is changed from 1300867974 to 1300867608.

The final list of good fills which includes F15312 and F15324 modified as explained above can be seen here.