The STAR trigger for the run 2001 is significantly more complex than the used in the first STAR run in 2000. Other than in 2000 it is not sufficent to simply scan 'min. bias' or 'central' data files to obtain a clean sample. During the commissioning phase various different trigger sets were run and tested. In what follows we discuss the 2 most essential trigger sets: productionCentral and ProductionMinBias. A trigger set (aka trigger group) consist of several triggers. The individual triggers in a set may be prescaled (and even if not, you might need to scale same because of different deadtimes).
Please note that the understanding of the various trigger sets is absolutely necessary for any physics analysis.
In what follows we try to provide background information which should help to understand the issues. This is a "living" document and will change as we gain experience in how to deal with the various triggers. Much of the 'insider' info is covered in the FAQ section below.
Figure 1 shows the central and min. bias trigger definition in the ZDC vs CTB plane. This is only a schematic few and doesn't contain all details but is helpful to show the trigger coverage of the various regions of the "boomerang". For more details you need to study table 1 and 2.
Note that the CTB units changed with the introduction of the central trigger (beginning of October). Originally the units were ADC counts, now the CTB signal is given in MIPS, where 1 MIP = 5 ADC counts.
WARNING: Currently the CTB ADCs in the data stream cannot be interpreted as for the 2000 data. The expert on this is Falk Meissner (FMeissner@lbl.gov). He summarized the problem in an email. We will try to fix the problem at an early stage (DAQ maker) soon.
Below is a detailed list containing the definitions of the trigger words for the minimum bias and central trigger set. The expert here is Jeff Landgraf (email@example.com) who compiled the list below and of course everybody from the trigger group.
|No B+Y||0||x||x||x||x||x||x||No trigger|
|CTB < 75||1||1||0||0||0||x||x||No trigger|
|1||1||0||0||1||1||1||UPC Minbias (No SVT)||(0x1001:4097)|
|75 < CTB < 33000||1||1||1||0||0||x||x||No trigger|
|33000 < CTB||1||1||1||1||0||x||x||hi-mult||(0x1101:4353)|
|B||Y||T||CTB15||CTB2000||CTB6600||ZDCa 85||ZDCCo||Vertex||Name||Trigger Word|
|No B+Y||0||x||x||x||x||x||x||x||x||No trigger|
|CTB < 15 (topology triggers)||1||1||1||0||0||0||x||1||x||TOPO & ZDC||(0x3002:12290)|
|1||1||0||0||0||0||x||1||x||TOPO Efficiency||(0x3011:12305) (ps:50)|
|15 < CTB < 2000||1||1||x||x||0||x||x||x||x||No trigger|
|2000 < CTB < 6600||1||1||x||1||1||0||0||0||0||No trigger|
|CTB > 6600||1||1||x||1||1||1||0||0||0||Hi-mult||(0x1101:4353)|
|Level 3:||L3 topology trigger filter operates on TOPO, TOPO & ZDC, and TOPO Efficiency events.|
Here's a list of useful links related to run 2001 triggers:
Q: What vertex cut should
I use to select unbiased events?
A: In the Au run, the ZDC was used to select a restricted vertex in the setups listed above. Unfortunately, the resolution of the ZDC is dependent on the centrality of the event due to slewing and ADC pulse height. Hence one must further cut on the vertex distribution to avoid centrality bias. The worst case 1 sigma cut is 10 cm inside from the online cut (see this page ). For most of the data set (as listed below) the online cut was at 35 cm, hence one must cut at |vertZ| < 25 cm. For the set which used an online cut of 25 cm, one must cut at |vertZ| < 15 cm.
Q: How do I cut in centrality
in centrally triggered data and minbias triggered data?
Please see this page on centrality cuts to find the multiplicity cuts that should be used in year 2001.
The central trigger cut on the ZDC at 10% central (see this page ). One can make a 5% cut on Nch and be unbiased, or one can accept the ZDC cut at 10%. One should not make a 10% Nch cut, since this only accepts ~8% and is tantamount to a double cut in ZDC and Nch. See this mail on hypernews.
Another issue is the UPC events in minbias which were enhanced because
of an online CTB cut to reduce the SVT readout. If you use the
centrality cut (numbers of Nch) provided by Zhangbu and Masashi, there
should not be problem. If you want to calculate the centrality
yourself, you have to take into account the UPC enhancement which
fluctuates without any control.
One should also cut on the trigger word to be sure not to get the hi-mult and UPC triggers: 0x1000 for ProductionMinBias, 0x1100 for productionCentral* for standard Au analyses.
One should also by default cut on l3 unbiased, or accept the event if no l3 is in the datastream.
bool isUnbiased = true; if ( mEvent->l3Trigger() && mEvent->l3Trigger()->l3EventSummary() && !(mEvent->l3Trigger()->l3EventSummary()->unbiasedTrigger())) isUnbiased = false;
Q: How do I select the hadronic central events out of the productionCentral trigger set which is a mix of UPC and hadronic trigger?
A: For each trigger type a specific trigger word is defined, as seen in Table 1 or Table 2. The trigger word is provided via StL0Trigger* StEvent::l0Trigger() and unsigned int StL0Trigger::triggerWord().
Q: Is productionCentral the only trigger set which contains hadronic central events?
A: The "standard" central trigger for 2001 called productionCentral was first implemented and used on Oct 4, 2001. The bulk of central triggers were taken using this set. There were two other central triggers sets defined before which were mainly used for rate studies. One is called hank and contains only 33k hadronic central triggers. Since this was used for various tests it has to be used with great care and you certainly have to study the run logbook if you intend to use it. The other trigger set was simply called central. Its definition is very close to productionCentral but it contains no UPC triggers. Around 100k good central events were recorded with this set. As of Nov 8 we were using the L3 trigger if the ZDC coincidence rate was sufficiently high. For rates larger than 1200 the trigger set used is productionCentral1200, for rates between 600 and 1200 the set used is productionCentral600. For rates below 600 the standard productionCentral set is used.
Q: Where do I find information from the Level3 trigger?
A: Access to Level3 information is provided via StL3Trigger* StEvent::l3trigger(). Most importantly, you can find information about each L3 Algorithm that looked at an event in the StL3EventSummary, available via StL3EventSummary* StL3Trigger::l3EventSummary(). A description and examples how to access this information can be found in the StEvent manual (click here for a copy) and on a dedicated web page.
Q: But I am not interested in special triggers, so why should I look at L3 information?
A: The events written to tape will also contain special events, that will in general introduce a bias. To correct for this, you have to check if L3 just passed the event through or triggered it because a special algorithm fired. The StL3EventSummary provides a function unbiasedTrigger() to do that. Example code can be found in section 3.10 of the StEvent manual (click here for a copy) and on a web page.
Q: There are many different version of min bias trigger sets. Are definitions of the individual triggers in these sets identical?
A: No! One has to be very careful. It is easiest to just stick to ProductionMinBias, unless one really needs the statistics.
Starting from run 2213006 to 2220010, the minbias trigger without vertex cut has trigger word 0x1000, from 2224041 to 2228020 it became 0x1001. From 2255031 to 2307028 it is 0x1002. For two runs 2218037 and 2218038, it is 0x1100.
For triggers with vertex cut the situation is not better. The standard MinBiasVertex trigger set and the ProductionMinBias (the sets with most of the data) have again 0x1000 for the hadronic min bias trigger word. However you might see a MinBiasVertexCTB75noSVT trigger set which has 0x1001. The situation concerning the vertex cut is also something to watch. All ProductionMinBias trigger set (and actually all central triggers) as well as MinBiasVertex sets from run 2251003 to 2308008 and later use a vertex cut of |z| < 35 cm. However, for the MinBiasVertex from run 2226003 to 2246008 the cut was |z| < 25 cm.
So the trigger word does unfortunately not uniquely select a physics set or cut. Even the 'last DSM' info doesn't. One has to be very careful.
Q: I am suspecting the reversed full field 200 GeV data has problems, please look at this plot (P01gl production). There is a bump at the high CTB end.
A: For the MinBiasVertex and ProductionMinBias, the CTB>33K was applied without any ZDC cuts (neither adc nor vertex cut) to accept high multiplicity events. This was done to check for an inefficiency of the ZDC for very central events. You can either (i) apply the vertex cut in your analysis to be the same as the online |Z| < 35cm or (ii) check the trigger word and exclude the Hi-mult trigger (see above).
Q: What exactly is triggerActionWord ? Does this give one information about the detector presence?
A: There is a document on the trigger web called Trigger DAQ Interface and on page 6 of that document is the following definition:
"Trigger Action Word (16 bits): These bits consists of and 8 bit mask detailing which detectors are participating in the event and must be read out by DAQ, a 4bit Trigger Command Word instructing each readout board of actions to be taken for this event, and a 4bit DAQ Command which defines 16 different DAQ actions related to that event.".
Further along in that document (pg 10) is a description of the trigger data that gets recorded. Included in the first part, event descriptor, is that Trigger Action Word.
so for example:
case 1: a trigger action word 0x9001 would be decoded as
9 - (1001) trigger action - special trigger (in this case, laser)
0 - (0000) daq action - daq has never used this space
01- (0000.0001) detector mask - only detector #1 participating (tpc)
case 2: a trigger action word 0x4023 would be decoded as
4 - (0100) trigger action - physics trigger (normal physics event)
0 - (0000) daq action - none
23- (0010.0011) detector mask - detectors 1,2 &6 (tpc, svt, & rich)
The bits are assigned as follows:
If by "Detector Presence in datastream for a given run configuration" you want to know what detectors were included in a given run, not just event by event (as I outlined above) but integrated over all events, then you should talk to Jeff Landgraf. If you want to know what detectors were live while an event was read out but took no action (i.e. in case 1 above, smd (detector 3) could have been live, but upon receiving the trigger action word 0x9001, it would see that the 3rd detector bit was not set and know that it was not required to do anything) then that "trigger information fifo" data is also stored in the trigger data.
Q: How do I recover real pp minbias triggers from the 0x2000 trigger word in fpd* setups?
A: In the table of pp triggers, one may notice that in the fpd* trigger setup names the trigger word 0x2000 is overridden if either the FPD or EMC towers fired, and becomes 0x2003 or 0x2004. In other words, this means that the 0x2000 trigger word in these setup names is not the same as the one in ppMinBias, but has holes cut into it (essentially it is minbias AND not FPD AND not EMC). In addition, scaledowns were applied to the 0x2000 trigger word, so it is not clear how to recover a full minbias dataset within runs of these trigger setup names. Unless one really needs the statistics it is much easier to just stick to ppMinBias for a minbias data set. However, if one needs to gather more statistics, please see this mail from Jan for an attempt at a solution using the scaler boards.