Tracker DCS Data Analysis Using ROOT

Introduction

Tracker DCS(Detector Control System) data includes :

CAEN Power Supply: T, V, I, Status ...

DCU: Silicon Detector Temperature ...

PLC: Humidity Value ...

All these data are archived to Oracle database. And because the large data volume, PVSS only archives when the value changes (5% deadband in default).

What we (University of Kansas) are doing is to retrieve tracker DCS data from the Oracle database and analyzing it in ROOT.

Basic Features

The tool is to import DCS data from the Oracle database and save it to root files. It then plot time history plots of voltage, current, temperature(Silicon Detector/Power Supply), etc. And our ultimate goal is to integrate the DCS data analysis software to DQM(Data Quality Monitoring) software so that user could study the corelation between DCS and DQM data. The version for CMSSW is now available at DCSAnalyzer.

Root File Retrieving

DB->Root dump is done bye CMSpage1, http://cmsdaq.cern.ch/cmsmon

To get MTCC DCS data,click DatabaseBrowser or go directly to http://cmsdaq.cern.ch/cmsmon/cmsdb/servlet/DatabaseBrowser From left menu bar, select CMS_TRACKER_DCS(MTCC account), and click TABLE, select table you'd like to see and submit a query, you can now save the qeury result as a root file or a text file, which can be used for furthur analysis. For PSU info, related tables are # FWCAENBOARDEASY # FWCAENCHANNEL # TKDCUFE # TKDCUFEMASTE # TK_PLC_DEWPOINT # TK_PLC_RELAY # TK_PLC_S

To get TIF DCS data, click cms_pvss_tk. Same method as above to get a root file.

One thing we have to bear in mind is the CMSpage1 save NULL as 0 into the rootfile. Current way to solve this is, everytime only queries only DPID(never NULL), Timestamp(never NULL), one vaiable interested (ACTUAL_VMON for example), and make sure to check the not-NULL box and put 1 in Order text box besides CHANGE_TIME. Then save the queried data as FWCAENCHANNEL_ACTUAL_VMON.root (note the file name standard: Tablename_VariableName.root). Here's a screenshot of submitting the query.

Front End

The tool has a startup frame that allows the user to access various plots. This front-end allow a user to select a specific PSU by input branchcontroller#, crate# and board#, then by defining the starting/ending time and variable(T,V,I) wanted, the front-end plot the vaiable time history (with connection tool applied, i.e., filling values(same as last value archived) in between 2 consecutive values) for all 4 channels(2 for temperatures) of the board.

The preliminary frame looks like so:

How to run it locally.

1. Dump root file using CMSpage1
2. download source code DcsDataAnalysis.cc ,PlotTimeHistory.c.
3. in your local download folder, input command
root -l DcsAnalaysis.cc
the GUI will appear.

Version Track

The version for CMSSW is now available at DCSAnalyzer. * Note files on this page are always latest version.

* TkDCSRootAnalysis Version 0.1.6, Changed to dynamic array size(Thanks Richard for doing this).

* TkDCSRootAnalysis Version 0.1.5, Changed connection alogrithm to adding just one fake value in between(Thanks Richard for doing this).

* TkDCSRootAnalysis Version 0.1.4, Fix a memory leakage problem.

* TkDCSRootAnalysis Version 0.1.3, DBtree name updated. Fix time scales for all 4 plots as user define.

* TkDCSRootAnalysis Version 0.1.2, DpName function updated, due to DCS system changed to a new PC.

* TkDCSRootAnalysis Version 0.1.1, added Branchcontroller selection box.

* TkDCSRootAnalysis Version 0.1.0, only CAEN power supply works. File Menu is not implemented yet.

Progress on MTCC tracker DCS analysis

No PLC, DCU info archived. CAEN info only high voltage info archived. MTCC DCS data now is deemed as incomplete, and since it's not recoverable, we've stopped analyzing it.

Progress on SliceTest (or TIF currently) tracker DCS analysis

* I use V0.1.2, check data between Jan.25 --jan.26. It looks nice now. LV and HV are turned on/off reasonably. The reason might be TkDCS system is changed to another PC in Dec, also a new RDB archiving version was used. Seems the new RDB archiving version gets rid of the error we observed in Nov.

* The preliminary frame shown above is actually based on TIF DCS data. We actually see some LV are turned on and kept on for several days. But it's strange LV1 is not ON when HV is ON. It could be a archiving problem or something else. We're still working on this. (fixed in v0.1.2)

Connection tool in TkDCSRootAanlysis Package

Since the information is only present upon change, the analysis tool fills in values over time periods when there is no change by presenting the previous reading. To analyze upward and downward trends, plots are made connecting these real and made up data points for easier reading. The Figure below demonstrates the output after this connection tool has been employed. Here, the data is plotted from one HV channel over the period of one day. The top plot shows the raw data, and the bottom plot shows the result after the connection tool is used.

The HV is seen turning on and off in this figure. Here, the data is logged over a short time as the HV ramps up and down and archiving is happening in smaller increments of about 10-20V change as seen in the top figure. A time scale (deltaT in GUI) is chosen for filling in the points when there is no change in value which is tunable upon implementation in the bottom figure(this is changed in V0.1.5 by filling just one fake point in between 2 data points).

Relevant Links

DcsLastValue

-- JieChen - 17 Aug 2007