Changes for page The FLASH HDF5 structure
Last modified by sndueste on 2025/02/06 10:55
Summary
-
Page properties (1 modified, 0 added, 0 removed)
Details
- Page properties
-
- Content
-
... ... @@ -18,56 +18,8 @@ 18 18 19 19 [[Contents>>doc:||anchor="Contents"]] 20 20 21 -== The new (starting2021)HDF5format ==21 +== The current FLASH HDF5 structure == 22 22 23 -\\ 24 - 25 -FLASH provides a conversion of its data acquisition (DAQ) to the commonly used [[HDF5>>url:https://www.hdfgroup.org/solutions/hdf5/||shape="rect"]] format. Correlated data are mapped by a primary index called **train ID**. Every data set has an individual index of train IDs to identify the data even when data are missing or deviate in update rate.If the data set of choice contains gaps, users have to decide how to treat missing values. DAQ channels are spread across various files with one file name pattern for each DAQ. This means users have to assemble data from different files if necessary. 26 - 27 -The hierarchy is realized using a human readable named HDF tree with each DAQ channel containing the data sets "//value//" and "//index//". Additionally, the HDF group //zraw// contains a tree of the original DAQ channel names. 28 - 29 -Reference implementation, which follows the concept of Python libraries like Pandas, Xarray, or Dask, is given below. 30 - 31 -\\ 32 - 33 -{{expand title="Discontinued HDF formats"}} 34 -== Comparison to FLASH's deprecated HDF formats == 35 - 36 -Before 2021, FLASH provided two different HDF formats formally known as //near-online// and //offline// HDF files. 37 - 38 -All data sets were aligned by the same global index by filling in missing data. Fast DAQ channels (pulse synchronous data - update 10 Hz) are filled in by NaN or 0, slow channels (not pulse synchronous data - update e.g. 1 Hz ) are interpolated by the DAQs libraries, by keeping the value constant up to the next change. 39 - 40 -While the near-online HDF files were converted live during the beamtime, the offline HDF files were manually compiled weeks/months later. Near-online HDF files were used by a provided API [[BeamtimeDaqAccess>>url:https://confluence.desy.de/display/FLASHUSER/Near-Online+data+analysis#Near-Onlinedataanalysis-BriefUserGuidetoBeamtimeDAQAccess||shape="rect"]], which also searches for DAQ channels in files. With assembled files "by run", all DAQ channels were existing in the same HDF file. While shorter runs usually fitted into one file, longer runs still had to be aggregated over several files. The creation of the assembled type of files still involves the use of fairly unstable DAQ Mex-functions As the environment required for using the Mex-functions is already deprecated, its continued existence is uncertain. 41 - 42 -The HDF hierarchy is almost identical in all formats. While in the the recent format each DAQ channel contains the data sets "value" and "index", the deprecated format had one data set for each DAQ channel. The deprecated formats supplies no time axis parameters for spectra data types. 43 -{{/expand}} 44 - 45 -== HDF excerpt: == 46 - 47 -\\ 48 - 49 -ADC data as example for **fast** **data** (10 Hz): 50 - 51 -\\ 52 - 53 -[[image:attach:image2020-11-16_15-26-28.png||height="250"]] 54 - 55 -\\ 56 - 57 -{{info title="Sample scripts in python"}} 58 -== Reference implementation (Python) == 59 - 60 -[[~[~[image:attach:binder_badge.png~|~|thumbnail="true" width="120"~]~]>>url:https://mybinder.org/v2/git/https%3A%2F%2Fgitlab.desy.de%2Fchristopher.passow%2Fflash-daq-hdf/master||shape="rect"]] 61 - 62 -(% class="Object" %)[[https:~~/~~/gitlab.desy.de/christopher.passow/flash-daq-hdf>>url:https://gitlab.desy.de/christopher.passow/flash-daq-hdf||shape="rect"]] 63 -{{/info}} 64 - 65 -\\ 66 - 67 -\\ 68 - 69 -== The discontinued (till 2021) FLASH HDF5 structure == 70 - 71 71 The photon diagnostic, electron diagnostic and beamline information as well as the information about the pump-probe laser and the infrastructure offered for users (GHz/MHz ADCs) can be included in one HDF5 file which is organized according to train IDs. The general structure is: 72 72 73 73 * Electron Diagnostic ... ... @@ -102,6 +102,14 @@ 102 102 103 103 \\ 104 104 57 +== The new (starting 2021) HDF5 format == 58 + 59 +Here is [[some documentation on the changes of the HDF5 format>>doc:FLASHUSER.HDF5 format]] that well be available in 2021 (work in progress) 60 + 61 +\\ 62 + 63 +\\ 64 + 105 105 == Most popular FLASH parameters and their names in HDF5, DOOCS and (raw) DAQ == 106 106 107 107 {{id name="DOOCSparameters"/}} ... ... @@ -198,7 +198,7 @@ 198 198 (% style="color: rgb(0,0,0);" %)**Discontinued GMD data recording / evaluation (VME + PhotonFlux ML server)** 199 199 200 200 {{code language="none"}}/FL1/Photon Diagnostic/GMD/Average energy/energy tunnel{{/code}} 201 -//always saved (PBD)// 161 +//always saved (PBD)//\\ 202 202 203 203 (% style="color: rgb(0,0,0);" %)DOOCS prop : {{code language="none"}}TTF2.DAQ/PHFLUX/OUT04/VAL{{/code}} (%%) 204 204 (% style="color: rgb(0,0,0);" %)DAQ channel: {{code language="none"}}PBD.PHFLUX/TUNNEL.ENPULSEIC{{/code}}(%%) ... ... @@ -456,7 +456,7 @@ 456 456 ===== electron bunch energy ===== 457 457 458 458 {{code language="none"}}/FL1/Electron Diagnostic/Electron energy/average electron energy{{/code}} 459 - //always saved (PBD)//419 +_always saved (PBD) 460 460 DOOCS prop : {{code language="none"}}TTF2.DAQ/ENERGY.DOGLEG/E_INTRA_MEAN/VAL{{/code}} 461 461 DAQ channel: {{code language="none"}}PBD.ENERGY.DOGLEG/E_MEAN{{/code}} 462 462 desc: electron bunch energy (average over the bunch train) ... ... @@ -464,7 +464,7 @@ 464 464 \\ 465 465 466 466 {{code language="none"}}/FL1/Electron Diagnostic/Electron energy/pulse resolved energy{{/code}} 467 - //always saved (PBD)//427 +_always saved (PBD) 468 468 DOOCS prop : {{code language="none"}}TTF2.DAQ/ENERGY.DOGLEG/E_SPECT/VAL.TD{{/code}} 469 469 DAQ channel: {{code language="none"}}PBD.ENERGY.DOGLEG/E_SPECT{{/code}} 470 470 desc: electron bunch energy bunch resolved ... ... @@ -473,7 +473,7 @@ 473 473 \\ 474 474 475 475 {{code language="none"}}/FL1/Electron Diagnostic/Electron energy/wavelength bunch train average{{/code}} 476 - //always saved (PBD)//436 +_always saved (PBD) 477 477 DOOCS prop : {{code language="none"}}TTF2.DAQ/ENERGY.DOGLEG/LAMBDA_MEAN/VAL{{/code}} 478 478 DAQ channel: {{code language="none"}}PBD.ENERGY.DOGLEG/LAMBDA_MEAN{{/code}} 479 479 desc: Wavelength calculated by the electron bunch energy (average over the bunch train) (FLASH1) ... ... @@ -667,12 +667,12 @@ 667 667 668 668 ==== Beamline info (FLASH2) ==== 669 669 670 -{{code language="none"}}/FL2/Beamlines/Attenuator/pressure 671 -(% style="color: rgb( 0,0,0);" %)//always saved (PBD2)// (%%)672 -(% style="color: rgb( 0,0,0);" %)DOOCS prop : {{code language="none"}}FLASH.FEL/ATT.GAS_DOSING/FL2.HALL/PRESSURE{{/code}} (%%)673 -(% style="color: rgb( 0,0,0);" %)DAQ channel: {{code language="none"}}FLASH.FEL/ATT.GAS_DOSING/FL2.HALL/PRESSURE{{/code}} (%%)674 -(% style="color: rgb( 0,0,0);" %)desc: set pressure in the gas attenuator (%%)675 -(% style="color: rgb( 0,0,0);" %)units: mbar(%%)630 +(% style="color: rgb(255,153,0);" %)(not yet available){{code language="none"}}/FL2/Beamlines/Attenuator/pressure {{/code}}(%%) 631 +(% style="color: rgb(193,199,208);" %)//always saved (PBD2)// (%%) 632 +(% style="color: rgb(193,199,208);" %)DOOCS prop : {{code language="none"}}FLASH.FEL/ATT.GAS_DOSING/FL2.HALL/PRESSURE{{/code}} (%%) 633 +(% style="color: rgb(193,199,208);" %)DAQ channel: {{code language="none"}}FLASH.FEL/ATT.GAS_DOSING/FL2.HALL/PRESSURE{{/code}} (%%) 634 +(% style="color: rgb(193,199,208);" %)desc: set pressure in the gas attenuator (%%) 635 +(% style="color: rgb(193,199,208);" %)units: mbar(%%) 676 676 \\ 677 677 678 678 {{code language="none"}}/FL2/Beamlines/FL20/Shutter/open{{/code}} ... ... @@ -691,13 +691,11 @@ 691 691 units: degree 692 692 \\ 693 693 694 -/FL2/Beamlines/Filter wheel/position wheel 2 695 -always saved (PBD2) 696 -DOOCS prop : {{code language="none"}}FLASH.FEL/FL20H.PH.MOTOR/MOTOR2.MOT3/FPOS{{/code}} 697 -DAQ channel: {{code language="none"}}FLASH.FEL/FL20H.PH.MOTOR/MOTOR2.MOT3/FPOS{{/code}} 698 -desc: Position of the BL filter wheel 2 - to correlate with the filter material please look [[here>>doc:FLASHUSER.jddd-linked help pages.Filter-Units.Filter wheels in FLASH1 and FLASH2.WebHome]] 699 -units: degree 700 700 655 +//always saved (PBD2)// 656 +\\\\**[[here>>doc:FLASHUSER.jddd-linked help pages.Filter-Units.Filter wheels in FLASH1 and FLASH2.WebHome]]** 657 +{{code language="none"}}/FL2/Beamlines/Filter wheel/position wheel 2 DOOCS prop : FLASH.FEL/FL20H.PH.MOTOR/MOTOR2.MOT3/FPOS DAQ channel: FLASH.FEL/FL20H.PH.MOTOR/MOTOR2.MOT3/FPOS desc: Position of the BL filter wheel 2 - to correlate with the filter material please look units: degree{{/code}} 658 + 701 701 \\ 702 702 703 703 NOTE: Aperture positions in the beamline as well as the positions of the beam steering mirrors are also saved. for more Info contact your local contact ... ... @@ -776,15 +776,13 @@ 776 776 777 777 \\ 778 778 779 -/FL2/Photon Diagnostic/GMD/Pulse resolved beam position/position tunnel x 780 -always saved (PBD2) 781 -DOOCS prop : FLASH.FEL/XGM.BPM/FL2.HALL/X.TD 782 -DAQ channel: FLASH.FEL/XGM.BPM/FL2.HALL:2 783 -desc: Besides the well calibrated averaged beam position information there is also the option to measure the beam position on a single bunch level. HOWEVER this methide needs a perfectly adjusted signal level (talk to your local contact !!) and also then the signal to noise is rather small and one needs some averaging ... BUT with this option one can determine if there was a spatial slope on a burst (say forst bunches were lower than the last ones or so ...) 784 -units : mm 785 785 786 -(x=horizontal, y = vertial) 738 +//always saved (PBD2)// 739 +\\\\{{code language="none"}}/FL2/Photon Diagnostic/GMD/Pulse resolved beam position/position tunnel x DOOCS prop : FLASH.FEL/XGM.BPM/FL2.HALL/X.TDDAQ channel: FLASH.FEL/XGM.BPM/FL2.HALL:2 desc: {{/code}}Besides the well calibrated averaged beam position information there is also the option 740 +{{code language="none"}}to measure the beam position on a single bunch level. HOWEVER this methide needs a perfectly adjusted signal level (talk to your local contact !!) and also then the signal to noise is rather small and one needs some averaging ... BUT with this option one can determine if there was a spatial slope on a burst (say forst bunches were lower than the last ones or so ...)units : mm {{/code}} 787 787 742 +{{code language="none"}}(x=horizontal, y = vertial){{/code}} 743 + 788 788 again the same parameter set is available for the **HALL GMD** 789 789 790 790 \\ ... ... @@ -1185,6 +1185,61 @@ 1185 1185 1186 1186 \\ 1187 1187 1144 +=== Samples how to read HDF5 with Matlab === 1145 + 1146 +The examples apply to HDF files with HDF tree version before vers. 0.3.0. 1147 + 1148 +(% style="color: rgb(0,128,0);" %)%% read in the needed data p=path; path(p,'D:\mess-daten\DAQ-data\hdf5\Gotthard') % add the actual folder to the path hdf5file='FLASH1_EXP-2016-03-16T1420.h5'; % data with Gotthard and VLS 1149 + 1150 +(% style="color: rgb(0,128,0);" %)% h5disp(hdf5file,'/','min') % to get an idea what is in the file % h5info(hdf5file,'/Experiment/Gotthard1/BL.0') % get info about the individual channel 1151 + 1152 +(% style="color: rgb(0,128,0);" %)% read a Number per 10 Hz pulse train: 1153 + 1154 +(% class="code" %) 1155 +((( 1156 +FEL_Wavelength_energy_server=h5read(hdf5file,'/Photon Diagnostic/Wavelength/Calculated by energy/wavelength' ); 1157 +))) 1158 + 1159 +(% style="color: rgb(0,128,0);" %)% read in a 1D array (spectrum) %GMD data GMD_Spectrum=h5read(hdf5file,'/Photon Diagnostic/GMD/Pulse resolved energy/energy BDA'); 1160 + 1161 +(% style="color: rgb(0,128,0);" %)% This reads in all the data . One can also limit the amount of data read to a subset of the data stored in the file. e.g. for image Data (Gotthard) 1162 + 1163 +\\ 1164 + 1165 +(% class="code" %) 1166 +((( 1167 +Start_event= 500 ; % define the first 10 Hz event 1168 + Number_of_events = 20; % how many 10 Hz events to load 1169 +))) 1170 + 1171 +(% style="color: rgb(0,128,0);" %)% read only part of the data: 1172 + 1173 +(% class="code" %) 1174 +((( 1175 +Gotthard_data=h5read(hdf5file,'/Experiment/Gotthard1/BL.0',[2 650 Start_event],[50 85 Number_of_events]); 1176 +))) 1177 + 1178 +(% style="color: rgb(0,128,0);" %)% start stop increment "manual" for 2 D data : [start bunch number in Gotthard data start pixelin spectrum start sample in 10 Hz trains], % [number of bunches in the Gotthard data number of points in the spectrum number of spectra ] 1179 + 1180 +(% class="code" %) 1181 +((( 1182 +VLS_Spectrum=h5read(hdf5file,'/Photon Diagnostic/Wavelength/VLS online spectrometer/PCO.ROI.X',[400 Start_event],[200 Number_of_events] ); 1183 + 1184 +))) 1185 + 1186 +(% style="color: rgb(0,128,0);" %)% start stop increment "manual" for 1D data : [start sample in the spectrum start sample in 10 Hz trains],[number of points in the spectrum number of spectra ] 1187 + 1188 +(% style="color: rgb(0,128,0);" %)%GMD data 1189 + 1190 +(% class="code" %) 1191 +((( 1192 +GMD_Spectrum=h5read(hdf5file,'/Photon Diagnostic/GMD/Pulse resolved energy/energy BDA',[1 Start_event],[40 Number_of_events] ); 1193 +))) 1194 + 1195 +[[Contents>>doc:||anchor="Contents"]] 1196 + 1197 +\\ 1198 + 1188 1188 == HDF5 and DOOCS == 1189 1189 1190 1190 Here is an outdated [[list with the available properties that are always saved (PBD) for FLASH1 as>>attach:FLASH1__DaqChannel2HdfNamePbd.xlsx]] HDF5 names and the corresponding DOOCS names