SNS/HFIR Software Developments

1
SNS/HFIR Software Developments

• Steve Miller
• Analysis Software

August 15, 2006
2
Organizational Structure October 1, 2006
3
SNS Recent Milestones Achieved

• SNS Facility CD-4 First beam on Target April
  28
• Backscattering Spectrometer first data May 19
• Reflectometers 4A and 4B first data July 19

4
SNS Facility CD-4 Measurements
Accelerator/Target Controls
BL-7 Data
5
BSS Data Reduction Diffraction Bank (Mica)


Instrument Geometry Step 1 pathlength Step 2
angle
1. TOF to Wavelength
2. Wavelength to d-spacing
Courtesy of Michael Reuter
6
BSS Inelastic Data
4-methyl pyridine N-oxide 5 kWatt, 3 hour, ¼
current detector/analyzer, T 3 K
Tunneling Peaks Visible!
Courtesy of Ken Herwig and Eugene Mamontov
Fit using DAVE Software
7
Reflectometer 4B Incident Beam Wavelength Spectrum
Courtesy of John Ankner
8
First measurement on BL4A July 21, 2006
Sample Ni 50 Å / Ti 50 Å multilayer SNS
source power 250 W Data collection time 2.5
h Preparation time needed 8 years TOF spectrum
integrated over whole detector Position
spectrum integrated over all TOF channels Same
sample measured at IPNS
Bragg Peak
Neutrons reflected from sample at 1 deg. grazing
incidence
Courtesy of Frank Klose
9
Magnetic Reflectometer Reflectivity

• 50 A Ni / 50 A Ti 20 Multilayer
• 2 kw low power

10
Instrument Commissioning Challenges

• Low accelerator repetition rates
• Design point 60Hz, current rates between 1 to 10
  Hz
• Results in long time intervals between pulses on
  target
• Keeping all neutrons with fast (25 uSec) sample
  rate gives a large number of TOF channels 5000
  to 10000 typical
• Reflectometry histogram files on the order of
  1.5GB each
• Currently only able to perform linear
  histogramming
• Low beam power
• Measurements contend with background
• Must acquire for longer periods of time
• Must contend with occasional accelerator glitches

11
Creating, Processing, and Storing Data

• Event Histogramming
• Detector to Pixel mapping
• Instrument Geometry
• Metadata extraction
• Create NeXus file
• Catalog and Store
• Reduce Data
• All subsystems functional to some degree

12
Data Reduction Architecture Overview

• Three levels of reduction at SNS
• Level 1 Driver is the overall mechanism that
  runs the data reduction process. It based on the
  requirements that are given by the instrument
  scientists.
• Level 2 HLR is the representation of functions.
  It unifies calls to retrieve data and to call low
  level functions. Levels 1 and 2 python.
• Level 3 (mostly C)
• DOM provides abstract layer for data
  manipulation.
• SCL is a toolbox of reusable primitive functions
  necessary for data reduction process.

Driver
HLR (High Level Reduction)
DOM (Data Object Model) SCL (SNS Common Lib.)
13
Visualizing Data via the Portal
MCA Data
ISAW Plot
NeXus tags
NeXus Files
metadata
14
Metadata Search via ICAT
Search String
Optional Search Fields
15
Networking Diagram
16
Instrument Commissioning Lessons Learned

• Detector and Instrument Definitions
• Non-trivial to determine
• May need an iterative approach to derive
  calibration factors
• Using NeXus
• Adds overhead for examining raw data
• Question about how best to associate updated
  calibration data with existing NeXus files. Note
  that data reduction can take an optional
  calibration file along with NeXus data.
• Create NeXus files during acquisition or
  on-demand?
• Event Data
• Had to adapt to provide histogramming tools
  sooner than originally anticipated
• Detector Mapping
• Needed to create and validate detector to pixel
  mapping files

17
Problems and Issues Encountered

• Python memory management issue
• Deleted ints and floats still remain in heap
  memory, thus memory footprint can grow very
  large. 6GB not uncommon for us and larger data
  to come.
• Causing us to rethink how best to utilize python
  for memory intensive applications
• Memory de-allocation fix in python 2.5 does not
  address this problem
• Does DANSE team have experience with this?
• Need for tools to explore pre-NeXus data
• Responded by creating prototype IDL tools
• More tools needed, such as for producing rocking
  curves
• Need more tools for working with event data

18
Pre-NeXus Data Examination Tools

• Instrument Scientists need tools to explore
  data.
• Currently producing prototype IDL tools to
  enable exploring and extracting data.
• Considering producing tools derived from this
  which will integrate with the portal.
• Using IDL VM to distribute applications

Courtesy of Jean Bilheux
19
Instrument Commissioning Pattern Emerging

• Analysis Software group provides
• the data plumbing
• Data reduction
• Visualization tools
• Utilizing existing analysis software tools (DAVE,
  ReflPak, ISAW, GSAS, etc.)
• The DANSE team can dramatically help SNS/HFIR
• Identify prototype instrument commissioning
  software tools
• Identify and produce Instrument Day-1 analysis
  software
• Keep an eye to the future developing portal-based
  advanced analysis software
• Urge DANSE Science team leaders to continue to
  work closely with ORNL instrument teams and the
  Analysis SW group!

20
Instrument Schedule
21
Instrument Commissioning Schedule by year

• Ongoing
• Backscattering Spectrometer
• Liquids Reflectometer
• Magnetic Reflectometer
• 2007
• ARCS
• Powder Diffraction
• CNCS
• 2008
• SANS
• SNAP
• Sequoia
• 2009
• Vulcan Engineering Diffraction
• Spin Echo
• TOPAZ SCD

• Take Home Message
• We should anticipate the desire to bring
  instruments on-line sooner in order to broaden
  the scope of the user program.
• Urge DANSE team members to continue to work
  closely with corresponding SNS and HFIR personnel
  (and vice versa).
• With higher neutron flux in later years, the
  push will be for first science, not just first
  data.
• Acting now can enable world class software to be
  available with first neutrons on instruments
  our desire and a worthy goal.