ILL-Archive - a collection of bytes from ILL History

An informal collation by Ron Ghosh, (ILL 1974-2008)

Recently a number of former colleagues from the early scientific days of the ILL have died. Before my own memories are a forgotten past I record here my personal experiences of the ILL from 1973 onwards. In addition I include information on instruments, data, and the exploitation of data over a period exceeding a generation, and where there is little possibility of maintaining such information on the official ILL website. The web files here are all written in basic HTML, some from the first versions in 1995, in order to avoid problems of future compatibility seen with html generated by more sophisticated editing systems. Without wishing to be too presumptuous, following software changes over 40 years leads one to be reluctant to adopt smart tools which have but a fleeting existence of a decade or so, and demand huge manpower investment to update to match the latest trends. These pages hence may not adapt well to the iPhone X+n.

The Institut Laue Langevin

Summary of the role of the ILL
Providing neutron beam instruments on a High Flux reactor for visiting scientists.
The ILL data archive Conservation of measured data.
Computing at the ILL
A brief summary of computers for control and data treatment at the ILL from 1973 to the present.
Software tools used for ILL data treatment
Some tools compatible with old and current data. Data fitting routines used in several packages are described here.
ILL standard formatted data
A review of data formats and the current archive at the ILL which contains data back to 1974.
Treating data from ILL instruments
Information on treatment of data from earlier times.

Le Réacteur Haut Flux (RHF) de l'ILL

Since 1972 the nuclear reactor at the Institut Laue Langevin has been the most powerful source of neutrons in the world for neutron scattering experiments in the fields of nuclear and solid-state physics, physical chemistry, and increasingly life sciences. Around the reactor some 35 instruments are located at the ends of beamtubes leading neutrons from the reactor core. These are then scattered from samples under controlled temperatures, magnetic fields etc, before being detected and counted. From these measurements atomic structure, positions in crystals, surfaces and liquids,and their motions can be studied. A unique feature of neutrons compared to X-rays is the specificity of the nuclear scattering mechanism, unique for each atom, and each isotope, One outstanding example is the very distinctive difference between scattering from hysdrogen, and its isotope deuterium. Deuteration in the neutron studies of life-sciences has been the dominant technique for elucidating structures of viruses and membranes, and the mechanisms of transport within these systems in a quite unique fashion. An additional property of the neutron is that it has a magnetic spin, which can be used uniquely to probe magnetism in crystals.

On 19th January 2017 the Institut celebrated the 50th anniversary of its foundation. The Historian Denis Guthleben gave a presentation giving insight to the events leading up to the creation of the Institut. This complements a book, Des Neutrons pour la Science, written by the first French associate director, Bernard Jacrot, written in 2005, translated 2013.

The web-site of the ILL contains much detail of current operations, but two generations after the start there is now nobody who maintains detailed knowledge of the early operations and experiments, apart from the detailed science in the many thousand publications derived from work at the ILL. These pages contain notes on the early data from carefully prepared experiments. These were measured sometimes on well tended instruments yielding exceptional quality data. A few software tools which span some 40 years, but are still in use outside the ILL are described in the second part of this informal archive.

The ILL data Archive

The reactor has been delivering data since the start of 1973. These have been archived, initially on magnetic tapes, which in turn have been transcribed for conservation at regular intervals, copied to other media, including CDROMs. As a consequence of the change to Unix from 1992 the existing VMS binary archive was transformed to ASCII files and compressed. For some instruments (eg SANS these data are compatible and may be treated with programs still available. Only data post 1995 have been catalogued. IDA, an interactive catalogue tool can be used to search for data taken after the restart in 1995. However most data are still, for the most part, readable on current computers. A summary of early measurements has been constructed from the available archives. This include some major breakthrough experiments from the novel instruments created initially at the ILL.

The aim of these notes is to include a description of these raw data. Initially the description was limited to a few internal reports, still available at the ILL. In 1995 the web server www.ill.fr was established and from the start it contained information on data formats which had evolved to use an ASCII text layout following the change to unix. Some of these descriptions are retained in these web pages. .

Summary of Computing at the ILL

There are few establishments which can still offer access to results of experiments over a fifty year time span. There are inevitable shortcomings, which may be partly explained, if not excused by the inevitable march of technology described below.

The first control and data recording computers at the ILL were chosen by a committee of two, one German, and one Frenchman. For multichannel experiments with large arrays of detectors, IN4, IN5, D7, PN1 (or, in the case of D11 a two dimensional counter) a 24-bit Telefunken TR86 computer (NICOLE), running a real-time operating system devised in Aachen was installed (1972-1978). For the eight step-scanning instruments two Telemecanique T2000 19-bit (12-bit address, 7bit opcode) process control computers (CARINE-1, CARINE-2) were installed; these could be programed in an interpreted Fortran, leading to quite novel control scans and orientation calculations. The drawback was the program swapping of the multiple processes back and forth to drum storage. In both cases data were transferred for treatment each morning on magnetic tape.

Mössbauer, the second director of the ILL, had seen the nuclear physics community benefit from interactive computing facilities. This lead to the introduction, in 1973, of a DEC-PDP10 central computer (which could even have been used for real-time control!). With some 12 terminals and even simple graphics terminals (DEC-GT40) it was heavily used from the start for data treatment. Hardcopy plots were only available from a Calcomp plotter, which was poorly adapted to displaying the time of flight and SANS data. The original system KI-10 was replaced with a faster PDP10 KL-1091S in 1981. When the projected updated DEC-10 Jupiter system was abandoned the ILL invested in a VAX-8600 from DEC in 1986, to share load of the data treatment activities. The PDP10 was pensioned off at the end of 1987 when a VAX-8700 processor was added to the VAX-cluster. These were replaced by DEC-alpha 64 bit processors during the 1990s. In 1992 a managerial decision was taken during the long shutdown to convert to unix systems; initially SGI and HP-UX systems were acquired. After 2000 Linux was adopted as a standard both for instrument control and central services.

Introduction of a few new control computers, PDP11s from DEC, had started from about 1975. The poor reliability and slowness of the NICOLE and CARINE systems led to these being replaced with one computer per instrument from 1978 onwards. Again there was pressure to use the French SEMS-Solar 16/40 computers, which had quite competitive hardware specification, but lacked software, compared to the PDP11s. Seven were acquired, but only 5 were put into service.
Networked data transfer started with D17 and D11 using serial line technology in 1977. The overheads on the central computer were quite high and the instruments data transfers were later multiplexed. The PDP11s were hence used on the majority of instruments until 1985 when 32-bit VAX11/730 computers were at a price to enable them to replace PDP11/40s from 1977. The software developed for these systems was re-used succesively as the cheaper microVAX in turn replaced other PDP11s, and became the standard control computer. From 1987 the installation of ethernet throughout the ILL was started With the development of VME electronics and unix/linux, networked electronics progressively replaced bus-based systems after 1995.

A more detailed review of computing and electronics, is available here

Software tools for ILL data treatment

These notes describe a number of tools, suites, and libraries which can be used to reduce and analyse ILL data. Calculation codes are in Fortran-77 or Fortran90, but tk/tcl scripts are also used to provide GUI access expected by the present generation. The open-source gnu compilers (available on all current systems) have been used for building cross-platform programs, and pre-built libraries and working programs are described. Most data access routines general plotting routines are in rlib, complemented by newer routines for accessing NeXus (HDF5) files in conjunction with the basic HDF5 library. Source codes also are available on request from the author reghosh (at) gmail.com .
  • Prop - program opener for Windows
    • manages sets of programs, printing, and plotting - bridging gap between Windows and unix.

  • PGPLOT graphics with ILL extensions
    • Tim Pearson's library has been adapted at the ILL to offer a common usage on Windows, Unix, and Macintosh systems. Typically ouput is either to a single graphics frame, or written to file (where a new filename is generated systematically, hence easily matched up with data output by treatment programs.)

  • Fitfun data fitting

    • a packaged library including a very powerful non-linear fitting routine, which calculates derivatives numerically; it is easily adapted to a wide variety of problems, especially useful for developing sets of programs with common features. The user provides a data read routine and a function calculation routine. Treating sequences of data is possible. Example programs including fitting SANS data, diffraction data, etc especially treating sets of data measured as a function of temperature etc.

  • Fitfun_m fitting multiple data sets

    • a development of the fitfun package which performs simultaneous fitting to several data sets, with shared or free parameters.

  • Clickfit, clickfit_m GUI for fitfun

    • Provides a GUI for the fitfun programs.

  • SANS suite for ILL data
    • This is a suite of programs for treating ILL SANS data (and importing from other sources). The final pattern for interpretation is derived from a number of sample, background and calibration measurements. It is important to perform the reduction systematically. The intermediate data can all be inspected, and a "last step" history is maintained throughout. Mistreatment of components can be easily identified. The final reduced data includes important metadata.

    These, and many other programs, can be on most of the current computing platforms. Further information can be obtained by emailing Ron Ghosh, reghosh(at)gmail.com Source code too can be provided, and also advice on obtaining working copies of Fortran and C compilers. The GNU compiler (first gcc/g77 then gcc/gfortran have been used on many systems with few source code changes.