ILL - Instrument control and electronics

Main page          Previous   ---   Top 9/14   ---   Next

The Electronics Group
Head engineers : (1969) Anton Axmann (initially Julich, subsequently HMI, Berlin), (1975) Winfried Drexel (ILL physicist), (1985) Reinhard Klesse (ILL engineer),  (1995) Alain Barthélemy (engineer), (2001) Frédéric Descamps (engineer), (2013) Paolo Mutti (ILL physicist)
Assistant engineer to 1975: Michael Taeschner
The rest of the staff is listed later in this page.

Note that, from 1993 the Electronics group was part of the "Branche INSTRUMENTATION" which had 4 services :
  Electronics : R. Klesse,
  Mecanics : M.Thomas,
  Computing: instruments, network Infrastructure : A.Barthélemy
  Computing: systems, and communications : M. LeSourne

In ????, the "Service du Contrôle d'Instruments" SCI headed by A. Barthélemy regrouped electronics and instrument control.

Approximate computer and component costs are available here.

Engineers: Walther Kaiser (3-axis), Alain Barthelemy (Diffraction)
Programmers: Mathurin Le Sourne, Michel Roure, Philippe Blanchard, John Allibon et Chantal Turfat
Technicians: Gérard Pastor, Jean-Jacques Delacroix, Armand Guellec

The Télémécanique T2000 process-control computer used for the step-scanning instruments, CARINE, was novel in that the multiple instruments were served by swapping separate processes in and out of memory from disk on demand, typically at the end of each counting step. There were two operational computers, with a third for backup and program development. Each system could run up to 6 instruments with an additional background process. The computers had a 19 bit word length, matching a twelve-bit positioning component and a seven bit operational code.

Each process could have a maximum of 32k words, and were swapped in and out from a magnetic drum.

The user’s terminal on the instrument was an ASR33 Teletype (printing at 10 characters per second); it was possible to set up a sequence of commands for execution. An additional button labelled APPEL requested the computer, interrupting the current activity.

To offer maximum flexibility a semi-compiled FORTRAN had been developed (Mathurin Le Sourne) and this was heavily used, especially on the 3-axis instruments, to set up scans in reciprocal space and energy. Typical counting times at each position was in minutes, hence the activity demands were limited, though the alignment times involved very short counts.

For diffraction, especially the four circle instruments, the counting times were very short and a huge fraction of measurement time was lost due to the slow interaction with the control system. This was a major factor in the advantages of updating the computers by more modern systems.

The counting electronics were based on single detectors, monitors and scalers, with time and monitor scalers available to control scans. The motors were controlled through the standard ILL control box using Precilec coders multiplexed to a single decoding electronics.

In addition to running the instruments one additional background process could swap and share the cpu; this offered development facilities using Fortran (and some graphics applications too were available e.g. for plotting D1B powder data). Once the PDP10 was available there was little development of such applications given the priority taken by the instruments.

Engineers: George Messoumian and Michel Grevaz (D11, IN4, IN5 and D7), Michael Taeschner and Helga Schwab (PN1)
Programmers: Rolf Hildebrandt (D7), Klaus Wotschack (PN1)
Technicians: Günter Anderlohr, Jean-Jacques Tschofen

For other instruments multichannel data acquisition was required:

- Time-of-flight spectrometers: IN4, IN5 (inelastic, quasi-elastic scattering)
- Diffuse scattering diffractometer: D7
- Small angle instrument: D11
- Fission fragment spectrometer: PN1 (Lohengrin)

The principal function was starting, stopping counting and data storage, with limited data visualisation of the histogrammed data. These were performed by a single AEG-Telefunken TR86 computer; the system was called NICOLE. Elsewhere at this time in Europe and America dedicated instrument control computers were being introduced to offer greater reliability and flexibility.
History has it that the system kernel was conceived at the university of Karlsruhe as a theoretical problem of real-time computing. The project was headed by Prof. Heinrich Herbstreith, follow this link to see how he describes it.

Interface electronics
The interface electronics was under the responsibility of Klesse, and Ertel looked after the CAMAC modules, notably the time of flight encoders. These were derived from accelerator physics designs, and consequently had a finite reset time, corresponding to awaiting the next valid accelerator pulse. For the rotating choppers at the ILL where there was never any uncertainty in the next trigger pulse from a rotor at 2000-20000 rpm and there was always the inconvenience of the incomplete recording of spectra with this time-gap. The time generation was set using the computer, or via thumbwheels; the units were half microseconds, (50 microseconds for D11) using octal (though decimal thumbwheels were used).

Computer
The computer was a 24 bit computer with two Burroughs disks (“Boris” and “Rebecca”) each with a fixed vertical platter about 1m diameter, driven with a belt drive, with 800 bpi magnetic tapes, card input and output, and a line printer. A second machine served for development work. Initially, before the Central Computer was installed, data tapes were sent to Saclay, being returned with a copy on cards. My first experience of this service unfortunately coincided with the tape being shredded at Saclay. Data processing at the ILL was then provided through an RJE link (the “Terminal Lourd”, printer and card reader) to the (IMAG) University’s IBM 360-VM computer. Later the tapes were taken off each morning and read on the PDP10 directly. Again unfortunate data losses occurred in the early days when the previous days tape was replaced again on the Nicole tape drive and overwritten.

Data handling
Each instrument was connected through several video-grade cables. These carried the event data (as parallel descriptors up to 16 bits) which were sent through a FIFO buffer on the instrument with a handshake on the computer. For D11, with count rates from the area detector which could exceed 100,000 events per second, data were accumulated in memory (4096 words). For the other instruments the mean data rate was much lower, and the data were updated periodically rewriting the incremented data back onto disk. When computer problems arose these data were safe, but the current data from the small angle interment were lost on restart. Fortunately these latter experiments were typically short measurements (less than 30 minutes), and the experimenters manned the instrument continuously, and hence were present to restart the computer. While D11 had a maximum count limit of 2²⁴-1, it was only discovered in the last months of operation that the limit for the data stored on disk for the time of flight instruments was 2²⁰-1. The less demanding time-of-flight experiments never attained this limit until an experiment to measure the scattering from highly absorbing liquid ³He in 1975. To observe the small signal counting was continuous for over two weeks. The beam monitor data exceeded this count limit several times and posed major problems in normalising the data subsequently.

The User Interface
The premiss for the design seems to have been built on a shared multichannel analyser. As a consequence the user was presented with a display (HP) CRT monitor, a keyboard, and a console with rotary switches to control x,y and z ranges, and a set of control buttons, together with a cursor intensity display (on most units in binary, later on one or two units in decimal). The buttons selected x,y preferences, and an isometric display (useful for the 2D detector on D11). The TR86 monitored the box and responded with the appropriate display. This required the counting electronics, the console box, ad its switches, and the monitor to be directly connected to the computer. Several 3cm diameter multi-core cables connected each instrument to the controller. These were later used (without permission) as a general support for later cable installations around the reactor and guide hall. These high grade cables terminated with LEMO connectors with around 100 pins. Many years later a number of these connectors in pristine condition were still in the spare parts stores of the electronics group.
The screen was divided into three regions: a graphical or text display region, a keyboard input line, and a line for error messages at the bottom.
In graphics mode the data from a scratch buffer was displayed in various forms and the scale could be change using rotary knobs on the console which were sensed by the processor. In text mode the text could be edited. This was used in the data save routines; the first twelve or so values were read from electronics (beam monitor, time etc.) and the remaining lines were comments. This was used from 1974 onwards to annotate the data with information not directly readable from the instrument, but which was of great use in later treatment of the count information. A cursor (haeltespunkte) could be set with thumbwheels, and the intensity content displayed (in binary).
Although CARINE had a silvered paper hardcopy in the machine room which was used often for D1B, few users of NICOLE made hardcopies of the screen there. Data tapes were changed each morning and copied to the Central Computer (a DEC-PDP10).

Keyboards
The keyboards had the German layout. This hence added variety to the American keyboards on the ASR33 teletypes used on CARINE, and the IBM029 card punches with French keyboards.

Programs
There were a number of small programs available for the user. Obviously the basic start/stop commands could be sent; the preset time or beam monitor values were set by thumbwheels on the CAMAC or sent as commands.
A standard program, PAPE, could transfer data from disk to memory buffer for display, or to tape. (Command STA,PAPE,BILD or STA,PAPE,BAND), or bring the text parameter block into memory for on-screen edition of supplementary information. About 64k words were available to buffer data for plotting in several histogram modes controlled by key switches.

The principal control for all instruments was the preset time or beam monitor limited count. D7 had a number of motors too which could be set, while PN1 could control magnetic and electric fields used in the fission fragment separator. On D11 even at this stage a number of timed rapid stop-flow experiments were satisfactorily performed.
Program development was performed on a second machine which was also used as a reserve of spare parts.

Reliability
The overall reliability of the NICOLE system was low, and successful operations depended to a large extent on D11 experimenters, who were present at the instrument night and day, to restart the system. The current D11 data (in memory) was lost, but the other machines conserved their data on the disks. The real-time system made it inherently difficult to reproduce problems. In one incident, in the twilight days of operation an error message appeared for the first time, leading to great celebrations in the machine room.
While D11 used 24 bit integer storage, during one of the last experiments on IN5 before the system was abandoned, an experiment ran for over a week (measuring the minute scattering from ³He). Normalising the data subsequently showed that the incrementation only treated 21 bits before “overflowing”.

Replacement
The NICOLE replacement was a project of the "Deuxième Souffle", with its replacement using individual PDP11-34 computers on each instrument being completed in 1978. As instruments were successively transferred the overall reliability inevitably improved.

Tschofen was trained in supporting the PDP11/40 harware. He demonstrated his aptitude when the computer on D11, as always under extreme pressure from the heavy scientific program imposed on it, failed one morning. With Metrix voltmeter in hand he programmed test sequences which he keyed in to the console, then following the logic sequence on the cpu board, identifying a misbehaving gate chip. After soldering in a replacement the measurements could continue before lunch. (Some years later he identified a fault on the PDP11/55 concentrator backplane wiring which had always eluded the DEC technicians)

For software, Roure was sent for training on the RSX11D system, and wrote the low-level handler for the CAMAC crate controllers. This included full multi-user protection to forbid interactions between the crates used by D11A (SANS, Ibel) and D11B (Diffuse scattering detectors, Kostorz). Later, under RSX11M, there was considerable simplification and these distinctions were rendered unnecessary.

Main page          Previous   ---   Top 9/14   ---   Next