Resurrection Home Previous issue Next issue View Original Cover

Computer

RESURRECTION

The Journal of the Computer Conservation Society

ISSN 0958-7403

Number 101

Spring 2023

Contents

Society Activity  
News Round-Up  
Queries and Notes  
Delilah Voice Secrecy System John Harper
The Development of the MOSAIC Computer Edward Smith
Simulation Modelling of Historical Computers[2] Roland Ibbett & David Dolman
The Birth of the Barcode Derrick Brown
50 Years Ago .... From the Pages of Computer Weekly Brian Aldous
Forthcoming Events  
Committee of the Society 
Aims and Objectives 

Top Next

Society Activity


EDSACAndrew Herbert

The focus remains commissioning the Arithmetic Unit and its interface to Main Control and the Order Decoding system.

Some progress has been made with the Transfer Unit (which decouples the machine’s input and output side busses).

There are ongoing issues with main store addressing, although this area is becoming more stable.

The initial orders subsystem (including the engineer’s switch panel) has been reworked and hopefully will be recommissioned in the coming weeks.

The current hold up is a recent failure of the “Signal Sequence Injector” Arduino-based box for injecting programs into the main store. This is possibly linked to the addressing and Transfer Unit issues noted previously. Hopefully this will be resolved shortly.

The TNMoC Colossus engineering team has started an investigation towards building a photo-electric paper tape reader for EDSAC (as per the original in 1951). Early experiments with period photodiodes and high speed mechanical drives based on cannibalised uniselector mechanisms look promising.

Ferranti Argus/Bloodhound Missile SystemPeter Harry

The pandemic put us in limbo for over a year as we are on a military site and access to our equipment was prevented. Secondly, once back in action the Argus 700 has been running reliably and has simply been left while we get on with other restoration work.

We have to leave our base at RAF Cosford by the end of January and our launch control post (with the Argus 700) and radar will be moved to the RAF Air Defence Radar Museum at Neatishead in Norfolk. The pressure has been on us for the last six months to complete the restoration work on both these items, not an insignificant task. We do run the Argus on a weekly basis to ensure it remains serviceable.

What next? The simulator (Argus 700) is now shut down but will be recommissioned in March ready for the RAFADRM museum’s opening; the museum is closed to visitors for the winter period. Plans are in place for the future maintenance of the Argus 700 system with local support at Neatishead. The team here will be active in providing backup and holding spares.

Once the dust has settled on our move I’ll have time again to focus on the Argus 700 and provide updates. I’ll be back online with the CCS in April.

We remain committed to maintaining the Argus 700 in a demonstrable form i.e. running the Bloodhound engagement simulator. We have a YouTube video of the Bloodhound simulator running, all managed by the Argus 700. It can be viewed at www.youtube.com/watch?v=5Xl1By8vh3k.

ICL 2966 Delwyn Holroyd

For some time the DCU fan has been making an unpleasant noise which is only getting worse. With the museum closed to the public during January it was the ideal time to remove the fan unit to replace the motor bearings. I also replaced the bearings on a spare unit at the same time, and both have been successfully tested.

SoftwareDavid Holdsworth

In November my reports promised to sort out all the material around my servers so that everything was on the “new” Raspberry Pi at TNMoC.

In recent days, I have had new ideas and written a simple software tool to speed the job along. So the job is not finished, but is now making good progress.

Elliott 803, 903 & 920MTerry Froggatt

Peter Onion reports that TNMoC 803’s +13.5V inverter is giving a bit of trouble, as it sometimes fails to start up when the machine is turned on from cold. This results in an immediate “turn-off”. Last time this happened the cure was to replace the “start-up” capacitor that biases the (self oscillating) transistors into their linear region for some milliseconds to get the oscillation started.

A MkII version of the Paperless Tape Station board has been built. This version uses a pair of Raspberry Pi Pico boards in place of a single PIC micro controller in the MkI version. This has made for much easier code development (in C rather than PIC assembler) which has allowed new facilities to added. The most important new feature is the ability to direct 803 output to the recently restored Creed 75 by setting the output selector switches on the Paper Tape Station control panel to “Teleprinter”.

During the testing of the MkII board a few Minilogs with weak outputs were found in the Paper Tape Station. These have been replaced with devices removed from some “scraped boards” that have had their gold plated edge connectors removed. Some investigations of the gates with weak outputs suggest the failures are due to increased leakage in the OC42 transistors used in the Minilogs. In some cases observations suggest that the leakage is in the collector-base junction as a “fast turn off” is followed by a partial “slow turn on”. In one case the gain of the OC42 appears to have fallen so low that a single input being up (0v) is not giving enough base drive to turn the transistor on. A test rig for Minilogs removed from the “scraped boards” is being designed. This will ensure only working ones are used to replace faulty ones removed from the PTS.

Peter Williamson reports that TNMoC’s 903 has had to be moved to facilitate radiator replacements at TNMoC, but it seems to still be working OK. The engineers’ display panel has had the PSU module replaced and powers up but is not as yet in service. It would be good to establish that this is working properly before we next need it to diagnose faults on the 903 itself.

Regarding the 920M, I believe that I have now completely traced all of the motherboard wiring, and I understand most of it. The purpose of most of the logic module types is also clear. I cannot claim to understand the timing circuits of the store or of the control deck, or even which way some of the timing signals go. The internals of many of the potted store modules will remain a mystery.

In Resurrection 95, I reported that Dr Erik Baigar had found some 920M training notes, which included a microprogram diagram, reduced to a single barely-readable sheet. The more readable parts helped me understand those parts of the microprogram in the hardware. I’ve now been able to reverse this process, tracing the hardware to draw the full microprogram diagram, as a single sheet of searchable PDF, nominally A3 but still readable on A4.

ICT/ICL 1900Delwyn Holroyd, David Wilcox, Brian Spoor, Bill Gallagher

PF50

The sorting of manuals, boosted by John Harper’s recent contribution, has turned up a matter that will require some investigation regarding the handling of interrupts from the console typewriter. The console typewriter on PF50 used a 6164 board (in common with 1901A and PF56) so this potentially affects all three emulators.

TV Device?

We have just been made aware of a relative of the GD1830 General Purpose VDU, a cluster VDU system called the 1831. This was a multi-headed set of small VDU display screens (4” x 2”) for data enquiry, with two models, one supporting up to eight terminals, the other up to 64. This appears to match an ‘orphan’ device in the E6RM source, known there as the TV device (PERI type 40). The code there matches the description of the eight terminal 1831. We have no programs or documentation for this (presumably) old device. However E6RM’s last release still had the device package included – anyone know more?

7210 IPB

David is now able to run the test program #N210 for long periods but there is still the odd gremlin which causes the tests to fail.

Our Computer HeritageSimon Lavington

Some modest progress has been made on compiling the systems architecture, instruction set, register details, etc., for a new subsection of the CTL Modular One entry. Information is being extracted from relevant CTL technical manuals and the resulting information presented in an appropriate OCH style.

I have been asked to review and catalogue a large collection of Ferranti manuals, internal reports and personal papers lent by the son of two long-serving former Ferranti employees (husband and wife) whose service with the company extended from 1953 to 1985. It is not expected that much in the way of additional material will be found which causes edits to be made to existing OCH entries.

IBM MuseumPeter Short

Current Activities

We have now received two complete AS/400s, including terminals, printer and magnetic tape drives, from an IT consultancy in Wales.

We are hoping that the museum can resume visits in the not too distant future. These will be coordinated by Hursley employees and in the main led by lab volunteers. Curators will be providing training.

Scanning of hundreds of hardware slides is continuing at home, allowing further expansion of the photo archive.

Photo Shoot

A Data Storage Specialist Communications Agency visited the museum just before Christmas for a photo shoot session. They are planning a social media campaign about old technologies later this year, focussing on data storage.

Law Society

A working IBM PC with mono display has now been collected from Hursley for the High Court exhibition in London, mentioned in our last report.

SSEM ReplicaChris Burton

The Baby replica has been running very well in recent months. However, shortage of volunteers occasionally means that with only one present, the machine can only be shown un-powered and not working. This situation means that even simple maintenance is not possible. Consequently, the big news is that SIM is about to start a drive for new volunteers which should be a great help, as we lost many of our regulars during the Covid years.

Top Previous Next

News Round-Up


In 2014 the Society organised a group visit to the Heinz Nixdorf Museumsforum in Paderborn, Germany www.hnf.de/en/home.html.

Such was the success of the trip that we have repeated it each spring, visiting computer history museums throughout Europe. In 2023, we are returning to our friends in Paderborn to see the progress that has been made and to experience, once again, the world’s largest computer museum.

The centrepiece of our visit will be our day at the Museum itself on Saturday 15th April. The plan is to foregather the previous evening for dinner in one of the excellent local restaurants (everything you may have heard about German cuisine is wrong – it’s great) and, in all likelihood, this will be repeated the following evening. All CCS and TNMoC members and their guests are warmly invited. Do join us. It will be fun!

Our previous hotel, the Arosa Hotel in the Westernmauer, a Best Western hotel, is our planned base for the weekend.

Hotel and travel arrangements are, of course, the individual responsibility of participants.

There is an airport at Paderborn, but it is quite a long way out and flights from the UK tend to be via Munich! Flying to Dusseldorf or Dortmund and thence by rail to Paderborn would appear to be a better bet. Or perhaps take to the rails all the way from St Pancras.

A few of us are planning an excursion by rail on Sunday 16th to (relatively) nearby Wuppertal to experience the well-known “Danglebahn” (more properly the Schwebebahn – a unique suspended monorail railway) and you are welcome to join us.

As usual, Dan Hayton will be our organiser. You should contact him at Daniel.hayton@bcs.org so we know you’re coming.

101010101

Our friend Herbert Bruderer has published some more of his researches. At ccsoc.org/bru11.htm may be found some musings on Colossus quoting Brian Randell and Tony Sale. He surveys very early work on compilers at ccsoc.org/bru12.htm. Finally, at ccsoc.org/bru13.htm he draws our attention to the recent discovery of a copy of the user manual for the Zuse Z4 relay computer which was thought to have been lost. An English translation is available. Where does he find the time?

CCS Website Information

The Society has its own website, which is located at www.computerconservationsociety.org. It contains news items, details of forthcoming events, and also electronic copies of all past issues of Resurrection, in both HTML and PDF formats, which can be downloaded for printing. At www.computerconservationsociety.org/software/software-index.htm, can be found emulators for historic machines together with associated software and related documents all of which may be downloaded.

Top Previous Next

Queries and Notes


Ferranti Cerberus

Simon Lavington writes – There is one project about which it would be interesting to learn more. This is the Ferranti Cerberus computer, developed at Lily Hill, Bracknell, as part of the ‘Green Ginger’ classified defence project in the early 1960s. There is no existing OCH entry for this computer. If anyone can help by providing pointers to Cerberus information, please get in touch with me.

ICT 1900 Information Wanted

And Delwyn Holroyd also enquires – We have a good deal of information on the 1900 series, but we are lacking in certain areas:

  • 1901/2/3– we have logics but nothing else
  • 1906/7 – we have nothing except FPU logics
  • 1906A/S – very little except brochures
  • Executives other than E4BM, EWG3 and E6RM

EXCHLF Manual

Nigel Benée’s request for information on this rather obscure Atlas programming language was met by [ed] who happens to have a copy of the manual on loan from Bill Williams. And what, you may ask, was EXCHLF? An extension of CHLF developed by Benedict Nixon at London University. CHLF? A collaboration between Mercury users CERN, Harwell, London and Farnborough to extend the life of existing Mercury Autocode programs. Given that Atlas (and ICT 1900) already had a compiler for Extended Mercury Autocode (EMA) the motivation for this project is unclear.

Elliott 503

Bill Purvis is currently collaborating with Eric van der Meer (in Holland) in reconstructing a modern representation of the Algol 60 compilers produced by Elliotts in the early 1960s. We are working with original binary tapes, supplemented by an incomplete set of source listings provided by Elliotts to all customers who had been issued with the compilers for the 803 and 503 computers. In doing this we have uncovered a bug which exists in the compiler tapes but not in the listings and are puzzled as to how this can have arisen. We would be interested in communicating with any of the original Elliott Algol developers and maintainers. There are aspects of how the compilers were put together and in trying to determine how the above mentioned bug can have crept in. Please contact if you can help.

Top Previous Next


Delilah Voice Secrecy System

John Harper
xxxxxxxxxxxI
combiner
Combiner Unit
key
Key Unit
key
Key Unit
Delilah was a voice secrecy system produced at Hanslope Park by Alan Turing and a REME officer, Donald Bailey during 1943 and 1944.

The units are a Key Stream generator, a combiner Unit, and a Power Supply. One of each was used at the sending and receiving end. The system works on mains electricity and is not very heavy so it can be moved around easily.

The first official record that has come to light appears to have been a report dated 6th June 1944 in which Alan Turing wrote that research began in May 1943. A Combining Unit appears to have been in existence at this time but the ideas about the Key Stream were just coming together. The main report appears to be undated, but it must have taken many more months to build a Key Unit and carry out the reported tests and measurements. One might speculate that the project would have run until the end of 1944.

What is common to many voice encipherment systems is a key stream that is unique for a given transmission. This is ideally in a ‘one-time pad’ form that is added at the send end and subtracted with an identical key stream at the receive end. It is believed that Alan Turing was shown under strict security the workings of an American Voice Secrecy system called SIGSALY when visiting the States (see en.wikipedia.org/wiki/SIGSALY etc.) and might have contributed to its design. SIGSALY systems were extremely large, about the size of a 1950s mainframe computer and extremely expensive. Even more expensive were the ‘gramophone records’ that held the unique key stream. These had to be distributed very securely to each end of the voice link and once used, destroyed. It is speculated but assumed that Turing thought that he could improve dramatically on this at a fraction of the cost. In Delilah his one-time pad was the setting of five letter transposition wheels like an Enigma Machine, plus a seven way patch panel. These modified the key stream. Not quite a one-time pad but as this key stream changed with every send and receive change, breaking an enciphered voice message that would change in minutes was considered at that time to be very secure. To add to this, wheels could be reversed, or alternatives fitted. As with Enigma the weak spot would be the ‘Wheel Setting Sheet’ which the enemy would wish to get hold of so that they could then decipher the messages.

The system is labelled MK 1 with the report detailing the need for further development. No doubt this work could have continued but the war was being won and people were looking forward to a civilian career. From what we read about Alan Turing he was also very keen to get back to computing. As it was the system was not developed to the production stage but what was a technical advance was, the very clever method of producing a one-time pad that was thought to be highly secure in 1943-44. The technique bears some similarity to a German Enigma machine where every message/voice transmission is unique. The setting for a given message was secure for the time of a given ‘talk’ so long as it was longer than a few minutes. When the sender went from send to receive the key stream changed ready for the next unique transmission.

The original Delilah Project came to the attention of the Bletchley Park Volunteers when one of our team, who regularly visits the National Archive in Kew, discovered a very large report. It consisted of 80 pages of text and formula plus circuit diagrams, blueprints, and oscilloscope images. We arranged to make copies of these using a camera. This was a start but later the GCHQ Archives department made available a full copy in much better quality than ours.

We became quite excited about this project because it was a major piece of work carried out by Alan Turing.

To be fair, Andrew Hodges had already ‘discovered’ the Delilah Project some years ago (Alan Turing – The Enigma – pages 273-276 inclusive). In this he writes about Don Bayley, Turing’s partner, and co-author of the report. Also, there is a picture and a brief reference by Dr. Robin Gandy in a BBC TV Horizon documentary directed and produced by Christopher Sykes some years ago. Although Turing was considered to still be based at Bletchley Park, he spent many months at Hanslope Park presumably because they had good workshop facilities away from the Hurly Burly of Bletchley Park.

With the help of retired Hanslope Park Foreign Office staff we were able to track down Don Bayley who was then living quietly in Yorkshire. I have visited him there and since corresponded.

No doubt the authorities were no longer keen to fund further, war-related, developments. Whatever the reason no further work was done on Delilah and it never went into production

Top Previous Next

The Development of the MOSAIC Computer

Edward Smith

Architecture Review

MOSAIC, the Ministry of Supply Automatic Integrator or Computer, was developed by two Post Office Engineers, Allen Coombs and William Chandler, based on version seven of Turing’s logical design for ACE. A major difference between this and the pilot ACE design was that a single MOSAIC instruction could simultaneously access two source addresses as opposed to the single address accessible by a Pilot ACE instruction. Like Pilot ACE and EDSAC, the machine used mercury delay lines as main memory and the designers were able to draw on their experience with the Colossus project, Turing’s logical design and the published experience gained on EDSAC with delay lines. The delay line technology meant that although individual delay lines could be directly selected, the availability of data was governed by the timing of circulation of data within the delay line.

block diagram
Figure 1 Block Diagram of MOSAIC

This description is based on journal and research reports produced by Coombs and Chandler. The machine used 7,000 valves mounted in 12 standard Post Office racks and its basic structure is shown right. MOSAIC used an increased number of valves to improve circuit simplicity, given that faults with valves were comparatively easy to resolve. The principal design goals were: the speed of operation and the necessity of avoiding any interference.

The topic is explored by first examining the store (memory), which is fundamental to understand the approach to programming. The paper then considers the structure of an instruction and how programs are written, before exploring how the hardware such as the Arithmetic Unit, the Control Unit, and the Input-Output mechanism are deployed.

The Mercury Delay Store

MOSAIC used ultrasonic pulses, timed at about 1µs, travelling along mercury filled tubes for storage. The architecture was based on a 40-bit word that could represent a 12-digit decimal number and was capable of adding two numbers together in 70µS. 1,000 bits could be stored before the first bit transmitted reached the receiver. The pulses were carried on a modulated 10MHz carrier, an approach championed by Turing. On reaching the end of the delay line the signal was converted to an electrical signal using a quartz crystal then reshaped, resynchronised and amplified prior to being fed back to the line input.

The machine’s storage provided a capacity of 1024 words main store mainly using 64 delay lines each holding 640 bits (16 words), with a pulse frequency of 570k pulses/sec. The passage of a single 40-bit word past a given point in a delay line was known as a minor cycle and the complete circulation of a pulse was known as a major cycle. A major cycle consisted of 16 minor cycles each of 70.2µs making the period of a major cycle 1.12ms. Data was read and written using the electronic part of the circulation path provided at the regeneration point. A transfer was defined as a “fetch data”, “operate” and “write data” cycle.

store
Figure 2 The Store

32 short delay lines each capable of holding one 40-bit number were also provided to give more rapid access to data. The delay lines had to be maintained to the same temperature (within 1°'C or less) using a thermally isolated container, surrounded by a cavity wall and installed in a separate room from the electrical equipment. The long lines were five feet long and held as four stacks of 16 lines.

Long delay lines were labelled as Lx, with only L1-L31 being used for instructions, while short delay lines were labelled as TSx. A double length delay line, DS, also formed part of the main store and was capable of storing an 80-bit number. L0 was used to control the master clock frequency automatically and was not available for general use, leaving 1008 positions in the main store.

Instruction Format

Figure 3 shows the layout of an instruction. As an example, the instruction to add the contents of short delay line 3 to those in long delay line 5, place the results in long line 35 and taking the next instruction address as long line 5 is binary 101/1100001/0/1010000/1/1100010/0/1011/10/10100. The function code of 11 is the add function. The instruction is executed immediately (Go bit set), with the instruction deferred, as indicated by the α bit being set by the 13 minor cycles specified in the timing field. The seven-bit addresses are 0-63 (L0-L63), 64-95 (TS0-TS31) and values greater than 95 have special meaning covered later.

instruction
Figure 3 The 40-bit Instruction Word

The instruction store was part of the main store and comprised 30 long lines and one short line. Access to these was obtained using the five digits of the Next Instruction Service (NIS), which identify the line, but not which of the 16 words is used.

Multiplication was specified as being signed or unsigned. Addition and subtraction could require suppression of the carry after the 40th bit, after the 80th bit or not at all. This facility involved four possible function settings for each. The three forms of comparison (“AND”, “OR” and “XOR”), two multiply options and the delay function gave 14 possible operations requiring four function bits. The functions (0100 and 0000) cut off all output from the Arithmetic Unit to Destination Highway.

The machine worked on a two-beat cycle; the first beat, lasting one minor cycle, examined an instruction extracted from the store, setting up the machine conditions to enable its execution. The second, the “Obey” beat lasted for a variable number of minor cycles allowing the instruction to execute. Towards the end of the Obey beat, the next instruction was extracted from the store ready for the next Setup beat. The duration of the Obey beat needed to allow its minor cycle to coincide with ensuring the next instruction was available to the next setup beat. Most instructions involved a number transfer taking a single minor cycle; however more elaborate timing sequences were used as shown in figure 4.

A capability known as the “discriminate facility” permitted conditional branching. Here an instruction X may be followed by instruction Y or Z depending on the result of the execution of X. The length of the Obey Instruction beat could be set by the four-bit timing number, which could take a value on n minor cycles, where n had a between 0 and 15. This worked in conjunction the two-bit characteristic which took the form of:

  • Immediate (α=0, β=0) – The current instruction X started immediately and was obeyed for n+1 minor cycles; taking Y which followed X with a gap of n+1 minor cycles as the next instruction. If a discriminate condition was specified, an extra minor cycle was inserted at the end of the transfer and the Z instruction at the (n+3) th minor cycle executed.
  • Deferred (α=1, β=0) – The transfer took one minor cycle and occurred n minor cycles after the setup beat. In the absence of a discriminate action the next Setup Beat followed immediately (minor cycle n+2); with discriminate an idle minor cycle was inserted as before.
  • Serial (α=1, β=1) – For this the Obey Instruction Beat is 16 minor cycles with a gap of n minor cycles before the transfer took place. Successive instructions could therefore be in successive positions of a long line. Discrimination worked as described previously as execution of either W or V one minor cycle later.
  • Forced Discrimination (α=0, β=1) was a type of immediate instruction, with discriminate automatically applied so that the next instruction was always Z.
timing
Figure 4 Timing Diagram for MOSIAC

When a new Setup was completed two binary counters Z and N were initialised. Z produced a pulse at the start of the obey cycle, and N, controlled by the timing number n, produced a pulse at the end of the nth minor cycle after the Setup Beat. The Obey Control circuit (based on the Z and N pulses, the characteristics and GO bit settings in the current instruction word and under the overriding control of the GO bit in the instruction word) provided an output pulse to the discriminate circuit to indicate the end of the basic “Obey” period. For immediate, forced discriminate and deferred instructions, the output pulse to the discriminate circuit occurred one minor cycle after the N pulse and was transmitted by the discriminate circuit either with or without a minor delay, initiating a new setup beat. If an instruction X was to provide the facility, then the destination address specified in X was either 97 (D PLUS) or 96 (D NOUGHT). If D PLUS was called the minor cycle delay was inserted, if the output number was negative (bit 40 = true) and if D NOUGHT is used, the delay was inserted if the output was non-zero.

If the GO bit was set the instruction executed immediately, if clear it suspended execution pending external authorisation to proceed (push button by the operator or from mechanical input gear). This could be overridden by an external “Stop” control, or the “Step” control, which permitted a program to be executed a single step at a time.

Programming

flowchart
Figure 5 Flowchart of an Example Problem

The five-bit NIS selected which of the 32 storage lines permitted to contain instructions should be used. The seven-bit fields used for sources and destinations could theoretically reference 128 lines, but no more than 100 were used, leaving 28 available for other uses. These include “D NOUGHT” and “D PLUS” for discriminator control. A number with all 40 bits set (equal to -1) could be supplied to Highways 1 and 2 by specifying Source “107” in the instruction word. Similarly, setting a source as “97” (P1) created an input with 1 in the least significant bit position followed by 39 zeros, i.e. +1. Source “98” (P2), likewise, generated an input word of +2. In a flow diagram the instructions were arranged in the order they are held in the delay lines. For example, the flow diagram in figure 5 describes a program to identify the highest factor of a number N, by beginning with a solution of a=N-1 and decrementing it until a value is found that is a factor of N. Each number “a” is repeatedly subtracted from N until a remainder of zero is seen, if the remainder is positive then “a” is again subtracted, if it is negative then the solution a =a-1 is tried. The program whose storage layout is shown here is assumed to be located in Delay Line 1. The first instruction may go into the first minor cycle and is of the “Immediate Nought” type. The next instruction source will be delay line 1 as it will be for all other instructions in the program.

layout
Figure 6 Storage Layout (left) and Logic Flow (right) for an Example Program

The next instructions are at minor cycles 2, 4 and 6 and are of immediate form. The next instruction has a discriminate aspect and may be followed by the instruction at minor cycle 8 when D PLUS is on (positive number) and minor cycle 9 when it is not. The instruction at minor cycle 9 is deferred by 7 and is therefore followed by the instruction at 9+2+7=18 i.e. minor cycle 2. If the outcome is a positive number the instruction followed is at the eighth minor cycle with destination D NOUGHT, which is also a discriminate instruction that will not execute for another 9 minor cycles. If the result of the operation is zero, the next normal instruction at position 8+2+9 =19 (wrapping to minor cycle 3), which contains the instruction “Record the number in TS2” is selected. The discriminated location for the non-zero result is one further on at minor cycle 20, which is location 4.

Subroutines could be implemented using the concept of a “Link” instruction; here the last instruction of the subroutine identified a particular location in the instruction store as holding the next instruction source to return to. The calling routine loaded this storage location with an instruction address, the link instruction, which returned control to the appropriate point in the calling program.

Hardware Description

The arithmetic unit had three main components: the function box, the delay unit and the multiplier.

arithmetic unit
Figure 7 The Arithmetic Unit

A delay function was provided to shift a number n positions towards the most significant end; thereby multiplying it by 2n, more easily and rapidly than the multiplier unit. The “delay unit” could delay signals by up to 40 pulse periods; the delay was encoded in the bits of “SOURCE 2”, since there were insufficient spare bits in the instruction word. Only Highway 1 was fed to the delay unit.

Most operations took one minor cycle of 70 µS, but the multiplier needed about 6mS to produce its answer and was sealed off for the duration of its operation. The multiplier required about six uninterrupted major cycles to complete its operation. Any subsequent instruction using the multiplier was forced to become a Wait instruction until the multiplier finished its computation. The other functions of the machine could continue and the result placed in a temporary store.

The “multiplier accumulator” was an 80-bit delay line, with its local recirculation path through the multiplier. It “collected” the answer to a multiplication and circulated it until a later instruction called for it by specifying “MAC” (95) as Source 1 or 2. It could also be used as an 80-bit store by feeding to Destination “MAC” (95). If Destination “MAC+” is specified (94), then the number in the Destination Highway is added to the contents of the accumulator. The unwired address “112” was often used as a dummy destination for multiplication.

Words and numbers were stored in dynamic form as ultrasonic pulses in delay lines but needed to be held in static form so that all bits were simultaneously and continuously available for electronic processing. This is done using an element known as the Staticiser; the inverse unit was known as the “dynamiciser” and turned a static array of 40 bits of information into a dynamic stream of signals.

control circuit
Figure 8 Block schematic of Control Circuit
block schemativ
Figure 9 MOSAIC Block Schematic

The Control unit involved two closed rings-the Current Instruction Staticiser, which held the current instruction, with its circuit, and the main Control itself. The main control loop was entered during the last minor cycle of an Obey Instruction beat, when the instruction being obeyed was held in the Current Instruction Staticiser. The next instruction, held in the delay line, identified by the NIS bits of the instruction in the staticiser, was transferred serially to a special delay line known as “INST” which is one minor cycle long, is not part of the main store and is associated with the control circuit, being used to route instructions to the Current Instruction Staticiser.

The setup control circuit came into operation and the setup beat commenced, causing the array of bits previously held in the staticiser to be replaced by those emerging from INST. The context of INST couldn’t be changed and once the next set of bits were established in the staticiser, the Set-up beat was over, and the staticiser contacts were disconnected from the input circuitry. The control unit used only about 500 valves of the 7,000 in the machine, and occupied one complete rack.

Input and Output

Standard punched cards, with 80 columns and 12 rows, were used for input and output, with the card reader sometimes acting as a slow-speed back-up store. Cards represented one 40-bit binary number per row. Punched cards were chosen for output. Data was initially produced in binary form, until the necessary decimal to binary conversion mechanisms were created. A reproducer was used as the output device and a modified tabulator for input.

I/O
Figure 10 Input and Output Controls

Cards could be read at 200 cards per minute; with any one row under the reading brushes for at least 5mS. To read a new card, the next instruction specified destination address 100 (“Hollerith Read”), which operated the card feed circuit and started a card moving under the reading head. The instruction to read a row, made use of the “Input Dynamiciser” as the first source address of 0000000 and as destination, the store location to be used for holding the information. This was a wait type instruction that was obeyed when permitted by the card reader. The transfer was followed by a new instruction being setup. If this called for the transfer of the second row of the card, which was not available for a further 20ms, it would be a wait instruction specifying Input Dynamiciser as first source. This process was repeated for the other rows of the card. In practice, the Dynamiciser is nearly always fed to the short line TS0 as an intermediate stage.

The Hollerith punch used a Staticiser and the cards were punched at a rate of 100 per minute, cards being punched using Destination 101 (“Hollerith Punch”), which started a card feed to the punch. The next instruction was of Wait type, and called for the required output to be transferred from its storage location address to the Output Staticiser (99); it was obeyed when a card was in position under the punches. The Staticiser needed to be reset to all zeros between uses. The “Staticiser Reset” was applied automatically after a punching, but for the first Read-out of a calculation, the starting program for all calculations required a Staticiser Reset instruction (Destination 98).

There were three types of instructions: firstly those preparing the machine, which were fed directly from Input to Control using an instruction of 40 zeros and were lost once used. The second type directly or indirectly control fed in the application program and its data; these were held in store temporarily, but eliminated as the true programme was fed in. The third type constituted the main program and were fed to the store but not used until feed-in was complete. The Hollerith Reader was a standard tabulator, that used a “Start” button to initiate a card feed. Start-up required engaging this and the “Clear” key, interrupting the circuit feeding INST and causing zeros to flow into Control. Source 1 “0” was the Input Dynamiciser, Source 2 “0” is a source of zeros, Destination “0” is “INST,” and Next Instruction Source “0” was short line TS0. This transfer was executed on receipt of external authorisation and after an integral number of major cycles following the end of the “Set-up” beat. After the “Start” operation, the Clear switch was depressed feeding zeros to INST and the Step button operated, causing the instruction in the current instruction staticiser to be obeyed. An “Obey” beat ensued and a “Set-up” beat, after which the staticiser held the “40-zero” instruction. When the Clear switch was restored, and the Start button operated, the Hollerith Reader started a card feed. When the first row of holes reached the brushes, the “40-zero” instruction was obeyed and the card read in. The word which the first row of holes represents was then set up on the current instruction staticiser, and was thus the next instruction to be obeyed. The first word on the first card would be a “Go” instruction to feed zeros to TS0, specifying TS0 as the next instruction source. By the time the second card was available, the “40-zero” instruction was circulating in TS0 and set up in Control. The second instruction on the card was a “Go” instruction used for preparing the machine, which was fed to Control via INST and used destination “Staticiser Reset” (98). TS0 was given as the next instruction source. The “40-zero” waited for the third word on the card. This applied to all initial words on the card, which would be fed to Control, clear various parts of the machine and then be lost, leaving the “40-zero” established in Control.

The last of the type 1 instructions ordered transfer of the next line of the card (type 2) to the area in store to be used for feeding-in the main program and to be stored temporarily. The type 2 instructions could be fed in by alternating them with type 1 instructions, each calling for next instruction source TS0. The first type 2 instruction was made to control the feeding-in of a batch of type 2 instructions, each of which could then feed in another batch. Hence, every row on the cards was made effective instead of every other row. One type 2 instruction was the “Hollerith Read” instruction, and after it has reached the store it must be executed each time a new card is required. When all the type 2 instructions were stored, the feed-in of the main program began.

Conclusion

This article has described MOSAIC, a computer built for the Ministry of Supply by the British Post Office and based on an advanced form of Turing’s ACE architecture. It has taken the reader through the basic architecture, the instruction format and provided an overview of how that format was implemented in hardware.

Top Previous Next

Simulation Modelling of Historical Computers[2]

Roland Ibbett & David Dolman
Being part 2 of the article part 1 of which can be found in Resurrection 100.

CDC 6600

The CDC 6600 was first demonstrated by the Control Data Corporation in 1964. The design team was led by Seymour Cray, who went on to design the CDC 7600 and later, after he left CDC to form his own company, the Cray-1. The 6600 was designed to solve problems substantially beyond contemporary computer capability. It achieved its performance by the use of parallel functional units, a small set of Scratch Pad registers, a three-address format (Figure 5) that allowed successive instructions to refer to totally independent input and result operands, instruction buffering and by off-loading peripheral handling to separate peripheral processors. Particularly interesting to students of computer architecture is the Scoreboard system used to control the operation of the parallel functional units..

6600 insruction formatg
Figure 5 : CDC 6600 instruction format

Figure 6 shows the HASE simulation model of the CDC 6600 during execution of the first of the two versions of the model available from the website. This version contains a program that demonstrates the operation of many of the CDC 6600 instructions implemented in the model. In particular, it includes sequences of instructions that show how the Scoreboard deals with instruction dependencies. The second version contains a matrix multiplication program that shows how the instruction stack works and how instruction unrolling of the scalar (dot) product calculation (the inner loop of matrix multiplication) is used to maximise register use.

simulation model
Figure 6 : CDC 6600 Simulation Model

In principle it ought to have been possible to represent the 6600 instruction set in the model using the INSTR construct but in practice there is no regular relationship between groups of functions and the register sets, so most instructions have to be decoded, and their registers selected, individually. Instructions are therefore represented using a STRUCT containing 5 integers and instructions are decoded in the Scoreboard simulation code using a switch statement. The code for each case in the switch selects the required registers and also contains text to be displayed in the Scoreboard icon showing the assembly code representation of each instruction.

On screen, active entities (other than memories) are coloured red, inactive entities are shown in light blue, while entities that are held up waiting for a packet from some other entity are shown in yellow. In the functional units in Figure 6 the functional unit colours appear correspondingly as dark grey, light grey and white. Also shown on the right in Figure 6 are the instructions in the sequence currently being processed. The first instruction is being executed in the Divide Unit. The second instruction has been issued to the first Multiply Unit but cannot start because it is waiting for the result from the Divide Unit to be returned to X7, as shown by the entry in the window at the left-hand end of the Multiply Unit. This is known as a “second-order conflict” in 6600 terminology, but would nowadays be called a “read-after-write dependency”. The third instruction has been issued to the Add Unit, but cannot start because of a “third-order conflict” (a “write-after-read dependency”) on X4, as shown in the window at the right-hand end of the Add Unit, i.e. it is ready to write its result to X4, but the current value in X4 has not been sent to the Multiply Unit. The design of the 6600 was such that register values are sent to a functional unit only when both are ready, i.e. not waiting for a result from a previous instruction; here X7 is not yet ready. The next instruction does not depend on any previous results, so has been issued to the first Increment Unit and is being executed. The Scoreboard is displaying the last instruction, which is yet to be issued. The window at the bottom of the Scoreboard is used when there is a “first-order conflict”, either because the instruction about to be issued requires a functional unit that is already busy or because its destination register is already waiting for a result from another unit (a “write-after-write dependency”).

The Manchester ‘Baby’

baby model
Figure 7 : The HASE Baby Model

On 21st June 1948 the University of Manchester Small Scale Experimental Machine, affectionally known as the ‘Baby’, became the first computer to execute a program (Kilburn’s Highest Factor program) stored in addressable read-write electronic memory. This memory worked by storing charge on the screen of a cathode ray tube, now known as a Williams-Kilburn Tube. Like the Baby itself, the main emphasis of the HASE Baby model (Figure 7) is to demonstrate the operation of the Williams-Kilburn Tube Main Store. This requires its icon to be large enough to clearly display the Baby’s 32 lines of 32 bits each. To allow for this, the Accumulator and CI/PI icons are much smaller. In the case of the Accumulator a further compromise is that the value is displayed as an integer rather than a pattern of 32 dots and in the case of the CI/PI icon, only the relevant bits of the CI and PI words are displayed, i.e. 5 bits and 8 bits respectively. The PI bits are only visible whilst PI is in use.

At the start of a simulation the program and its data are contained in two separate files held in the Input entity. In the current version of the model, these files can contain the code and data for four programs (there are currently three in the model). A parameter of the Input entity allows the user to select the program to be run. During the first clock period of the simulation the relevant sections of these two files are transferred into the Main Store entity. Entries from the program file are converted from the human-readable form in the input file into Baby format and transferred to the Main Store starting at location 1 (because the first action in the Baby was to increment the value (initially 0) in the CI register). The remaining Main Store locations are then filled with values from the input data file. Each word in Main Store is then displayed on the screen in focus/defocus format, with a small (focussed) dot representing a 0 and a large (defocussed) dot representing a 1, and with the least significant digit at the left-hand end.

Reading a line in a Williams-Kilburn Tube involved writing a 1 to each bit location in the line. If the location contained a 0, there was a redistribution of charge on the screen and this induced a voltage in a pick-up plate placed close to the screen. If the location contained a 1, there was no redistribution of charge, so no voltage was induced. In the HASE model a line being read from the Main Store is shown changing to all 1s and then changing back to its original value, though this is purely a feature of the display; the Main Store array remains unaltered and is simply read by an assignment statement.

A further difference between the model and the Baby itself concerns the way in which the serial nature of the Baby is visualised. Watching 32 bits transfer between entities at a rate of one per clock period would be extremely tedious, so words and addresses are each transferred in a single HASE clock period with their integer values represented by strings of 0’s and 1’s. The time taken to read a full word from the Main Store, plus the fly-back time, constituted a ‘beat’ of the machine. Four beats were required to execute an instruction. In the model these are:

add +1 or +2 to CI, copy CI to staticisor
read Main Store, send instruction to PI
copy PI to staticisor, fetch operand from Main Store
execute the present instruction

The timing bar at the bottom of the Project View pane shows which beat is active as each instruction is executed.

The black dot below the Subtractor represents the neon lamp that lights up when a Stop instruction is executed.

Of the three programs contained in the model, Program 1 demonstrates the operation of all the Baby instructions, while Program 3 displays a simple message on the Main Store icon. Program 2, for which the code (in Chris Burton’s modern mnemonic version) and data are shown on the right in Figure 7, is the 18th July 1948 version of the Highest Factor program, as described by Geoff Tootill, along with some of his explanatory comments. The number under investigation, a, is set to 4537 (=13x349) in the input data file, this being one of the numbers used in the first tests of the Baby on 21st June 1948. The initial trial factor, b1, for that test was 4537 but in the model it is set to 360 to avoid a very long simulation run. At the end of a simulation run, the highest factor of 4537, 349, is in Line 27 of the Main Store. Of course users of the model can set the input values to numbers of their own choice.

Conclusion

Simulation models of historical computers that provide visual demonstrations showing how their various architectural features operated offer effective alternatives to preserving, in working order, what were in many cases very large pieces of equipment. Having simplified models in an application such as HASE allows anyone interested to down download a model and observe it in action.

Future

It is intended to update the text of this article from time to time as and when the system develops. Go to www.ccsoc.org/hase0.pdf for the latest version.

HASE simulation models of a variety of computer architectures and architectural components have been created, some for research investigations of large-scale computing systems and others for use as teaching and learning resources in lectures, for student self-learning or for virtual laboratory experiments. The HASE website (www.icsa.inf.ed.ac.uk/research/groups/hase/) provides access to the models and to the Java code for HASE itself. Each model has its own supporting webpages describing the system being modelled, as well as the model itself, and from which the source files for each model can be downloaded. These files can be used as inputs to HASE, which has options to load a project, to compile a simulation executable and to run the simulation. Running a simulation produces a trace file which can then be used to animate the on-screen display of the model to show data movements, parameter value updates, state changes, etc.

Contact details

Readers wishing to contact the Editor may do so by email to
or by post to 124 Stanley Road, Teddington, TW11 8TX.

Members who move house or change email address should go to
www.computerconservationsociety.org/membership/membership_general.htm. Those who are also members of BCS, however, need only notify their change of address to BCS, separate notification to the CCS being unnecessary.

Queries about all other CCS matters should be addressed to the Secretary, Rachel Burnett at , or by post to 80 Broom Park, Teddington, TW11 9RR.

Top Previous Next

The Birth of the Barcode

Derrick Brown
This article covers the development of the world’s first barcode system to be used in a commercial environment. The problem that was to be solved is described together with the solution, in which the barcode played a crucial role. Although 50 years has flown past, elements of the system including the barcodes remain in use to this day.

The two engineers carried a large heavy box and a bulky bag into the director’s office. They unpacked the contents revealing a plain steel box with a display panel and a dial-telephone cradle but without the usual telephone handset. Instead a largish tubular device that they called a ‘reading wand’ sat on the telephone cradle, connected by a wire to the box. This, they explained, could read a code made up of black and white stripes. A striped code was produced, about twenty centimetres long. The director and myself made up the audience. We watched as one of them wiped the wand across the coded stripes. The number represented appeared in the display. Another striped code was produced and wiped, with the number appearing in the display. Then another. Then the code was wiped backwards, the number appearing correctly. The audience was impressed. The director took the wand and wiped the code himself.

“Of course”, he said, “they may receive rough treatment” and with that he vigorously wiped the wand back and forth across a code. There was a bang and a flash from the box and a pungent burning smell.

“Oh well”, he said, “these things rarely work the first time!”

Despite this little set-back he was sufficiently impressed to give me his support for on-going development and this led to the world’s first commercial barcode application.

This event took place alongside the River Thames, in the Blackfriars, London Head Office of Sainsbury’s (JS), following some months of discussions and site meetings between myself and a small unit of the Plessey company, based at Poole, Dorset. This business had been behind one of about 60 responses from an enquiry that Sainsbury had put out for a proposal for a method of capturing data in remote locations. Most of the unsuccessful responses were fanciful, ridiculous, or pure science fiction. This was 1971 and data was collected by keying. Mark sensing was also being used in specialised applications but had severe limitations. Many of the solutions put forward relied upon keyed data, in some form or another. The solution from ICL, for example, involved mobile teletype terminals with trailing power cables – but the brochure was beautifully produced! I ruled this out along with all keyboard-based solutions. The requirements specification demanded a data collection system that was error-free, that could be used by anyone without training, with equipment that was easily portable by a person. But how had this come about, one may ask?

Well, I had fallen on my feet. I had been running a project at Sainsbury’s to develop a system to produce the picking lists – the information for the distribution depots to assemble and transport goods to Sainsbury’s supermarkets. I had achieved this and the system was up and running on our recently-delivered ICL 1900 series computers which replaced two Emidec 1100 paper-tape machines – the last such working machines in the country. I had installed data transmission systems to transmit the picking lists from head office to each depot, using the STC GH210 off-line equipment, running on private telephone lines at 2400 bauds – the fastest then available. ICL had nothing to offer to achieve this. But Sainsbury’s – at the time the biggest supermarket group in the UK and the retailer with the highest turnover per square foot of selling space in the world – had a problem. It was not a new one. All retailers need to have control of their stock so as to not have too much stock, as it tied up capital and space, but with a minimum of out-of-stock items. Getting this balance right was the key and the current system at JS was not working.

The amount of each item delivered to the stores each week was calculated by the large, central computer system, based on the previous sales and the current stock. The latter had to be counted laboriously by the supermarket staff, sent to the head office in London where the computer went through its complicated mathematical routines to produce the forecast requirements for the following week. The requirement – e.g. a number of packs of item xxx, say, baked beans, 10oz, in a pack of 24 tins to a pack – was calculated for every product, in each of the supermarkets. Not only that but that xx packs would be sent on say Tuesday and yy packs on Friday. Fine, but the forecast was invariably wrong and the delivery dates were often too early or too late. Stores were overstocked on some things and out of stock of others – for days. There can be no perfect forecasting system that copes with real life. For years the forecasting system had been tweaked by experts and used the Box-Jenkins factors among other routines. But the system never was, and never could be, satisfactory. The supermarket managers had no power or ability to do anything. Each supermarket had a backroom stock, and stock had to be moved from there to the selling area. Sometimes stock would be in the backroom but couldn’t be found and was not available on the shelf. Customers get upset when products are not available, and Sainsbury’s board was very sensitive to customers’ demands. A new system was desperately required.

A high-level committee was formed and I was appointed as the I.T. representative. I was the most junior, being the only non-senior manager/departmental director on the committee. Meetings were held and much head-scratching went on. Visits were made to Europe and the USA in the search for ideas. There was nothing that could be found that solved the problems. Eventually a ‘dream solution’ was described. Firstly, all the stock would be on the supermarket shelf – no backroom stock at all. This meant that the huge job of moving stock around would be obviated. A shelf space would be allocated for each product, based on an average two-days sale (this was smaller than that then used). At the end of each day’s trading, the staff would assess how much of each product was required to fill the hole now in the shelf space. This information would be sent to the head office and the requirement would be processed and sent to the distribution depots, assembled, loaded onto trucks and sent out overnight to arrive at the supermarket the next day. Staff would move the stock straight from the truck to the shelf. These days it would be called a ‘just in time’ system, then it was just a dream! If it worked the supermarket need never be out of stock as long as the distribution depot had stock, and there would be large savings in labour – no movement of stock from backroom to shelf. The main advantages would be reducing out-of-stocks and improving Sainsbury’s reputation but it was realised that another advantage was there for the taking. Sainsbury’s was so successful that a common complaint was overcrowded stores. If the stock backroom could be eliminated, then this space could become selling area!

The required system logistics were challenging to say the least. There was nothing on the market remotely suitable to capture data in several hundred remote locations, quickly, without error, and with the capability to transmit the data to the head office. There would be 250,000 item orders per day, and this was planned to increase dramatically. Even an error rate of 0.001% would result in a number of lost orders. Hence the call to companies and private inventors and the demonstration of the ‘breadboard’ model. Perhaps surprisingly, there were a number of vocal objectors to the proposed barcode idea. It was new, unproven, and the equipment to make it useable didn’t exist. I saw it immediately as a solution, if not the solution. Perhaps my enthusiasm backfired. I had arranged the demonstration as a means of getting the I.T. director on-side, and it worked.

I led an initial team of two systems analyst-programmers, although this quickly grew. A pilot system to test the basic philosophy was set up with a crude card system to collect data with supermarket staff calling out the results on a voice telephone to head office staff, who then completed a mark-sense card for input to the depot picking system. This proved that the basic idea of ordering sufficient product to fill a hole in the supermarket shelf actually worked, and that it was easy to provide sufficient information for anyone, with hardly any training, to assess the requirement. Suddenly the store could order products that were needed, deliveries were made within hours, and the shelves were re-stocked. Despite the drawbacks of the very basic system that we had put in place the board of directors was delighted with the results and demanded that we immediately expand the system to more supermarkets! I resisted this proposal, saying that it was a side show and would deflect us from the main game, but not surprisingly lost the argument. At one time we had 16 telephone operators at head office taking down the data from 70 supermarkets. It also was the means of establishing the changed distribution processes which was quite challenging for the depot management. In the meantime we were searching for the data collection solution

reading trolley
The Mark 1 Reading Trolley

At this time several companies had been trying to develop a successful barcode reading system. Several attempts had been made with no great successes and in 1971 there were no systems on the market. It wasn’t until 1973 that the Universal Product Code was agreed upon using an IBM barcode design but the uptake was slow. By 1977 there were only two hundred stores in the USA using the UPC codes as suppliers were slow to co-operate by printing barcodes on products. However, the UK Plessey company had developed its own two-bar code design earlier but had found no takers for it ... until Sainsbury’s came along in 1971. They hadn’t really any idea what to do with it, hence the reading ‘wand’ was on the cradle of a telephone handset, along with a dial! Yes, they could replace the dial with a keyboard, and yes, the barcode could be reduced in size. The barcode and the reading wand were reduced in size, the code could be read in either direction and reading tests showed that errors were so few, because of the cyclic redundancy check employed, as to be non-measurable. I wanted belt and braces though, so we introduced a further check – a modulus eleven check digit on the stock number that we would be using. As there was no printing process that we could acquire we had to purchase the barcodes then allocate them to the stock units, rather than using the Sainsbury’s standard I.D. system. Over the first several years of operation with many millions of barcode readings we never detected any reading errors. The data, once read, had to be recorded and a mini-cassette was used for this. Power was a problem as there were no small batteries available to do the job. Two lead-acid car batteries were used at first so making the portable device a rather clumsy trolley. The NASA space programme soon resulted in smaller batteries so it wasn’t long before the mark 2 unit appeared – a satchel unit. When lithium batteries became available the mark 3 unit became a hand-held device.

system diagram
The Full System Diagram (both images courtesy of Sainsbury's)

Plessey designed the unit to suit our requirement and designed a transmitter to be used in the supermarket to transmit the data over the standard telephone network. The transmission was sent when the head office called up the unit and response from the supermarket was all automated. At head office a receiving unit dialled the supermarkets, collected the data and created a standard magnetic tape that was transferred to the ICL computers running the picking list system. Each supermarket was equipped with two data collection devices and one transmitter. All the staff had to do was to plug the collection devices to the transmitter. Nothing else, so nothing to go wrong, but see below! Answering the phone, switching on, tape re-winding, sending the supermarket id (very important) was automated. Each transmission took a couple of minutes. As more supermarkets were switched to this system the calling up operation at head office was also automated, using a front-end processor that auto-dialled the supermarkets using 32 lines.

The very first supermarket to trial the system was a north London store in Islington. Barcodes identified every product place on the shelves. A part-time shelf filler was briefed and sent to assess what was required and capture the data. He achieved this without any questions or delays. The trial went on for some weeks and it did produce a problem that took a while to solve – sometimes part of the order was not transmitted. It was caused by static – when the trolley was wheeled close to a cooler cabinet a spark sometimes jumped as the trolley earthed to the cabinet. The spark produced a signal – a ‘rewind tape’ signal so all orders recorded to that point were lost! It had us all scratching around for some time before we found the issue. Another time a cleaner unplugged the transmitter to plug in her vacuum cleaner so the store unit was unplugged and could not respond to the head office call! These issues were quickly fixed and the system was an outstanding success, with stock-outs being virtually eliminated, stockholdings reduced and labour costs radically reduced, even more than had been anticipated. So in August 1972 an article appeared in The JS Journal – the company house magazine – that described the system that by then had been installed in 30 supermarkets and was rapidly being installed in a company-wide programme, overall 250 supermarkets. It was described by the company Managing Director, Mr J.D. Sainsbury as the greatest step forward that the company had made in its more than 100 year history, even more significant than the change from counter-service to self-service!

It took more than two years to prepare, equip and switch all the supermarkets to the new system. The directors then demanded that all product groups be included, not just the 2000 grocery products. These other groups had special requirements and the system was slowly extended to accommodate these. And today, 50 years later, you will still see shelf barcodes in every Sainsbury store, now numbering more than 1,000, identifying up to 10,000 products in each. Not to mention their appearance in almost every other walk of life!

At the time the application described here made quite an impact so that I was invited to give a number of presentations to interested parties, firstly to branches of the BCS then further afield. The UK retail trade picked up on it with presentations in London, followed by international conferences in Amsterdam and Paris.

So who would have thought it? A piece of technology still being used, no, seriously embedded, in today’s society, after 50 years! Is there anything else connected to the computer industry that is still in use after this long? Millions of barcodes are read around the world every hour of the day. And the first ones were installed and working in the U.K in 1972! So 2022 was the Golden Jubilee of the barcode!

A CCS seminar on this matter was given in January 2023. Go to ccsoc.org/jsbc1.htm where a link will be found to a recording of the event. This article is also scheduled to appear in a forthcoming edition of IT NOW.

Derrick was born in London and lives in Melbourne, Australia. He joined J Sainsbury where he worked from 1969 to 1976 in systems development. He then became Information Services Manager for Plessey at its Ilford head office from 1976 to 1979, running a profit-centre computer services business. Derrick is long retired. He has authored “An Unexpected Development” Austin Macauley Publishers – ISBM 9781787100398 (paperback) / ISBM 9781787100404 (e-book). The contemporary article he wrote for JS Journal can be found at ccsoc.org/jsbc0.htm.

Top Previous Next

50 Years ago .... From the Pages of Computer Weekly

Brian Aldous

Europe’s Met Centre in the UK: A £9 million weather centre equipped with a giant computer will be established at Shinfield Park, near Reading, following a decision on the siting of the centre, taken in Brussels on March 5th by representatives of 19 countries which have agreed to cooperate in scientific and technological research. When fully operational the centre will undertake medium term – four to 10 day – forecasts on behalf of the 19 countries, providing estimated savings of £80 million a year for those throughout Europe whose business is weather dependent. The centre will be an independent body, administered by the EEC Commission, but initially, for a transitional period of two or three years, it will rely heavily on support provided by the Meteorological Office at Bracknell which is equipped with a 360/195. According to Mr Patrick Meade, director of services at the Met. Office, the centre will be accommodated in the Bracknell area while the necessary buildings are being erected at Shinfield Park. During this period staff of the new centre will have access to the Met Office’s 360/95 and will make use of its worldwide telecommunications system. Mr Meade told Computer Weekly that the centre would require a far more powerful computer than that currently installed at the Met Office. Studies carried out by the European countries suggest that a system capable of carrying out 50 million instructions per second will be required. This compares with the eight to 10 mips capacity of the 360/195. (CW 15/3/1973 p1)

ICL launches addition to Key-Edit series: : A new key-to-disc system which is available with as few as four keystations at a rental of £370 a week has been introduced by ICL. It is a member of the Key-Edit series to be known as the Key-Edit 50. Like its bigger brother, which will now be known as the Key-Edit 100, it uses a PDP-8/E as the central processor, but with a maximum of 24K core. The major new feature is re-designed software, a necessary requirement if the core maximum was to be adhered to. One result of this is a multibatch grouping feature, said to be unique to the new system, which enables the supervisor to group up to 200 batches of the same type. The supervisor can list them, write them to tape or delete them with one command instead of 200. A basic Key-Edit 50 system comprises processor, memory, fixed head disc of 1.4 million characters capacity, a magnetic tape unit of 556 or 800 bpi, a supervisor console and four keystations. (CW 29/3/1973 p1) )

In-store PoS system installed: : With the opening of the new Bentalls department store at Bracknell. Berkshire, NCR has installed what it claims to be the first complete “in-store” point-of-sale terminal system in the UK. As reported exclusively in Computer Weekly, Bentalls has opted for the NCR 280 retail terminal and 45 of these units, together with two NCR 723 data collectors, have been installed. At present, sales data collected by the system is processed at an NCR computer bureau in London, but Bentalls has plans for an in-house computer, likely to be an NCR machine, before the end of year. The terminals, which each have a 512 byte memory, are programmable to enable a company’s specific transaction requirements to be handled. Also incorporated in each terminal is a small display screen, a magnetic tape cassette unit, a receipt printer, and a sales bill printer. It is anticipated that light pens will be used in the near future to read tags and labels marked with bar codes. This will enable management to obtain a more detailed sales analysis than at present possible. (CW 12/4/1973 p17) )

Automation plan for local radio station: : The London commercial radio station, Capital Radio, which is due to take to the air within the next 12 months as a general entertainment channel, is finalising negotiations with EMI for an automation system worth more than £20,000. Capital is acquiring an EMI 903 system equipped with a solid state direct access memory which can be directly interfaced with either a computer or teletype system. The 903 can provide a full 24 hours of continuous broadcasts – the memory being programmed to select on a time or sequential basis pre-recorded material. The operator, using a 16 position keyboard, can insert a complete broadcasting programme for a day in the memory, building it up on a minute by minute basis if required. EMI says that the 903 can be equipped with a plain language logging system which automatically produces a printed record of an entire broadcast schedule after transmission; this is considered particularly useful to verify the transmission of commercials. The system will cue all random access devices in advance by searching the programme time-file for up to six hours ahead of the material currently playing. (CW 19/4/1973 p3) )

2903 spearheads ICL’s drive to win new markets: : As first predicted in Computer Weekly on March 15th, ICL have launched an important new product aimed at the lower end of the market which, although compatible with the 1900 series, is eventually likely to become part of the New Range product line. Called the 2903, the new system. which is to have its first public showing at Hanover Fair which opens today, is a microprogrammed “soft machine” that currently looks like a 1900 Series system. It is disc-based, has video operator’s console and direct data entry capabilities, provides IBM compatible RPG2 as the main software language, as well as COBOL and Fortran, and has a wide range of applications packages available, including bill of materials, order entry, stock control. Pert and Prosper. (CW 26/4/1973 p1) )

Logica to sell ARPA know-how: : Access to the expertise which has gone into the development of the ARPA network – the world’s most advanced computer network – will become available, at a price, to European organisations. It follows an agreement between Logica with its French associate, SESA, on one hand and Bolt, Beranek and Newman Inc, on the other. BBN, a US consultancy with a great deal of know-how in the communications field, has been involved in the four-year-old ARPA project virtually since its inception and has been responsible for development of the communications protocols and the Interface Message Processor (IMP) and Terminal IMP (TIP) packet switching systems, on which the network is based. The network, which has been developed by the Advanced Research Projects Agency of the US Department of Defense, currently links more than 40 large and medium sized computers installed throughout the US, at sites such as MIT, Harvard, NASA’s Ames Research Centre – home of ILLIAC IV – Stanford Research Institute and the Rand Corporation. An offshore link to Hawaii has also been established and agreement has been reached on the establishment of a link to Norway with a connection to London University’s Institute of Computer Science. Under the new agreement Logica and SESA will have access to BBN’s communications know-how. Consultants from both companies will spend several months with BBN at Cambridge, Massachusetts, familiarising themselves with BBN’s work on the ARPA project, possibly studying future developments such as the transmission of data via satellite and the further enhancement of IMP and TIP systems. (CW 3/5/1973 p1) )

Banks sign for SWIFT service: : The international banking message switching network, SWIFT, achieved official status last week in Brussels when 240 leading banks in 13 countries from Europe, Canada and the US became founder members of the non-profit-making Society for Worldwide Interbank Financial Telecommunications. Based on switching centres in Brussels and Amsterdam with concentrators in 14 other countries, the SWIFT network will provide its members with a private international telecommunications network which will enable the banks to transmit international payments, statements and other related messages. The objective of the new service, which is scheduled to become operational in 1976, is to provide more convenient and reliable transmission capabilities than mail and telex facilities and to make possible reductions in the banks’ message handling and processing costs by the introduction of a number of measures of standardisation. (CW 10/5/1973 p1) )

Greater London Council extend London traffic control system: : As a further development of the GLC’s traffic control system, an additional 200 signal intersections are to be controlled by next March, extending the area of the system radially outside its present approximate boundary marked by a line linking the main London railway termini. The system has been operating since mid-1972, following a pilot project covering 70 signals in West London, starting with 23 signals in the Baker Street and Marylebone area. Computers for the system were supplied by Siemens under a £759,000 contract awarded in 1970 against two British tenders from Plessey and GEC-Elliott Traffic Automation. Two Siemens 306 computers, one with 32K and the other with 48K of core, were installed in 1972 at the GLC’s offices in County Hall. Both machines have now been moved to a new traffic control centre at New Scotland Yard, and the 32K 306 has been enhanced with a further 16K of core. In addition, a further four Siemens 81/52 display consoles are being installed. Work by Siemens is being carried out under the terms of the initial contract, which also planned for the control of 1,070 of the 1,500 traffic signals in London by 1976. Environmental control for a new computer suite at New Scotland Yard to house the Siemens system is being provided by Precision Air Control Ltd under a £33,000 contract. (CW 17/5/1973 p20) )

Memorex launches floppy disc drive: : A new OEM floppy disc drive, the 651 Flexible Disc Drive, has been announced by Memorex UK Ltd. This is an improved version of the 650 which was announced last year but was never strongly marketed. The improvements include a faster track-to-track access time of 10 milliseconds, with 10 milliseconds settle time. The average latency time is 80 milliseconds, and the transfer rate from unit to unit, 250 kilobits per second. Another new factor is the FD/IV extended life disc, giving a minimum life of 50 million passes, evenly distributed. Data can be written in either sector or index mode, starting with 132 byte records with 32 records per track (on 64 tracks), up to one record of 4,880 bytes per track, giving a maximum capacity of 312,500 bytes or 2.5 million bits. Applications envisaged include auxiliary storage, remote terminal data acquisition, data logging, key-entry recording, point-of-sale recording, accounting machine storage and programmable calculator storage. Other uses are as a microprogram loader in writeable read-only memories, and as a diagnostic program loader. (CW 17/5/1973 p20) )

Calcomp launches Model 7000 flatbed plotter: : A new flatbed plotter, the Model 7000, has been introduced by Calcomp Ltd as the most sophisticated system in its range. The company says it will provide drawing quality comparable to that of a skilled draughtsman at speeds of up to 42 inches per second, with a resolution of 0.002 inches. To overcome the problem of quality as good as that done by generating lines of constant width and ink density, a new pressure inking system using four pens has been developed. The four pen reservoirs have separate air pressure tubes. An electromechanical system senses plotter speed and regulates the pressure to each pen as a function of plotter speed. The pressure ranges from a vacuum to one high enough to provide uniform lines at 42 inches per second diagonally on paper or synthetic materials. The operator can select from a range of performance modes by a simple cartridge load. If he has a good tape, the quality mode assures him of line quality as good as that done by hand, or alternatively, the performance can be adjusted to suit scribing or film cutting. The operator can scale the plot by factors of 2, 1, 0.5, 0.25 and 0.125. (CW 1/3/1973 p5)

Top Previous Next

Forthcoming Events


Members and others are welcome to attend CCS Seminars but these are also available via Zoom.

London Seminar Programme

16th Feb 2023 The Mobile Revolution: the Early Days of Cellular, 1983 - 1993. A Personal View. John Carrington
16th Mar 2023 Programming an ICT 1900 – and how we wrote Maximop for QMC’s ICL 1905E. Arthur Dransfield
20th Apr 2023 Kathleen Booth – UK Computer Pioneer Roger Johnson
18th May 2023 QR Codes for Beginners Tony Davies

London meetings take place at the BCS — 25 Copthall Avenue Moorgate EC2R 7BP starting at 14:30. The venue is near the corner of Copthall Avenue and London Wall, a three minute walk from Moorgate Station and five from Bank.

You should use the BCS event booking service to reserve an in-person place at CCS London. Go to www.computerconservationsociety.org/lecture.htm . The service must be used for remote attendance.

For queries about meetings please contact Roger Johnson at

Manchester Seminar Programme

28th Mar 2023 to be advised
30th May 2023 Alan Turing’s Manchester Jonathan Swinton

Manchester meetings take place at The Manchester Metropolitan University, Chester Street, Manchester, M1 5GD – Room E0.07 in the John Dalton East Building starting at 18:00.

Details are subject to change. Members wishing to attend any meeting are advised to check the events page on the Society website.

CCS Website Information

The Society has its own website, which is located at www.computerconservationsociety.org. It contains news items, details of forthcoming events, and also electronic copies of all past issues of Resurrection, in both HTML and PDF formats, which can be downloaded for printing. At www.computerconservationsociety.org/software/software-index.htm, can be found emulators for historic machines together with associated software and related documents all of which may be downloaded.

Museums

Do check for Covid-related restrictions on the individual museum websites.

SIM : Demonstrations of the replica Small-Scale Experimental Machine at the Science and Industry Museum in Manchester are run every Wednesday, Thursday and Friday between 10:30 and 13:30. Admission is free. See www.scienceandindustrymuseum.org.uk for more details.

Bletchley Park : daily. Exhibition of wartime code-breaking equipment and procedures, plus tours of the wartime buildings. Go to www.bletchleypark.org.uk to check details of times, admission charges and special events.

The National Museum of Computing At present opening days are somewhat irregular so see www.tnmoc.org/days-open for current position Situated on the Bletchley Park campus, TNMoC covers the development of computing from the “rebuilt” Turing Bombe and Colossus codebreaking machines via the Harwell Decatron (the world’s oldest working computer) to the present day. From ICL mainframes to hand-held computers.

Please note that TNMoC is independent of Bletchley Park Trust and there is a separate admission charge. Visitors do not need to visit Bletchley Park Trust to visit TNMoC. See www.tnmoc.org for more details.

Science Museum : There is an excellent display of computing and mathematics machines on the second floor. The Information Age gallery explores “Six Networks which Changed the World” and includes a CDC 6600 computer and its Russian equivalent, the BESM-6 as well as Pilot ACE, arguably the world’s third oldest surviving computer.

The Mathematics Gallery has the Elliott 401 and the Julius Totalisator, both of which were the subjects of CCS projects in years past, and much else besides.

Other galleries include displays of ICT card-sorters and Cray supercomputers. Admission is free. See www.sciencemuseum.org.uk for more details.

Other Museums : At www.computerconservationsociety.org/museums.htm can be found brief descriptions of various UK computing museums which may be of interest to members.

North West Group contact details

Chair Bob Geatrell: Tel: 01457-868700.
Email: <
Secretary Alan Pickwick: Tel: 0161 973 6796.
Email:
Top Previous Next

Committee of the Society


Chair  Dr Doron Swade MBE FBCS:
Secretary  Rachel Burnett FBCS CITP Hon D. Tech:
Treasurer  Arthur Dransfield CEng FBCS CITP
Chairman, North West Group   Bob Geatrell:
Secretary, North West Group  Alan Pickwick:
Editor, Resurrection  Dik Leatherdale MBCS:
Website Editor  Dik Leatherdale MBCS:
London Meetings Secretary  Dr Roger Johnson FBCS:
Membership Secretary  Bill Barksfield CEng MBCS CITP:
Media Officer  Dan Hayton MBCS FRSA:
Digital Archivist  Prof. Simon Lavington FBCS FIEE CEng:
Awards Sub-Committee Co-ordinator:  Peta Walmisley:

Awards Sub-Committee

Rachel Burnett (Chair), Roger Johnson

Museum Representatives
Science Museum  Rachel Boon:
Bletchley Park Trust  tba
National Museum of Computing  Kevin Murrell FBCS:

Project Leaders
SSEM  Chris Burton CEng FIEE FBCS:
Bombe  John Harper Hon FBCS CEng MIEE:
Delilah  John Harper Hon FBCS CEng MIEE:
Elliott 8/900 Series  Terry Froggatt CEng MBCS:
Software Conservation  Dr Dave Holdsworth CEng Hon FBCS:
ICT 1301  Rod Brown:
Harwell Dekatron Computer  Delwyn Holroyd:
HEC-1  Kevin Murrell:
DEC  Kevin Murrell:
Our Computer Heritage  Prof. Simon Lavington FBCS FIEE CEng:
ICL 2966/ICL 1900  Delwyn Holroyd:
Analytical Engine  Dr Doron Swade MBE FBCS:
EDSAC  Dr Andrew Herbert OBE FREng FBCS:
Bloodhound Missile/Argus  Peter Harry:
IBM Hursley Museum  Peter Short MBCS:
Data Recovery  Delwyn Holroyd:
IBM 360/20  Adam Bradley:

Co-opted Members
      Prof. Martin Campbell-Kelly FBCS CITP FLSW:
      David Morriss FBCS CEng CITP:
Top Previous


science museum logo TNMoC logo SIM logo

Computer Conservation Society

Aims and Objectives

The Computer Conservation Society (CCS) is a co-operative venture between BCS, The Chartered Institute for IT; the Science Museum of London; and the Museum of Science and Industry (MSI) in Manchester.

The CCS was constituted in September 1989 as a Specialist Group of the British Computer Society (BCS). It thus is covered by the Royal Charter and charitable status of BCS.

The objects of the Computer Conservation Society (“Society”) are:

  • To promote the conservation, restoration and reconstruction of historic computing systems and to identify existing computing systems which may need to be archived in the future;
  • To develop awareness of the importance of historic computing systems;
  • To develop expertise in the conservation, restoration and reconstruction of historic computing systems;
  • To represent the interests of the Society with other bodies;
  • To promote the study of historic computing systems, their use and the history of the computer industry;
  • To publish information of relevance to these objectives for the information of Society members and the wider public.

Membership is open to anyone interested in computer conservation and the history of computing.

The CCS is funded and supported by a grant from BCS and donations. Some charges may be made for publications and attendance at seminars and conferences.

There are a number of active Projects on specific computer restorations and early computer technologies and software. Younger people are especially encouraged to take part in order to achieve skills transfer.

The CCS also enjoys a close relationship with the National Museum of Computing.


Resurrection is the journal of the Computer Conservation Society.
Editor – Dik Leatherdale
Printed by – BCS, The Chartered Institute for IT
© Computer Conservation Society
Top Previous


Valid HTML
   4.01 Transitional