The Jim Austin Computer Collection

Powers-Samas Tabulator


This is an example of a mechanical computer. It is probably dated before 1950, but I have no details. There is no name plate on the machine.

The machine is owned by the London Science Museum, it was lent to ICL in 1988 then given to the collection by mistake.  The machine is now on display in the Computer Science department at York University.

The machine was altered for display, probably by ICL. It came from Fujitsu in there offices in Bracknell, found and donated by Hamish Carmichael.

The machine is used to read and print the contents of cards. It can also add up the values on the cards and print the totals out.

A note from Dennis Hart on the machine...


"Your photo looks like a 36 column sorter in use till the early 50's and was replaced with the 40 column system the card size was about the same with extra columns and an extra row to allow for full alpha punching.I was a field engineer with Powers Samas from 1951 to 1971 having worked on 21 col 36 col 40 col 65 col and 80 column systems then untill about 1990 I worked as an independent engineer servicing the old mechanical equipment for customers who were not ready to go electronic.Your photo of the tabulator looks like a 65 column model again replaced with the 80 column and early electronic equipment I think the 40 column systen out lasted the 65 & 80.I converted many tabulators from £.s.d. to decimal on the early 70's.

 I do not know any way to accurately date the sorter. Check to see if it is a 36 column. lift the glass cover lid over the sensing pins and on the left chrome bar there will be a scale 1-36 or 1-40 as I said if it is 36 then these went out in the early 50s the 40 column range I worked on as late as  1990/91-but have now long since scrapped all my parts books and operators manuels. If the card has wrecked under the sensing pins U must wind the pins up by hand so that they are clear with small driver inserted between the feed rollers-to open them up slightly use a wreck knife (which I dont suppose you have) to remove the card I might still have one. These old machine ran at about 400 cards per min the later ones at about 650 so the feed settings have to be very accurate-the card pick up knife 4.5 thou and the gap for the card to go through abouit 9.5 thou but why did it wreck in the first place dirt incorrect feed settings etc. Many public libraries all over the country used these sorters for keeping check on lending books"

Dennis Hart 13 July 2008

The following is a copy of an email from Robin Hill, who used this type of machine and nicely places it into context:

"The machine is quite complex. It takes cards in from this side, the cards are shown in blue in the middle (see below). These come out on the other side of the machine.

I started vac work with Vickers Armstrongs (Aircraft) Ltd at Weybridge in 1953 (while at London Univ '53-'56).   After graduation in 1956 I joined them full-time as a statistician in the Design Office before helping set up an Organisation & Methods Department.  Attended at that time was what I believe was the first 8 lecture evening course in systems analysis and design at City University - using Fortran examples. Some time after that F R Brown gave a week's course at Oxford on  Mathematical Forecasting as Scientific Management began to emerge.  It was in the later '60's, using Cobol and a rented ICL 1901 machine with 8k (perforce enhanced to 16k!) magnetic core memory that workable factory scheduling systems were implemented for Birds Eye Foods (the company's first computer application). 
The aircraft industry were heavy on office automation and mathematical calculation and particularly receptive to new inventions because of the science-based nature of the business, not to say the predominance then of cost-plus defence contracts!  In the 1950's/'60's Powers Samas was a wholly-owned subsidiary of Vickers Group, the subsidiaries in Aircraft, Shipbuilding and Engineering were therefore handy test-beds for Powers-Samas  equipment with their heavy data volumes and extended coding structures.  I transferred to Powers Samas Accounting Machine (Sales) Ltd from Weybridge as a sales engineer in August 1958, located at the principal London Sales Office at 88 High Holborn, which also housed the main Powers Training Centre.  It was rather more than a year later that Powers Samas was subsumed into ICT (later ICL).  I left to join Peat, Marwick Mitchell as a consultant and became an early member of the British Computer Society in about 1961.
If memory serves Hollerith had invented punched cards for use in the US census in the 30's,  based on a card containing 80 vertical columns of thirteen positions in which the figures 0 - 9 alone or, combined with three top edge overpunching  positions, gave alpha-numeric representation. (I think their over-punches were labelled X,Y and Z, but can't now be sure).  This equipment was marketed by British Tabulating Machines (BTM) in UK. It relied upon rectangular perforations allowing 80 electrical brushes to make/break contact as each card fed top edge first.   The columns of the card were allocated to 'fields' of data which, with a descriptor, were usually pre-printed onto the card.  Each job thus had its predefined dataset comprising one or more cards into which the data were punched. When 'read' by machine, electrical connection prompted actuation of standard machine features such as a printhead or counter via a removable code panel using double-ended leads to set up appropriate connections to translate the sensed codes. While ostensibly offering flexibility for connection changes, such was the complexity of wiring needed for most jobs that pre-plugged panels were kept for all but the simplest routines.  By their nature the mechanical components of the Hollerith machines were relatively expensive to produce, with squared holes which needed to be broached.  Positions allocated needed to be sufficiently spaced and sized as to allow good electrical contact without interference. In contrast, no doubt for patent reaons, but also perhaps reflecting British and Vickers mechanical engineering heritage, 'Powers' machines opted for round perforations and mechanical machine actuation. Round holes could be more cheaply drilled, and closer proximity achieved without interference. This meant that the standard Hollerith-size card could be used to carry 90 data columns - a marketing plus.  Later this was further enhanced with the successful introduction of intermediate punching effectively giving 180 columns.  
The Tabulator was, of course,  central to any punched card installation - the central processor.  The front card hopper on the Powers Tabulator fed cards top edge first into a machine track between drilled blocks containing as many rods as there were hole posions in the card.  This permitted rods to be forced down wherever coincident with a punched hole in the card. This was achieved by locating bove the reading block, in contact with the tops of the matrix pins, a removable Y-shaped  'connection box' (equivalent to the Hollerith plug board) which was hard-wired spcifically to the job. The box had at the base as many rods as were needed to read the positions within the used data fields, so that, when forced down, appropriate features of the machine - printheads, counters or control links were physically set as a reaction to the moving tops of the connecting box rods. Thus while many connection wires were straight-through, some sensed holes needed to allow multiple actuation, while some multiple code-punching needed to be combined to achieve a single purpose.  Designing the system including 'programming' the tabulator was the sales engineers job, while soldering the  'conn-box' forest of cranked rods to meet the design requirement was down to the skill of the Powers Engineer who was thus the doyen of the machine room.  
Each machine cycle first fed a card, next forced down the rods and pins to set  the mechanisms (e.g. throwing up a hundred or more foot-long character bars), then executed the settings (e.g. simultaneously banging hammers on  print heads to force them onto a typewriter ribbon to impress onto the paper beneath), and then feed out the card as the next fed in.  Given all the metal on metal, the cranking of hundreds of rods, and hammering of printheads,  noise levels in a machine room with a dozen or more tabulators (plus sorters etc.) can perhaps be imagined. Mechanical wear and tear was heavy requiring constant engineering maintenance, while card consistency and handling were also critical - extracting a 'card wreck' from the machine and trying to piece it together sufficiently to be able to reconstitute that data was the constant fear - bad enough on tabulators, but worse on high-speed sorters where one hiccup could destroy many cards and several hours work.  A special flexible steel hook-ended wreck knife was available (a must for your collection!) A later tabulator refinement was the replacement of mechically thrown character bars with multiple wire character imaging printheads to increase speed and reduce noise levels and wear.  Later tabulator models also sought to replace soldered rod conn boxes with bowden cable boxes to allow greater flexibility for change, but by then electronics were already having an impact. 
The Powers tabulator could be fitted with an ancilliary mechanical device called a Summary Punch.  This was wheeled up to and attached to the back of the Tabulator so that when totalling was actuated a mechanical connection actuated the Summary Punch,  punching the results into a fresh card.  Up until that point multi-stage processing of data  necessitated tabulating all cards several times with or without intermediate resorting.  With a card for each line entry this could be a massive burden in high volume applications. A further development, and the first step towards stored program computing, was the Powers Electronic Multiplying Punch (EMP).  Instead of punching a mechanically derived result relayed from a Tabulator, this was able to read and retain factor data from the lead card of a set fed into it conventionally and apply it to variable data in succeeding transaction cards via valve-based electronics, punching the results into those cards while also summary punching identified intermediate or tralier cards of the set.  However, the basic speed limitation of machine cycling remained. 
During the mid '50's and '60's Powers had begun addressing electronics more seriously at Whyteleaf in Surrey.  Its first attempt at a stored program computer was the Program Controlled Computer (PCC).  The first machine I think went to Bristol Council, while the second came to Vickers at Weybridge in about 1957/8.  Only machine code was available and programming was by means of copper rivets installed into a removable fibre board  which  I seem to recall allowed about 22 program steps.  I/O was punched cards with processed data recorded to a magnetic drum, which involved timing each read/write instruction.  Like the EMP this had valve electronics but resembled a diesel rail engine with a cooling system which sounded like Niagara Falls. The machine was never likely to be viable.  In parallel, Stress engineers at Weybridge  were making some progress with an early Elliott machine applyiedd to mathematical computation.  I know no details except that it was possibly an 803 or it may just have predated that model. 
For tackling the vast potential of the beckoning Electronic Age Whyteleaf had turned its attention to grander things and a sales briefing meeting was organised in about 1957 to equip us to sell the Atlas - a 'proper' large-scale computer design which, despite the glamourous artists impressions, so far as I know was never actually constructed.   Unlike the ill-fated Elliott 405 of the time of which one was actually installed at great expense at Watford for Radio Rentals, but this never processed a byte in anger before being trashed.  Aside from pioneering analysis and programming difficulties, a serious limitation for it, as for all machines then was I/O. 80 column punched cards were the preferred choice for pre-loading programs but punched paper tape quickly superseded punched cards for keyed data using teleprinter 8 channel tape.  Teleprinters were also much commissioned for punched data output to feed line printers (developed from tabulators) which  tended to be buffered off-line. The introduction of magnetic media and internal storage quickly transformed this scenario as did such techniques as print spooling.
Returning to punched cards I seem to have heard that Powers had at one time tried the introduction of a half-length card of 45 columns but cannot be sure.  It did have a downsized (in both dimensions) card in the '40's with 36 columns, but this later changed to 40 columns. This was particularly successful in business, rather than engineering, situations, especially accounting and retail markets, which were transaction-volume rather than data-complexity driven.  Its commercial success waned as the electronic revolution dawned, but it was greatly loved by staff and customers alike.  So much so that somewhere round 1960 IBM (successor to BTM) tried to introduce a 36 column range (System 11?) but was unable to bring it successfully to market. 
The most lowly piece of equipment in the punched card range  (after the wreck knife) was the hand-punch, a small flat-bed device with a ratcheted sliding bridge carrying button-topped spring-returned dies - one for each of the thirteen punch positions, with a tangential ratchet-actuating button.   Cards were inserted singly with one hand as the other drew back the bridge to the first column,  the keys were then operated with fingers of the second hand in rapid combination ending with the ratchet key to move the bridge to the next column.  This required battery-hen mentality, extreme manual dexterity and enough finger strength to push each die fully home (otherwise  'hanging chads' resulted which we all know about from Florida!)  
The automatic punch was a sit-at machine which fed cards in succession automatically from a hopper along an extended bed end-to-end, while the keypad caused electrical actuation of the dies reducing the danger of imperfect perforation.  A typewriter-type tab mechanism enabled skipping over fields where all were not required to be punched and for carriage return. Some female operators were capable of quite suprising punching speeds.  I cannot now recall particular records but they were perhaps double the planning average used when calculating equipment needs which, I seem to recall, was 11,000 key depressions per hour - and repetitive strain injury was unheard of in those days! (not to say though that it didn't occur).
Sorters were piano keyboard style machines with a hopper to the left feeding under a read head onto a roller-bed which carried cards over the top of hoppers at high speed. On feeding, the sensing of a perforation caused the appropriate trapdoor to flip up ready to 'capture' the card as it reached that pocket, unperforated cards dropping into the last hopper.  Sorting was in ascending order of column of the field containing control data - one pass for each numerical column, or two for alphabetical or coded data. Pick-up accuracy and ability to remember just where one was in multiple sorts were, of course, vital - especially if a high volume sort was entailed.   
These recollectionsl may be incomplete or faulty in some respects as quite a few years have rolled by, but I will examine your site more thoroughly, and search my (often consolidated and moved) belongings for any remaining relevant memorabilia, and perhaps respond again in a more considered way."
Robin Hill, Norrköping, Sweden, 2005.


Here are the photos of the machine when  got it:



The machine is run from a motor in the box at the bottom. The handle on the right (above) is used to hand crank the machine.

The box with the rods in (see below, right)  is the programming box for setting up the machine.



Above left shows the printer mechanism.

The picture below shows how complex the machine is.