• If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • You already know Dokkio is an AI-powered assistant to organize & manage your digital files & messages. Very soon, Dokkio will support Outlook as well as One Drive. Check it out today!

View
 

TAC and PANIC

Page history last edited by Alan Hartley-Smith 3 years, 5 months ago

Computers

Home

TAC

To give it its full name Transistorised Automatic Computer, used in both air defence and industrial roles. It is an outstanding example of the capabilities of the staff of the Research Division, in particular for those who knew him, of Digby Worthy.

 

   

 

Datasheet

 

Wikipedia

 

Archive

 

Closely associated with TAC were the S3101 and S3102 Tabular Display equipments for the presentation of changing alphanumeric and symbol data either as a direct read-out from the computer or controlled from a manual keyboard. See descriptive article here. The data display capability was also used to provide "marked" PPI radar displays.

 

The original application in the Fur Hat air-defence application is covered in this article.

 

Wylfa Head

 

The Wylfa Computer Room in June 1970.

 

It has recently come to light that the ex-Wylfa TAC located at The National Museum of Computing at Bletchley has been restored to run a test program - there is a slide show of the work to date and a detailed exposition here. There is an appeal for assistance with continuing the work to complete restoration so if anyone can help please contact  . There has been a mention of the long operational life in a recent BBC item.

 

 

Steve Kay would like to thank all of you that supplied your knowledge and information about the TAC coax connectors used for the I/O interface. I have now identified them as being a variant of the military Pattern 12 connector and have sourced nine plug/socket pairs from electrojumble.org. Roy, the website owner tells me they were used on several projects when he worked at NPL, including Alan Turing's Pilot Ace computer. To replace the photographs I previously posted showing the coax connectors, here are some photographs of the TAC at TNMoC.

 

 

 

The text on the A4 description reads:

 

Marconi TAC

 

The Marconi TAC, or Transistorised Automatic Computer, was designed by Marconi in the late 1950s as a fast general purpose computer for realtime applications.

 

It was followed up by the MYRIAD, before the Marconi computer interests were merged with Elliott in 1968 to form Marconi Elliott Computer Systems, later renamed GEC Computers.

 

This machine, one of only five built, was used at Wylfa Nuclear Power Station on Anglesey to handle the monitoring and alarm system. It was installed in 1966 and provided continuous service until early 2004. It ran non-stop from 1968, which could make it the longest running computer in the world.

 

The TAC was originally fitted with a magnetic drum storage device, display screens, and Friden Flexowriter electric typewriters.

 

The TAC has a 4096 words of core store memory, each word being 20 bits long, and used paper tape for program loading.

 

Unusual for the period, the TAC was a micro-coded machine with a complex instruction set. Addition takes 22 microseconds and multiplication 92 microseconds.

 

Wylfa operated a second identical TAC. Computer 2 is now preserved at the Jim Austin Computer Collection.

 

 

 

I wrote a simple memory test program - here's a listing plus video of it running (the incrementing display shows the memory address currently under test, the test pattern is set on the switches):

 

;***********************************************************

;* Marconi TAC Program                                     *

;* Input from console switches defines test pattern        *

;* Start location hard coded as 0000050 in location 0014   *

;* Stops on error and location is output on console lights *

;* Delay added to display memory location incrementing     *

;* Steve Kay 25 August 2015                                *

;***********************************************************

0000 0 67 0010  Jump to start of program

0001 0000000  Variable storage (test pattern)

0002 0000000  

0003 2 67 0000  Stop and jump to 0000 ready for restart by pressing single step key

0004 0000000

0005 0077777 Variable storage for Delay counter

0006 0000000 

0007 0000001 Decrement value for Delay counter

0010 0 76 0001  Set RSR to 0001 Console switches

0011 0 02 0001  Input to 0001 from Console Switches

0012 0 76 0002  Set RSR to 0002 Console Lights

0013 0 72 7777  Count next location up to 7777

0014 0000050 Current test location

0015 0 27 0000  Clear AB

0016 0 10 0001  Fetch location 0001 to A

0017 0 71 0014  Modify next instruction with location 0014

0020 0 14 0000  Store in location 0 from A

0021 0 71 0014  Modify next instruction with location 0014

0022 0 12 0000  Subtract location 0 from A

0023 0 03 0014  Output from 0014 to Console Lights

0024 0 51 0003  Jump to 0003 if AB is non-zero

0025 0 27 0000  Clear AB

0026 0 10 0005  Fetch location 0005 to A

0027 0 12 0007  Subtract 0007 from A  This command and next provide a delay loop of around 1 second

0030 0 51 0027  Jump to 0027 if AB non-zero

0031 0 55 0013  Else Jump to 0013 if B trigger not set

0032 0 67 0010  Else Jump to 0010 to repeat outer loop

 

TAC_videos 003.mp4

 

Here are some photographs of the TAC at TNMoC with the doors open. I have annotated them to show the functional areas. We are looking at the rear of the cabinets here, so you can see the backplane wiring.

 

 

 

 

 

Most of the PCBs do not plug in, they are hard wired. This photograph shows the backplane where the wires come from the PCB, pass over a "comb" to separate the wires and then solder to tags.

 

 

The high speed logic boards used micro-alloy germanium transistors - mainly type DAT1A like this one:

 

 

These were made by Semiconductors Ltd., a joint venture between Philco in the USA and Plessey in the UK. These were the first transistors to be manufactured in a fully mechanised process. A young man named Clive Sinclair heard that Semiconductors Ltd. were disposing of large quantities of out-of-spec micro-alloy transistors into landfill - he offered to buy them for 6d. each in boxes of 10,000. In the end he bought 1.5 million of them - these were all tested, graded and labelled by him and two other people by hand. He sold them as MAT100, 101, 120 and 121 transistors at either 7/9d or 8/3d each or supplied them in his kits. The full-spec DAT1A transistors supplied by Semiconductors Ltd. were priced at £1 13s. 4d. each in 1964.

 

Here is a PCB - unusually this is a plug-in type. It is a high speed logic board with six 3-input gates. These are DTL (diode transistor logic). You can see the white diodes in groups of three and the gold microalloy transistors. The black moulded components labled NLM are Non-Linear Elements (or Modules). It seems that these are a cross between a thermistor and a varistor and protect the expensive microalloy transistors from thermal runaway and voltage surges.

 

 

University of Manchester - John Rylands Library

 

Jim Austin Computer Collection

 

PANIC

Input from Don Ward

One machine seldom mentioned was PANIC (Periodic and Noise Interception Computer), used on the PD system. A TAC was installed at Malvern for the trial PD system. PANIC was a development from TAC but was reduced from 20 bit to 18 bit words. The magnetic core store was also reduced from TAC's massive(!) 4096 words to 2048 words. This was the RRE's idea to reduce the cost! It was designed specifically for PD and never used anywhere else. The design team included Derek Canfield and Brian Mellor and others.

 

Computers

Home

Comments (13)

Ian Gillis said

at 4:13 pm on Feb 15, 2016

Page checked

Chris Mower said

at 1:28 pm on Aug 14, 2018

Not sure if this is the same machine but the system that I learned to program on installed at RAF Staxton Wold was programmed using a language called PANEC (not PANIC). It was part of the Passive Detection system and from what I can remember it looked very much like the system in the photos. There were three consoles with two active and one arbitrator. Mind you this was 45 years ago so I could be wrong, but they do look similar.

Tony Bartoletti said

at 11:02 pm on Jul 1, 2019

Chris is correct, the computer was called PANEC.. "PANIC" is what we did when the thing wouldn't work.
During RX12874 training at RAF Locking, we were informed PANEC was derived from the Marconi Myriad. Myriad I was a desk computer whereas Myriad II was a two cabinet rack mounted computer as were PANEC computers. I do not recognise the internals of the TAC cabinet and conclude PANEC was more likely based upon the Myriad II.

Our PANECs at RAF Staxton Wold (there were three of them) each had a ferrite core store of 4096 18 bit words. PANEC was an octal computer with the six MSDs being for instruction and the 12 LSDs for data.

Our training notes referred to PANEC as the GP for general purpose computer, hence the computer hardware was not classified.

Ian Gillis said

at 8:46 am on Jul 2, 2019

Note from Peter Bain in reply to Tony Bartoletti:
I'm sorry to contradict you but I believe Don Ward is correct in what he said, he was an expert in PD. I also believe that PANIC predated MYRIAD I and MYRIAD II came even later. I have never heard of PANEC, was this an RAF name?

Ian Gillis said

at 8:48 am on Jul 2, 2019

Note from Owen Hawkes in reply to Tony Bartoletti:
As someone who installed and commissioned PD alongside Don Ward on all UK sites, and spent 1969 supporting the RAF on all sites, I can assure you that Don is correct. The PANIC computer started life with only a 2K store and had to be upgraded when it was realised extra code was required during flight trials. PANIC was built using only discrete components with no micro circuits and definitely preceded the MYRIAD.

Tony Bartoletti said

at 11:06 pm on Jul 1, 2019

Correction "six MSDs being for instruction and the twelve LSDs for "address" not data.

Ian Gillis said

at 9:02 am on Jul 2, 2019

Note this para from https://marconiradarhistory.pbworks.com/w/page/53908166/Computer%20Equipment%20for%20Display%20and%20Data%20Processing :
One machine seldom mentioned was PANIC, used on the PD system. A TAC was installed at Malvern for the trial PD system . PANIC was (I think) a development from TAC but was reduced from 20 bit to 18 bit words. The magnetic core store was also reduced from TACs massive(!) 4096 words to 2048 words. This was the RRE's idea to reduce the cost! It was designed specifically for PD and never used anywhere else. The design team included Derek Canfield and Brian Mellor and others. The mind boggles at what we did with a 16K Myriad compared with today's memory.

Tony Bartoletti said

at 10:48 am on Jul 3, 2019

So, using terminology of back in the day, we were given "duff gen" as to the lineage of PANIC/PANEC was derived from TAC and not Myriad, coupled with my assumptions based upon the loss of 2 bits and cabinet form factors.

Owen Hawks confirmed PANIC/PANEC originally had a 2K store that was later upgraded to 4K which fits with my PANIC/PANEC Block Diagram showing a 4K store. We repaired PCBs to component level and I recall soldering in new transistors when gates failed.

Steve Kay of TNMOC provided this information of TAC that accounts for 20 bit vs 18 Bit.

The 20-bit word of the TAC is used for instructions like this:

0 0 000 000 000 000 000 000
Stop Check Order Address

Question: On a project of this size and given there would only be nine production examples, would the reduction in store and lopping off 2 bits really have reduced the cost by any significant amount?

I still cannot account for PANIC vs PANEC, but in training at RAF Locking, I recall our instructor Roy Street if any of you knew him, emphasized it was not PANIC. In the RAF it was always known as the PANEC computer although we seldom used this name, but referred to a specific computers by name CC, TC, or DFC.

Question: Why did Marconi call it PANIC? Is it an acronym or code name, because it is a bit of and odd name when TAC is an acronym of the machine function and Myriad meaning a great number?

Can you confirm the name of the programming language as we only knew it as machine code?

Store Test
C.C.

A
30 0010 :0000
34 7777
76 0012
05 0011
76 0112
06 0001
62 7777
02 0000 :0010
00 4013
END

It has been fascinating talking to you all and thanks for letting me into to your circle, and updating me with the correct information.


Ian Gillis said

at 3:50 pm on Jul 4, 2019

Reply by Don Ward 4/7/2019: "I have only just read all the chat about TAC and PANIC. The logic used in both of them was the same. TAC boards were wire ended PANIC boards had pins which were connected to the chassis pins with wire wrap links. The PANIC name came from Periodic and Noise Interception Computer, there was no name for the computer language, it was a simple code ie A=A+ FRED where the compiler would allocate a memory location to FRED.
Because of the limited memory I had to write a two computer compiler where the compiled code was assembled on one and the dictionary of addresses was constructed on the other.
Myriad was long after PANIC!
Don"

Tony Bartoletti said

at 3:03 pm on Jul 5, 2019


Ian please thank Dan for his input.

From time to time, problems arose with the PCB edge connector that were resolved by re-seating the board or gently cleaning the contacts with an eraser. Very rarely we had to replace a wire wrap.

We loaded the main programs for CC, TC, and DFC computers from the Sperry Drum Store. Gentleman please hold your breath... it 's capacity was a massive 1 Megabit or about 55K words!

I recall a head failed in the drum where we had to resort to loading programs from punch tape. No one had any idea how to repair it as we were only told how to read programs from the drum, but not how to repair it. After obtaining a new head and with your support, we repaired the drum by the classic British workman approach of one airman doing the work with around 10 onlookers.

Failures were always hardware and we never has to diagnose code problems. Rather than trying trouble shoot a problem using a main program, we would look at the error conditions and write a short test program often by direct input from the front panel. I recall a problem with CC, that I diagnosed to a failure of LS, AB 6 and it was only 6 that failed other values were fine. I traced the problem to a PCB connector pin where I cleaned the one pin that resolved the problem.

That was a great Control Panel you built for us as we able to diagnose problems to the Micro Instruction level. The PANIC panel layout had some differences to TAC noticeably the Modify Triggers switch and Micro Instruction step switches.


Tony




Ian Gillis said

at 3:35 pm on Jul 6, 2019

Comment by Alan Matthews in reply to Tony Bartoletti:-

Those old PCBs 6R3s 8R2s etc. often had connection problems –particularly when taken in and out of the equipment during commissioning..
This was, I believe, because over time, dust collected on the gold plated edge connectors which was disturbed when the board was removed and fell between the contacts when the board was re-inserted. The standard “fix” was as you say, to use a green rubber to clean the gold plated edge connectors on the board. The problem then was, if this was done several times the gold was damaged which made matters worse!!
To get round this in the 60s wire wrap was used. This provided great reliability but was a real pain when 32 wrap connections had to be redone if a board was changed.
Scheduling the wiring was also a bit of a nightmare because only 3 wraps could be put on a pin and if say a lower wrap had to be taken off then the other two and their wiring to other boards had to be replaced –a sort of domino effect resulting in long repair times.
When designing the rack wiring the above problem had to be taken into account and wrap counts made for the pins.
Fur HAT used mostly wire wrap connectons and once working had fantastic reliability – most of the system having virtually no faults for about 30 years.
Later on proper, but more expensive, connectors were used on both the boards and in the racks and these stopped the dust getting on the contacts.

Tony Bartoletti said

at 3:04 pm on Jul 5, 2019

Sorry Typo "Don"

Chris Mower said

at 8:54 pm on Dec 20, 2020

Is it possible that PANEC was the RAF name for the programming language? As I said earlier it was a long time ago but the name PANEC seems stuck in my recollection of the front cover of the language manual that i used.

You don't have permission to comment on this page.