This is Ceefax (1975)

Towards the end of 1974, the BBC television service in the UK began experimental test broadcasts of Ceefax. Ceefax was first developed by the BBC in 1972 as a text based information service, which was encoded into the vertical blanking interval of the standard analogue TV signal. The generic name for Ceefax is of course teletext.

Originally it was intended to provide a subtitling service for the hard of hearing, but it was soon realised that it could provide a multi-page “magazine” of news and other information.

All this was achieved before the time that 8-bit microcontrollers appeared on the market, so it was done with help of minicomputers and custom hardware.

The video below is a colour public information film produced by the BBC in the early summer of 1975 to announce the new service. Part 2 follows later.

Teletext produced a colour text and block graphics display of 40 characters wide by 24 lines. It used 3 bits for the colour, to give 6 colours plus black and white.

Wikipedia has more technical information here:

The experiments at the BBC would almost certainly be done on a minicomputer, with custom created equipment to insert the teletext into the video signal.

In one shot of the Ceefax control room there is an ASR33 teletype.

A paper tape of the page contents was produced in the editorial office and then rushed down 2 floors of Television Centre, so that it could be loaded into the “core store” in the Central Apparatus Room.

This article from 2012 describes some of the history of Ceefax at the BBC.

This one provides a slightly different technical perspective.

As a 21 year old electronic engineer, straight out of university I joined BBC Research Department, some 11 years after this film was made. Some of my departmental colleagues appear in the film.

The BBC disbanded their Ceefax service in 2012 - 40 years after its inception.


From 1980:

Gwyn Morgan, CEEFAX publicity and promotions: 'Our old computer, Esmeralda, has gone. She was really rather rudimentary and not very reliable. In her place we have Selene, which is three computers linked together. One handles material produced by the team of CEEFAX journalists. It also provides them with a library which can store up masses of information for future use. The other two computers act as the broadcasters. We really need only one, but if anything goes wrong we can switch over to the spare.

'The other important development in the last couple of years is that on the receiving end the manufacturers have improved their techniques to such an extent that they are now offering CEEFAX sets which cost only about 50p extra in rental a week. They are also beginning to compete in advertising these sets, the result being that, whereas a year ago there were only 7,000 sets in use, in January this year there are now well over 40,000. What’s more, people are using the CEEFAX service for an average of 25 minutes a day - a fair time.

I wonder what (mini) computers these Esmerelda and Selene were.

I think we probably have to mention the BBC Micro too, with the memory-efficient Mode 7 using a teletext chip. Some people will recognise the font as teletext, and some will recognise it as BBC Micro.

Here’s a version of the Engineering Test Page in your browser.

There’s an ongoing reverse engineering of the Beeb’s teletext chip here.


I know that Research Department had a PDP-11 ( I nearly salvaged it from the skip in 1993).

I am also seeking information on “Esmerelda” - I guess she may have been a DG Nova or something else of that era - rack mounted. I’ll as some of my former BBC colleagues if they can shed any light.

The BBC has a rich online library of it’s engineering archive. Here’s a mention of Ceefax - back in May 1973.

1 Like


“Esmeralda” was apparently a PDP-11 and supported 6 text input terminals, that were manufactured by Aston Electonics - who specialised in character generators that could be gen-locked to the TV signal.

Aston Reference: Aston Broadcast Systems

Reference to Esmeralda being a PDP-11

Wikipedia:Reference desk/Archives/Computing/2010 September 11 - Wikipedia.

Use of computers by BBC for the 1977 Election (where Aston got their big break as live on-screen text entry terminals) - Page 10


Aston became major producers of character generators from 1974, but they initially were CAPS only.

An earlier development for a character generator that would produce “pleasing” characters on a viewers TV set was called ANCHOR. It was designed and produced in-house by BBC Designs Department. The BBC had a history of producing a lot of their equipment in-house before putting it out to contractors.
Many of the equipment suppliers were set up by ex-BBC engineers!

ANCHOR did not use conventional pixel graphics, say as an 8x8 matrix. Instead it synthesised them from lines, curves and slants - generated from analogue signals. These in turn were masked by ANDing with a sutable blanking signal. Thus an O could be turned into a C by blanking a hole on the right hand side, and a C could be turned into a G by adding the horizontal bar, Slanting characters like K M N W could be synthesised from sawtooth waveforms.

The system appeared in 1970 and was used extensively throughout the early 1970s until more capable character generators appeared. Page 15 for article on ANCHOR.

Page 10 for interesting article on BBCs central London computing facility.

That is one aspect of the earlier days of computing that seems inconceivable these days: many companies, institutions, whathaveyou, had their own research/engineering departments which did groundbreaking work. These days, IT departments just order Windows x86 PCs for the desks and pay AWS bills for the cloud. Lack of ambition. Or just mindless cost-cutting, more likely. Concentrate on the “core business”.

I think economy of scale, and commoditisation, have a lot to do with it. In 1971 you might build an in-house CPU from logic gates but in 1981 that would be folly.

One thing about character generation is that the broadcast media had much higher standards and expectations than early computer users: an 8x8 matrix really wasn’t going to look good compared to televised artwork. So, for 16x20 characters, you need either relatively large ROM (large for the time) or some ingenuity, such as Mullard’s (later Philips’) rounding logic.

Broadcast engineering presented a lot of new challenges, and the 1960s and 1970s were periods of great developments and expansion such as the introduction of colour TV in 1967, and the requirement for standards converters so that different nations could receive and view TV pictures from all around the world.

Some countries adopted the NTSC system from North America, some adopted PAL, from Europe, and some countries like France and her colonies had their own SECAM system. To convert from one TV system to another required a lot of racks of custom built equipment.

The Apollo Program, the Olympics of 1964 (Tokyo) and 1968 (Mexico) were the driving forces behind electronic standards converters. Previously 16mm and 35mm film had been the means of transferring program images from one format to another.

With the advent of cheaper and larger semiconductor memory in the 1970s, line stores and frame stores eventually became practical, allowing digital sampling and interpolation of video information.

Minicomputers were just becoming sufficiently useful for research into image processing techniques from the late 1970s.

My department had a PDP-11, followed by a VAX and then a pair of MicroVAX machines, before moving towards “3M” workstations in the early 1990s. The VAX had custom built semiconductor framestores interfaced to it so that short sequences of images could be processed.

When I joined in 1986, there was fundamental research into HDTV coding and decoding techniques putting a high demand on the computing resources.
As an HDTV picture contained at least 4 times the pixel data as a standard definition picture, how do you compress the bandwidth sufficiently for broadcasting without visible loss of image quality?

Image processing routines were written in FORTRAN for FFT and DCT algorithms. Small 64x64 pixel blocks of monochrome images would be batch processed overnight, or even over the weekend - such were the demands of the coding tasks.

Eventually by about 1988/89 an experimental HDTV CODEC had been built from real hardware, capable of running at realtime video speed - it took up about three and a half full height racks.

The CODEC was asymmetrical by design, most of the complexity was put into the coder - which would be a large expensive machine owned by the broadcasting operator. The decoder would be considerably simple and cheaper, and would be built into the viewer’s set top box.

The BBC’s role was to try to maintain picture quality for the HDTV viewer, against a lot of competition from the European set top box manufacturers, who just wanted the decoder to be commercially viable, and boost lagging TV sales.

I remember there was an atmosphere of distrust between the BBC engineers and their European commercial partners.

Out of this research from 35 years ago, came the various MPEG video compression standards, now universally used on all computing devices.

Once only possible using several racks of custom TTL and ECL circuitry can now be performed on a fingernail sized ARM device costing a few dollars.


The Ceefax rabbithole beckons…

This is a podcast from 2016 discussing the early days of Teletext.

On the panel were Mort Smith who was the first BBC salaried employee of Ceefax, the world’s first teletext service and Peter Kwan who has been working with teletext since 1999, and has been involved in television much longer.

There’s some good technical detail from 28:00 onwards

Ceefax formats have been adopted by “Teefax” an online group for teletext enthusiasts.

We mustn’t forget telesoftware either!
The Teletext Museum - Telesoftware

I’m not sure if the Basic is specifically BBC Basic, but I suspect so. There’s an archive here of a collection of disks of telesoftware. Disk 8 contains, among much else, the game GLOOP which can be played in-browser here:
Super Green Gloop Gulper (AKA Gloop Gulp)

The original teletext screen was 40 x 24 characters or 960 bytes.

If each character was a bytecode instruction - you could have a ~1K program.

If each character was a call to a routine, you could have a 1000 routine program

1 Like

Hee hee - almost a One K challenge… but is teletext 8 bits or only 7 bits per character?

Channel 4’s teletext output was much less sober. But Bamboozle, the quiz, was family-friendly, and here is a possible nod to retrocomputing:

Speaking of retro, we should perhaps mention today’s Block Parties, annual celebrations of visual art in the medium of teletext. See also here and previously:

Teletext Blocktober (and 2020 block party)

Edit to add this teletext art links page.

I stand corrected: The original Teletext Computer was an “LSI-2” made by Computer Automation programmed by “Logica” It had a hard disk ( perhaps 10MB I cannot remember.

1 Like

With that clue, I found this, which might be relevant, about the preparation of subtitles for presentation on teletext:

The subtitling of television programmes to aid deaf viewers, using teletext, has begun on the U.K. BBC channels. The BBC has been developing methods of mixing words with pictures using teletext. Full subtitling, captioning and Plantype shorthand transcription methods are being used. Teletext is used so that large minorities of viewers can be offered a service that the majority will want to do without. In all the methods used the updated message frame can be inserted anywhere into the normal sequence of Ceefax pages which cycles every 23 seconds. Some three million viewers who are either deaf or hard of hearing are expected to receive the programmes.

For example, in a recent programme about 200 messages were prepared by rewriting from the transcript of a programme commentary and scheduled for transmission with the aid of the Telecine machine already used for automated subtitling for foreign films. Based on a Computer Automation Naked Mini LSI 220 with 16 Kilobytes of store and two floppy discs the Telecine system outputs to the Ceefax Digital Equipment Computer PDP-11/34.

(Contemporary Cybernetics. (1980). Kybernetes, 9(1), 1–7. doi:10.1108/eb005536)

1 Like

If “Esmeralda” was a LSI-2 she was replaced in late 1979 by “Selene” that consisted of three PDP-11/34 units, when Logica were called in to develop a future proof system with some system redundancy and a larger character set to cover most EU languages.

1 Like


IIRC the BBC tried an experimental early version of telesoftware that had a “bit-modulated” flashing graphic character block in the bottom right hand corner of the screen.

It would transfer a fairly generic BASIC source code file. You needed a photo-diode and a suction cup to stick to the correct place on the screen, to download the over-air file(s).

I think this was first shown on the Computer Programme around 1982-83, I was still at school then.

I’ll have to trawl Youtube to find a link to a video.

EDIT - Getting my memories confused - see later reply for more correct information.


just FYI I wrote the BBC Micro Teletext software for Acorn. I still have the source code on a floppy somewhere. I also wrote some of the programs that were distributed over the air by the Telesoftware service. One was a search engine that scanned all transmitted pages in parallel, and another was a utility to draw a teletext page in one of the bitmapped modes so that a page could be printed out using any screen dump software (which at the time never handled MODE7 graphics).


telesoftware wasn’t limited to a single screen’s worth. It was encoded using a protocol that spanned multiple subpages and programs could be of arbitrary length. That said, the protocol design was absolutely terrible and in terms of ISO layers, spanned about 3 layers! It could have been considerably simpler if it had been designed by someone who knew what they were doing (eg atob or uuencode format).

Thanks for bringing that up: I remember that too, but any attempt to find a reference to it online has come up blank. Think it may have been demoed on Tomorrow’s World


I raked through my fading memories, and I believe that offscreen download was a service provided as part of the Channel 4 program “4 Computer Buffs”.

They refer to it at the beginning of this program from 1985

(You also have an interview with a rather prickly Clive Sinclair on this episode).

A lot more information in this clip:

There is also discussion of the “light pen port” on YCombinator here:

As well as “4 Computer Buffs” - I am sure that the BBC transmitted a similar software download scheme on possibly “Making the most of the Micro” or possibly “Micro Live”.

1 Like