Hello and welcome. My name is Russell Trafford-Jones and thanks for joining me. This is Broadcast Focus, the show where we look behind the scenes of broadcast media and entertainment industry by chatting to the people that make it happen. This is a small independent channel, so thanks to everyone who supports it by liking, subscribing and of course watching. We cover both technical and non-technical topics, so as long as it relates to media, it will be here. Now, today's topic is inextricably linked with broadcast. It's a unique form of distributing information in a pre-internet age. To tell us more and to dig into the details, here's Peter Whitesall. Hi Peter. Hi. Welcome to the show. Thanks for spending some time with us. It'd be great fun to talk about what we have done and how we've been broadcasting digitally to homes in the UK and across Europe since the 1970s. I think the more that we dig into this, and of course we're talking about teletext, about CEEFAX, however you know it by. I've done a number of jobs, mainly in the projects area in the BBC and in November 1989, I took up a job managing a team of people that were looking at what we would refer to as the central areas, the MCRs and that sort of thing of broadcasting. One offshoot into that was looking at teletext, which had been going since the mid-1970s. Actually, on my first day, I received an invite to a BBC CEEFAX relaunch. I showed it to my boss and said, well, I've had nothing to do with it, should I be going? She said, well, yes, go along and have some good fun because it will be the last thing you'll ever do for CEEFAX. I then spent the next 15 years or so off and on being involved in teletext alongside all sorts of other things. You mentioned how things from a simple system can move to a more complicated thing. Certainly my experience and the things that I'll be talking about me doing in the CEEFAX teletext area helped me greatly when we came in, actually, strange to think about it, year 2000-ish to work with audio and video over IP, both packetised systems and working where you've got different signals of the same type, but all carrying different information muddled up. So let's have a look at what we're talking about and make sure that we're all on board. So first of all, this is what it would have looked like on a TV. Let's just get the names straight. So we've got CEEFAX and teletext. What's the difference? Back in the early 1970s, the BBC research department, I think mainly sort of blue sky research and what do we do with these lines on the television picture that we're not actually carrying picture? And they thought of a system of putting digital data in there. And it was a matter of seeing facts. So that's why it's called CEEFAX. Actually, at the same time, the Independent Broadcasting Authority developed something very similar to that. The two ideas merged together, involved others, and they had a system called Oracle and we had CEEFAX and the overall technology is called teletext. So the aim of it really is to deliver information on a TV. You press the teletext button and up instead of the picture would come this, which is the index page. And you just type in the numbers of what you want to. We're on page 100, so it says top left there. So if we go to 101 and it's searching through the pages, so to speak. And here we are. So if I want to look at any of those news articles, then I would select that number. So 104, etc, etc. And then if it is a long article, it can cycle through different pages. We have a look at that. Let's have a look at some other things to see what else the service has got on it. And so this is a simulation or is this because it's not being broadcast now, is that right? This actually is rather clever. It is. The data, I believe, is being taken from the BBC website. Yeah, because actually one thing that CEEFAX did was you don't have very much text that you can put on the page. Therefore, you've got to write succinctly. So that actually has been carried over into the website. So it's picking that information up. It's then formatting it in teletext format, and it's then putting it out in the vertical blanking interval of a signal into another computer that pulls it up and remembers it here. Ah, yes. Now, as well as text, there is a variety of graphical, simple, chunky graphical characters it can do. Therefore, you can produce a weather forecast. And here you can see that we've got a multi-page set. We've got three pages, as you can see, by the three of three at the top right there, which are all transmitted on the same number. So it's got these variations that we can see how the weather is going to change over the next few hours, few days. We've got a 501. Oh, yeah. Entertainment. 501 was sort of an electronic version of radio times-y type information. So entertainment news. And then the most viewed pages other than 101 is 601, which is BBC One listings. In practice, there actually is another magazine called Seven, which was held in reserve by the BBC. It was used for other things like UCAS, the University Clearing System. You could get all that information out to all the anxious schoolchildren at home instantaneously. It was used for that. And of course, the most well-known, well-displayed teletext number was 888 for the subtitles. And we'll talk about that later. But let's concentrate on the text system. So the purpose of this is this is going to be one of a number of episodes where we're just going to look at how it works. We're going to go through this PowerPoint presentation was originally put together for SMPTE, was it not, Peter? Yeah. To celebrate the 40th birthday, I put this together because it was a technology which everybody was sort of familiar with, but was not anything like the current technology. Now 10 years further on, and hence a bit of a demonstration of the front, which just showed people what it actually looked like. As I said, the BBC invented CEEFAX, the IBA invented Oracle. It was the first occasion when the broadcasters and the industry got together to formulate a fairly complicated specification, which we now see and we just take as normal for DAB or for DVB or things like that. Yeah, indeed. So you've already alluded to the fact that in the unused video space, so to speak, before the actual picture, there's gaps. And this is one of those things that can sit in that area, isn't it? Yes. But it's all standard definition system. So you have 65 lines of which 576 used for picture. That leaves you with about 50 lines which are not occupied. And roughly speaking, 32 of those can be used for this data. Because I think we were mentioning a bit later, some of the lines right at the beginning and right at the end of a field are full of equalising pulses and similar things, which mean you cannot put data on it. So basically, it's all very simple. One packet of information per television line and one packet has five bytes of status information, a payload of 40 bytes. So we've got 40 characters per row of text. 40 characters on a line, do you say? Yeah. So the first three bytes are a clock running and framing. And then the next two bytes are a magazine and row address. So have a look at what the MRAG is, because it's quite interesting. We've got eight magazines, magazine zero becomes magazine eight. So we run one to eight rather than zero to seven. Is magazine like the block of 100? Yes. The magazine is the hundreds character. So that is coded with three bits. And then the row address, that is for each row going down the page. So at the top, there's a header, which is full of lots of information like the service name, and usually a clock or something like that. Packets or rows, one to 24 are the page, one to 23 being the normal text. And then there's some additional information normally dropped on the bottom. Packet 26 doesn't exist. And we then are in a realm of packets that are not directly displayed. So packet 26 is actually used when we've got complicated language sets to add diacritical marks for instance. Packet 27 enable us to tell the receiver to cache various pages to link to them. Packet 28, if you really want to get into that, is all to do with changing colours and then totally nothing to do with text, but the coding space was there. The last two slots of packet 30 and 31 are used for data broadcasting rather than text broadcasting. Okay. So you can hide some extra data in there. Yes. The one thing to note is that the page number is only defined in the header. So we have got a very large coding range, in fact, more than 100 pages per magazine if we really wanted to. And the way in which we signal those one of three, two of three, three of three pages is by using a subcode, which means that you can actually get an awful lot more uniquely identifiable pages in than just the plain 100. And of course that was famously used, was it not, the ability to have hexadecimal pages for quiz games. Yes. Channel 4's text service had a lovely quiz called Bamboozle, which was probably deemed to be as destructive as most children playing games on their phones now, with children sat in front of the television answering various quiz points on that. The header, most importantly, defines the page address, i.e. the tens and units side, because actually the magazine and row address does not define what the page is. So you have a header followed by other packets that pick up and assume the page address. If you're controlling the page, there are the control bits, like you can erase it. There's a type of page called a newsflash. Just think of that these days as push notification to your phone. You could set your television up that if a newsflash came in, it would be displayed. In the BBC service, that was either on page 150 or page 151. There are things to do with suppressing the header. There are times when that might be useful. If a page has changed, it's useful to signal that, so that if the receiver, if you have a television set that actually happens to store the pages, you can flag the fact that it's updated. We'll talk about how pages are transmitted, but if there is an interrupted sequence, you can signal that, and that also has effects into the receiver. And for some peculiar reason, you might want to inhibit the display. This is actually useful on some uses which have gone out of favour, where we were actually transmitting software, a BBC basic programme, and you didn't want that to be seen. And then the two ways of transmitting magazines, and then it can hold a set of different national character sets. Therefore, for instance, the French character set does not have the at symbol in it, and mapping the hash sign and the pound sign, and adding different currency characters in, as well as diacritical marks. So it really is an international system. So that's what the page looks like. A displayable text over 24 bits. Here you can see the graphic we used to make big letters saying BBC Facts, and an important point there in red. If you lose the header, the page will be overwritten. That means that you do need a fairly good transmission route. But this isn't going to work too well? Yes. A lot of the time with the typical reception we got, actually it wasn't too bad. It's just if you lose it, it will tend to garble the text. One of the reasons the page number is more protected is that we don't really want to use it. And then the navigation off the bottom, the fast text, as it was called, that enabled the receiver to cache six pages, which could then be shown with text to say things like next page, or news. You can see it at the bottom there. It says headlines, sport, regional TV, A to Z index. So instead of pressing the button, typing in the number, you just press red. Yes, which is where the red button comes from. It was felt that, particularly as at the time very few vertical blanking lines were being used for CEEFAX, it was all a bit slow, and therefore the idea to cache a page was a good thing. And therefore, the idea of putting coloured buttons on your television remote control came in, which is where, as I said, we get the red button. That's why there was a red button to press. Nearly always the next page. But let's have a look at the display characters you can get. It has RGB and compliments and black and white as the colours available. So you've got eight colours available. Red, green, blue, cyan, magenta and yellow. You've got a large text character set. It's like old typewriters. It's non-proportional, which means that every line is equally long, no matter what different, if it's awful of I's, it would be the same length as if it's awful of M's. And then you have got the ability to run the mosaic characters, the graphics characters, in both a joined up form and a separated form. So we've got separate or contiguous mosaics. You can flash the characters as well. And for things like subtitles and actually how the newsflash worked, you could define a box that would be cut into the programme video. So this box study thing is at the bottom? Yes. On this test page, what they actually had there was very clever. They defined the box and they actually alternated between two pages. One had the word steady and one had the word flash. And the flashing one would flash. It was actually determined by the television set, which was a bit interesting. That test page actually had got was very useful because the divide sign and solid blocks are all ones and all zeros. So that actually could test out the fact that if you've got any errors, you could see immediately what was going on. There's always a great debate as to what your sampling frequency should be and locking things to the horizontal and vertical television waveforms. The whole thing works out of we're actually running a waveform which is close on seven megabits per second, which is sort of beyond power frequency allocation. And it also is longer than a active line, a full line with 64 microseconds and blanking with the sync pulse in it was 12 microseconds. So this actually extends outside what is normal active picture, which is why there's a lovely wording shown there in the spec, which says the two leading data ones may be absent or reduced in amplitude compared to subsequent data. But actually it's what happens in Rec.601 SDI. And then they defined levels of 0 being 0 volts and 1 being about two thirds peak white adjusted so it didn't cause too much buzzing and other things on sound or interfering with the picture. And then 17 lines per field ending up with, if you use 600 lines or thereabouts, you have got a two and a half megabit per second signal, which is something that we actually did use on one occasion. Full field text actually was used in Singapore as part of their Trestel system. You could have your own terminal dial up, but the signal came back to you, not down the telephone wires, but over the air. It really is a very flexible system. There are two ways in which you've got a stack of pages and you want to transmit them. The way in which things started off was a simple way, magazine serial, where you had a heap of pages and you just transmit them as fast as you could on as many lines as you got available. It meant that every page came up with about the same frequency. There were things like perhaps transmitting page 100 more frequently, but basically the idea was that you just had all the pages, equal access time. When you had a few pages, that was not too bad, but as you got more pages in, it got more complicated because the editorial staff would obviously like people perhaps to get the news more frequently than they would get the football results or something like that, which is why in magazine serial, you might transmit magazine 1, 2, 3, 1, 4, 5, 6, etc. and just balance things out. The way in which almost all teletext services in the world have broadcast is magazine parallel. The packets in each magazine are allocated to a number of VBI lines on which they are output. Thus, depending on how many pages in each magazine and how many lines to transmit them you have given per magazine, you can determine how long it is between the two transmissions of the page. If you're quite good at mathematics, you can actually set those cycle times quite closely. The target number was always about 14 seconds, which was about the time it took you to read the page. So, if you're running on a multi-page thing like the television listings, you've sort of read the whole page before it automatically changes. Basically, if you think about it, you've got 24, 25 rows of text. So, that's one line per field, that's one page a second, you've got two fields, so that's two pages per second per line pair. So, if you know that, you can judge how many pages you've got. So, if you're transmitting eight pages a second and you've got 80 pages, that's once every 10 seconds. There's another little trick in there, which is if the row's all spaces, you don't transmit it. And that actually makes quite a noticeable difference to how fast a transmission actually appears to run. Yeah, because there are quite a few pages that aren't very dense in that respect, particularly the index pages. Yeah, well, yes, and you've got all of this sort of thing going on where the index pages you want to transmit more frequently. On a page like that, you see you have got one, two, three, you've got half a dozen packets that won't be transmitted there. It's things like that that just give that little bit of edge to what's going on. While things speed up, we then hit the other major feature that slows things down. And we'll talk about this probably in part two, the page erasure interval. After a header, there must be a field interval before other displayable packets. That came from the early days when it took the television set a finite amount of time to work out what page you were wanting to have a look at and then setting itself up to capture that page. What that means is that if you have, let's say, you have got a 26-packet page and you're going out on eight VBI lines, you've got three fields, that's 24 packets, absolutely chocker, all packets used. And then the next field, field three, has two packets used and six not used. That is not very efficient and it slows things down. And we'll be hearing in part two how I managed to find ways around that to improve the speed of HeliText. And actually, there's a slight trick there in that non-dispayable packets can actually be put in at the same time. And again, packet 27, which is the one that describes the numbers for the fast text RGB buttons along the bottom, if that's transmitted early, that gives the television set time to pick it up. And just to mention two areas of the data broadcasting. The first one is to do with data associated with the television service. Packet 830 Format 1 has two main bits of information. One is the name of the service or actually the name of the programme and United Coordinated Time. Packet 830 Format 2 is what's used for programme delivery control, which is the way in which video tape recorders were turned on and off automatically. They basically take up slots every 200 milliseconds. So we have Packet 830 Format 1 appearing exactly on the second transition and then 200 milliseconds later and later and later the instructions to control television sets. And then there is data broadcasting. Think of this as a radio modem. The great use for this was things like the betting prices into bookmakers or the transmission of Duff credit card numbers. In fact, the service level agreement the firm that did the credit card numbers had was that between the Barclay card or Visa notifying the company to every single outlet having got the information that credit card was fraudulent, stolen or whatever, was two minutes. In fact, various outlets were actually closed down because they couldn't get a decent television signal to receive that. So let's talk about two sort of standard bits of equipment that we put in the video path to add teletext to the signal that's going typically to the transmitters. One is a thing called an inserter. That basically is those days a serial line with some data on it and it then inserts the properly formatted data on the right lines of the video that you've got coming in so that you've got video coming in, data coming in and video and teletext going out. The next bit of equipment is a thing called a data bridge where rather than inputting data, you input a video with teletext on it in the first place and therefore you then add that to the video that's coming in and then there'll be video with teletext going out. Obviously, actually, if you can control what VBI lines you're putting the teletext on, you actually can have teletext on the video in and you just add more teletext. These are the two main things but for convenience actually in most systems a data bridge was what is being used because once you've got the teletext it's very easy to just connect everything together using standard video rather than data circuits. We've got a manual here for a teletext inserter. That's a very, very ancient... 1883 copyright. Yeah, yes. Well, that's where that particular company MRG started and in fact where it comes the address on it is the owner and managing director's house which shows how like so many of these things it is broadcast manufacturers certainly relatively recently have all been fairly small sort of almost cottage industry. MRG Systems which was the manufacturer of most the teletext processing equipment I would use and actually provided the system for teletext insertion and systems for teletext limited. Aha! Got this bit on teletext display codes. Yeah, yes. These are the graphics, are they? These are the graphics alongside the alphabetical codes and with the first two columns being the control codes for turning on and off graphics and putting flashes and start boxes and similar things there. Interesting. Much loved. Basically I'll be mainly concentrating on transmission systems but obviously there's an editorial system which was in the olden days just a very boring terminal to do graphics. You actually would type in a letter knowing that you'd just before it to put in a control character saying change the display technique between alpha and graphics. So in this case you put in an A, do you, and then it becomes a two dot? Yeah, basically if you put in a small A you get a top left bottom right out of the six blobs that made up the graphics character. J was very useful which is the right hand side of the screen and that's how it was programmed. As teletext grew so very popular and because of it being used in almost every single television set the display chips were readily available and therefore when the BBC microcomputer came along that enabled the viewer to use a very efficient way of displaying basic text for programming which also had the advantage that the teletext editors could have BBC micros to do their editing which meant that they could actually do stuff reasonably rapidly and then when PCs came along you were able to actually almost scan an image in and then get the computer to make it look beautiful and there's an awful lot of teletext art around. So having got these nice pages and the like how do we actually deal with this issue around magazine parallel operation and how do we allocate lines to magazines in the VBI? So if we have a look at what the VBI actually is, the vertical blanking interval, the lines above the picture extends from line 6 to line 22 in field 1 and 318 to 335 in field 2. You've got very very used to adding 313 to the number that you first thought of. Obviously having 17 lines does not fit very well into computerized systems therefore most of them work on line 7 and 22 and they're complements. So we've got 16, typically 16 lines to play with and some decoders don't work on certain lines. This actually was one thing that had come about because in the 1980s in particular the companies making equipment tended to look at what was being transmitted rather than the then specification to work out what they should be doing and there were some issues around the use of those extra three lines which is why modern specifications got rid of it. Yeah, we're talking about the bit at the top. Okay, so we've got these 16 line pairs to play with. What have we got? Well, the data broadcasting, whether it's packet 830s or straight packet 30/31 information, text pages, which is the main CEEFAX or Teletext Limited's magazine, subtitles on page 888 or something similar. The maximum number of subtitles that we've ever put in one vertical blanking interval has been 13 different languages because we've actually got the ability to do about 15 different language character sets and BBC for one of its international services was using 13 of those. But there are other things that lurk in the vertical blanking interval. Test signals, timecode, you need quiet lines on the transmission system to see how much noise it's picking up. There's ghost cancellation signal when we're using PAL Plus and all sorts of other things. So actually, you haven't got as many lines as you think for the use of Teletext. And then of course, how do you divide up? And we talked about this right at the beginning, news, stock exchange, TV listings, and the BBC using for UCAS or for things like election results. And when you were doing things like that, you would actually wind down the amount of space that you are allocating to other things so that certain services would get through quicker. Okay, now this was something that the 1984 Broadcasting Act put in. It had the concept of an additional service, which was independent from the programme service. So CEEFAX and Teletext Limited were nothing to do with what was going out. So that independent of EastEnders or the news or whatever. So you've got data broadcasting, BBC have used the trademark data cast, and you have ancillary, which is associated with the programme service. So subtitles, programme listings, and programme delivery control. It says there on ITV and Channel 4, the ITV programme company had a line pair or so of space that it could put out its own programme listings, usually provided by a company called Intel Fax. So at the end, you were carving up individual lines and kind of giving them to different... Yes, I mean, we will come on to, I think, probably, maybe in part two, how we actually carve it up. But this whole idea of additional and ancillary services was very, very important, because it maps into what happens in digital television, where you've got a whole host of ancillary services, subtitles, audio description, very intimately related to the programme service, while the rest of the digits sort of flow around that, being totally independent of what programme is being transmitted. And so, given the six magazines and eight VBI lines, how do you match it? Now, if we've got 80 pages, and these are the basic ways in which we do this. If you've got two line pairs, the 80 pages run 20 seconds. Three line pairs gives you the magical number of 14, which we all quite liked. And perhaps going up to four line pairs is a bit too quick on 10 seconds. So that's how you begin to start to build up your thing. Now, if you've got six magazines, each of eight pages, that does not easily fit there. I can combine the number of pages and magazines into a stream, so that I can have a stream with magazines A, B and C, each having those numbers, making up, say, 60 pages. And then two line pairs, four pages a second, cycle time's about 15 seconds. You would actually transmit that, as I've shown there, with about once, at least once a second, you had got a page of the particular magazine that you were looking for going on. That's because the page head has a time on it, and it looks very funny if you don't see that if you transmitted lots of A's for, in that case, that would be two and a half seconds, the clock wouldn't move on any of the other pages. And that looks strange. So you've transmitted one page, one magazine, next magazine, perhaps the same magazine, and the next magazine, so it all rolls by. It also means that people feel there's a much more equal cycle time. This is where it all begins to get that little bit more complicated, and how pages are transmitted. But that's really where I think it's useful to draw a line for the basics of what teletext is, how it is used, the fact that it was so novel, the fact that you come in and turn on your television set, and by the time you got back into your chair and pressed the text button on your remote, you've got the news headlines up. Yeah, or the programme guide, of course. So the first EPG, so to speak. That's a very interesting topic, because actually, it was a form of data broadcasting in teletext that was the first international standard for electronic programme guide. And we had a lot of fun explaining to other broadcasters what an electronic programme guide was, even though they might have it in teletext, again, to do with the amount of information that's carried. Teletext in the UK was a great success. And for instance, Teletext Ltd, were actually responsible for well over half the holiday bookings. You didn't go to the travel agent, you went and sat at home, looked at these multiple sets of pages coming through, and phoned up the travel agent's call centre. And that really was how it was impacted, and being a very great commercial success. That one year, Teletext Ltd paid more corporation tax than the budget for CEEFAX. So we've been benefiting from a website. I want to bring up a second website now, and we'll just give credit where credit's due. So this one was an interesting one. It's somebody who's been trying to recover teletext from VHS recordings. And it kind of gives a glimpse into not only the jumbling of characters that you mentioned at the beginning with this one, but also you can actually see what it looks like. And you can actually spot the differences in the restored characters where it's been incorrectly done and correctly done here. So that's been really interesting. That's zxnet.co.uk. But the one that we have been looking at is the NMS CEEFAX. Nathan Media, it really is preserving what teletext looked like. And there's a Raspberry Pi that produces teletext, which is designed by a chap called Peter Kwan, who is one of the software writers at MRG Systems. And it's all quite good fun and quite simple. So do have a look, nathanmediaservices.co.uk. You can try it out for yourself and read more about it. I'm sure we'll see this website again next time we talk. I just want to give credit where credit's due there, because it's really good. Talking of credit where credit's due, I want to say thank you to Fiona and Becky from PageMelia PR, who help keep this channel going and the episodes coming thick and fast. If you enjoyed what you saw today, then please let Peter know by clicking the like button. Subscribing is always good. There'll be plenty more on its way. We'll be picking up again shortly, and we will continue talking about the details of teletext, putting it together, and helping preserve the knowledge. Because as Peter has alluded to, many of the things that were done are being done again in a different way, perhaps at 800 gigabits a second now. But nonetheless, many of the principles are the same. And it will always be the case that what we are doing now will be relevant in the future too. So thanks Peter for your time today and I look forward to speaking to you soon. Bye.