SO YOU WANT TO BUILD A STANDARDS CONVERTER…………
Some thoughts and reflections on generating 405 line signals.
No-one who ever tried it said that making a 625 to 405 converter was easy. That privilege is reserved for armchair designers.
HALL OF FAME
Several people have climbed the hill and built converters. It all started with the professionals.
When BBC2 started on 625 lines in 1964 there was an obvious and pressing need to convert TV signals from 625 to 405. The only method available before then was optical conversion. This is just a rather grand way of saying that you point a 405 camera at a 625 monitor. It does work, and remarkably well if you use good equipment. It was used in early international exchanges starting with a pioneering 1952 relay from Paris to London. After BBC2 started the BBC planned to originate all programmes on 625. Initially 625 to 405 conversion was done at Television Centre but later all distribution was done at 625 leaving conversion to 405 to be done at each transmitter. This needed a reliable all-electronic solution that could run unattended for long periods.
The first BBC analogue design was demonstrated in 1963 and was developed into the rather prosaically named CO6/501. It was a daring and complex design, dividing each line into 576 elements, each of which was stored in a separate inductor/capacitor network. A pair of electronic commutators scanned round these storage elements. An updated version, the CO6/501A, was also built by Pye for the ITV network. By the way, these converters are huge, each occupying a pair of 6ft high rack cabinets. The same BBC teams also designed the incredibly complex 525 to 625 analogue converters in the late 1960s. The extra complexity is due to the different field rates which meant storing a whole field rather just a single line. The whole converter occupied seven rack cabinets.
By 1970 it looked as if the 405 service would outlive the original analogue converters. Digital video techniques were then becoming feasible and the BBC Research Department demonstrated an experimental model in 1971. This was the basis for the CO6/509 which was less than a quarter of the size of the CO6/501 and much more reliable. Modern digital chips at that time included 4 bit TTL adders and in particular some fast 256 bit shift register chips that could be used in sets of six to store a whole line of video. The design was still very complex, particularly the analogue to digital converter (ADC) which was a large board full of very hairy circuits.
These digital converters were employed at BBC transmitters until the end of the 405 service in 1985. ITV somehow soldiered on with the analogue converters and I suspect that the 405 pictures were getting pretty bad towards the end due to maintenance difficulties.
The IBA developed an experimental digital 625 to 405 converter at about the same time as the BBC. The design was conceptually very similar too. It was used temporarily to feed the channel 9 transmitter at Croydon. Its real purpose was to prove the technologies for the famous DICE - Digital Intercontinental Conversion Equipment. This was the first digital converter between 50Hz and 60Hz TV systems.
Then there was a long gap. Nobody really needed to design a new converter until enthusiasts wanted to run their 405 sets after the shutdown of transmissions in 1985. By this time a converter would be relatively simple for professional designers and would become increasingly so into the 1990s. Unfortunately the potential market for such converters would not really justify the attention of a commercial company so the field was wide open to enthusiasts to have a go. Unfortunately, even highly competent enthusiasts rarely have access to the skills, techniques and components that are often taken for granted in the professional world. In compensation they do have plenty of dedication, initiative, ingenuity and patience.
His was the first amateur design. Sheer hard graft and innumerable hours of work went into this converter. When he started, the vital video analogue to digital converter (ADC) was available as a single chip but they were scarce and expensive. His converter has evolved over several years. All done with hand built boards full of TTL. I take my hat off to him.
He started his converter after David Boynes and probably had more experience of design techniques but his several boards full of hand wired TTL would not be fun for anyone to copy. I remember documenting this design for publication, fully realising that nobody was likely to replicate it! This design still stands as the only amateur built converter with a 4 line interpolator.
Jim Daniels and the Pineapple
The Pineapple converter arrived at about the same time as the Dinosaur. I would not claim to know which really came first. This was a deceptively simple design with several cunning features. Uniquely it stored a full frame of video which allowed tricks such as freeze frame. The interpolator was a brilliant concept. The main complexity of an interpolator is in the multipliers. These used to be either complex or expensive to do digitally. This design used a pair of digital to analogue converters (DACs) and did the interpolation very simply in the analogue path.
Dave Grant, Mike Izycky and the Dinosaur
The Dinosaur converter became the one to have. As with the Pineapple, the designers of the Dinosaur had the benefit of modern components. They made and sold a compact and thoroughly engineered package that could be used reliably by any enthusiast. Alas no longer available and secondhand units are sought after.
Some people had the affront to complain about the price of the Dinosaur converter. The only reason why complex consumer electronic equipment is cheap is because it is made by the million. When you are doing total production runs of 10 or even 100 the one-off costs of PCB layout loom large. And components are much more expensive in dozens than in thousands, let alone millions. Add in the hand assembly and test and I can assure you that the Dinosaur converter was a bargain. Dave and Mike would have done better financially by keeping their money in the bank.
Malcolm Everiss and Domino
Since the extinction of the Dinosaur there were several years when it was not possible to buy a converter. Several people have talked about doing designs for sale but only one has emerged as a product. At the time of writing this (July 2002) I have the first production unit on my bench for review. (see p???) It includes a channel 1 modulator as standard. Like the Pineapple it uses framestores. This means that the output syncs will always be clean and continuous regardless of any corruption of the input signal. Malcolm has devised a some good techniques for simplifying the design including the use of low cost PIC microprocessors for timing generation.
WHAT’S THE DIFFERENCE
Now we know who has done it let find out how. Let’s look at the how 625 and 405 signals differ. I’m only going to look at the baseband video; I’m assuming that you have 625 video available and can make or scrounge a System A modulator.
SYSTEM A I
Lines per picture 405 625
Fields per second 50 50
Interlace 2:1 2:1
Line frequency 10125Hz 15625Hz
Line length @98.8ms 64ms
Line blanking 17.5 to 19ms 12ms
Front porch 1.5 to 2ms 1.6ms
Line sync width 8 to 10ms 4.7ms
Field blanking 13 to 15.5 lines 25 lines
Number of broad pulses 8 5
Number of equalising pulses None 5 + 5
Broad pulse width 38 to 42ms 27.3ms
Black level 0V nominal 0V nominal
Sync tip below black 300mV 300mV
White above black 700mV 700mV
Before we plunge into the details it's worth a quick look at two items that could enhance the pictures on your 405 line receivers. Equalising pulses and setup. Bad interlace was a perennial problem on many 405 receivers and equalising pulses were invented to make the sync separator's life much easier. It would not hurt and could easily help if equalising pulses were added to the 405 signal. Maybe 5+5+5 as in the 625 system though it is just possible that some sets would fail to lock with only 5 broad pulses. 8+8+8 would not fit in the field blanking interval. The pre-equalising pulses are much more important than those after the broad pulses. Some versions of the 405 standard included a setup or pedestal which raises black level above blanking level. The NTSC 525 system includes setup even today. Its purpose is to help receivers blank the video during flyback. This is not a problem with modern sets but many older ones suffered from visible flyback lines. It would not be difficult to include setup on the 405 line output of a converter.
The good news is that the field rates and interlace are the same for both systems. If they were different then a converter would be much more difficult. The main difference is the number of lines and if you try to put fewer lines into the same field length then those lines will inevitably be longer. This leads us to the first main problem. We must redistribute 625 lines, each 64us long into 405, each 99us long. With the benefit of hindsight this time redistribution is conceptually simple to do.
It is almost essential to divide each line into a number of separate elements, now commonly known as pixels. At least 500 are needed to preserve picture detail and modern digital TV equipment uses 720. Roughly one in every 3 lines from the 625 line input will not be written into the store.
Analogue storage was used in the early BBC converters and it might be possible to build a modern equivalent with charge coupled delay lines if they are still available. Not a practical option.
The BBC digital converter used 3 separate line stores. Think of them as 3 buckets, one being filled, another being emptied and a 3rd one to ensure that you never have to read and write the same store at the same time. This approach was the only practical one in 1970 but now we have dual port memory. This can be written and read simultaneously and is a thoroughly practical and low cost method. A good example is the NEC UPD485505. I would not recommend the standard FIFO memories made by many companies. Their control requirements are a nuisance in this application.
A frame store seems like overkill but allows the output sync to be steady at all times even when the input is disrupted. Also allows freeze frame, the possibility of a simple test card generator and few other tricks. With modern parts such as the Averlogic AL422B this is no more expensive than using line stores and would be my favoured approach.
This is the other main problem. To smooth out those jagged edges we need that dreaded word. Interpolation. The concept is a bit harder than time redistribution but let’s have a go at the theory of a simple 2 line interpolator. If you want to generate a new output line that is half way between two input lines then you need a 50:50 average of the input lines. The proportions are varied according to the position of the output line.
The BBC did theoretical studies backed up by practical trials to show that 2 line interpolation is vastly better than none at all. 3 or 4 lines are better than 2 though you will be hard pressed to see this, except by direct comparison on carefully chosen test signals. More than 4 lines is just not worth doing. I would choose a 2 line interpolator for a simple and practical design.
THEORY INTO PRACTICE
Now we know what has to be done let’s see how to actually do it.
INPUT and OUTPUT
Curiously enough, the difficult bit is no longer the time redistribution and interpolation. When all that had to be done with boards full of TTL and small slow memories it was hard. Modern programmable logic and memories have reduced it all to 2 or 3 chips of which more later. What we need to do first is convert the 625 input to a digital signal and reverse this at the 405 line output.
The BBC design in the CO6/509 is a scary piece of equipment. Thank goodness we don’t have to it that way any more.
Single chip ADC
The first 8 bit video ADC, the TDC1007 made by TRW, was introduced in about 1978, cost about £200 per chip and was welcomed with open arms by the professionals in companies such as Ampex and Quantel. Its modern counterpart is a small low cost, low power device. The generic 1075 device was first made by Raytheon and does a good job for under £5. There are plenty of other parts such as the Philips TDA8708 which includes video clamping and automatic gain control ahead of the ADC itself.
Along with the ADC we also need to amplify and clamp the video, separate H and V syncs and derive a sampling clock with a phase locked loop (PLL). If we cannot be bothered to do this then we can buy an….
Several manufacturers produce multistandard decoders on a single chip. They take in analogue video on any colour standard and deliver decoded digital signals. They were originally designed for “stunt” modes in TV sets, video on computers and suchlike. I’ll use the Philips SAA7113 as I know it well but Analogue devices, Brooktree, Harris and others have all made some. Most cost under £10.
All you need to do is connect the analogue input via a capacitor, add a small handful of external parts and hey presto you have 8 bit video and a 27MHz clock. The main snag is that you really need a microprocessor in the system to program all the registers in the decoder via an I2C port.
If you are one of that very small band who want to explore 405 NTSC colour then a decoder chip is definitely the right approach.
Converting digits to analogue is much easier than the other direction. A discrete DAC is nothing like as scary as a discrete ADC but I’m still glad we don’t have to do it that way now.
Single chip DAC
Curiously the problem here is that single 8 bit video DACs are really rather old fashioned. They have been largely superseded by triple DACs and ultra fast single DACs. The Philips TDA8702 is a simple device that is probably still available. Then we need to filter the output and insert syncs. Sync generation is quite a few TTL chips or an easy bit of logic inside a suitable programmable logic device.
I hope that most of you are still with me. I know that this sort of thing is second nature to design professionals but it is definitely not easy. But you weren’t seriously contemplating designing a converter. Or were you?
THE DIGITAL BITS
We need to choose the sample clock speed. On the 625 side I would go for 13.5MHz without much thought since that's what the whole TV industry uses. It is possible to run the 405 side at the same clock speed but this would involve a complex interpolator to change the number of samples per line. It is much easier to follow the route of all previous converters and keep the same number of samples per line for both input and output. The 405 output clock is then 13.5MHz * 405/625 = 8.748MHz. In a linestore based converter this will be a voltage controlled oscillator phaselocked to the input clock via a funny bit of logic. In a framestore design it will be a free running crystal. Before you go to get a crystal cut note that twice this frequency is 17.496MHz which is tantalisingly close to 17.734475MHz, the standard PAL 4fsc, which is readily available off the shelf. If you cheat a little and make the 405 horizontal blanking 12 pixels or about 1.3ms too long it will work out just nicely and I doubt if anyone will notice. The BBC used 12.65625MHz as the input clock which is OK but I would not recommend going any lower. The only other frequency you might use is 14.75MHz which gives square pixels and is standard for some decoder chips.
Once you have decided on line or frame storage the basic design decisions almost take themselves. If you have access to modern programmable logic devices (FPGAs) made by companies such as Xilinx and Altera then there is no contest. A single device costing under £20 can easily contain all the logic including the line stores. The Xilinx XC2S100 is an example. It will also connect to a framestore chip. Besides the low component count, the other great merit of this approach is that you can develop and change your design ad nauseam without physically rewiring anything. If you are not able to access this sort of technology, and the manufacturers do not exactly encourage amateur users, it is amazing what can be done with EPROMs, PIC processors and standard logic chips. What you will doing is using ingenuity and dedication instead of a complex and possibly expensive design environment. If you want a taster of what's possible with modern programmable logic download the Xilinx Webpack software free of charge from www.xilinx.com and give it a try. You could then buy or make the programming probe and have a go with a small cheap device such as the XC95144XL. You will be amazed at how many '161 counters, '138 decoders, '283 adders and '374 registers can be absorbed into one part. The main disadvantage is the multi-legged surface mount packages.
ACTUALLY BUILDING A CONVERTER
Right. We’ve designed our converter. Well I have anyway. Now let’s build it. There are 2 routes and both can be unpleasant. The pioneers favoured hand made boards full of standard logic. These are wretched things to build and debug. A PCB would be feasible but expensive.
For a modern converter design a PCB is essential. All those surface mount chips need very fine connections and that PCB is likely to be a precision multilayer job. These are commonplace in professional circles but not something you can make yourself. If you climb the learning curve you can design the PCB with one of the low cost PCB CAD packages and have it made by a specialist manufacturer. You will find it expensive in small quantities. Unless you are good at watch repairs you will also need a surface mount workstation to assemble it.
Now do the costings. Add up all the parts, including a box and power supply. Amortise the costs of PCB layout and manufacture and see what it comes to. If you do this honestly, for the likely (small) quantities, you will be alarmed by the answer.
The author once adapted a design that he did for a client to do 625 to 405 conversion. It worked quite well but like so many projects it never got finished. It was not suitable for production or sale since it was based on a fairly expensive piece of professional equipment.
So far I have only written about conventional methods of converting 625 to 405. Now lets look at the alternatives. I mentioned optical conversion. I have tried this venerable method and it can be remarkably good. I have even used it to convert 405 to 625. It does need careful setting up and is prone to moiré patterning from the interaction between the different line structures. The simplest way to avoid this is to defocus the camera very slightly. Better ways include spot wobble on the monitor or reducing the height on both camera and monitor so that the line structure disappears. All this is well within the scope of an enthusiast.
Some have suggested using computers or digital signal processors (DSP). Conceptually there is no problem. I’m a pretty lousy programmer and I could write you the conversion algorithms in a few lines of a high level language such as BASIC. These would take a 625 line image, already in the computer, and convert it to a 405 line image, also within the computer. And there’s the snag. You still need to get the picture in and out and you have to ensure that the computer can keep up with the data. Video comes at you continuously so real time means exactly that – you cannot put your hand up, take time out and catch up later. So when you say you have built a converter with a cast off 486 PC I won’t believe you. Modern DSP chips are entirely capable of doing the digital parts of the processing though you will still need to get the signals in and out. Maybe I'm prejudiced but I would rather use programmable logic than DSP for this job.
I have not really mentioned modulators. There are at least 2 published designs which are feasible for the enthusiast. One was (optionally) built into the Dinosaur converter and the Domino converter includes one as standard.
405 NTSC colour is an interesting subject. A few experimental receivers were built for the BBC trials in the 1950s and some of these survive. If you use a decoder chip as the input system the digital complexity is not much greater than for monochrome. There is also no conceptual problem about designing the colour output. Annoyingly, standard colour encoder chips could easily do the strange subcarrier frequency needed but not the 405 syncs. I would probably use a largely discrete component design which needs a lot of parts but is not too hard to do. A clever approach would be to do the whole NTSC encoder digitally in programmable logic. I reckon this will need rather more programmable logic resource than the whole of the rest of the converter.
Designing converters is only for the knowledgeable or the brave. Building them commercially is probably a certifiable activity. I salute all those who have tried and succeeded.
REFERENCES AND FURTHER READING
Many of these references are certainly not light bedtime reading.
Television Engineering, Principles and Practice Vols 1 - 4. S W Amos and D C Birkinshaw. A useful if dated general reference. Available from many libraries and secondhand.
Television Standards Converter using a line store. P Rainger and E R Rout. Proc IEE Vol 113 No 9, September 1966. An excellent detailed paper on the BBC analogue converter.
Television Standards Conversion. S M Edwardson and C K P Clarke. Wireless World, January 1987. A review of standards conversion over 30 years.
Fifty years of high-definition TV transmission. R C Hills.
IEE Journal Vol 56 No. 1, January 1987. A historical review, mainly about transmitters.
Digital TV line standard converters. Wireless World, May 1987. A brief introduction.
Digital line store standards conversion: Determination of the Optimum Interpolation Aperture Function. G M Le Couteur. BBC Research Dept Report 1973/23. Rigorous and authoritative mathematical and experimental treatment.
Digital Line Store Standards Conversion: Preliminary Interpolation Study. J O Drewery, J R Chew, G M Le Couteur. BBC Research Dept Report 1972/28. Theoretical study of interpolation.
Digital Standards Conversion: Interpolation Theory and Aperture Synthesis. C K P Clarke and N E Tanton. BBC Research Dept Report 1984/20. Another rigorous treatment of interpolation including field rate conversion.
NB: None of these last three papers is for the faint hearted. They are quite mathematical, involving Fourier analysis and other university level subjects.
IBA Technical Review, Vol 3. Digital Television. Ed. Pat Hawker. Excellent description of the IBA experimental 625/405 converter.
IBA Technical Review, Vol 8. Digital Video Processing - DICE. Ed. Pat Hawker. Although this deals with 525/625 conversion much is relevant to 625/405 problems.
CQ-TV is the bulletin of the British Amateur Television Club. It is available only to members. There have been many useful designs for SPGs, test pattern generators etc. A full index is available from the Club.
TELEVISION (Previously Practical Television) has published many relevant articles. Here are a few:
System A Modulator. David Looser. October 1984.
Recording 405-line signals. Gareth Foster. October 1983.
Goodbye to 405 Lines. Pat Hawker. January 1985.
Channel 1 modulator. Jeffrey Borin. March 1989
405 MAC, a new approach to HDTV. Jeffrey Borin. April 1988.
Practical 405 - How To Run Your Historic Receivers. Jeffrey Borin. November/December 1988
COMPARISON OF RF STANDARDS
SYSTEM A I
Nominal channel width 5MHz 8MHz
Sound carrier frequency relative to vision -3.5MHz +6MHz
Vision modulation polarity Positive Negative
Sync tip carrier Zero 100%
Black level carrier 30% 76%
White level carrier 100% 20%
Sound modulation AM FM
Ratio of peak vision carrier voltage 2:1 3.3:1
to unmodulated sound carrier
BAND I CARRIER FREQUENCIES
Channel Vision MHz Sound MHz
1 45.0 41.5
2 51.75 48.25
3 56.75 53.25
4 61.75 58.25
5 66.75 63.25