DATV in simple terms - Part 1
This article was first published in CQ-TV issue 208
In this short series of articles, I am going to try to explain digital television in layman’s terms and without getting too involved in mathematics. The reasons for this are that I am a layman and I have never been any good at math! I have however, a reasonable grasp of the techniques used in digital TV systems and plenty of experience in electronics.
Almost without exception, when I mention digital ATV people switch straight off because they think it is too complicated for them to understand and they have a perfectly good analogue system already so why bother changing. While it is true that digital is complicated, it is no more so than analogue but it requires you to look at things in a different perspective. It is probably easier to visualise the effect of something you can read on a voltmeter or oscilloscope rather than a massive chain of numbers but when it comes to a complicated video signal, the distinction starts to blur. Consider for example the problems of ‘group delay response’, ‘chroma phase error’ and ‘transient overshoot’, are you well versed in their causes and prevention? Probably not, but these are likely to be the main causes of distortion in your analogue signal. The basics are straightforward but the devil is in the detail. The good news is that digital is much simpler because none of the mentioned distortions exists and once the first principles are understood, everything else follows the same rules. Digital is digital, there are no variations, deviations or anomalies as analogue has.
In order to make the introduction as painless as possible, I am breaking this mini tutorial into easy steps. The first step is to explain how a digital signal is produced from an analogue one. The process is called ‘Analogue to Digital Conversion’ or ADC for short.
ADC has been around for as long as numbers. That may seem a strange statement but if you recall your earliest memories as a child when asked how old you were, you probably said ‘four and a half’ or something like that. You mentally converted an analogue quantity, your age, into a numeric value. I am sure you will immediately recognise many more examples that you, and generations before you have made without even realising it. All an electronic ADC does is convert an amount of something, usually a voltage, into a number representing that amount.
Before delving into measurement techniques, we need to define a few terms which will be used later. The first is ‘resolution’, this is the size of the number (not its actual value) which will hold the measurement. If you prefer, you can regard it as the amount of accuracy that it is possible to measure to. A higher resolution means the analogue quantity can be represented more accurately. Because digital systems work in binary, the resolution is usually a multiple of two, for example 2, 4, 8 or 16 bits are common. Each bit that is added doubles the resolution. You might like to think of resolution as the number of markings along a ruler, the more there are, the easier it is to take an exact reading from it.
The other important term is ‘conversion time’. This is a measure of how long it takes to produce the digital value when an analogue input is presented. Different types of ADC take significantly different times to complete their tasks. A longer conversion time means it takes longer to take the measurement and produce the result. The longer it takes to return a value, the fewer values can be returned in a given time period. This pace is generally referred to as the number of ‘samples per second’. If for example you were measuring tide height at the seaside, taking one sample per hour would give a set of figures that nicely demonstrate the rise and fall of the water height. If you changed to sampling once per minute, the readings would show more clearly where the highest and lowest tides occurred, they may have been missed before if they fell between the hourly readings. If the samples were taken still more frequently, say once per second, the individual waves would start to show as well as the background increase and decrease of the tide. As you see, the more rapidly the samples are taken, the more detail can be extracted from the measurements.
Back to talking electronics; if a signal is sampled frequently enough, the samples can be converted back to a close approximation of their origin. Because modern ADC devices can convert extremely quickly, it is not only possible to convert fast enough to show the outline of a signal, it is possible to take samples within a single cycle of a signal. For reasons deeply buried in mathematics, (and therefore not to be explained in detail here) it is necessary to take at least two samples of a signal to be able to recover anything meaningful from it. Digitising a 2MHz signal would for example require sampling at a rate of at least 4MHz and preferably more.
Digital TV signals are actually a stream of numbers read out of an ADC circuit. After processing, transmitting, receiving and a bit more processing the numbers are recovered in their original form and fed to a device that does the opposite of the ADC. This is perhaps confusingly called a DAC or Digital to Analogue Converter. It takes numbers in and produces a voltage out. (fig.1) Like their ADC counterparts, they also have a resolution and a conversion time and these factors also have to be taken into account when conveying signals.
It is difficult to demonstrate the effects of different resolutions and sampling rates in a printed form like this but in a live picture it is very obvious. I have used a computer graphics package to manipulate the same image and try to show the effect each has so you can see the resulting pictures in print. (fig. 2) Note that the different resolutions have no effect on the sharpness of the picture, they only show as a limited number of shades while the sampling doesn’t effect the shades (except that they can’t be seen so well when the detail is lost) but it does affect the sharpness.
As the resolution increases or the sampling rate increases, the quantity of numeric measurements also increases. The actual rates used depend on how many measurements can be stored or sent to their destination given the practical problems arising from a limited transmission bandwidth. There is a trade-off between quality and quantity which is decided by the needs of the broadcaster.
There are many ways to measure a voltage and return a digital value from it. Lets look at a few of the common ones and see if they are suitable for use in a video application. Note that any of these methods could be used but some would require circuits running so fast or components of unwieldy values that they would be impractical to utilise.
1. Dual slope integrators.
I’m sure we have all measured the voltage across a charged capacitor at some time and see it gradually fall as the charge leaks away. What we saw was a voltage falling as time passed by. If we timed how long it took for the voltage to reach zero and knew how fast the charge was leaking away, we could work out how much was there to start with. That’s the principle of a dual-slope integrator. A timer operates while a series of switches route the voltage to charge a capacitor, then isolate it and then remove the charge. The timer period is a numerical representation of the amount of the original charge. The higher the voltage, the longer the period and hence the larger the count is. The count is used as the ADC digital data output.
2. Flash converters.
These are named because of their speed, not because of any optical phenomenon! They use a bank of comparators to compare the input with a pre-determined list of voltages. The voltages generally come from taps along chain of series resistors. A reference voltage is applied to the top and bottom (often zero at the bottom) of the chain so at each resistor junction there is a voltage proportional to its place along the chain. As the resistors are normally of equal value, the voltage steps are all the same size. Each of these voltage references is connected to one input of a comparator and the voltage being measured is applied to all the other comparator inputs in parallel. Any comparator seeing more on its measurement input than its reference input will switch output state, the others won’t. All the comparators seeing more on the input than from the divider chain will switch one way, the remainder the other way. By looking at all the comparator outputs to see which is in each state it is possible to derive a number equivalent to the input voltage.
3. Successive approximation converters.
These are similar to flash converters in some ways but use a different method to decide the output measurement. Instead of using lots of comparators each connected to a tap along the resistor chain, these us a single comparator and a series of switches to connect it to different taps. The switches are cleverly arranged to minimise how many are needed and in general these ADC use much less silicon area. To be efficient with only one comparator, the switches have to be operated in such a sequence that they find the voltage in the minimum number of tries. Starting at the bottom tap and working up is not an option, especially when there may be several hundred taps to try. The method they use is rather like a number guessing game, they start by connecting to the tap half way down the chain and the comparator says ‘more than’ or ‘less than’ when it compares the tap with the input voltage. The next step depends on the outcome of the first ‘guess’. If it decided the voltage was below half way, it next tries the mid-point between zero and half way tap, in other words the quarter way up tap. The comparator again decides whether the voltage is above or below that tap. If below it again compares half way between that tap and zero, if above it compares with that tap and the mid-chain tap. Each time a decision is made and each time a new half-way is decided until the number of bits of resolution are used up. Each time a switch is operated, the difference between the voltage on the resistor chain and the one being measured is halved, hence the name ‘successive approximation’. The way the digital output is created is very simple; if the input was higher than the guess, a ‘1’ is output, if lower a ‘0’ is output. When all the guesses are used up the ones and zeroes make up the digital output data. As a bonus, as the bits are guessed one at a time, the output can be read serially while the conversion is in progress.
4. Ramp and compare converters.
These are very simple in operation and actually use a DAC, a counter and a comparator to perform the ADC function. The idea is to start with a small but known voltage from the DAC and compare it to the one being measured. The voltage is then gradually increased by incrementing (adding one to) the number fed to the DAC. When it reaches or just exceeds the unknown one, the comparator tells us to stop and the number fed to the DAC at that point is used as the output. Although it’s a measure of the voltage being created, we know from the comparator that it must match the one we are measuring. You might think of it as a pair of weighing scales, the unknown on one side and the known on the other. When the scales balance, the known side matches the unknown one.
Those are the four most common ADC types. There are many more, including ones that use the voltage to tune an oscillator then measure its frequency and ones that use pulse width techniques, but these are well outside the scope of this article.
Each type has its good and bad features, finding the best compromise, especially when dealing with TV signals is all-important. Let’s investigate the features of each type:
Simple to build
No critical components
Speed depends on voltage being measured
|Complicated to build
Uses precision parts
Serial output available
|Complicated to build
Ramp and Compare
Simple to build
|Speed depends on voltage being measured
For slow-scan applications, the speed probably isn’t that important although it’s worth mentioning that some dual-slope integrators on the market will struggle to achieve five samples a second. Generally, though, the sampling rate will be fast enough to exceed that needed for transmission.
As the scan rate increases, so does the need for faster sampling rates and the slower converters start to fall out of the race. Another important factor is how constant the sampling rate is. Take for example the ramp and compare type; if the DAC gave outputs in steps of 1mV and the voltage being measured was 2mV, the conversion would stop at a count of two. If we are increasing the DAC count at fixed intervals, two intervals will have elapsed. On the other hand, if the voltage being measured was 250mV the count would have to reach 250 so that many of the intervals would be taken. For a standard 8-bit ADC the range of conversion times can be as much as 255 cycles and as little as one cycle. With a range so wide it can be difficult to maintain the timing restrictions placed on a digital transmission.
In practice, the flash and successive approximation types are most used and with the high integration of modern integrated circuits it is possible to fabricate all the component parts of these, including the resistor chain on one chip.
The speed needed for TV depends upon the highest frequency in the video waveform, in other words, the video bandwidth. Given that a minimum of two samples are needed per cycle, the sample rate must be greater than 2 times that frequency. If we assume a 5.5MHz bandwidth, (for PAL, a bit less for NTSC) the sampling must take place at no less than an 11MHz rate. Bear in mind that each sample is more than one bit wide, in most cases 8-bit resolution is used, so if sent serially an 11MHz sampling rate becomes an 88MHz bit-rate. Successive approximation ADCs conveniently give a serial output, for the other types an additional shift-register is needed to convert the parallel data into serial format.
That concludes the first part. We have seen that any voltage can be converted to a numerical value and that the sampling rate determines how closely the voltage is tracked while the resolution determines how accurately it is measured. We have looked at a few methods of doing the conversion and discussed the speeds we need to use in order to convey a quality picture. In the next article, we will look at the clever tricks we can do with numbers and why they are more versatile than their analogue equivalents.
Part one of this three part document is below.
For parts two and three click these links -> Part 2 Part 3