Fullscreen HTML5Med Res (??MB)HTML4

Most readers are familiar with analog oscilloscopes but these are being rapidly supplanted by digital storage oscilloscopes. These can capture and display waveforms with a much wider range of frequencies and they are also better at catching “one-off” glitches and fault conditions. By BRYAN MAHER Until about 1970 there was no satisfactory method of displaying infrequent or once-only events. Yet transient electrical signals and errors, intermittent faults and glitches, are very common in all types of electronic and computer equip­ment. They may occur perhaps once a day or even less, yet they can wreak havoc and no analog oscilloscope can display them. Some deliberate actions, like explosive shots or failure testing of mechanical components, generate (through transducers) once-only signals. We need the ability to find such signals when they occur, to capture them on the screen and to display and analyse them after the event. Analog storage CRO tubes were used 68  Silicon Chip for a decade or so but they had lots of disadvantages. We had to wait until someone thought of digital storage. Basic digital storage scope Then in the late 1960’s some enlightened person combined an analog-to-digital converter and a computer iron core memory with a conventional oscilloscope. From this marriage the Digital Storage Oscilloscope (DSO) was born. The block diagram of Fig.1 illus­trates the basic idea. In modern instruments, the blocks on the left side of Fig.1 constitute the signal acquisition section. There the Sample/Hold unit quickly takes many short samples of the analog signal as it occurs and these samples are immedi- ately digitised in the Analog to Digital converter (ADC). The system stores those digital copies in binary form in a random access memory (RAM). This first part of the operation is illustrated in the functional diagram of Fig.2(a). After the event has passed, you can read out from the memory that captured data and display it on the scope screen as an approximate copy of the original analog signal. That second function is illustrated by Fig.2(b). You can see two important differences between analog and digital oscilloscopes. Firstly, the analog scope can only display the signal while it is occurring. In contrast, the digital storage scope displays a reconstructed copy of the input signal some time after the event has passed. And that data can be held in the memory for as long as required and displayed as many times as you wish. Secondly, the analog scope must display (write or trace) the signal as fast as it occurs. For high frequencies and rapid transients, that requirement demands very expensive cathode ray tubes using electrostatic deflection. By contrast, in the digital scope only the signal acquisition and digital processing sections need to be fast enough to follow the live signal. Those areas include the analog preamplifier, the sample/hold, the ADC and the write to memory functions. Once the data which represents the signal is written to and held in memory, the display section can read out that data and display it on the screen at a conveniently slower pace. Therefore cheaper and slower display tubes using magnetic deflection and raster scan are perfectly adequate. Furthermore, because they can also display continuous waveforms, modern digital storage scopes are now supplanting analog scopes and providing lots of measurement functions as well. Digital scopes are ideal for capturing occasional glitches that would never be seen on an analog scope. This little glitch (circled) on an otherwise normal square waveform could cause untold intermittent problems in digital circuitry. (Yokogawa photo). Sampling Fig.2(a) depicts the taking of 500 Fig.1: in a digital storage oscilloscope the incoming analog signal at left is sampled, converted to a digital code, then stored in memory. Some time later that data is read from the memory, converted to a raster display and shown on the screen. Fig.2: functional diagram of a simple digital scope. The record­ing process (a) converts 500 samples of the analog signal into digital words which are written to the memory. Then that data can be read from memory and displayed as a reconstructed copy (b) of the original signal. September 1996  69 new data are fed into the memory, overwriting the old record. But if the signal never recurs, that first record is all you will ever get, so you keep it in memory as long as you wish. But most importantly - because you have it safely recorded in the memory - you can continue to re-display that waveform for as long as you choose. And many modern digital scopes allow you to print a copy of the screen display as well. Updated display Fig.3: when sampler switch IC1 conducts, capacitor C charges to the instant­ aneous voltage of the analog signal. When IC1 switches off, capacitor C holds that sample voltage while the A/D convert­er encodes it into an 8-bit digital word. Whether the input signal repeats or not, the display is updated, perhaps every 20 milliseconds. This means that the whole record of digital words held in memory is again read, converted to raster format and displayed on the screen, as illustrated in Fig.2(b). This frequent updating (between 30 and 150 screens per second for a simple display) together with the fairly long screen persistence used, gives the appearance of a continuous signal. The sampling, A/D conversion and writing to memory functions should run fast enough to adequately capture every wriggle, spike and harmonic in samples of an input signal at regular slight spread of the electron beam will time intervals. Fig.3 illustrates the merge those dots into a continuous essential components of a sampler, trace. where IC1 is a fast electronic switch. The full set of digital words held in To take each sample, a logic control memory is called one Record, which pulse applied to pin 12 causes IC1 represents all the information you to conduct between pins 10 and 11. know about that analog signal. If the During the few nanoseconds (or less) event repeats, each time the oscillothat IC1 is conducting, the capacitor scope is triggered the sampler and ADC C charges, through resistor R, to the collect a new record of samples. These voltage of the analog signal at that moment. At the end of the control signal pulse, IC1 ceases conducting but capacitor C continues holding that charge. The ADC quickly encodes the voltage value held in capacitor C by generating an equivalent digital word of eight bits. The clock control circuits promptly cause that word to be written to a unique address in memory, as indicated in the functional diagram of Fig.2(a). On each subsequent clock pulse, the instrument repeats the cycle: sample-hold-convert-storein-RAM. This continues until 500 samples are taken and the corresponding 500 digital words are stored in memory. That is sufficient data to reconstruct an approximate copy of the analog signal on the screen. In simple systems this display will be an array of dots, one point Frequency, period and other waveform measurements are an inbuilt feature of for each sample taken, as most digital storage oscilloscopes. This HP 54601 model has four input channels Fig.2(b) indicates. But the and a bandwidth of 100MHz. 70  Silicon Chip a high frequency analog waveform. Otherwise the reconstruction of fast rising or falling edges will be poor. For example, the steep fall at the right hand end of the waveform shown in Fig.2 demands that many samples be taken at a fast rate to record the true wave shape. Sampling Interval is the time between one sample and the next. This is the inverse of Sample Rate, which is also the frequency of the clock pulses. For best resolution and widest bandwidth, the sampling interval should be very short and the process should be repeated at a very fast sample rate. Some modern digital oscilloscopes can take 5,000 million samples each second, or 5 Gigasamples per second, written as 5GS/s. They can fill a 500-point record in the memory in one tenth of a microsecond! The Tektronix TDS320 digital storage oscilloscope has 100MHz effective bandwidth on each of the two input channels. The sampling rate is 500MS/s and the memory holds a record of 1,000 points. The 8-bit vertical resolution in real time mode can be extended to 11 bits with repetitive signals using averaging techniques. Vertical sensitivity extends down to 2mV/div, with an accuracy of 2%. This instrument can capture up to 86 waveforms/sec and make a wide range of automatic measurements. Hard copy output to a printer is a standard facility. Fig.4: in a flash A/D converter, comparators give high or low output depending on whether the analog signal is above or below the DC voltage tapped from the resistor string. IC3000 decodes this data into a digital word. Real oscilloscopes use 255 com­parators and 256 equal resistors to encode the analog sample into an 8-bit word. Real time bandwidth To display one-shot events, digital storage oscilloscopes must operate in Real Time Mode. This means that the samples of the analog signal are displayed on the screen in the same order as they are taken and one trigger event must initiate the total acquisition. These conditions are implied by Fig.2. By a Trigger Event we mean either a voltage change in the analog signal which is sufficient to actuate the oscilloscope trigger circuits or an external signal applied to the scope “Ext Trig” terminal. Real time digital oscilloscopes have two measures of bandwidth. Firstly, the analog bandwidth is the -3dB frequency limit of the analog preamplifier stages. Secondly, the sampling rate also sets an upper frequency limit. In the next chapter we will see why Nyquist’s Rule requires a sampling rate more than twice the frequency of the input signal. So we define the digital real time bandwidth as a frequency less than half the sampling rate. The Effective Real Time Bandwidth is the lower of the quoted analog and digital bandwidths. Flash A/D converters The Flash A/D converter is a very fast circuit which can encode an analog signal as a binary digital word on parallel output lines. For September 1996  71 Fig.5: the 4-bit A/D converter allows only 16 decision levels, which is too coarse a result. Real time scopes use 8-bit systems, giving 256 decision levels, so the steps in the display are fine enough to be acceptable. simplicity, we will look at a 4-bit version, shown in Fig.4, although 8-bit ADCs are standard on digital scopes. These ADCs are referred as “flash” because they are very much faster than the older “successive approximation” types. The circuit shown in Fig.4 can create a 4-bit digital word to represent each positive analog sample which is less than +5V. It is called Unipolar because it accepts only single polarity signals. A 4-bit digital word can represent one of only 16 different voltage levels. So Fig.4 contains (16-1) = 15 analog comparators, IC1 to IC15. A comparator gives a logic high output if the signal at its positive input exceeds the voltage at its negative input. And it gives a logic low output in the opposite condition. The full output from the Sample/ Hold circuit is applied to the positive inputs of all comparators in parallel. In addition, a stable +5V reference source sends a constant current down a series string of sixteen equal resistors, R1 to R16. Each comparator has its negative input connected to the corresponding tap on this string. Decision levels Each resistor develops a voltage drop of +5V/16 = 0.3125V. As Fig.4 shows, the negative input to IC1 is held constantly at +0.3125V; IC2 negative input is at +0.625V, etc..... up to IC15’s negative input, which is held at +4.6875V. These specific values are called the sixteen Decision Levels of this 4-bit circuit. Suppose at some moment that the analog sample (from the sampler in Fig.3) has an amplitude of +0.756V. In Fig.4 this voltage appears at the Because of their very fast sampling rate and inbuilt waveform storage, digital scopes are ideal for viewing irregular and infrequent pulse waveforms. This 150MHz model from Hewlett Pack­ard can view waveforms with risetimes as short as 1.4ns. 72  Silicon Chip positive inputs of all comparators. So in both IC1 and IC2 the positive input voltage exceeds their negative inputs. Therefore the outputs of IC1 and IC2 both go to a logic high level. But all higher comparators, IC3 to IC15, find their +0.756V positive input is less than their various negative inputs. Thus they all give logic low outputs. The outputs of all comparators in Fig.4 feed to 16 digital latches in the assembly IC2000. From thence 16 parallel lines feed to IC3000, the Digital Logic Unit. Here a complex tree structure of logic gates converts the data on the 16 input lines to digital code on four lines, as a 4-bit digital word, which is then written to the memory. We use MSB to mean the Most Significant Bit and LSB to mean the Least Significant Bit, of parallel digital data lines. Table 1 shows the sixteen possible digital words in a 4-bit system produced by the A/D converter illustrated in Fig.4, together with the decision level voltage corresponding to each step. Notice that the difference between the +5V reference and the highest acceptable input, +4.68750V, is equal to the contribution of the LSB, which is +0.3125V. Quantisation noise Imagine, just for a moment, that we constructed a digital storage oscilloscope using 4-bit digital words, generated by the ADC shown in Fig.4. As this circuit has only 15 comparators, it has only 16 voltage decision levels (including zero), as listed in Table 1. The circuit represents each analog value by a quantised number, which is equal to the voltage of the decision level immediately below. So in a 4-bit system, only 16 variations in the input analog voltage are recognisable. Fig.5 shows those sixteen levels. Also depicted in red is an analog input, actually 500 samples, so close together that they look like a continuous signal, which is varying between zero and about 3V. Immediately below this, is its quantised reconstruction which would be displayed on the screen of such a 4-bit oscillo­scope. That lower stepped waveform is the closest approximation our 4-bit system could make to the input signal. Just released from Tektronix, this TDS220 100MHz oscilloscope has two input channels. It has been designed to behave as much as possible like an analog 'scope, to the extent that the actual sampling rate being used at any time is not shown on the screen. The other big change is that it uses an LCD screen instead of a raster-scanned CRT. This makes it very compact – it is only 110mm deep. As you can see, the 4-bit waveform would be awful. Between points g, h, i, j, k & m, the analog signal varies through six different voltage values. But all of these fall between two adjacent decision levels, +1.5625V and +1.875V. Because any analog input can only be represented by the decision level vol­tage immediately below, all those points are called +1.5625V by the ADC. The voltage increment between decision levels is (1/16) 6.3% of screen height, which is obviously much too coarse! When displayed on the screen, you would never know the real value of the input between times g & m. All points in that area would be displayed on the screen as +1.5625V, because they all would result in the same digital word, 0101. This loss of vertical resolution in the display is an error called quantisation noise. This results in a stepped display on any digital scope, in stark contrast to the smooth contin­uous trace on an analog scope. To make these vertical steps or increments so small that the display looks like a smooth continu­ ous trace, we need much more than 16 decision levels. 8-bit flash ADC To achieve that aim most digital oscilloscopes use an 8-bit A/D con- TABLE 1 STEP V (Analog) Binary Word 0 0.0000 0000 1 0.3125 0001 2 0.6250 0010 3 0.9375 0011 4 1.2500 0100 5 1.5625 0101 6 1.8750 0110 7 2.1875 0111 8 2.5000 1000 9 2.8125 1001 10 3.1250 1010 11 3.4375 1011 12 3.7500 1100 13 4.0625 1101 14 4.3750 1110 15 4.6875 1111 Table 1: the 4 Bit Natural Binary Code; Reference = +5.00V. September 1996  73 Fig.6: the summing op amp IC2 translates all analog samples from their (-5V to +5V) range, up to new (0V to +10V) range, by inverting them and adding +5V. These are now accepted by the flash A/D converter and encoded to offset binary code. SAMPLE VOLTAGE DIGITAL WORD At A At C Output at F +5.0000 ZERO 0000000 +4.9609375 +0.0390625 0000001 +3.8671875 +1.1328125 00011101 +2.500 +2.50 01000000 ZERO +5.00 10000000 -1.6406250 +6.6406250 10101010 -2.500 +7.50 11000000 -4.9609375 +9.9609375 11111111 Table 2: Offset Binary Code verter for standard real time operation. The circuit is identical to that shown in Fig.4, except that it provides 256 voltage decision levels and contains 255 (256-1) linear com­parators. The series resistor string consists of 256 precision real-value resistors. Despite the resulting increase in cost, complexity and size of the converter, this larger 8-bit system is necessary to achieve adequate vertical resolution. The voltage increment between decision levels is 1/256 or 0.4% of the screen height, so the slight steppiness in the trace is much more acceptable. In this 8-bit version of Fig.4, IC2000 now contains 256 digital latches. These are joined by 256 parallel lines to IC3000, which contains about 3200 transistors in an enormous tree structure. This converts signals on 256 parallel lines to an 8-bit digital word on 8 parallel output lines, which feed to the RAM. 74  Silicon Chip Critical large scale integration (LSI) techniques are needed to manufacture such A/D converters and maintain accuracy. Bipolar A/D conversion Flash A/D converters are all called unipolar, because they respond only to positive signals. This means that they cannot directly accept bipolar analog samples, which range through negative and positive values. To fix that problem, we translate (ie, lift up) the samples of the analog signal into an all-posi­tive range. Fig.6 shows one form of voltage translator which we insert into Fig.1 between points A and C. It consists of an inverting summing op amp IC2, placed between the bipolar analog sample signal at A and the unipolar A/D converter at C. The op amp gain is equal to -1 from either input A or B to the point C. The -5V DC reference voltage at B, when inverted in IC2, adds +5V DC to all signals which are applied at the point A. Signals at A may be between -5V and +5V. As Fig.6 illustrates, that whole range is simultaneously inverted and lifted up by +5V. It is linearly translated to a new signal range, between 0V and +10V. For example, a +5V signal at A is inverted to -5V and has +5V added, to become 0V at C. Or a -5V signal at A is inverted to +5V and has +5V added, so is translated to +10V at C. Then, to cope with these higher signal voltages, the reference voltage in the 8-bit flash A/D converter is set at +10V. With this signal translation before A/D conversion, the system can encode bipolar analog samples. It produces 8-bit digital words in the Offset Binary Code. Table 2 shows a few of the 256 entries in this code. Using a +10V reference, the increment between decision levels is 10V/256 = 0.0390625V. Other codes exist which could also be used. Reconstructed display Fig.2(b) illustrates the reading of data from memory and its conversion and display on the screen, in a simple system. Each digital word of 8 bits is called one byte and occupies one memory address. Two separate pieces of information are associated with each word stored in memory. Firstly, the address of each word in memory corresponds to the horizontal coordinate (ie, sample number 1 ..... sample number 500) of that point on the waveform. And secondly the digital value of each word held in memory indicates the vertical coordinate of the corresponding point on the screen. This is the best approximation the digital system can make of the voltage of that sample of analog input. In the next chapter we will describe the intricacies of raster display, where a simple presentation consists of a set of 500 points on the screen, like those shown in Fig.2(b). Because the display consists of 500 points, the smallest horizontal increment is 0.2% of screen width. The width spread of the light spot merges the 500 discrete points into a continuous trace. SC