Fullscreen HTML5Med Res (??MB)HTML4

While the cathode ray tubes used in most analog oscillo­scopes use electrostatic deflection, the display tubes used in most digital scopes are virtually the same as in computer moni­tors and TV sets; they use magnetic deflection via coils on the neck of the tube. By BRYAN MAHER The magnetic deflection cathode ray tubes used in digital storage scopes are cheaper, shorter and more rugged than the electrostatically deflected tubes used in fast analog scopes. In a cathode ray tube (CRT), magnetic deflection of the electron beam is achieved by wrapping two sets of copper coils around the outside of the tube neck, as depicted in Fig.1. The horizontal deflection coils are mount­ed above and below the neck of the tube and current flowing in them generates a magnetic field which passes vertically downward through the tube. Electrons in the beam are deflected in a direction at right angles to this magnetic field; ie, across the screen. By contrast, the verti­cal deflection 76  Silicon Chip coils are mounted one on each side of the tube neck. Currents flowing in them produce a horizontal magnetic field which deflects the electron beam up or down. The two sets of deflection coils are held in one assembly called the yoke and its function is exactly the same as the yoke in a colour TV set. Fig.2 shows a more pictorial arrangement of the coils. Typically, the horizontal deflection currents are ±500mA and flow in coils having 13 millihenries inductance. The vertical deflection currents are typically ±150mA flowing in coils of 40mH inductance. Let us consider a basic digital scope having 8-bit resolu­tion. We featured a simplified description of how the digital storage scope acquires signals and stores them as digital data in part 5 of this series, in the September 1996 issue of SILICON CHIP. Now let us see how that data is displayed on the screen. To operate the deflection system, two sweep generator cir­cuits run at different frequencies, both derived from a crystal master oscillator. The horizontal (or line) system produces sawtooth waveform currents at exactly 28,800Hz in the horizontal deflection coils (shown as H,H in Fig.1). The resulting magnetic field deflects the electron beam from the left side of the screen to the right and back again in exactly 34.7222 microseconds (µs), as illustrated in Fig.1(c). The forward trace from left to right takes about 33µs and the fast retrace (flyback) takes the remaining 1.7222µs. At the same time, the vertical (or frame) system sends a sawtooth waveform current at exactly 60Hz through the vertical deflection coils. The magnetic field which results deflects the electron beam downwards comparatively slowly, taking 16ms to work its way from the top of the screen to the bottom and 0.666ms to fly back to the starting point at the top. Fig.1: the magnetic fields due to currents flowing in coils (a & b) external to the tube neck deflect the electron beam to cover the whole screen (c) with 480 raster lines. With both deflection systems operating, as we see in Fig.1(c), the electron beam starts at the top left corner of the screen. From there it traces in the phosphor a pattern of 480 fast horizontal lines, as it moves (comparatively) slowly down, to arrive at the bottom right hand corner. From there the fast retrace (flyback) returns the beam to the starting point. Auxiliary circuits always blank off both the vertical and hori­zontal retraces, so we show these as dotted lines in Fig.1(c). If the electron beam was turned on during all forward trac­es, you would see the whole screen covered in a pattern of 480 fine horizontal bright lines spaced about 0.25mm apart. In Fig.1(c) we illustrate this but for clarity we have drawn only a few lines. This is the raster and it is similar to the background line pattern you see on your computer display if you turn the brightness up at night. A standing bias, applied between the tube grid G1 and cathode K as shown in Fig.3, may hold the electron beam near cutoff, making the screen dark. A voltage drive applied to the cathode K (or grid G1) can then overcome this standing bias to illuminate the screen. Greater deflection angle Say a CRT has 16kV acceleration potential. We recall from past episodes that such a tube, when electrostatically deflected, will have deflection angles inversely proportional to the acceleration voltage. But a similar tube magnetically deflected will have deflection angles inversely proportional to the square root of the acceleration voltage. So the magnetically deflected tube can deflect its elec­tron beam through an angle four times greater. So we see why digital scopes commonly have wider screens and shorter tubes. However, because of the coil’s inductance, direct magnetic deflection is limited to frequencies below about 100kHz. Bit-mapped raster scan Most digital scopes have a frequency response that ranges up to 100MHz or a great deal more. To make that possible, an indirect method called “bit-mapped horizontal raster scan dis­play” is used. This is completely different from the direct electrostatic deflection used in analog scopes. And it is more complex. Fig.3 shows an abbreviated block diagram of a digital stor­age oscilloscope. The left hand half of this figure is the acqui­ sition section, which includes the input attenuator, analog preamplifier IC1, sampler IC2, A/D converter IC3 and the fast RAM (random access memory) IC4. In this chapter of our story we concentrate on the right hand half of Fig.3, the display section, which includes IC5, IC6, IC7 and the tube. When the digital data representing the input analog sinew­ ave is recorded and stored in the fast RAM IC4, then the display section can begin its magical work. Firstly that data is read out from RAM IC4 into the display processor IC5. This is a micropro­c essor which has running within it a scan conversion point plot­ting algorithm. This rearranges the data into a display image in scan line order, which it promptly writes into a second memory IC6, called the bit-map frame refresh buffer, which you see in Fig.3. In the basic digital scope we are describing, this buffer consists of an array of 307,200 semiconductor memory cells, electrically arranged into a two-dimensional planar matrix of 480 horizontal rows and 640 vertical columns (480 x 640 = 307,200). Fig.4(a) gives some idea of this scheme, though here for clarity we have drawn a much smaller number. Each memory cell in this buffer holds one bit: that is either a logic high potential or a logic low. And in Fig.4(a) we have drawn a 1 in some of the cells to indicate those cells which contain a logic high potential. In the remainder of the cells we have drawn a 0 to indicate those which hold a logic low. In Fig.4(a) you can clearly see a waveform in the pattern of 1s and this is called the bit map. This is an image of the original analog input waveform. March 1997  77 DEFLECTION COILS WRAPPED AROUND TUBE NECK TUBE NECK CRT FLARE Fig.2: the two sets of deflection coils are held in one assembly called the yoke and its function is exactly the same as the yoke in a colour TV set. This diagram shows a more pictorial representation of the coils. Now we aim to convert that blueprint of electrical 1s in the buffer into a corresponding display on the CRT screen. Once the processor IC5 has filled buffer IC6 with data forming the bit map, two different but intimately related actions commence simultaneously and run in synchronism, like two kids in a three-legged race. Displaying the bit map The deflection circuits cause the electron beam to commence from the top left corner of the CRT tube screen and trace out the full screen raster, line by line, as described above. During most of this time the beam electron current is reduced to nearly zero by the negative bias applied between the tube grid and cathode, which Fig.3 illustrates. So almost all of the screen is dark. At the same time, the system addresses all cells in the bit map refresh buffer IC6 and the bit value contained in each is read out. Starting at the top left corner, the system addresses the cells and reads their contents; cell by cell, from left to right and row by row. First, each cell in the whole top row is read, then those in the next row, and so on, until the bottom right corner is reached. Cells are addressed across a row of IC6 at the same speed as the electron beam is deflected across the tube screen. The final addressing of the cell in the bottom right corner of IC6 and the reading of the bit it contains coincides with the elec­ tron beam arriving at the bottom right corner of the screen. Displaying the signal In the basic digital scope we are describing, the single bit read from each buffer cell is simply a voltage, either logic high level or logic low. If a TTL system is used, logic high means about +4V and logic low about +0.5V. As each cell is read, its voltage is amplified and inverted by the following video amplifier IC7, whose output signal drives the cathode of the CRT tube in Fig.3. (Alternatively, you could drive the grid but without signal inversion.) Each time a logic 1 is read from a cell in the refresh buffer, Fig.4(a), the video amplifier IC7 inverts and amplifies this to a large negative voltage pulse, typically -30V to -60V. Applied to the CRT cathode, this is big enough to overcome the G1-K standing bias. Thus the electron beam is turned fully on momentarily. This produces a bright spot of light on the screen at a point corresponding to the address of that logic 1 cell in the refresh buffer. Each bright point is called a pixel (for picture element). In the same way, many pixels are displayed on the screen (Fig.4(b)) in a pattern which copies the disposition of cells containing logic 1 bits in the refresh buffer matrix (Fig.4(a)). But on a screen typically 135mm wide, each pixel is only 0.2mm apart, so normal spot width-blurring usually merges strings of these dots into continuous bright lines. If the sampler cannot provide enough points, firmware routines can fill in by adding more bright dots in straight line approximations or Sin(x)/x geometric curves. That trace we see on the screen in Fig.4(b) is a copy of the bit map in the buffer IC6. This is itself a copy of the original analog signal applied to the scope input socket. This is the raster scan method in action: the digital scope is indirectly displaying your input signals on a Fig.3: in a digital scope, IC1, IC2, IC3 and IC4 form the fast acquisition section. IC5, IC6 and IC7 then form the rasterising display circuits. 78  Silicon Chip Fig.4: a bit map (a) of the input waveform is drawn logically in the memory cells of the refresh buffer. Data read from this map turns on the electron beam (b) at points corresponding to the pattern of logic 1s in the bit map. magnetically deflected cathode ray tube screen. a refresh rate of 60Hz. That’s why we call IC6 the refresh buffer. Displaying a one-shot signal Video frequency Now let’s assume that the input to your digital scope was a one-shot; ie, a non-recurring signal. In the fleeting time that signal existed, it was sampled by IC2, digitised by IC3 and recorded in RAM IC4 and held there indefinitely. After the signal had gone and the sampler had stopped, the output section of your scope (the right hand side of Fig.3) then performed all the wondrous miracles we saw above. The reading of the whole buffer IC6 and the drawing of one raster on the screen displaying the waveform both take exactly 16.666ms. Digital scopes commonly use a tube with a P4 white phosphor, which has a compound 150/480µs persistence time, after which the trace fades away. To maintain a stationary picture on the screen, the scope must continually refresh the trace illumination by repeating the display process; ie, read the bit map stored in buffer IC6, amplify the signal in IC7, and drive the tube cathode to turn on the beam to re-illuminate the display. The system repeats this whole action every 16.666 millisec­onds; ie, at To perform these wonderful feats, all 307,200 cells in the buffer memory must be addressed and read every 16.666ms. So cells must be read at (16.666ms/307,200) = 54.2535 nanosecond inter­vals. This produces a serial stream of single bits passing to the amplifier IC7 at (307,200 x 60) = 18,432,000 bits/second. Because this bit stream produces a visible display on the screen, we call this an 18.432MHz video frequency. And we call IC7 the video amplifier. Notice that all this time the sampler IC2 and the A/D con­verter IC3 have stopped. This is not because they are lazy or slothful. It’s because you previously filled RAM IC4 with one record of data from a one-shot input signal, now long gone. So your scope continually refreshes the screen with the copy of that departed signal held in IC4. You are truly using the storage capabilities of your DSO to the full. Recurrent signals When you apply a continuously recurring high frequency signal to the input of your digital scope, the busy sampler very quickly takes a record of 500 (or more) samples of the signal. The A/D converts these to digital format and stores them safely in RAM. Then while the sampler has paused, that data is read from IC4, converted by IC5 to a bit map and stored in the refresh buffer IC6. Now the system reads that buffer and displays its contents on the screen raster at the much slower display speed. Once it does that, the system clock may reactivate the sampler and A/D converter, to take another record of samples and store them in RAM. These can be then read from the RAM, converted to a complete new bit map which includes any changes in the input signal and displayed on the screen, replacing the old. At fast sweep speeds, such as 2µs/ div, the sampling of one record of the input signal may take only 20µs. But in convention­al digital scopes the sampler pauses for about 20ms while the display processor and refresh buffer do their clever work and display the waveform on the screen. So typically you will see only one cycle in every thousand cycles that flow in your circuit. The elusive occasional glitch interference that you are searching for may escape detection. March 1997  79 Fig.5: block diagram of InstaVu acquisition architecture in the Tektronix TDS784 scope, which can capture 400,000 waveforms/second on one channel. Your scope would be capturing only about 50 waveforms per second and missing the rest. Alternatively, instead of deleting the old display on each refresh, the electrical variable persistence control gives you the option to accumulate old and new data points in the bit map, and hence on the screen. These can be kept over many acquisi­tions, or over some period of time between 250ms to 10 seconds, or infinitely. In this way, infrequent events can be found and displayed. Fast acquisition To increase your chances of seeing that occasional problem pulse which This is a 3MHz signal depicted on a Tektronix TDS784A digital colour scope in InstaVu mode. Here a runt signal is clearly visible, made doubly so by the colour display (although not reproduced in this B&W photo). 80  Silicon Chip is troubling your electronic system, more expensive digital scopes use proprietary methods to raise the rate of waveform capture. The Tektronix TDS400 series digital scopes can acquire 200 waveforms/ second in infinite persistence mode. In each 16ms period they capture and overlay three or more updated versions of the input waveform in the refresh buffer. This is then written to the screen at the 60 frames/second refresh rate. So you see a greater percentage of all the real cycles which flow through your circuit. But top analog scopes like the Tektronix 2467B or 7104 can display up to half a million waveforms per second, showing 90% of all cycles of your signal, because they have very short holdoff times. They show rarely occurring events dimly for emphasis and are very good at finding elusive faulty pulses! To produce digital storage scopes with equal capabilities, Tektronix introduced the very clever TDS700 series. They can capture and display more than 400,000 waveforms/ second when running at 1GHz using 500 sample points per acquisition, in one channel InstaVu Mode. How is this done? First let’s consider why you can’t just raise the rate at which the conventional digital scope rasterises and displays the signal. We saw that to display 60 updated versions of the chang­ing input signal each second produces a video signal of 18.432MHz. Could we just raise the refresh rate by a factor of 7,000? Would (7,000 x 60) frames/ second capture 420,000 wave­forms/ second? The answer is NO! To do that, a conventional architecture must read the buffer cells in IC6 at (307,200 x 7,000 x 60) = 129,024,000,000 bits/second, giving a video frequency of 129GHz. And the raster would need a vertical or frame rate of 420kHz and a line or horizontal frequency of 201.6MHz. No CRT tube cathode can respond at such a video frequency and the inductance of magnetic deflection coils prohibit such fast sweep rates! So Tektronix produced a revolutionary design. InstaVu acquisition mode For their high performance 4-channel TDS700 digital scopes, Tektronix manufactured a patented high speed dedicated processor and cache mem­ ory. It includes 360,000 transistors formed using 0.8 micron technology into a 304-pin CMOS IC called a Demux, which dissipates 2.5 watts when running at full speed. This is integrated into the acquisition system, duplicating the raster forming capability there, so keeping the required video frequency within manageable limits. Also a section of the very fast main memory is used as a refresh buffer. Here it builds up display images from thousands upon thousands of passes of the signal, including those glitches you seek. And the acquisition section can calculate its own trigger positions. This architecture, shown in block diagram form in Fig.5, is radically different from any other digital scope. The acquiring of more and more samples of the input signal almost never stops. Even while the screen display is being updated and refreshed, the sampler continues acquiring more points of the signal. In this way any elusive glitches, line reflections, jitter or bad pulses have a very high probability of being found by the sampler and shown on the screen. Making good use of available memory bandwidth, the raster­ iser operates on a 16ns clock. It can draw four complete acquisi­ tions at once into a 500 x 256 x 1 bitmap. Drawing is done in top to bottom, then left to The Tektronix TDS784 scope has 1GHz analog bandwidth and each channel samples at 1GS/s. In single channel operation all sam­plers interleave to achieve 4GS/s sampling speed. In InstaVu acquisition mode, this scope acquires 400,000 waveforms/second. The scope has a liquid crystal shutter to provide a colour display and it has an unsurpassed ability to catch and display rare glitches in signal waveforms. right fashion, so each data point in an acquisition need be fetched only once. Each read-modify-write cycle operates on 64 pixels at a time. Each cycle is 32ns long. Data is fetched in groups of eight bytes. Any column of the bit map, 256 pixels high, can be raster­ised in 32 to 128ns. When operating with one input signal in InstaVu mode, each of the four channels take turns acquiring that single input. Three channels can continue acquiring while the formed raster is unloaded in the fourth channel. This architecture raises the performance to 400,000 full screen (500 point) acquisitions and rasterisation cycles per second on one channel. This data rate represents 220,000,000 pixels/ second. The speed is limited by the trigger system rearm circuits as much as by the acquisition/graphics section. The Demux IC demultiplexes and processes the data from all four A/D converters working together on the one signal and ras­terises the acquired data. Also it performs digital signal pro­cessing for local programmability, mathematical algorithms and trigger position calculations. The firmware only intervenes every 10,000 samples to copy out the complete raster which shows the behaviour over that time. Then the acquisition section shifts out a complete bit-mapped image to the video amplifier at the modest frame rate. But as the display shows almost every cycle that ever passes through your circuit under test, the result is equivalent to a continuous running picture of the live signal. The display is so lively that signal aberrations are seen instantly. You have the confident feel of an analog scope yet also have the storage and mathematical powers of digital scopes. Colour gradations highlight sections of the traces which occur less frequently. You can show the continuously repeating part of the signal in red, with brilliant blue highlighting the occasional glitches. If the scope is left in variable persistence mode for many hours, more than 10 billion acquisitions can be amassed if neces­sary to find an elusive faulty signal. The vertical frame rate and the horizontal line rate of the raster display are approximately as described before. References: Tektronix Technical Brief SC 12/94.XBS.15M.55W-10341-0. Acknowledgements Thanks to Tektronix Australia and staff member Ian Marx for data and illustrations. March 1997  81