Fullscreen HTML5Med Res (??MB)HTML4

Pt.8: More Advantages Of Digital Scopes Digital storage scopes excel over analog scopes when displaying multiple inputs or very slow signals. Some DSOs can provide grey scaling or colour gradations to accentuate signal changes while averaging many recurrent waveforms improves the trace and the accuracy of mathematical calculations. By BRYAN MAHER It is common to use an oscilloscope to display two (or more) different signals simultaneously, usually to see the timing relationships between them. Let’s compare how this is done on analog scopes and digital scopes and then we’ll see why the DSO is superior. In Pt.4 (August 1996 issue) of this series, we described how an analog scope can display two input signals in either alternate or chop modes. The switching principle is shown in Fig.1. Briefly, two input signals, channel 1 and channel 2, are individually attenuated and preamplified in A1 and A2. Then a fast electronic switch, IC1, switches back and forth between channels 1 and 2, to select which signal is displayed on the screen. 86  Silicon Chip In alternate mode, IC1 selects channel 1 signal during all of the first sweep, then switches to channel 2 for all the second sweep. Then channel 1 is displayed again on the third sweep, and so on. This is not practical at slow sweep speeds, because the first waveform fades away before the CRT beam has time to draw the second. At faster sweep speeds, the screen persistence con­tinues to show the first waveform while the next sweep displays the second. Although they are actually being displayed alternate­ ly, you see both waveforms on the screen continually. But because each input signal triggers its own sweep of the scope independently, all time relation between the two waveform displays is lost. Comparative timing measurements between traces in alternate mode are meaningless. What should you do? Chop mode You could select chop mode on your analog scope. Now IC1 rapidly switches back and forth between the two channels, typically at a rate of about 1MHz. The screen displays many chopped up segments of both waveforms, as one of the scope screen photos in this article shows. We’ve shown a special case here to show the chopping action. As you can see, both traces are chopped up. This chopping mode is usually not evident because the waveform frequency and chopping speed are unrelated. Normally, all those little segments are blended into two continuous wave­ forms on the screen. One input signal triggers all sweeps, so comparative timing measurements made between the traces in chop mode are valid. But here a second disadvantage of chop mode becomes evi­dent. The screen doesn’t show what happens in waveform 1 while the scope is busy displaying the next short segment of waveform 2 and vice versa. Half of each signal is invisible. In this way you could miss seeing elusive glitches. Fig.1: two channel analog scopes have a fast electronic switch (IC1) to select between channels either at the sweep rate which is called alternate mode or at about 1MHz, which is called chop mode. If you were to set an analog scope to this low sweep speed (and on most analog scopes, you can’t), you would just show one bright green spot, slowly meandering up and down and taking 100 seconds to cross a dark screen. It won’t make much sense. But this sort of waveform is routine to a digital scope. After the signal has executed two full cycles, they will be stored complete in the memory. Then the whole waveform will be continually displayed on the screen, refreshed at the 60Hz rate. You can observe the linearity of the ramp signal by eye or measure it if your DSO supports a mathematical differentiation routine. Results of changes or adjustments can be seen after the next two cycles are complete. Grey scaling Fig.2: A digital oscilloscope displays multiple inputs by indi­vidually preamplifying, sampling and digitising every input signal. The four sets of data are stored in separate areas of RAM before being displayed. So neither alternate nor chop mode is ideal. What other choice is there? Two-gun CRTs having separate electron beams were tried but their mechanical alignment proved impossible. Cossor split-beam tubes displayed two inputs validly at any speed but were limited to two signals only. Today, to investigate timing diagrams in digital circuits, you might need four simultaneous input channels at fast sweep speeds. The only satisfactory answer is to buy a digital storage oscilloscope. Multiple inputs Digital scopes can successfully display two, three or four separate input signals simultaneously, at any sweep speed, using a very different technique. The block diagram of Fig.2 gives us an idea of how it’s done. Each input signal passes through its own attenuator and analog preamplifier, shown as A1 to A4. From there, each signal is individually sampled and converted in separate A/D converters A/D1 to A/D4. All the digital data from each channel is separately stored in different areas of the fast random access memory (RAM). The process of reading the contents of the RAM to its dis­play on the screen is complex, especially in Tektronix scopes using InstaVu mode. Suffice to say that neither chop nor alter­nate procedures are used, and the whole of each waveform is displayed on the screen. Everything recorded in the RAM is faithfully shown; nothing is lost. The process operates equally well at all sweep speeds, slow or fast. All timing measurements made on the screen and the phase relationships observed are accurate. In displaying multiple input signals, a digital storage oscilloscope is vastly superior to all analog scopes. Low frequency displays If you need to display long pulses or ramp signals, you’ll find digital scopes much better than analog scopes. Say you want to observe a ramp signal with a period of 50 seconds. Setting the timebase to 10s/div, the scope would take 100 seconds for one sweep across the screen. That would display two full cycles of the waveform. In the past, your trusty analog scope easily displayed com­pound signals, for example live TV waveforms or digital data which contained intermittent faulty pulses. Your display was brighter in those parts of the signal which repeat more frequent­ly, because at those points thousands of traces were overlaid. Sec­tions of the waveform which continually changed or occurred less often thus appeared less bright. These brightness gradations let you identify rarely occur­ ring spurious interferences or runt pulses. On the screen they looked different from the normal repetitive signals. Point one in favour of analog scopes! But the simple digital storage oscilloscope we discussed in last month’s issue (Pt.7) can’t do this. Remember that is had a 1-bitmap refresh buffer and as such, it could not display signals at varying intensity. The one-bit output has only two possible values, digital high or low. These correspond to the points on the screen being illuminated or not; on or off. So in that simple sort of DSO we saw in the previous chapt­ er, everything has the same intensity on the screen. But ideally we want a digital storage scope to be at least as good as analog scopes were in showing compound signals. With that in mind, we would like 16 levels of brightness in the trace. Frequently recurring parts of the signal should be bright­ er than infrequent anomalies and faulty pulses. April 1997  87 These two analog oscilloscope photos show the same pair of sign­als depicted in alternate mode (left) and chop mode (right). The problem with alternate mode is that because each alternate sweep is separately triggered, the precise time relationship between the two waveforms is lost. In the chop mode, by contrast, the two signals have sections chopped out and this can lead to glitches being missed in the display. To achieve this aim, digital oscilloscope designers en­ larged the bit map refresh buffer to store four bits (instead of one previously) in each of its memory locations. We imagine this structured as four planes of memory elements, as illustrated in Fig.3. Each plane is like the single bit memory map depicted last month and contains 307,200 memory cells, arranged in 480 rows, each row containing 640 cells. In each plane, each cell contains a single digital value, 1 or 0; ie, either logic high or low. As we saw in the previous chapter, the XY address of each cell corresponds to one particular point on the CRT screen raster. In Fig.3, all four planes of the refresh memory are ad­dressed in parallel. For example, the top left cells in all planes have the same address. So when the system reads the top left address of the re­fresh buffer, it reads the contents of the top left cell in each plane simultaneously. The output is then 4-bit data (one bit from each plane) carried on four parallel lines A, B, C, D. That 4-bit digital data is used to control the bright­ness of the spot on the screen, by changing the G1-K bias potential on the CRT cathode. But that tube is an analog compon­ent, so it requires a varying analog voltage signal on its cathode to alter the electron beam current and trace brightness. Therefore, the 4-bit digital data read from the bit map refresh buffer on lines A, B, C, D must be converted to an 88  Silicon Chip analog signal in the digital to analog (D/A) converter (IC7). D/A converter IC7 contains four CMOS switch elements, SA, SB, SC & SD, powered by an accurate +5V reference. Each switch produces output exactly equal to +5V if its input is logic high or exactly 0V if its input is logic low. The resistor group between IC7 and IC8 forms an R-2R ladder attenuator; resistors mark­ ed 2R have twice the value of those marked R. The combination of IC7 and the resistor ladder produces an analog voltage proportional to the value of the 4-bit digital data fed into IC7. This signal is then raised to a high level and in­verted by video amplifier IC8, which is DC-coupled and must have a high input impedance. This high voltage analog signal from IC8, applied to the CRT cathode, controls the electron beam current and thus the screen illumination at that point. This is called Z-modulation. Thus the trace brightness at each pixel is set to a value repre­senting how often that element of the signal appears at the scope input. Four-bit digital data can take only 16 different val­ues. So this scheme allows the trace on a digital scope to dis­ play compound signals in 16 different levels of brightness. This is called “grey scaling”. When the display processor IC5 meets a regularly occurring part of the input waveform, it writes a logic high at the appro­priate memory address in all four planes of the bit map refresh buffer IC6. When read from the refresh buffer, the output data on the four parallel lines A, B, C, D will be 1111. The D/A converter IC7 converts this to the maximum analog voltage and the CRT produces the brightest spot at the corresponding point on the screen raster. Now let’s suppose a spurious pulse appears only sometimes at the scope input. Sensing this fact, the display processor IC5 might write a logic high to the corresponding address only in memory plane A of the refresh buffer IC6, and write a logic low to the same address in planes B, C and D. On the next refresh cycle, when that data stored in the refresh buffer is read, the digital data on output lines A, B, C, D will be 1000. This corresponds to a screen dot of half bright­ ness. This indicates that that part of the signal appears less frequently; so you suspect it’s some spurious blip or a faulty pulse. There are a number of variations on this theme in modern digital scopes. When variable persistence is selected, rapidly changing waveform points can gradually decay through 16 levels of brightness. Some cheaper models support only two levels of grey scaling. You might ask where’s the advantage of digital scopes, when all analog scopes naturally showed brightness scaling? The answer is that DSOs support grey scaling at all sweep speeds equally. But normal analog scopes, at top speed, are flat out providing a visible trace even on repetitive signals, with no potential left for scaling. Colour display Some digital scopes can show a colour graded display, with different colours indicating how frequently some part of a com­pound waveform repeats. The very high frequency 50GHz Tektronix 11801B uses a 228mm diag­onal screen with a vertical raster scan. The display resolves 552 pixels horizontally and 704 pixels vertically, from a palette of 262,144 colours. Early colour scopes used colour TV technology. The tube contained three electron guns and the familiar tri-colour phosphor and beam convergence shadow mask. But a monochrome CRT is capable of a much sharper trace than any TV tube with a multi­ple colour phosphor. Therefore many modern colour digital scopes use a white phosphor CRT, overlaid by a three-layer liquid crys­tal colour shutter. An example of this is the Tektronix model TDS684B which provides horizontal raster scan on a 177mm screen featuring full colour grading from a palette of 256 colour lev­els. Signal averaging Analog signals may be corrupted by extraneous interference which results in a noisy display. Worse still, noise in the signal reduc­es the accuracy of mathematical operations performed These two scopes from Tektronix both use a white phosphor CRT, overlaid by a 3-layer liquid crystal colour shutter. Both models are showing colour graded displays, with different colours indicating how frequently some parts of the waveforms repeat by the oscilloscope. The way around this is to feed the noisy signal through your digital scope many times. Then you display the average of many passes of the repetitive input signal. Each pass will contain different noise, but random (white) noise averages out towards zero. So the average of a number of passes of the same signal will be more like the original uncor­rupted waveform. Say your digital scope takes a record consisting of 500 samples at each pass of the signal. We saw previously how the A/D converts each sample to an 8-bit digital word which represents the Fig.3: for grey scaling, the bit map refresh buffer contains four memory planes A, B, C & D. In each plane, each cell stores one bit. So four planes store 4-bit data. IC7 and the R-2R ladder form a D/A converter. IC8 is a linear amplifier. April 1997  89 Repeated from the February 1997 issue, these two oscilloscope waveforms show how the use of averaging can remove much of the noise in a repetitive signal. These two digital screen printouts show the menu setups necessary on a Tektronix RDS 360 digital scope, in order to obtain a two-level greyscale signal. The video signal is an off-air TV chan­nel. Note the use of “vector accumulate” and “contrast” menu options. The main trace is a normal video line signal while the background signal accumulation shows the variation in signal of a period of 1.5 seconds. Note the faint spurious sync signal in between the two main sync pulses. This faint signal is a ghost of the sync pulse. Such a faint signal is unlikely to be shown on an analog scope. nearest voltage decision level below the sample voltage. In real life more than two passes of the signal are aver­aged to obtain smoother results. Averaging four passes of an 8-bit signal yields 10-bit digital data. And eight passes results in 11-bit data. Many scopes let you choose the number of passes that will be averaged; eg, 2, 4, 8, 16, etc up to 2048. But they only keep the result of 11 bits and discard any further overflow. Of course, all normal averaging requires the signal to be repeti­tive. High resolution mode Some of the Tektronix TDS series 90  Silicon Chip scopes also feature a clever system called Hi-Res Mode which allows averaging, to reduce interference and noise, even on one shot signals. In these scopes the sampler always runs at the maximum speed. In normal mode, if you choose slow sweep speed the scope cannot use all the millions of samples taken. So only enough of the samples are kept to form the best display and the rest are thrown away. But in Hi-Res Mode the excess samples are kept in a section of the memory. There each group of 16, 32 or 64 contiguous sam­ples are averaged to form one point on the display. Such a point can be accurate to 12 or 13 or more bits. This process is repeat­ed over all the waveform until a whole screen-full is set up, then displayed. The slower the sweep speed in use, the more excess samples are available for this fast averaging. But of course when you select top sweep speed, Hi-Res Mode is unavailable, because all samples taken are needed to form the normal display. References: Tektronix: Technical Brief SC 12/94.XBS.15M. Acknowledgement Thanks to Tek­tron­ix Australia for data and for some of the illustrations used in this article.