Design Feature: July 4, 1996

Digital storage oscilloscopes (DSOs) provide considerably more than simple views of waveforms. The instruments perform such functions as limit testing and calculating FFTs. Consequently, understanding DSO features and performance well enough to wisely choose an instrument takes more time than many prospective purchasers feel they have. Buyers need to look beneath the banner specs, however. Considering only the number of channels, the bandwidth, the sample rate, and the price is a prescription for eventual disappointment. Although researching a measurement tool can seem like a bother, the effort saves time and money in the long run.

Part of the reason that many engineers dismiss scope selection as trivial is that, on the surface, DSOs seem similar to the analog scopes that EEs have used and loved for decades. Although a DSO and an analog scope perform the same basic function, the two instruments differ in many important ways. You can use a DSO without understanding how it works, but an understanding can keep you from misinterpreting what appears on the screen.

In all but a few DSOs, the input signal first passes through an attenuator and a preamplifier. These sections, which are similar to the corresponding sections of an analog scope, are among the DSO's few purely analog functions. The signal's next stop is an ADC, which has no counterpart in an analog scope. Most, but not all, DSOs introduced in the last few years have an ADC for each channel. The ADC is one of the more expensive parts of the scope, and a manufacturer can save money by having two or more inputs share an ADC.

With a shared ADC, a high-speed analog switch connects several signals in turn to the ADC input. When the highest frequency present in each signal is a small fraction of the rate at which the ADC samples each signal, shared ADCs cause no problems. But, when a signal's frequency content exceeds about 10% of the rate at which the ADC samples each signal, a shared ADC can make simultaneous events appear to occur at slightly different times.

Ganging up to sample faster

Many DSOs that have an ADC per channel reassign unused channels' ADCs to active channels, thereby increasing the sampling rate. A four-channel scope might digitize at 500M samples/sec when you use all channels, at 1G sample/sec when you use two channels, and at 2G samples/sec when you use a single channel. The converters take samples at staggered times, and the scope interleaves the conversion results.

Some channels of newer DSOs lack certain elements. You often see the number of channels in such scopes listed as "2+2." The 2+2 terminology is common in analog scopes, where it usually means that two of the four channels lack input attenuators and so can handle only relatively low-voltage signals.

In DSOs, 2+2 does not have a standard meaning, and vendors use the term in varying ways. In one definition, all four channels have full attenuators and input amplifiers, but, despite the four inputs, there are only two ADCs. In such scopes, you can switch any two of the input signals to the ADCs. This architecture differs from the shared-ADC architecture described earlier. In a 2+2 scope that has four inputs and two ADCs, the ADC stays connected to the signal you select. In the shared-ADC scope, the ADC switches rapidly among the input signals.

The sampling theorem

DSOs are sampled-data systems. The sampling theorem states that to recover the original signal from an ensemble of samples, you must take, on average, more than two samples during the period of the highest frequency component whose amplitude you consider significant. Nevertheless, many DSOs specify a -3-dB bandwidth that is considerably higher than the sampling rate.

Normally, such a situation would lead to aliasing: introduction into the reconstructed signal of spurious low-frequency components not present in the original input. However, many DSOs avoid aliasing by equivalent-time sampling (ETS). ETS works only when signals are repetitive. To be compatible with ETS, a signal need not repeat at regular intervals, however. That is, the signal need not be periodic, provided that its shape remains the same on each repetition.

There are two types of ETS: random and sequential. Some vendors call random ETS "random-repetitive sampling." Random ETS is much more popular than sequential. Although sequential-ETS scopes offer bandwidths as high as 50 GHz, most such scopes are quite specialized. These DSOs acquire one sample per input-signal iteration, incrementing the sampling point through the waveform on successive iterations. This process makes for slow acquisitions, however. Sequential ETS translates input-signal frequencies downward. Mathematically, the process is equivalent to mixing a signal at one frequency with a second signal at a frequency slightly lower than the first. Everything following the sampling gate operates at a modest frequency.

Handheld DSOs: powerful yet small

As with PCs, when you replace a DSO's CRT with an LCD, you open a world of packaging possibilities. In portable PCs, although the laptop form factor has become dominant, PC marketers have learned that the laptop package doesn't please everyone. In the scope world, the layout of work spaces varies more than that of PC work areas. Much fewer scope users than PC users sit squarely in front of their screens. Because even the best thin-film-transistor LCDs can't equal CRTs' wide viewing angles, LCD DSOs are not for every user. Still, dramatic improvements in flat-panel-display viewability and rapidly declining costs will make LCDs ever more prevalent in benchtop DSOs. Hewlett-Packard already offers color LCDs as options to monochrome CRTs in its 54500 series of high-performance benchtop scopes and in its 54620 logic-analyzer series, an outgrowth of its 54600 DSO line. Gould announced its first benchtop LCD DSO several years ago. If you let your imagination loose for a little while, you can think of many package shapes that take better advantage of a flat screen than does the rectangular box of a typical benchtop DSO. At first, there seems to be little sense in simply replacing the CRT with a flat-panel display and wasting most of the space behind the panel that the CRT formerly occupied. However, the box shape is well-accepted because it does the job—not merely on the bench but also in systems. So, most vendors have been reluctant to stray far from the familiar shape in packaging LCD DSOs for benchtop use. When you package a handheld instrument containing a fairly large LCD, the requirements and limitations are quite different. Nevertheless, at this writing, only one vendor—Hitachi—has introduced a small, portable LCD DSO packaged as anything other than a flat box with the screen on one of its larger surfaces. The Hitachi units, beginning with the $2795 VC-5430, are in clamshell packages, which resemble miniature laptop PCs in which knobs and signal connectors replace the keyboard. Moreover, unlike the displays of other handheld LCD DSOs, Hitachi's displays appear in vibrant color. Many differences Notwithstanding their almost universal use of the flat-box shape, handheld LCD DSOs exhibit many differences in control placement and operation and in the orientation of the instrument during use. But, the most important differences lie in performance and features. Tektronix's introduction last year of the THS 700 series of TekScopes radically changed the fairly new and rapidly growing market for handheld DSOs. Tek asserts that the TekScopes were the first products to bring benchtop-scope performance to handheld LCD DSOs. The units include a real-time-sampling ADC per channel. The top-of-the-line THS 720 provides 100-MHz bandwidth. In addition, the TekScopes offer some features not found in conventional benchtop DSOs—channels that are ohmically isolated from each other and from the chassis, for example. The isolation voltage is 600V dc or peak ac. Pricing begins at $1795. The TekScopes also offer a bright, high-contrast, monochrome display. Before the advent of the TekScopes, most designers of monochrome-LCD DSOs had opted for long battery life over brightness. To achieve its high-brightness displays, Tek somewhat compromised battery life, to 1¾ hours. This year, archrival Fluke Corp announced new Model B versions of its ScopeMeter Series II line. These units also feature bright, high-contrast displays but make no compromise in the four-hour battery life between charges. The ScopeMeters can also operate from alkaline cells. Pricing for the new versions begins at $1495.

Random-sampling DSOs, which are more common, typically use fairly fast ADCs. For example, many scopes with a 150-MHz, -3-dB bandwidth take 25M samples/sec. These scopes' sample-to-sample interval, or sampling period, is 40 nsec. Because the waveforms you observe with such scopes usually last longer than the sampling period, the scope can acquire more than one sample each time the waveform repeats. Nevertheless, the acquired waveforms usually also contain high-frequency components whose duration is much shorter than the sampling period.

Because the ADC clock runs freely, synchronizing the sampling points with the signal is impossible. This lack of synchronism is important and beneficial; it forms the basis for the "randomness" in random sampling. Whether or not the scope's sweep-trigger circuits detect a trigger event, the ADC continues to digitize the input signal. The sweep-trigger circuits do not trigger sweeps in the way analog scopes' sweep-trigger circuits do. But, from a user's perspective, the effect is the same.

The scope first stores the digitized samples in a circular buffer, which some vendors call the capture memory. As new samples arrive, they overwrite the oldest samples. The sweep-trigger circuit's detection of a trigger event stops the digitized samples from entering the capture memory. However, the ADC continues to make conversions.

When a trigger condition stops new samples from entering the capture memory, the memory contents are frozen until the scope copies the data into a second memory, known as the waveform or processor memory. You can choose to freeze the capture memory when it contains a record of what happened just before the trigger event, immediately after the trigger event, or both before and after the trigger event—the most common choice.

The ability of random- and real-time (RT)-sampling DSOs to record what happened before a trigger event is a powerful feature. Many analog scopes and some sequential-sampling DSOs delay the signal in a lumped or distributed delay line to provide a pretrigger view. However, this view is limited compared with that possible in random- and RT-sampling DSOs.

A random-sampling DSO also includes a time-to-digital converter that digitizes the time between the trigger event and the next sample. The accuracy with which the scope digitizes this time interval is crucial to proper reconstruction of the original waveform. Although it measures the timing of a sample that does not enter the capture memory, the scope can use the sample-clock period and the delay between the trigger and the first discarded sample to infer the position in the waveform of all samples in the capture memory.

As successive trigger events occur, the waveform memory accumulates an ensemble of digitized voltages, only a small number of which occur during any single waveform iteration. Each sample is time-stamped with its relative position in the waveform. The scope produces a waveform record by placing these samples in order, based on their position in the waveform. Positioning the samples relies on information about the time of each sample relative to the trigger event.

When the waveform contains enough samples—that is, when there are no significant gaps between samples—the scope can display the captured waveform. In a scope having 150-MHz bandwidth and a 25M-sample/sec ADC, the maximum sweep speed might be 5 nsec/div, or 50 nsec/sweep. If the display's horizontal resolution is 500 points, the waveform memory should accommodate at least 500 samples. If the memory contains just 500 samples, the effective sample rate is 0.1 nsec/point or 10G samples/sec, 400 times the actual sampling rate. The displayed waveform comprises samples gathered during hundreds of waveform iterations.

A DSO in your PC

A DSO need not be a benchtop unit or a handheld device. Some units have no displays of their own. Units configured as plug-in boards for PCs are but one example. Others include VXIbus modules. Although PC-based DSOs don't seem poised to displace more conventional units from dominance in labs or in the field any time soon, the boards are finding a niche in specialized systems, such as those used in manufacturing test. What is surprising is the sophistication that these units pack onto single, fairly small boards. For example, both Emulation Technology and Gage offer boards that perform random-repetitive sampling. Emulation Technology's prices start at less than $1000. In the random-sampling mode, Gage's CompuScope 2125/ETS effectively takes 2G samples/sec on repetitive waveforms. The two-channel board offers 125-MHz bandwidth and also operates in the real-time mode, providing as much as 1 Mbyte of waveform memory. As with benchtop scopes, when you use only one channel, you can interleave samples from both of the board's ADCs. Thus, the real-time sampling rate becomes 250M samples/sec. This performance comes at a price though; with a 256k-sample memory, the board costs $5495. Keithley Metrabyte offers several units that do sequential ETS and one that takes 1G sample/sec in real time, far and away the highest real-time sampling rate of any PC-based DSO board. This board, the single-channel DAS-4301 costs $6995 with an 8k-sample memory. Like 200G samples/sec PC Instruments' boards also perform sequential ETS at effective rates to 200G samples/sec. The boards offer bandwidth as high as 300 MHz with 50(ohm) input impedance (100 and 200 MHz with 1 M(ohm)). The company makes dual- and single-channel boards at prices from $1495 to $2995. If you start with a one-channel board, you can upgrade to two channels within one year of purchase for the list-price difference plus $295. Link Instruments offers boards having as many as four channels. The four-channel DSO-28464 takes up to 200M samples/sec in real time. The board offers 125-MHz bandwidth and 64k-samples/channel of memory (128k samples/channel when you use two channels). Moreover, the board includes a 16-channel, 100-MHz logic-timing analyzer and FFT/spectrum-analyzer software. The price is $3299. A two-channel version costs $1999. Although 8-bit resolution is common in benchtop and handheld DSOs and in most PC-based DSO boards, Signatec's PDA12 PCI bus boards resolve 12 bits and provide an auxiliary bus that can transfer 200 Mbytes/sec. The top-of-the-line board takes 50M samples/sec on each of two channels or 100M samples/sec on one channel. Memory depths are 256k and 1M 12-bit samples per channel. Pricing starts at $5500 for a version that takes 40M samples/sec on two channels and 80M on one.

First this... then this...

Placing the samples in order and creating the display take time. Even if the transfers of samples from the capture to the waveform memory take place in zero time (they don't), displaying the waveform takes much longer than the product of the number of samples and the minimum interval between samples.

In the example, the scope reconstructs the waveform from 500 samples, some taken 40 nsec apart but many taken at much longer intervals. Therefore, the minimum acquisition time is 20 µsec, but the actual time is longer. Once the 500 points are in the waveform memory, putting the display on screen usually takes milliseconds. Generally, DSOs do not update their displays more than about 100 times/ sec.

Compared with random sampling, RT sampling is conceptually simple. In an RT DSO, the digitized samples can go directly to the waveform memory—if the memory is fast enough. Just as with a random-sampling DSO, the ADC makes continuous conversions, the memory acts as a circular buffer, and a trigger event stops new samples from overwriting the oldest samples in memory.

Like a random-sampling DSO, an RT DSO can display pretrigger information. When the memory contains the desired ensemble of samples, the scope converts the samples to the display's pixel format and places this information on screen. Most DSOs use a raster-scan type of display, and converting the samples to the display format is called "rasterization."

RT DSOs offer several advantages over random-sampling units: RT scopes provide complete records of single-shot transients that ETS scopes can't capture. An important subclass of transient events is the low-duty-cycle anomaly in a normally repetitive waveform. Because ETS displays comprise an amalgamation of samples from multiple waveform iterations, the displays mask such transient anomalies.

An RT scope obtains a complete record from a single waveform occurrence. Therefore, small differences among successive waveform iterations need not superimpose what appears to be noise on the signal. The differences are not noise but are a real property of the signal. Still, many users interpret the display as indicating noise in the scope. Thus, an RT scope can appear quieter than a random-sampling instrument.

RT scopes also have drawbacks, however. For the most part, RT scopes cost more than random-sampling scopes that provide equivalent bandwidth. Nevertheless, RT scope prices begin at attractive levels. For example, Tektronix's TDS 300-series pricing begins at $2295. RT scopes usually take fewer samples per cycle of high-frequency phenomena. Consequently, despite the use of such DSP techniques as reconstruction filtering, the scopes don't always provide as much waveform detail as random-sampling units do. In addition, RT scopes cannot increase their effective sampling rate beyond their real sampling rate, so they are more easily subject to aliasing.

Despite their conceptual simplicity, RT scopes are more difficult to build than random-sampling DSOs that capture waveforms of equal bandwidth. The major complexity of random-sampling DSOs lies in the rather well- understood digital do-main, whereas that of RT DSOs lies in the very fast ADCs, which many designers only half-jokingly insist employ black magic.

Reconstruction filtering

Theoretically, you can reconstruct an analog waveform that you have sampled incrementally faster than twice the highest frequency present in the waveform at a significant amplitude. Actually, getting a decent picture of the waveform requires sampling at least four or five times the highest signal frequency. Scopes that sample at four or five times the highest signal frequency must use reconstruction filtering, a technique that uses DSP algorithms to infer waveform values at points between samples. To avoid such filtering, a scope must sample at more than 10 times the highest signal frequency, and the results still don't usually look as good as those obtained by filtering samples taken less than half as often.

Using waveform reconstruction, an RT version of the 150-MHz bandwidth random-sampling scope in the earlier example requires ADCs that capture at least 600M samples/sec on each channel. This speed is 24 times that of the ADCs used in the random-sampling scope. Without reconstruction filtering, the required sampling rate jumps to 1.5G samples/sec. Most DSOs use ADCs that resolve 8 bits. ADCs that take 600M 8-bit samples/sec are within the state of the art. At least one company builds 8-bit flash ADCs that take 2G samples/sec. DSOs that take 4G samples/sec—and more—in real time do so by interleaving multiple slower ADCs.

Reconstruction filtering is computationally intensive and, in most cases, penalizes the scope's display-update rate. Nevertheless, most scope designers feel that the trade-off is worthwhile. Reconstruction filtering need not complicate a scope's hardware design. By allowing the use of slower ADCs or by making possible greater bandwidth with only a small increase in the complexity of the signal-conditioning circuits that precede the ADC, the technique simplifies the scope's challenging analog design.

Note that, when it performs repetitive sampling, a scope generally does not need to use reconstruction filtering and does not do so. Instead of inferring waveform values between samples, the scope obtains real values, albeit from different iterations of the waveform.

No guarantees

Despite RT scopes' advantages in transient capture, the use of RT sampling does not guarantee that a scope can catch transient phenomena. One reason is that, because most DSOs update their displays fewer than 100 times/sec, the scopes are "blind" much of the time.

An RT scope operating with a 50-nsec sweep time and a 100-update/sec screen-update rate displays only 0.0005% of the activity of the system under test. If you now postulate that the system under test misbehaves only a small faction of the time (say, 0.01%), you can arrive at an extremely pessimistic estimate of the likelihood of the scope's catching the anomalous behavior. In this case, the product 0.0005%×0.01%=5×10^-10, or one chance in 2 billion. If the system under test cycles every 100 µsec, or 10,000 times/sec, this calculation predicts that the scope can catch a malfunction every 2×10⁵ sec, or about every 2.3 days, of continuous testing.

Although the probability that the scope can catch an anomaly during any sweep is low, estimates such as this one are unduly pessimistic. In this case, because it updates 100 times/sec, the scope monitors 1% of the system cycles and catches about 1% of the anomalous cycles. If the system cycles 10,000 times/sec and 0.01% of the cycles are anomalous, a malfunction occurs once/sec, and the scope catches a malfunction roughly every 100 sec.

Sometimes, you wait

In too many cases, though, the wait to catch a failure is much longer than 100 sec. In 1994, with the introduction of the proprietary InstaVu feature, Tektronix dramatically improved the speed at which DSOs capture transient failures.

One reason that InstaVu is so successful is that the message behind it is clear: Instead of capturing 100 waveforms/sec, Tek scopes in the InstaVu mode capture as many as 400,000 waveforms/sec. (Only faster analog scopes rival this number.) Because of the higher duty cycle, InstaVu scopes can capture phenomena in 1 sec that conventional DSOs might wait an hour to acquire. InstaVu scopes can capture so many waveforms per second, because the scopes use a proprietary IC to rasterize the data as soon as the samples are digitized. All waveforms acquired between screen updates are, thus, stored in a pixel map and displayed at the next update.

The InstaVu message is compelling, and Tek has marketed it superbly. Competitors, which cannot copy InstaVu because of Tek's patents, have been hard-pressed to articulate messages that prospective buyers find equally compelling. Nevertheless, InstaVu isn't the only way to simplify acquiring elusive transients. For example, the message behind LeCroy's exclusion-triggering feature is that, instead of displaying more waveforms, scopes with exclusion triggering display the waveforms that count.

Exclusion triggering keeps the scope from capturing normal waveforms; triggering occurs only on the waveforms whose time parameters are abnormal—and the scope can automatically determine what is abnormal. Moreover, the scopes store complete waveform records from which you can determine waveform parameters and perform statistical analyses. The scope displays the anomalous waveforms individually or in superimposed views and can display histograms of the parameters. Sometimes, the waveform statistics reveal several distinct failure modes that require separate corrective measures. With InstaVu, if you want to analyze or keep copies of waveforms, you must return to the conventional real-time-acquisition mode and acquire additional records.

In the exclusion-trigger mode, a scope need not update its display after each acquisition. You can choose to have updates take place only after the acquisition of many anomalous waveforms. This display mode keeps delays associated with unnecessary screen updates from slowing waveform acquisition.

DSOs can commonly perform both RT and repetitive sampling. But, how a scope should handle the changeover between modes is a subject of spirited debate among DSO designers. Scopes such as Hewlett-Packard's 1G-sam-ple/sec 54615B ($5595), which provides 250-MHz bandwidth when sampling in real time and 500 MHz when performing ETS, and the 2G-sample/sec 54616B ($6495), whose bandwidth extends to 500 MHz with either sampling technique, automatically switch to RT sampling as you decrease the sweep speed. HP maintains that users don't care which acquisition mode a scope uses.

Not enough samples

However, when a scope is doing RT acquisition, users must keep certain facts about sampled data systems in mind: All DSOs have a finite memory depth. As you decrease the sweep speed, the scope moves the waveform samples apart, so that a fixed number of samples can cover the entire sweep period. In many DSOs, the ADC always samples at the maximum speed, but, as you lower the sweep speed, the scope discards an increasing fraction of the samples before they reach the waveform memory. From the user's viewpoint, the scope might just as well not take the samples it discards.

If a scope simply discards samples, the highest frequency the scope can display without aliasing declines as the interval between the retained samples increases. The scope can, however, avoid aliasing by using peak detection. Some vendors call this mode "min-max," because, unlike normal peak detection, the mode detects both the highest and lowest values that occur during some interval.

Some scopes use an analog approach to mix-max mode, but an all-digital implementation is possible. A single X-axis position on the screen (one column) acts as a proxy for a group of samples. The scope examines every sample in the group and retains the highest and lowest value samples, illuminating an array of pixels in the column that spans the full range of sample values.

You might think that scope manufacturers would make the min-max mode the default—that they would force you to take some action to select a mode that allows aliasing. Usually, however, the opposite is the case, because the min-max mode has a down side: It produces "fat" traces, which many users misinterpret as symptoms of a noisy scope. So, if your scope switches from ETS to RT sampling with little or no warning and does so in a way that permits aliasing, you must be vigilant, lest your scope display aliased low frequencies that don't exist in the input signal. A DSO's potential for such misleading displays makes some EEs nervous about digital scopes and prompts others to cling tenaciously to their old analog scopes.

Glitch capture

Glitch capture is another name for an all-digital min-max mode. When a glitch brings a signal outside its normal envelope for a period that includes at least one ADC sample at the highest rate, a scope offering this mode displays the glitch. However, even scopes that include this capability don't necessarily display all glitches that you might want to see. For example, suppose that one full cycle is missing from what is otherwise a continuous sine wave. If the sweep speed is such that you can't discern individual cycles on the screen, the display does not reveal the missing cycle.

The main alternative to min-max mode is deep memory. The deeper the scope's memory, the slower you can make the sweep before the scope must start to reduce its effective sample rate. Memory depths vary over an enormous range, from 500 points (or even fewer in some handheld DSOs) to 8 million in some of LeCroy's 9300-series units.

Interestingly, when you display an entire long record on a single screen, most deep-memory scopes use display technology similar to that of the min-max mode: Each display column serves as a proxy for a group of samples, and the scope illuminates all pixels in the column that correspond to sample values in the group. A deep-memory scope does not discard samples, though. Therefore, the scope lets you detect, for example, the aforementioned missing cycle in an otherwise-continuous sine wave. Because the memory contains the complete waveform record, you can zoom in on short segments and view them in detail.

Deep memory

LeCroy has been extolling the advantages of deep memory for many years. Deep memory certainly adds to a DSO's cost, but manufacturers are recognizing that the feature's value outweighs the expense. LeCroy's 9300 series still offers the deepest memory in benchtop scopes. One model, the $27,490 9374L (500-MHz bandwidth), provides 2M-sample/channel memories that combine into an 8M-sample memory when only one channel is active. This feature nicely complements the interleaving of unused ADCs with ADCs of active channels. As you turn channels off, the memory depth and sample rate increase together. At a constant sweep speed, therefore, the records represent a fixed length of time.

If you order deep-memory option 1M, $20,445 buys 130k samples/channel in Tektronix's top-selling, 500-MHz-bandwidth TDS 744A. For $37,490, you get the same memory depth in the 1-GHz-bandwidth TDS 784A. Both units are four-channel InstaVu scopes. These scopes also let you assign ADCs and acquisition memory from unused channels to channels that are in use. So, if you use only one channel, the memory depth is 500k samples.

The ability to display long records isn't the only benefit of deep memory, though. A popular feature of many mid- and top-of-the-line DSOs is the ability to calculate FFTs. An FFT's resolution, or how well it separates energy at closely spaced frequencies, depends on the number of samples used in the calculation. LeCroy offers the option of processor memory deep enough for 6M-sample FFTs. Despite capturing waveform records that are considerably longer, some competing DSOs use only 10,000 or 50,000 samples in FFT calculations. When users of these scopes want high-resolution FFTs, they must transfer waveform records to a computer and make off-line calculations. If the previous FFT determines the conditions of the next test, this added step significantly slows testing.

Long memory, low price

A little over a month ago, HP came roaring into the deep-memory arena. The company's announcement of the 54645 series brought 1M-sample/channel memory to a price range in which it had not previously been available. HP offers two versions of the 54645. The 54645A is a two-channel scope priced at $3495. The 54645D ($4995), is both a two-channel scope and a 16-channel logic-timing analyzer. Both versions' 100-MHz-bandwidth analog channels take 200M samples/sec. Thanks to the deep memory, the scopes can capture 5 msec of real-time data on both channels, even at the fastest sweep speed. Moreover, the scopes also perform random-repetitive sampling.

In these units, HP introduced its MegaZoom technology, an elaboration of a design the company first introduced in the original 54600-series DSOs. A major goal of the 54600 series is to duplicate the live feel of analog scopes. Explaining how successful HP has been is not easy; to appreciate the results, you need to use the scopes—preferably side by side with competitive units.

Describing HP's implementation is easier. The scopes incorporate several µPs including a proprietary IC. The multiprocessor architecture divides the functions—display management, front-panel controls, and internal housekeeping—among the dedicated µPs. HP's design is in the vanguard of manufacturers' efforts to improve DSO responsiveness. Other vendors may not adopt a multiprocessor approach, however. Competitors claim that, because extremely powerful µPs have become commodity items, new designs may need only a single standard µP to achieve order-of-magnitude improvements in response speed at affordable prices.

User-interface issues

Though important, response speed is just one of many issues that relate to DSO user interfaces. Although you can buy many DSOs without trying—or even seeing—them first, user-interface issues suggest that you can put the time you spend evaluating DSOs to good use.

User-interface issues include factors besides speed of response, display resolution, color vs monochrome presentation, and what's on the menus. Two of the biggest issues relate to the means the scope provides for archiving waveforms. With analog scopes, the scope camera reigned supreme. If you found a waveform that you felt was significant, you'd photograph it and keep a copy in your lab notebook. You might also make photocopies to circulate in reports or technical articles.

Scope cameras started to become obsolete with the advent of DSOs having RS-232C or IEEE-488 interfaces. Such interfaces let you download data to a computer for manipulation, analysis, display, and storage. Lab notebooks and reports are now much more likely to contain computer-generated printouts than scope photos. Even five years ago, the software that generated the printouts was probably something you wrote yourself. Now, the software is likely a commercial package, possibly from the scope vendor, but often from a third party.

Floppy-disk drives have now begun to eclipse cabled connections for getting data from DSOs to PCs. The reason is convenience. Even in R&D labs, where networks of PCs and workstations abound, carrying a diskette from a scope to a PC is usually easier than making a connection between the back of the scope and the back of the PC. Also, if you use diskettes, you needn't worry about discovering that all of the PC's ports are in use.

PCMCIA memory cards are another popular option for transporting data from scopes to PCs. Although memory cards are much faster than floppies, and the cards' greater cost per megabyte of storage is unimportant in applications in which you reuse the cards, floppy drives remain more popular than PCMCIA slots on desktop PCs. And, desktop machines still outnumber notebook PCs in labs. So, for the moment, PCMCIA cards' principal use with DSOs is for transporting data to portable PCs.

Even if you store waveforms on your PC's hard drive and use a commercial software package to generate printouts for reports, you may still want to review lots of waveforms during an experiment. For this purpose, you may want a scope with a built-in printer. Although most such printers use the same sort of flimsy, heat-sensitive paper that fax machines use, you may find a sheaf of printouts that you can annotate and immediately review more convenient than a series of files that you can view only with the aid of the scope or a PC.