It’s well known that one person may find a particular digital storage oscilloscope (DSO) difficult to operate while another will not. EE-Evaluation Engineering ran a hands-on comparison of three leading DSOs to gain insight into this ease-of-use issue.
Some basic rules governed our evaluation of scope operation:
- All scopes submitted were current production models with off-the-shelf options. No special software was installed.
- No company representatives were on hand for any of the testing.
- All test conditions and comparison criteria were solely our choice.
Preliminary conversations with representatives from Tektronix, Agilent Technologies, and LeCroy identified 500 MHz to 1 GHz as the most popular bandwidth range for professional oscilloscopes. Tektronix provided a TDS5104, LeCroy a Wavepro 7100, and Agilent an Infiniium 548382B. These are 1-GHz bandwidth, four-channel, color-display, long-memory, Windows-based DSOs. Because the major specifications were identical, I could concentrate entirely on the operational intricacies of each product.
Initial evaluation revealed common traits such as a long power-up time. Because a Windows-based scope really is a computer that runs a DSO application, power-up can take two minutes or more. Several different displays appeared on each scope during this time, but little effective user-feedback was provided. The best advice is to turn on the scopes and be patient.
Because the three DSOs are very complex, statements about their performance must be carefully qualified. For example, similar memory lengths and time-base settings were used when evaluating control latency. People who work with real-time analog scopes or simple DSOs will find this aspect of all three instruments disappointing.
The LeCroy 7100 trace shift and V/div selection controls were the least responsive. The Tektronix and Agilent trace positionings were much faster than the 7100 as was reaction to changes to the V/div switch setting.
Other characteristics the three DSOs share include cost, size, and complexity. These are professional-level DSOs, and each costs more than $10,000. On the other hand, the functionality of the instruments runs very deep, and there are few applications they cannot address.
User Interface Philosophy
Modern scopes are flexible in the ways they allow users to work. In all three DSOs, there are multiple overlapping and interacting control interfaces available. In addition to front-panel knobs and buttons, LeCroy and Tektronix have a touch screen that allows you to make selections from pull-down menus, for example. It also is the mechanism that allows you to draw a box around part of a waveform to define the zoom criteria. Agilent, instead, uses a mouse.
Each company maintains a distinctive operational philosophy that governs exactly how its DSOs will behave in particular circumstances. One part of that philosophy is related to the Windows operating system. Basing a scope on Windows helps reduce user confusion by providing discoverability. Windows is relatively intuitive, and after working with it for a while, you learn what to expect.
In each of the scopes, there is a core of capabilities, but only a few are exposed at a time, depending on what you are trying to do. I found that the fastest way to learn how to use these scopes was to examine the menu selections and press the buttons. Exploring acquisition, math, and display capabilities in detail is the best way to understand what makes these professional DSOs tick.
Under the Hood
In their simplest form, modern scopes conform to the block diagram shown in Figure 1. All digital scopes always have behaved in this manner, but many years ago, the software blocks contained fewer functions. The important thing that has occurred to shape the current generation of professional scopes is a change of emphasis within the boxes.
No hardware has been shown in detail because, although acquisition speed and memory depth still dominate instrument pricing, hardware only represents a small part of the overall performance. Acquired data merely is the raw material from which a scope’s software derives and displays information in almost unlimited formats.
There are some links among the hardware and software blocks but not necessarily only the ones presented to the user. For example, in some analysis modes, the amount of memory dedicated to an algorithm limits the size of the acquired records that can be accommodated.
Similarly, there are interactions within each box. An often-overlooked example is the memory-length reduction forced at high time-base speeds. What may have started as a user-selected 1-MB memory length can be reduced silently to 500 words or less.
Also, very few displayed traces are related directly to acquired data. If the memory length used for an acquisition is longer than 500 points, intelligent compression must take place before the captured trace data can be displayed. Compression is the normal mode of operation in a long-memory oscilloscope, not the exception.
Users also must be aware of functional similarities. Use the Sesame-Street approach and ask yourself which of these things are kind of the same? You may think of the A intensified by B display mode as having to do with separate time bases and alternate sweeps, but it doesn’t. In a modern scope, it’s just a special arrangement of math traces positioned to look like the old real-time scope A/B mode.
Software adjusts the position and size of some portion of the acquired channel data and presents it as a B trace. That sounds like a description of the zoom function. How about the magnify trace math function? Yes, these three operations all do similar things but are controlled in ways that suit users with different backgrounds and preferences.
Horizontal Axis
Regardless of the number of esoteric features provided, a DSO must acquire input signals and display waveforms. To understand the operating philosophy underlying each scope’s control interface, I decided to explore how time-base speed, memory length, sampling rate, and displayed time were related. The details considered are among the conceptual control lines in Figure 1 that link software and hardware behind the scenes.
Tektronix 5104
Tektronix has adopted a display-centric philosophy of operation. The basic idea is to present a display that spans a selected amount of time. After setting the required time with the time-base knob, you can turn the resolution knob to alter the amount of memory corresponding to a captured trace. The approach works well, but there are several operational modes that affect the interaction of the two controls in detail.
The first thing you may notice is that the time bases are arranged in a 1-2-4 sequence rather than the traditional 1-2-5 sequence. There are even 1.25-ns/div and 250-ps/div speeds at the fast end of the range. The user who estimates waveform timing from the display graticule must mentally determine the correct factor corresponding to one 20% subdivision.
Zooming is intuitive. With the touch screen on, use your finger to draw a box on the screen. Choose zoom from the pop-up menu, make either the 80-20 or 50-50 split-screen selection, and you have a nice display of both the main and zoom traces.
After drawing the box, you are given the choice of having it control the zoom, histogram, or measurement operation. The last function, measurement gating, is very important. Without gating, automatic measurement routines operate on the entire trace or perhaps just the first pulse in a string of pulses. Gating allows you to control where the measurements are taking place.
Two general-purpose knobs control the zoom factor in both the vertical and horizontal directions. The vertical and horizontal positions can be assigned arbitrary values, but time/div and v/div follow the 1-2-5 sequence. Even if you enter an arbitrary value via the on-screen number pad, it will not be accepted. Arbitrary vertical scaling is accommodated in the vertical setup menu where each channel can be assigned an arbitrary sensitivity.
Operation of the time-base knob appears simple at speeds slower than 10 ns/div. Turning the knob to go faster than 10 ns/div, the fastest setting where 500 real points are displayed, results in horizontal magnification. At 10 ns/div with a single channel selected, the scope is running at its maximum 5-GS/s sampling rate. It can acquire 500 points in the 100-ns time that corresponds to the 10-division full-screen display.
From 10 ns/div to 200 ps/div, the fastest available speed, the acquired samples are drawn with increasing dot-to-dot spacing. At 200 ps/div, 10 real samples are shown with 490 interpolated points. Selecting intensified dots as the display style causes only the real samples to be displayed.
At sufficiently slow time-base speeds, it isn’t necessary for the scope to run at the maximum 5-GS/s sampling rate. For example, a 100,000-word memory length combined with a 250-kS/s sampling rate corresponds to a 400-ms time period—10 divisions at 40 ms/div.
However, to maintain this memory length at a 20-ms/div time-base setting, the sampling rate must double to 500 kS/s. The question is, what happens when the combination of time-base setting and memory length requires greater than the maximum 5-GS/s sampling rate? Eventually, the memory length is automatically shortened, but that’s not the first thing to happen.
Instead, the memory length is maintained, but fewer real samples are acquired, interspersed with interpolated samples. When this occurs, the interpolate (IT) text symbol is displayed to qualify the time per point readout. As a result, it is possible to have seemingly inconsistent memory-length, time-base, and sample-rate values. The answer is interpolation.
What you see displayed may not be affected much by these underlying mechanisms if a sufficiently large memory length has been selected. The most obvious clue that complex effects are occurring in the background may be the changing appearance of a zoomed trace under various memory length and fast time-base conditions.
Selecting FastAcq rather than the normal horizontal mode adds another layer of operational qualification. This is a separate acquisition and display mode that provides very fast screen update rates and uses Tektronix’s digital phosphor oscilloscope (DPO) technology (Figure 2).
With a relatively short memory length, hundreds of thousands of acquisitions can be processed per second, a rate approaching the best analog scopes. This means that infrequently occurring anomalies can be found much more quickly than with an ordinary DSO that may only handle a few hundred acquisitions per second. With FastAcq deselected, the DPO technology provides color or intensity grading when compressing long memory acquisitions to 500 displayed points.
Pressing the front-panel FastAcq button also selects auto equivalent time sampling (ETS). The FastAcq mode is limited to 1-Mword memory length and cannot be zoomed. You can select FastAcq together with measurements or the deskew operation or histograms only for one- or two-channel operation. If you have three or four channels enabled, FastAcq will be disabled when choosing measurements, deskew, or histograms.
The highest non-ETS sample rate used in FastAcq is 1.25 GS/s. The factor of four between this rate and the maximum 5-GS/s rate in the normal horizontal mode, combined with the maximum ×50 interpolation factor, result in a multiple of 200. As shown in Figure 3, at 2 ns/div and faster normal-mode time-base speeds, and for the same memory length, the time base in the normal mode is 200× faster than that indicated in the FastAcq mode.
For time bases slower than 20 ns/div in the FastAcq mode, ETS is deselected. Changes in the sampling rate and the memory length result in an additional mode-to-mode factor of 10 at normal horizontal-mode time-base speeds slower than 4 ns/div. In other words, for these speeds, the normal-mode time base is 2,000× faster than the displayed FastAcq time base for the same memory length.
ETS can be separately selected and operates at combinations of memory length and time-base speed requiring greater than a 5-GS/s sample rate. At slower speeds, operation reverts to normal sampling, and the memory length is increased. The maximum equivalent sampling speed permitted is 250 GS/s. Special acquisition modes such as high resolution and envelope also can be used with ETS.
Because the maximum ETS factor of 50 (250 GS/s ¸ 5 GS/s) corresponds to the 50× interpolation factor allowed in the normal horizontal mode, ETS operation generally follows the Figure 3 curve labeled Tektronix Normal.
FastFrame is the Tektronix term for a special mode of segmented memory acquisition. Its main advantage is a very short dead time between acquisitions, subject to the availability of triggers. FastFrame, roll mode, and FastAcq are mutually exclusive. You can set the number of segments that will be acquired in the FastFrame mode, but they only can be viewed one at a time.
LeCroy 7100
From the first Model 9300, LeCroy DSOs have reflected their high-speed physics roots. Emphasis was placed on trace manipulation in early models through the provision of distinctive A, B, C, and D computed traces. The Model 7100 I evaluated has extended this type of abstraction to the point that zoom fits quite naturally among many more obscure mathematical operators.
Zoom can be selected and operated via front-panel controls although labeling is minimal. Using the touch screen, you also can draw a zoom box with your finger. A pop-up menu asks you to select the function trace that will be associated with the zoom operation.
The more intuitive control method for a seasoned LeCroy user starts from the math menu. In this menu, one of eight function traces can be chosen together with the relevant operator and the source that is to supply the data points to be operated on. The dominant role of post-acquisition math and analysis operations suggests data-centric as a good description of LeCroy’s scope-interface philosophy.
Clearly understanding that waveforms or traces, signals, and channels are different kinds of things is important before working with any of the three scopes I evaluated, but particularly the 7100. Signals are connected to the inputs of the separate channels. Data is acquired by the hardware associated with a channel. The waveform corresponding to the c1 marker, for example, is an amplitude vs. time graph of the data acquired by channel 1.
Because the function-trace approach has been abstracted to a very high level in the 7100, f1, for example, could be a filtered copy of data from a channel. It could be a zoom of that channel’s data. Or, it could just as easily be a graph of a complex mathematical relationship among live channel data and stored trace data and other function traces.
Having many function traces available allows you to zoom three separate sections of the same channel, for example. If you do, a split graticule display and color-intensified areas on the main trace clearly show what’s happening. Each zoomed area can be assigned arbitrary vertical and horizontal scaling. This means that zoomed traces can be compressed horizontally and vertically as well as expanded. There also is a multizoom facility that allows you to group traces that you want to zoom by the same amount.
The horizontal axis is controlled in two ways. You can select a maximum amount of memory to be used. If a time-base speed is selected that would require a sampling speed greater than the 20-GS/s rate available for a single channel, the memory length is reduced. Although 48-MB was installed in the instrument I tested, 10 MB was selected as a maximum length so the characteristics of this mode would better match those of the other scopes shown in Figure 3.
The time-base rates follow a conventional 1-2-5 sequence. Memory length choices are decimal values that almost always result in a full-screen waveform display. There are two exceptions. When the full 48-MB memory is selected, the waveforms are displayed 9.6-div long. If a 25-MB or 2.5-MB memory length is used, at some time bases, it will reduce to 20 Mb or 2.0 MB, respectively.
In the other method, you select a fixed sampling rate. For example, if 5 GS/s were chosen, as slower time bases are selected, more and more memory will be assigned to a channel. When the maximum amount of memory is used, no time bases slower than 1 ms/div are available. This mode is useful for data-analysis routines that require a constant time between samples.
Time-base speed, memory length, sampling rate, and total displayed time have a fixed relationship to each other. What’s interesting is that the designers of the three test scopes chose different ways to resolve the boundary conditions of that relationship.
As Figure 3 shows, Tektronix adopted a 1-2-4 time-base sequence with small changes to mostly binary memory lengths at slower time bases. In contrast, LeCroy used a conventional 1-2-5 time-base sequence and made small changes to some decimal memory lengths at certain time bases. The Model 7100 does not interpolate except when displaying a magnified trace.
Equivalent time sampling in the 7100 is termed random interleaved sampling (RIS). It is not available in the fixed sample rate mode. Both Tektronix and Agilent instruments have a maximum 250-GHz equivalent sampling rate. The upper RIS limit is 200 GHz. In any DSO, the usefulness of the ETS/RIS mode depends on the stability of the trigger source and the scope’s own trigger jitter.
In the 7100, you can improve the resolution of acquired data by operating on it with the enhanced resolution (ERES) function. Tektronix and Agilent provide a separate high-resolution acquisition mode. N successive samples are averaged to produce a single point that goes into the actual channel memory. The process is repeated for successive groups of N samples, where N depends on the time-base setting and memory length.
The Agilent mode operates at the highest sampling rate and provides a kind of real-time digital filtering. The effective sample rate is lower than the digitizer rate by a factor of N, and the bandwidth also is reduced. In the Agilent Help section, the bandwidths are listed that correspond to each combination of memory length and time base.
Tektronix provides minimal details of the 5104 high-resolution mode, commenting in the Help section only that it requires twice the normal-mode memory length and confirming that successive samples are averaged. There is no information included about bandwidth reduction or the relationship between extra bits of resolution vs. memory length.
The LeCroy ERES function is implemented by a series of specially designed finite impulse response (FIR) filters with impulse-response lengths from two to 117. This means that the output record length will be shorter than the input by from two to 117 data points. The filters provide from 0.5 to 3.0 bits of improvement in resolution and reduce bandwidth by factors from two to 64, respectively. Because the system of function traces allow both single- and double-nested operator assignments, a single function trace can display an ERES version of the zoom of channel 1, for example.
In addition to the mathematical function traces, the 7100 also supports generalized measurements or parameters as LeCroy terms them. You can select up to eight measurements to perform on source data provided from channels or function traces. The live parameter data derived by the measurement algorithms can be the source data for histograms, trend graphs, or simply graphs of the parameter vs. time. These secondary histogram, trending, and recording operations produce outputs in the form of waveforms that can be combined with other function traces.
It does take time to perform the intensive processing required to display mathematical operations on very long data records. The Model 7100 uses LeCroy’s fast X-Stream technology. It is, no doubt, much faster than previous scopes, but long memory math operations can become cumbersome.
The upside is that you have almost unlimited flexibility in the operations available. In addition to the built-in functions, you have the capability to import special routines from MATLAB and Mathcad. You even can write your own algorithm in Visual Basic and import it as a math operator.
All of this functionality can be daunting, so a special web editing page displays a diagram of the relationships among channels, function traces, and parameters (Figure 4).
Agilent 54382B
With an 8½” VGA display, The 54832B looks like a conventional scope. The Tektronix and LeCroy instruments have 10½” higher-resolution displays, and the front panels measure 17″ × 10½” and 15½” × 9½”, respectively, compared to Agilent’s 16½” × 7¾”. Also, the simple shape of the Agilent case adds to the conventional scope idea. In contrast, Tektronix has a shallow 11″ deep case with a large rear corner cutout that houses connectors and the handle.
The 54832B does not have a touch screen, so a mouse is required to access much of the extended functionality. Without the mouse, the DSO operates as a basic scope with delayed time-base and cursor measurement capabilities.
In the past, measurement-centric was a term that encapsulated Agilent’s user-interface philosophy. From at least the days of the long-obsolete Hewlett Packard (HP) Model 1980 ATE DSO, decimal numerical values have been a distinctive feature. For example, the values represented by vertical and horizontal display divisions are completely variable. In an Agilent scope, you don’t necessarily deal in display divisions. It’s more natural to work with numbers to several decimal places.
This is a benefit if you assume that any measurements or scope settings are going to be used by an attached computer in an automated test setup. On the other hand, odd time/div values are difficult for users although automatic pulse measurements and cursor readouts effectively have made division counting a thing of the past. Of course, there are the usual 1-2-5 time-base settings, but these are just special cases of a completely general time/div capability.
Other scopes measure trace parameters, but the 54832B has five columns of statistics to qualify each measurement. In contrast to Tektronix and LeCroy, several front-panel buttons are dedicated to selecting measurements and controlling cursors. Agilent, at the time HP, was instrumental in developing the IEEE Standard on Pulse Measurements and Analysis (1977) and remains a leader in rigorous scope metrology.
As an example of present measurement capabilities, the 54832B features drag-and-drop control. Rather than using cursors or drawing a box to define the waveform segment to be measured, you simply drag the relevant icon to the appropriate place and drop it there (Figure 5).
Because of the mixture of features in the 54832B, however, measurement-centric no longer tells the whole story. The treatment of time also is distinctive. In particular, pretrigger delay takes precedence over the time-base and memory-length settings. The fastest sampling rate and the longest available memory will be used to position the reference point and its associated 10-div time window at the requested time ahead of the trigger event.
For example, if you have selected an 11-s pretrigger delay, single-channel operation, and a 200-µs/div time base, the sampling rate will be 500 kS/s for any memory length greater than 2 kwords. The total time for which the input signal can be acquired is 8.2 Mwords × 2 µs/word = 16.4 s. Turning on more channels reduces the maximum available memory to 4.1 Mwords, and the sampling rate automatically decreases to 250 kS/s.
Post-trigger delay in the 54832B operates as a simple delay from the trigger to the acquisition reference point. Pretrigger timing also is referred to the trigger, but because the trigger hasn’t yet occurred while the signal is being acquired, the data must be captured for at least the selected pretrigger time. This is why the memory and sample-rate settings are overridden to satisfy pretrigger delay settings but not for posttrigger delays.
In contrast to Agilent, Tektronix and LeCroy simply limit the amount of available pretrigger time to 10× the selected time-base speed. These scopes never allocate more than the total screen display time to pretrigger delay. They allow the left edge of the displayed trace to be positioned up to 10 divs ahead of the trigger.
Unfortunately, because LeCroy measures pretrigger time from the display to the trigger, it is a positive number. Posttrigger delays up to a maximum of 10,000 divs are negative values. This convention is the opposite of that adopted by Tektronix and Agilent.
As Figure 3 shows, Agilent supports either manual or automatic control of the memory length. In the manual mode, the selected memory length does not change. Like LeCroy, the 54832B uses a conventional 1-2-5 time-base sequence. In the automatic mode, the memory length reduces as the time-base speed increases, stopping at a minimum of 64,000 words.
Agilent has developed the MegaZoom ASIC that provides long memory compression in hardware. The result is fast screen update rates, independent of memory length. Color-graded displays appear to use a separate mode. The maximum memory length is 100k, and the displays run very slowly.
Measurements also slow down the screen update rate, but it remains at a very usable five or six per second even with four measurements and an 8-Mword memory length. Other than these qualifications, you are free to use one, two, or four channels without a speed penalty because separate Megazoom compression hardware is provided for each channel.
The time-base memory-sampling rate total time equation is solved by displaying only part of the memory. A long bar above the waveform graticule shows which part of the memory currently is on screen. Turning the trigger position control can bring any part of the acquired data on screen.
In the 54832B, zooming is done by using the mouse to draw a box on the screen. As in the LeCroy 7100, an arbitrary amount of horizontal and vertical expansion or compression can be applied, and positioning is straightforward. Unfortunately, only the zoomed traces are displayed, not the combination of main and zoom. To add to the possibility of user confusion, you can zoom traces to eight levels. Right-clicking on the mouse brings up a menu that allows multilevel unzooming, but there is no indication of how deeply the current display is nested.
Four separate mathematical function traces are provided, and zoom can be rigorously accomplished by selecting a source channel, the magnify operator, and a destination trace. This method of working supports multiple operations on the same channel data although the list of available operators is much shorter, and the operators are more conventional than in the LeCroy scope.
Agilent also provides a digital form of A and B alternate trace operation that caters to legacy users and substitutes for zoom when working without a mouse. This mode expands part of the main trace to provide the pseudo-B trace. Like conventional analog A/B operation, the B time base is constrained to be equal to or faster than the A time base. A box is displayed around the portion of the A trace that corresponds to the B sweep, mimicking the analog A intensified by B mode.
Regardless of the operating mode, the user must be aware of the unconventional display graticule. Each major division is subdivided vertically into four parts rather than the traditional five. This format was introduced many years ago in the original 54100 series, so it is familiar to experienced Agilent users. However, it is a difference that new customers may not notice until they have made several measurement estimation errors.
Conclusion
After a couple of weeks of intensive user-interface evaluation, my conclusion is that most application problems could be solved by any of the three scopes. Of them, it’s no surprise that the LeCroy 7100 has the most extensive post-acquisition processing capabilities and detailed Help pages.
This is a high-horsepower scope for the professional scientist, physicist, or engineer who requires precise data acquisition and reduction. Incidentally, it displays some pretty impressive pictures too. Of course, it’s possible to use the 7100 to troubleshoot a circuit, but doing so probably would require about 1% of the scope’s analysis capabilities.
The Tektronix 5104 is a very capable general-purpose scope for the design and development engineer. This DSO has a rich mix of ETS, high-resolution, zoom, and the DPO FastAcq mode, especially useful when working with complex electronics signals. The different operating modes allow you to approach problems from multiple directions.
Sufficient measurement, math, and analysis functionality is included so you can pursue your investigation to a conclusion. On the other hand, there can be no mistaking that the 5104 is a visual feedback instrument first and a sophisticated computing system second.
Agilent’s 54832B has capabilities that fall between the other two instruments. With an 8½” display, it makes no attempt to provide the very detailed menus and display formats of the two other scopes with larger displays. Alternatively, you do have the freedom to drag the small selection windows to convenient locations. You also can choose translucent or transparent menus to obscure less of the waveforms.
Requiring a mouse to access the total functionality may restrict the DSO’s portability. For example, it could be awkward to operate the instrument when it was standing on its rear feet on the floor, but more than 90% of the time, a scope is used on a bench.
Functionally, the 54832B is closer to Tektronix than LeCroy, and this is logical because it targets general-purpose design and development. Nevertheless, the detail given in the Help pages confirms that this DSO takes signal processing seriously. For example, a block diagram clearly shows how the high-resolution mode operates, and bandwidth and resolution are stated for different memory lengths.
Because there are many factors that suit any DSO to a particular application, you need to evaluate prospective purchases on the job. Plug in a salesman’s demo scope and display the traces that represent your own signals. See how easy it is to position pairs of waveforms to measure the time between events. What about zooming? How long does the scope’s memory really have to be? What measurements and mathematical operations could save you time in repetitive test applications?
These are just some of the more important considerations that can help you decide which scope to buy. Especially in the current economy, pricing is very competitive. Should one model appear to be an exceptionally good value, make certain it really is by using it yourself. If you’re not totally clear about a mode of operation, continue experimenting until you are. Modern DSOs are visual-feedback instruments, but there’s a great deal more going on than meets the eye.
FOR MORE INFORMATION
on the Tektronix TDS5104
www.rsleads.com/307ee-208
on the Agilent Infiniium 54832B
www.rsleads.com/307ee-209
on the LeCroy WavePro 7100
www.rsleads.com/307ee-210
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2003 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
July 2003