October 8, 1998

Shortly after Tektronix announced its digital-phosphor oscilloscopes (DPOs) (EDN, June 18, 1998, pg 44), archrival LeCroy Corp started to ask what the big deal was. LeCroy claims that its scopes' analog-persistence display mode is similar to that of Tek's DPO displays and that each company's scopes are superior in certain situations. EDN decided to give the competitors a chance to state their cases to you. We solicited a statement from LeCroy and gave Tek the opportunity to respond. You can read the correspondence by following the links below, then send in your own comments. We'll post the best of your comments here in the weeks to come.

Each company engages in some hyperbole in describing its products. For example, Tek says that a DPO can acquire 200,000 waveforms/sec and can acquire 500,000 samples. LeCroy points out that a DPO can't do both at the same time. For its part, LeCroy says that its scopes can capture 40,000 waveforms/sec and can provide the same color-graded or intensity-modulated displays that Tek's DPOs can. A feature of these displays is that recently illuminated pixels appear brighter than those that were illuminated earlier. This characteristic simulates the behavior of analog scopes' phosphors. What LeCroy says—but only if you ask the right question—is that, in the mode in which the pixels "age," its scopes don't capture 40,000 waveforms/sec.

Still, this sort of "how-many-angels-can-dance-on-the-head-of-a-pin?" discussion is beside the point. The important question is which scope most quickly provides the most important information in your application. Here, the two companies agree. Both say that you need to obtain demonstration units and evaluate them for several hours using representative signals of your choice. Happy viewing.

Read the conversation so far:

1. LeCroy's statement: "Oscilloscope myths, or debunking DPOs," by Michael J Lauterbach, PhD

2. Tektronix's Rebuttal, by Christopher Martinez

3. LeCroy's response to Tektronix's rebuttal, by Michael J Lauterbach

Read reader comments:

Contribute to the discussion yourself:

To comment, use the form below OR send email to Senior Technical Editor Dan Strassberg at ednstrassberg@cahners.com. If you send email, please put the words "GLOVES OFF: 10/08/98" in the subject line—otherwise your message will be ignored.

Back to top

Oscilloscope myths, or debunking DPOs

Michael J Lauterbach, PhD, LeCroy Corp

I got a piece of mail last week from Tektronix, a company that was formerly a technology leader in the analog-scope market. The company's bold claim was that its digital-scope product was "faster than the fastest analog scope." This claim appeared to be based on a triggering rate as fast as 400,000/sec. The mistake is probably the result of simple ignorance, because the vendor has not produced any new analog-scope technology for several years. The fastest analog scope (LeCroy's model LA354) triggers as many as 1 million times per second.

Under normal conditions I would probably have deposited the item in my wastebasket and gotten on with my work. But recent coverage in EDN and other publications has sensitized me to the debatable "great-new-technology" claims for DPOs. So, I write this piece to gently chide the editor and readers to be cautious in believing everything a PR person says.

Here are a few of the exaggerations I have heard and some relevant facts concerning each:

• A DPO can trigger 200,000 times/sec and capture as many as 500,000 samples per trigger.
  
  The vendor doesn't mention that these two figures are incompatible. (You can either trigger 200,000 times/sec or capture 500,000 samples, but not both.) Also, the trigger rate of 200,000/sec applies only if you are capturing data on a single channel. (The other three scope inputs must be inactive.) And, you must have the timebase set for 50 nsec/div. At other timebases, the scope switches into equivalent-time mode or to a substantially slower trigger rate.
  
• Only DPOs can bring data directly to the pixel map; other scopes use "postprocessing" techniques to make persistence displays.
  
  A not-too-technical marketing guy probably invented this gem. Actually, lots of scopes bring data to the pixel map without subjecting the data to any processing. On the other hand, if you want to look at signals using the DPO's high-resolution or averaging-acquisition modes, the scopes leave the DPO mode, forcing you to destroy the raw data and to view only the postprocessed data. (Other scopes provide enhanced resolution or averaging while still preserving the raw data.)
  
• DPOs provide a continuous data stream to the pixel map; other scopes have long pauses between acquisitions.
  
  Actually, the truth of this statement depends a lot on what type of signals you are trying to see. DPOs do a creditable job of capturing short, simple signals and transferring them, one at a time, to the pixel map. But if you want to see parameter statistics, use waveform math, or simply use all of the scope's acquisition memory, trigger rates can slow to a crawl. The DPOs optionally provide as many as 8 million points per channel of acquisition memory but take more than 10 sec to acquire and display one waveform of that length! Even data records of 100,000 to 500,000 points can cause substantial pauses between triggers. Fast scopes that handle the complex signals of communications and µP designs are more than 10 times as fast as a DPO in handling million-point data arrays. On the other hand, scopes designed for handling long data arrays aren't always optimal for looking at short, simple signals.
  
  A LeCroy scope can load as many as 2000 signals (2000 separate occurrences of the trigger) into the acquisition buffer at a trigger rate of 40,000/sec and then dump the whole buffer into the pixel map. This process causes a longer pause than in a DPO but has the advantage that parameter statistics, waveform math, and other features are available. Also, to view the "culprits" (aberrant waveforms) that may be causing problems in a circuit, you can go back and examine each of the 2000 separately triggered events in a normal display (without the clutter of a persistence display).
  
• DPOs acquire 1000 times as much data as other scopes do.
  
  This statement is a substantial exaggeration. You can look at this claim in several ways. One is that, in DPO mode, the longest available acquisition is 500,000 points. A LeCroy scope that performs state-of-the-art acquisition seamlessly acquires 16 million points in a single record with zero dead time and puts all of the unprocessed data into the pixel display. This amount of data is 32 times as much as a DPO can display. To capture the same number of samples (16 million), a DPO must acquire 32 500,000-point records. In doing so, the DPO exhibits more than 90% dead time. Second, DPOs operate in non-DPO, slow-triggering mode or DPO, fast-triggering mode. In the DPO mode, the scope can trigger about 1000 times as fast as in the non-DPO mode. Most other high-performance scopes also have two acquisition modes—one for normal triggering and the other for fast triggering. I made my own comparison chart of the time between triggers using all of the timebases from 50 nsec/div and longer. I started at 50 nsec/div because, on faster timebases, the DPOs use equivalent-time sampling (many triggers to build a single waveform)—a mode that can cause problems for engineers who work on fast signals.
  
  I found that if I had the Tek scope in DPO mode and the LeCroy scope in sequence mode (the faster trigger mode) and that if I set each scope to have the same sampling rate on the same timebase, the DPO's dead time between triggers was shorter on about one-third of the timebases, and the LeCroy scope's dead time between consecutive triggers was shorter on two-thirds of the timebases. You can perform this "time trial" in many ways, such as using different amounts of memory, asking the scope to also do some parameter measurements, or turning on some zooms to show waveform details). You can do your own speed test using your typical scope operating conditions.
  
• DPOs always find the problems in your signal more quickly than any other scope does.
  
  I found this claim to be dead wrong. Maybe my experience is atypical, but a DPO failed miserably in finding the problems in the set of signals I have been using for training engineers about troubleshooting signal problems. These signals were not esoteric ones designed to make the DPO look bad. They included runt pulses, slew-rate problems, intermittents, timing errors, and others. The data problems were obvious. I don't know the source of the problem, but data integrity is an important quality in a scope, and problems bear checking.

My recommendation—the same one I have been giving to engineers for years—is that if you want to buy the newest oscilloscope technology, don't put off getting two or three samples from different vendors. (Any reputable vendor will leave a scope with you for a few days.) Take two or three hours to compare the scopes side by side using your own signals. Don't use the vendor's demo board. Because you are likely to use the scope for two to three years, this time will be a good investment. Also, if you look at long, complex signals, try out the methods for looking at zoom details, parameter measurements, waveform math, and other tools for handling these waveforms.

Michael Lauterbach, PhD, is director of product management at LeCroy Corp (www.lecroy.com), Chestnut Ridge, NY. He started with LeCroy 17 years ago as manager of engineering services. He holds a BA in mathematics and physics from Carleton College (Northfield, MN) and a PhD in high-energy physics from Yale University (New Haven, CT).

Back to top

Tektronix's rebuttal

Christopher Martinez, Tektronix Inc

Michael Lauterbach's comparison of the capabilities of the Tektronix DPO and LeCroy Corp's analog real-time (ART) oscilloscopes is intriguing, if somewhat baffling. Although the temptation to respond to Lauterbach's sarcasm is strong, it is highly resistible. You might benefit more from a discussion of the fundamental issues Lauterbach raises.

Since the introduction of the digital storage oscilloscope (DSO) in the late 1970s, the objective has been to make the DSO as much like the ART oscilloscope as possible. And the need for merging these two measurement approaches is more compelling as electronic signals become faster and more complex. More than ever, engineers and technicians need the storage and measurement capabilities of a DSO combined with the real-time, information-rich display of the ART scope.

Special DSO display modes, such as persistence mode, were an initial acknowledgment of this demand, but they fell short of the goal. Although persistence modes create a highly detailed display, they also require extensive postprocessing, eliminating any real-time oscilloscope response. Further, during the processing time, the DSO cannot acquire new information. Consequently, conventional DSOs ignore large chunks of signal activity while the oscilloscope transfers and processes the previously acquired data.

Four years ago, Tektronix pioneered the InstaVu acquisition technology that supports waveform-capture rates that match those of premier ART scopes. As a result, InstaVu acquisition eliminated the problem of the DSO's relatively slow waveform-capture rate. However, this technology provided only two dimensions of signal information. Unlike an ART scope, InstaVu does not indicate how often one event occurs with respect to another.

With InstaVu as a foundation, Tektronix invested years of research to expand this technology. The result was the DPO. These oscilloscopes display, store, and analyze signals in real time using three dimensions of signal information: amplitude, time, and the distribution of amplitude over time. The DPO can continuously acquire and display three dimensions of information because its parallel-processing architecture integrates the display and acquisition systems. The key differentiator is that DPOs produce an ART-like real-time display. Finally, using one instrument, engineers and technicians can view the most elusive and subtle signal behavior and measure and analyze all waveform activity (Figure 1).

The key to Tektronix DPOs' performance is the DPX proprietary waveform-imaging processor, which acquires and manages waveform information. Manufactured in 0.65-µm CMOS, this highly pipelined processor, with 1.3 million transistors, is tailored for high-speed waveform acquisition and image generation and display, as well as for memory management. Because DPX works in parallel with the DPO's main processor, it eliminates time-consuming postprocessing and provides real-time display capabilities. Further, because each acquisition channel has its own DPX waveform-imaging processor, Tektronix DPOs exhibit no performance degradation when you use multiple channels. (Conventional DSOs and single-beam ART scopes suffer such degradation.)

Tektronix's DPOs address the demands of a market that will expect DPOs to be available in a variety of configurations from a variety of sources. Consequently, Tektronix has elected not to make "DPO" a trademark.

You cannot use the standards of yesterday's technologies as the exclusive basis for evaluating new approaches to test and measurement. For example, Ken Olsen, the CEO of Digital Equipment Corp, could see no reason 20 years ago why any individual would own a computer. Today, there are 300 million PCs in the US.

But more to the point, I invoke the wisdom of Groucho Marx, who asked, "Who are you going to believe—me or your own eyes?" Experts and journalists have drawn their own comparisons and come to their own conclusions. Ultimately, of course, customers will determine the fate of the DPO in the market, so this brouhaha is not about public relations, but about the public, and that is as it should be. In this, finally, I agree with Lauterbach. For more information, visit www.tek.com/ Measurement/scopes/dpo.

Christopher Martinez is the worldwide business-development manager for the Design, Service and Test Business Unit of the Tektronix Measurement Division (www.tek.com/Measurement). A 21-year veteran of Tektronix, Martinez holds a BS from Marylhurst College for Lifelong Learning (Marylhurst, OR) and an MBA from the Oregon Executive MBA, University of Oregon (Eugene, OR).

Back to top

LeCroy's response to Tektronix's rebuttal

Michael J Lauterbach, LeCroy Corp

I was somewhat suprised at an important factual error in the Tek letter. It states:

"Further, because each acquisition channel has its own DPX waveform imaging processor, there is no degradation in performance when multiple channels are used (unlike a DSO or single beam ART)."

This is easily observed to be untrue. If any user looks at the triggering rate of a DPO (by looking at the time spacing of the Trigger Out pulse), you can see the deadtime between triggers is about 5 µsec when using one channel (200,000 triggers/sec), but the deadtime inceases to about 9 µsec when using two channels and 17.5 µsec when using 4 channels.

As a comparison, a LeCroy DSO has a minimum deadtime between triggers of about 25 µsec (40,000/sec trigger rate) when using sequence mode. This deadtime is the same when using 1 channel, or 2 or all 4.

So the statement in quotes above is wrong in two ways—concerning both the DPO and the DSO...

Back to top.

Reader comments

Maybe Mr Martinez can explain why it is so important for the pixels on a DSO display to "age," as do the images on an analog scope display. Were it not for the aging feature, it seems as though LeCroy's "sequence mode" would be quite equivalent to Tek's DPO mode. Comments?

—Donald O. Stusskey, One Communications

Contribute your own comment
Back to top

Tektronix's Chris Martinez responds to Donald O. Stusskey’s implied question, “Were it not for the aging feature, it seems as though LeCroy’s “Sequence Mode” would be quite equivalent to Tektronix DPO mode.”

We have tried to contact Mr. Stusskey to ensure that we fully understood the context of this question but to date Mr. Stusskey has not been available.

Without knowing more about the application we could describe LeCroy’s memory segmentation feature “Sequence Mode” as being useful in applications such as Radar, Lidar, Ultra-sonics, or other applications where it is useful to capture multiple sequential events with high sample rate over long time durations. Tektronix oscilloscopes also provide the ability to segment the available memory and capture each individual event in succession. Tektronix calls this feature “FastFrame™ TimeStamp where the entire memory can be partitioned for storage of waveform segments from 50 to 50,000 points. This segmentation allows the user to store as many as 1,489 frames (with the optional 8M of memory) of waveform data at a burst trigger rate of 80,000 frames/second.

Unlike FastFrame Time Stamp or Sequence Mode, DPO is a continuous acquisition technology that speeds waveform capture rates up to a maximum of 200,000 waveforms a second. Further, DPO technology provides distribution of occurrence information (intensity grading), much like an analog oscilloscope, i.e. in real-time. The combination of fast waveform capture rate and distribution of occurrence information provides both confidence in capturing signal variations, including glitches and other intermittents, and being able to view dynamic complex waveforms in real-time. Neither FastFrameä Time Stamp or Sequence Mode provide this type of continuous insight.

Additional information about FastFrameä Time Stamp acquisition techniques and applications can be found at http://www/tek.com/Measurement/App_Notes/dpo/fastframe.

Additional information about DPO architecture and capabilities can be found at http://www/tek.com/Measurement/App_Note/dpo/architecture.

Contribute your own comment
Back to top

Since I usually wish to view only single shot waveforms, I have never found any of the persistance views useful. I also hated ART's because they never worked well with single shot waveforms. When I got my first HP digital scope, HP1631, I was in love, but then HP kept changing the user interface on version after version of their scopes, never had enough memory, and their zooming was terrible. Tek and LeCroy both have a consistant user interface through multiple versions of their products, Teks user interface is probably better the LeCroy's. But LeCroy offers MEMORY and the BEST EXTENDED WARRANTY (AND NIST CAL) AND SERVICE IN THE BUSINESS. Lecroy upgrades the software (does anyone else ever update software with their maintenance agreement?), fixes hardware problems we were unaware of, and calibrates in a timely manner. I haven't tried Tek's service but HP's is awful.

—Gary Lameris, PMI Food Equipment

Contribute your own comment
Back to top

—Gary Lameris, PMI Food Equipment

Contribute your own comment
Back to top

I am both a designer of Digital Storage Oscilloscopes (DSO) [For Gould Instrument Systems in the UK] and a user. I tend to err in favour of Dr. Lauterbach's view of the world in general. Whilst the new Tek DPO technology is good, it is not earth shattering. They are really only playing catch-up with the concept that a DSO should be all things to all men, a view which we have held since we first introduced DSO's to the market place in 1974.

We first introduced our analogue-like display technology in June 1996, with the release of the Gould Classic 6000, and have introduced more members to the family since then. You can read all about our patented TruTrace technology on our web site at www.gould.co.uk. The big four DSO manufacturers: Gould, HP, Tek and LeCroy (given in no particular order), all produce good scopes and some people prefer one user interface or feature set to another. I am not interested in getting involved in some sort of muck slinging contest in this respect; it is the customer who has to decide what is the best product for him/her.

Personally I like Tek products, and I have lots of them. In general each product has its own set of good and bad features. On my bench is a Tek 2465A 350MHz analogue scope. Nobody would be able to rip it from my clutches and yet if you stick a 4 div 200ps edge into it and shift it partially off the screen, it rings really badly. So what? It's a fast overload and you can't do that with it. Fair enough.

On the other hand we have a Tek TDS524A 500MHz 500MS/s colour DSO. This is a very nice piece of kit. It does not suffer from pulse degradation when you shift the trace off screen in the same way as the Tek 2465. But when you want to do single shot acquisitions, peak detection and the like it can be quite a hunt to find the feature. Then again, if you try to run it at 5 seconds per division it is a real dog. This is one of these unspecified features. There is a mode which Gould scopes do, which many others don't do, and that is technically known as slow refresh. After a trigger on a slow timebase, we start updating the display before the sweep is complete. Many scopes, including the TDS524A do not. What this means is that you have to wait for the whole timebase sweep period before you see anything on the screen! If you ever use a DSO on slow timebases, this is something that would make you tear your hair out. I have in the past been using the TDS524A and had to shove it to one side and get a Gould scope out, so I could look at the slow timebase events sensibly.

Many DSOs don't even try to implement Roll mode (this is for slow timebases again and it puts the scope into a chart-recorder type mode where the most recent data enters the screen at the right and rolls across to the left). These features, along with the various max/min (peak detection) and display decimation techniques allow the DSO to operate correctly over a a very wide range of timebases, which is necessary for general purpose lab equipment.

Many user interfaces are unacceptable to me as a user. If I have to use selection keys before I can move the trace on a 4-channel scope I get very frustrated. It slows me down too much. HP scopes have been very poor on this in the past and I would never use them, even though we have had them knocking around. On the other hand the new HP Infiniums are very well thought out and are undoubtedly the best scopes that HP have ever made. My only reservation on the user interface being the idea of using a mouse and what the scope is like when you don't use one. I personally have a bench full of equipment, including an 18GHz HP spectrum analyser, a Tek 5GHz sampling scope, the Tek 2465, a Gould 475, and various power supplies, generators etc. There isn't space to fit a soldering iron (it lives on top of the Gould 475 currently) let alone a mouse. It would also be interesting using a mouse when running the scope on a trolley or, as we sometimes do, standing on its back feet when we are trying to fit even more gear in the area. This is very much a personal choice based on the working environment that the scope will be used in.

But I digress. A point made by Dr. Lauterbach is that Tek claims features that are mutually exclusive. Well that's marketing departments for you! The banner specs you see are the best obtainable and are often filled with caveats. LeCroy are renowned for quoting "peak-music-power" type bandwidths. We had one here, fresh out of the box, where the LeCroy final calibration numbers given were lower than the published spec! (They were only marginally lower it has to be said, but we still had a good laugh about it.) It is also not unusual for manufacturers to quote a banner bandwidth and then for the user to find that on the highest sensitivity ranges the quoted spec is less. The Tek web site publishes the TDS784D spec as 1GHz (but it is reduced to 500MHz on the 1mV/div range). This doesn't make it a bad scope, unless you didn't read the spec and are disappointed.

If you want a really misleading number look at the resolution quoted on averaging schemes for 8 bit scopes. People are quoting 12-bit resolutions. I had a long fight with our marketing department to stop them quoting such meaningless numbers, but others are quoting them, so unfortunately I lost. But I did manage to get them to publish a meaningful number, the effective number of bits (ENOB) as 10.8 bits. Now this is quoted on a 200MHz product, but you are not in 200MHz bandwidth mode when you get your 10.8 bits. These are mutually exclusive numbers and anybody with any common sense would realise this. The point is that resolution enhancement by averaging is one thing, and improving the differential and integral non-linearities is something else altogether. ENOB shows up this improvement and to be valid this should be done on a single acquisition sweep, not the average of 32 acquisitions or any such thing.

Marketing departments will always quote the best numbers they can get away with. It is up to the user to sift through the BS and get to the truth. Both Dr. Lauterbach and Mr Martinez agree that the user should try before he buys. This is the most important thing. Get it in and use it. Don't just play with it on the bench. Try to use it for its intended function, in its working environment. This separates the wheat from the chaff.

—Leslie Green, Gould Instrument Systems

Contribute your own comment
Back to top

—Mathew A. Dirjish, Electronic Products

LeCroy's Michael J Lauterbach responds: Mat raises an interesting point. There is a big difference between "real time display," such as is found in an analog scope that can update its screen 1,000,000 times per second and "real time measurements." New modes in digital scopes allow for fast trigger rates and transfer of the data quickly to a chip that holds the pixel map—but the data does not go immediately to the screen. LeCroy scopes can make measurements on the data acquired in these special modes but many other scopes can't. I wonder what the user community would use as a definition of "real time measurements."

The fastest digital scopes update their screen 60 times per second. Would a scope that could update parameter measurements with fresh values 60 times per second qualify as real time measurements? I think most users would say "yes." But there are some scope users who would like to perform measurements that are more complicated than simple pulse parameters—multiplying waveforms, taking ratios, doing FFT's to look at harmonics, etc. Those types of measurements take longer—especially if the signal is long/complex. LeCroy scopes have up to 64 Mbytes of RAM that enables fast calculations but even with that type of horsepower it is difficult to make complex measurements in "real time." I would be interested in reader comments about how much interest they have in this subject. Would they vote for faster processors and more RAM in high performance scopes (even if it meant a few more dollars in cost)?

Contribute your own comment
Back to top

EDN Senior Technical Editor Dan Strassberg responds: Well, one real-time function that scopes from Gould, HP, LeCroy,and Tek perform is real-time triggering, and capture of pretrigger events when anomalies are detected. In scopes that provide this feature, triggering can take place at virtually any time independent of other functions that the scope is performing. I doubt that there is a way of improving on that level of immediacy unless someone can invent a scope that says, "you didn't define this anomaly as one you wanted me to trigger on, but I think the signal may be of interest to you so I'm going to trigger on it anyhow." Since the anomaly-trigger function does not depend on digitizing the incoming information, I don't see such an "I (that is, the scope) am smarter than you are, Mr Engineer" scope becoming available any time soon.

An intermediate step would be a scope that could make parametric measurements on waveforms that it does not retain in its capture memory, as well as on waveforms that it does retain. This would allow you to retain only "interesting" waveforms, while obtaining statistics on all waveforms, retained or not. This proposal may or may not be technically feasible. In fact some scopes may operate this way already.

Contribute your own comment
Back to top

Thank you Mr. Lauterbach. I am glad someone saw the point I was trying to make in the form of a simple question.

Contribute your own comment
Back to top

If Dr Lauterbach's response addresses the question that you apparently were asking in the first place, it appears to me that one or both of you have inadvertently changed the meaning of the phrase "real-time" as it applies to DSOs. I have always taken "real-time" in a DSO to be the alternative to equivalent time, which refers to the use of sequential (or the more common random) sampling of multiple incoming waveforms to produce one displayed waveform. Both equivalent-time techniques allow the undistorted acquisition of waveforms containing signal frequencies that would be undersampled at the scope's "real-time" sampling rate. When discussing DSOs, everybody understands real-time in this sense. It would be extremely unfortunate if someone were to try to give the term a different meaning. The resulting confusion would do nothing to foster confidence in instruments that too many engineers already mistrust. The most descriptive term I can think of for the aspect of DSO performance to which you were apparently referring is immediate display update, a property that no DSO offers. I believe that immediate update is an attribute that DSO designs can only approach asymptotically.

—Dan Strassberg

Contribute your own comment
Back to top

From the two e-mails I just received it seems that I took the meaning of Mat's question in the way he intended (asking about "real time measurements") but that other people would interpret the question differently. Dan's definition of "real time" as it relates to acquisition methods is certainly the classic definition used in the industry ("real time" as opposed to "equivalent time" sampling by the ADC). And furthermore he is right that digital scopes can only approach real time display—or real time measurement—asymtotically. Nonetheless, I still think it would be interesting to know how many readers/scope users would like to have "immediate" (can I use that word instead of "real time" ?) measurement results within a fraction of a second each time the scope triggers.

—Michael J Lauterbach

Contribute your own comment
Back to top

| EDN Access | Feedback | Table of Contents |