Those 6- and 8-GHz digital storage oscilloscopes (DSOs) that were sufficient a year or two ago are now becoming overrun. Silicon and motherboard designers' scopes seemingly have lost the ability to handle the lightning-fast speeds of the latest and next-generation computer buses.
For instance, new high-end PCs feature a Serial ATA (SATA) bus for drives and a PCI Express motherboard. As a result, they promise faster-than-ever speeds for chip-to-chip, graphics, and other I/O functions.
In addition, cell phones, cordless phones, and other consumer and military devices operate in the extreme UHF to microwave frequencies, requiring DSOs with wide bandwidth and high sampling speeds. Designers are looking toward bandwidths of 10 GHz and beyond in their scopes to keep up with the devices they're creating (Fig. 1).
Not all design engineers need a light-speed DSO with a $50,000 to $130,000 price tag, though. Motor drives, audio devices, industrial I/O boards, light dimmers, and other relatively "low-tech" devices probably require a DSO of less than 100-MHz bandwidth. But engineers still dream about practical features that make their jobs easier, even with the less glitzy sampling speeds and bandwidth features. Scope manufacturers are fulfilling these dreams in their latest products.
DSOs offer several features not available in analog scopes. These features are sometimes no more expensive than the analog scope you may have used eons ago. In the dark ages, engineers used an awkward Polaroid camera to freeze a waveform on the screen—somewhat tricky if they were waiting for that all-elusive, fleeting transient—and the triggering wasn't set just right. Along came analog storage technologies, which made it possible to trap a transient for an extended time on the screen, but not forever.
Today, the DSO detects and displays the transient and writes it to memory, hard drive, or network file location. DSOs serve up a variety of options for triggering. Some provide the means of analyzing data internally on Microsoft Windows-based platforms or externally on the software of your choice (Fig. 2). The possibilities are almost limitless.
TRIGGERING NEEDS IMPROVEMENT Many designers believe triggering is a pretty important feature in a DSO, but the feature needs further development. Stan Katz, senior embedded designer at Control Technology Corp., lists flexible triggering and easy setup of triggering options among the functions and features he would value in a new DSO. He also cites the ability to "trigger on multiple events on different inputs, e.g., trigger on axis-2 positive edge or on axis-1 negative edge, whichever happens first."
Senior staff engineer Kevin Crawley of Keithley Instruments would like to see improved cross-channel triggering and keying on analog levels, as well as a trigger that would automatically save to a file each time an event happens. Thus, data could be gathered during overnight testing.
Some scopes already have these triggering and data-saving capabilities, remarks Boyd Shaw, a product manager with Yokogawa Corp. of America's Test and Measurement Division. In his company's scopes, an "OR" trigger works across several channels. It would, say, trigger any time the signal on channel 1 rises above 2 V, channel 2 falls below 7.8 V, or channel 4 rises above 6.2 V.
The user sets rising/falling edge conditions and trigger-level conditions independently for each channel of interest. The scope then continuously monitors each input signal and compares it to the triggering conditions set. Afterward, if and when the trigger condition is met on any channel, the instrument triggers (i.e., saves that segment of data).
Second, an "Action on Trigger" performs one or more activities every time a trigger occurs (Fig. 3). Actions might include saving the data to a floppy drive or printing a hard copy of the screen image. If the instrument is attached to a network via Ethernet, it could save data directly to a network drive or send an e-mail indicating that a trigger occurred. The ability to automatically save data each time a trigger occurs is ideal when running tests over extended periods.
According to Jerry Murphy, Agilent's manager of mixed-signal oscilloscopes, customers often design equipment with a combination of digital and analog signals. But using a logic analyzer to monitor signals in these devices is like using an elephant gun to kill a fly, he adds.
Giorgio Decker of Elkron was one of these customers. "We had a big problem in our design regarding transmission synchronization on many devices," he says. "Usually, we analyzed these signals with a logic analyzer. But in debugging serial synchronous interfaces (data in, data out, and clock), it was really not easy since the problem was random and there was no way to trigger on it. We had to look at eight different channels in which the anomaly could appear in any one of them without a timing relationship."
Borrowing another engineer's 54621D mixed-signal scope solved the problem (Fig. 4). Unlike a logic analyzer, the scope was easy to set up. The scope also made it possible to see the waveform related to eight transmission lines simultaneously in a continuous mode with a slow time base (a transmission every second). It also stored the image when the problem happened. The scope's MegaZoom feature even let Decker see a single bit in the entire pattern, allowing a complete analysis of the problem.
The ability to trigger on patterns is an optional part of Tektronix's new Pinpoint triggering system, says Colin Shepard, vice president, oscilloscope products. The name Pinpoint summarizes the precision with which the new trigger system can isolate individual events or combinations.
Customer demand brought about this system, and it's been applied throughout Tek's high-performance scope family. Pinpoint triggering on the TDS7704B provides a jitter spec of less than 1.2 ps rms and less than 170-ps glitch capture for isolating events. The system offers full capabilities and flexibility on both A- and B-triggers.
BANDWIDTH AND SPEED AT WHAT COST?
Stu Streiff, analog measurement architect at National Instruments, has been involved in a project called T-Clock. This hardware design, coupled with software drivers, synchronizes the modules in NI's PXI data-acquisition chassis (Fig. 5).
While NI purchases 3-, 5-, 6-, and 8-GHz scopes for developing applications like T-Clock, it buys several 500-MHz scopes for its debug applications. Like everything else, speed requirements are going up at this level, too. Streiff's complaint is the supposed 50% price jump between four-channel, 500-MHz and 1-GHz scopes—from $10,000 to $20,000.
Mike Lauterbach, director of product management at LeCroy, says the price difference in its WaveRunner series is less drastic. It ranges from $11,250 for a four-channel 500-MHz scope to $16,250 for the 1-GHz version. Keep in mind that the 1-GHz operation requires an advanced level of expertise and materials in its design compared to the 500-MHz version.
Lon Hintze, Agilent's product manager, says that users of high-speed scopes may need to jump product lines to get the features they need at 1 GHz. For example, it may be necessary to go to a Windows-powered scope, which can double the price. Hintze's rule of thumb for price comparison of high-end scopes is to allow about $10,000 per gigahertz.
Another crucial feature is the ability to control jitter in high-performance scopes. Hintze notes that jitter, the horizontal equivalent of the vertical (noise floor), has been halved in Agilent's new generation of high-end scopes, reaching up to 13-GHz bandwidth. The fact is that more bandwidth inherently brings more noise unless steps are taken to reduce the noise.
Microwave-design techniques are necessary to get better bandwidth and noise figures at these bandwidths. So, Agilent has used microwave thick-film technology in its probe amplifier, built a Faraday cage around its preamplifier, and changed its analog-to-digital converter (ADC) from bipolar to CMOS. The resultant noise figure at 8 GHz is 3.4 mV rms at 100 mV per division, down from 6.8 mV for older 8-GHz scopes.
How much bandwidth is necessary? Obviously, not every designer works with high-speed serial buses. But for those who do, Agilent has computed the numbers. PCI-Express (Ver. 1 at 2.5 Gbits/s) has a fundamental frequency data signal at 1.25 GHz, third harmonic at 3.75 GHz, and fifth harmonic at 6.25 GHz; PCI-Express (Ver. 2 at 5 Gbits/s), 2.5 GHz, 7.5 GHz, and 12.5 GHz, respectively; and SATA III (at 6.0 Gbits/s), 3.0 GHz, 9.0 GHz, and 15 GHz, respectively.
Measuring a 100-ps rise/fall time with 10% accuracy, you'll need a 4.0-GHz scope; 50-ps rise/fall time, an 8.0-GHz scope; and 30-ps rise/fall time, a 13.3-GHz scope. Better accuracy demands wider bandwidths, translating to a requirement of an 11.2-GHz bandwidth to produce 3% accuracy at 50-ps rise/fall time and 9.6 GHz to produce 5% accuracy at 50 ps.
THE PROBE IS THE EYE
A probe must deliver an undistorted waveform to the scope's front end, regardless of the scope's bandwidth, sampling rate, or triggering sharpness. Thomas Patterson, electrical design engineer for DRS Test & Energy Management, would love to see probes better shielded from electromagnetic radiation (EMR). NI's Streiff adds that to get a good pulse response, bandwidth is important. But his experience has been that probes, not circuitry, all too often limit the scope's bandwidth.
No doubt about it, poor shielding and an improperly designed probe will hinder a scope's ability to display a true image of the signal. Tektronix market development manager Chris Loberg says that Tek's customers have clamored for a more flexible, wider-bandwidth, lower-load probing architecture. This has been realized in Tek's Z-Active low-loading probe, which offers greater than 12.5-GHz bandwidth with a fast rise time (25 ps, 20%/80%). The probe can be used with Tek's 12-GHz TDS6124C and 15-GHz TDS 6154C DSOs.
According to Hintze, the trick is measuring the signal without disturbing it, especially at these extreme bandwidths. First, the probe must be shielded to prevent interference from signals in space and from adjacent circuits. Second, the capacitive loading must be as small as possible. Agilent gets it down to 220 fF, which still isn't zero but works with 12-GHz scopes at 50-kV input resistances (Fig. 6). At 12 GHz, a probe amplifier should be capable of a noise figure of 2.5 mV rms (referred to input) and have a differential dynamic range of 3.3 V p-p.
DREAM FEATURES AND UIs
A feature that doesn't behave the same across different brands of scopes is the "auto" button. Jamie Hopper, senior applications engineer at Datatronic, would like an auto button feature that's truly an auto button. "Many scopes require the user to have an in-depth knowledge of that particular scope," he says. "Many times, waveforms do not show accurately without internal settings getting changed. I spend more time learning the scope than I do just looking and measuring a couple of waves."
Xerox faced similar issues. Its Asset Management Group in Webster, N.Y., was tasked in finding 100- to 500-MHz scopes that would be upgradable, portable, and easy to use. Different engineers and technicians would use a single scope, leaving their settings to be changed by the next user. The solution, according to John Sullivan, project lead, was Tek's TDS5000B (Fig. 7) with the MyScope customizable user interface (Fig. 8).
The scope is basically a complete computer, making it easy to add new applications and software-based analysis tools or to expand memory. Upgrades are possible without replacing the box. Applications include PCI analysis for PCI bus development, jitter analysis for advanced timing of lasers, fast frame testing for capturing intermittent events occurring over long timeframes, power analysis, and USB analysis. Users can create their own personalized "toolkit" of oscilloscope features, making it unnecessary to search through menu after menu to repeat similar tasks or relearn how to drive the scope after a break from the lab.
Agilent's Murphy says the "auto" button issue, which affects the scaling of input channels, can be remedied by looking at all of the channels and determining which channels have activity. Channels with no activity should be turned off. Active channels should be scaled to fit the screen, the timebase should be set to give two cycles of the fastest period observed, and the trigger should work off the high-number channel. Philips did this elegantly in the Scopemeter, using a separate microprocessor to do the autoscaling.
Stephen Coan, design engineer at Relcom Inc., says that color display is a highly valued feature. Longer record lengths and the ability to extend record length with a reduction in channels are important as well. The problem is the tradeoff between color and speed, he says.
Agilent's 54600 series emulates the performance of an analog scope with very low dead time and a deep Z-axis. The trace is brighter when the beam is moving slowly, or in the case of a digital scope, when the data density is higher. In the analog model, when the beam is moving faster, the trace isn't as intense. To get the responsiveness, a monochrome CRT display provides about 1000 points across the screen. With a color display, the resolution comes down to about 500 points across the screen.
Longer record lengths or memory lets users maintain higher sampling rates and records for longer periods of time than a scope with less memory. According to Yokogawa's Shaw, his company's Model DL1700E has up to 8 Mpoints of memory per channel, allowing every waveform point to be saved in memory. Acquisition update rates are 30 times/s for a 1-Mpoint/channel acquisition.
For Wayne Stanley, design engineer at CTS Interconnect Systems Corp., must-haves include remote bus access, large capacity for real-time acquisition, internal waveform processing, and a compact design. The first two features can often be had with Windows-based scopes. GPIB is still a popular way to connect to PCs to get data, but many scope vendors offer a faster USB 2.0 connection.
Most new Windows-based scopes are computers as well, and they offer the amenities you'd expect from a computer. Look for built-in hard drives to store your acquired data, software to analyze your data (several Tek scopes include a trial version of NI's LabView), and Ethernet ports to connect to your local-area network. There also will be built-in FTP and Web servers so you can transfer files and monitor your data remotely or set up your scope and control it (just like the early HP network printers), as well as built-in printers, and of course, color displays.
If size matters, scopes like Yokogawa's DL 1700E have the footprint of a sheet of paper and weigh in under 12 pounds. If you really need small, Fluke's 190 Series Scopemeter (Fig. 9) has increased waveform memory from 1200 to 3000 samples per channel, while the 190C (color) models have added fast-Fourier-transform spectrum analysis functionality, two new trigger modes, and cursor-limited automatic measurement capability.
ANALYZE YOUR NEEDS
Before you purchase, look at traditional scope features; then digital features such as sampling rates, memory, analysis capabilities, and acquisition rates; and finally, built-in computer features—if you decide on a PC-based scope.
Be sure you know your application. How much bandwidth do you need? How many channels? What do you need in triggering? Is your application mixed signal? What kind of noise floor do you need, and how much jitter can you withstand?
Consider what you'll do with your data. How will you analyze it? Can you export it to computers? Remember that a PC-based scope can be compared on PC features, so think about the software that comes with it and the cost of additional software.
Scope vendors do listen to their customers. If there's a dream feature missing, you'll see it in the not-too-distant future.
Control Technology Corp.
CTS Interconnect Systems Corp.
DRS Test & Energy Management
Yokogawa Corp. of America