What is the essence of a scope? Certainly, features such as display resolution and brightness, ease of use and responsiveness, and sampling speed, triggering, and bandwidth are important. Some digitizing products offer most of these attributes but are better described as data acquisition systems or transient recorders. Some scopes don’t provide all the listed features but they obviously are scopes, albeit slow or hard to use.
Newton’s Telecom Dictionary defines a digitizer as “a device that converts an analog signal into a digital representation of that signal.” This function is “usually implemented by sampling the analog signal at a regular rate and encoding each sample into a numeric representation of the amplitude value of the sample.”1
The same source defines an oscilloscope as “an electronic testing device that can display waveforms and other information on a TV-like cathode ray tube” and adds that scopes are basic fixtures in sci-fi movies. And actually, it’s not often that you see a mad scientist monitoring his monster’s vital signs with a digitizer.
Oscilloscopes differ fundamentally from digitizers because they are analog instruments used in test applications. The display of waveforms is emphasized to the extent that the images that scopes produce are even used in movies.
Digitizers are modules embedded in a VXI or similar test system controlled by the system computer. A small digitizer physically and electrically embedded in a laptop computer can provide scope-like benefits, but the virtual user interface may not be as easy to understand and use as real knobs and buttons. On the other hand, data analysis and report generation can be accomplished on the same instrument. Digitized waveforms are presented on a digital display.
Oscilloscopes
Triggered-sweep analog scopes still are sold in the hundreds of thousands of units each year. They are an extremely cost-effective and fast means of gaining insight into the relationship between two or more input signals. The signals have to be nearly repetitive for an analog scope to be useful.
What if they’re not repetitive at all? Or what if the signals are so slow to change that the waveform displayed with a 1-s/div time base is just a slowly moving bright dot? These are two of the special cases that digital storage oscilloscopes (DSOs) solve.
Because a DSO’s sampling rate can be much higher than the frequencies contained in the signal of interest, single-shot transients can be captured without the need for repetitive, similar waveforms to refresh the display CRT phosphor.
In a DSO, the digitized data is continuously replayed from semiconductor memory as quickly as required to generate an image of the waveform on the display device.
Early DSOs displayed waveforms on electrostatic CRTs via digital-to-analog converters (DACs) that drove each axis. Many of these instruments also operated as conventional analog scopes. Engineers didn’t fully trust the DSO function and wanted to be able to switch to analog to verify that what they were seeing was correct.
And with good reason. There are whole classes of signals for which analog scopes offer superior performance. True, techniques that use trace intensity or color to increase a DSO’s displayed information density go part way toward addressing the shortcomings. But, the very high trigger rate; high displayed vertical, horizontal, and intensity resolution; and instantly responsive controls of an analog scope are a hard act to follow. This is especially true when viewing modulated, complex signals that challenge a DSO’s data acquisition, compression, and display capabilities.
A DSO’s forte is displaying low- or high-speed signals where an analog scope is limited by the need for a special type of CRT. An analog scope excels at combining ease of use, high time and amplitude display resolution, and the display of modulated waveforms where the envelope characteristics are important. If you need to archive or further analyze digitized signals, a DSO is the only practical solution even though an analog scope may have a better waveform presentation.
The largest improvement in DSO performance has been the increase in the sampling rate that ensures most signals will be adequately oversampled. Consequently, waveforms displayed on modern DSOs usually represent the input signal rather than being an interesting but otherwise totally false alias. Of course, signals have become faster too, and aliases still can occur. Nevertheless, the slow sampling rates of early instruments often did result in aliasing, causing engineers to spend hours trying to find nonexistent problems.
Today’s DSOs use sophisticated digital signal processing to balance the user’s need for long memories and a fast display update rate without creating spatial or temporal aliasing artifacts. The problem is exacerbated by the simultaneous requirement for high acquisition rates.
Signal integrity is one of the more important buzz phrases of the ‘90s, coinciding with the explosive growth in digital data communications and wireless services. Fast trigger circuitry allows many DSOs to capture only exceptions to nominal signal characteristics, so the scope has to process far less information. But because a well-implemented communications link has very low error rates, huge numbers of messages must be examined before a trigger will occur. Consequently, a DSO’s acquisition rate also attracts scrutiny.
Many DSO design trade-offs are required because scopes have remained visual test instruments. They are display-centric by definition, but their acquired waveform data also can be transferred to external devices such as printers or computers. As more computation has been added to DSOs to enhance their functionality, the results of on-board measurements and analysis also have become available at the data I/O ports.
Scope Errors
Signal integrity relates to the scope as well as to its input. Most DSOs have a gain inaccuracy of 1% to 1.5%, but this is specified at DC. A scope’s overall Gaussian roll-off ensures good transient response although absolute gain at high frequencies is seldom specified. DSO-displayed relative gain accuracy is affected by the preamplifier, the attenuator, and the analog-to-digital converter (ADC), but not by the display system unless an analog-scope electrostatic CRT is used. Analog scopes generally have 2% to 3% gain errors because additional errors are contributed by the deflection amplifiers and the CRT.
LCD panels and magnetic-deflection CRTs are driven at TV-like rates with a composite signal that includes all cursors, graticule lines, special characters such as trigger markers, menu text and graphics, and the waveforms. Consequently, waveform-to-graticule relative accuracy is not affected by the linearity of the display device.
The absolute accuracy of the graticule lines may be in error, but the signal always will be properly registered with respect to each line. In contrast, the graticule in an analog-scope electrostatic CRT is etched into the glass, so any deflection amplifier or CRT nonlinearities add to the total gain error.
Most DSOs have 8-bit resolution, but a few provide 10 or 12 bits. For biomedical signals that contain unknown amounts of offset, for example, the larger dynamic range of a 10- or 12-bit system is advantageous. It’s also required if a complex signal must be expanded vertically to examine small details.
Higher resolution is claimed for averaging modes. These modes often rely on an amount of random noise to drive the process, resulting in a typical square root (n) relationship: acquire the signal 16 times to improve the signal resolution by a factor of four.
A more rigorous approach adds an amount of specially weighted noise to dither the ADC input. At each successive sample period, the ADC input is offset slightly from the actual input signal value. The amount, direction, and distribution of offset (dither) are predetermined to ensure that the averaged, higher-resolution data is not spectrally distorted by the dithering process.
One problem with dithering is that the peak-peak voltage range of the ADC is reduced by the amplitude of the dithering signal. Because the amount of resolution improvement achieved is proportional to the size of the dithering signal and the number of samples being averaged, there is a limit to what can be accomplished.2
Digitizers
Digitizer output data is analyzed to discover trends, anomalies, parametric distributions, and maximum and minimum values. A digitizer’s design may emphasize signal fidelity, data transfer speed, or numbers of channels, better suiting it for some applications than others. But technical or cost trade-offs don’t have to be made with the display section: there is none built in.
Johnnie Hancock, a program manager at Agilent Technologies, summed up the scope-vs-digitizer issue very neatly: “Scopes are most useful as troubleshooting tools, usually in an R&D environment. On a scope, you can view the signal in real-time and determine how a DUT is performing. Waveform digitizers work best when the problem already is known and more specific information is needed.”
Digitizers and DSOs are based on ADCs. For slower, very-high-resolution applications, sigma-delta converters can provide more than 20 bits of resolution—1 part in more than 1,000,000. Of course, noise and amplifier nonlinearities can quickly reduce this performance to 16 bits or less, but even 16-bit resolution is 256× higher than a scope’s usual 8-bit resolution.
A typical 20-MS/s digitizer has 0.5% basic accuracy, a 68-dB common- mode rejection ratio (CMRR), and 12-bit resolution. In contrast, the channel isolation specification for a representative 100-MHz, 8-bit DSO is greater than 40 dB from 0 to 20 MHz, and accuracy is 1.5%.
Digitizers are further optimized for time-domain or frequency-domain operation by the type of input band-limiting filter used. Filters with an abrupt passband/stopband transition, such as Chebyshev designs, reduce aliasing aggressively, but they introduce ringing in the time domain.
Frequency-domain filter-related errors can be corrected, resulting in an accurate spectrum. A less abrupt Bessel filter will give a much better representation of the signal in the time domain without requiring correction, but it will not reduce aliasing as completely.3
Scopes Used as Digitizers
The fastest scopes and digitizers usually use parallel flash converters and provide 8 bits of resolution. Eight bits or 256 digitizing levels are sufficient to present a relatively smooth, easily understood waveform display. So, why not use a DSO as a digitizer, especially for high-speed signals where it’s difficult to achieve more than 8-bit resolution in either type of instrument?
In fact, this often is done with good results, although there are exceptions. Scopes are discontinuous acquisition instruments; digitizers may not be. There usually has to be somewhere to put the data that a scope has captured before it can capture more. Some scopes avoid this problem by merging successive waveform acquisitions into pixel maps presented at a TV-like frame rate. Their acquisition and equivalent display rates are high, although their data format makes further external analysis awkward.
Other than this special case, scopes only can acquire and display signals continuously at very low speeds. Above about 0.1-s/div sweep speeds, the visual effect of the waveform scrolling across the display as it is captured begins to blur. At higher speeds, successive bursts of data repeatedly are acquired, processed, and displayed with significant dead time between bursts.
A digitizer may be capable of continuous 100-MS/s or higher throughput, limited only by the memory bus speed available. For example, the Acqiris DP210 and DC240 PC Data Acquisition Cards can transfer data at up to 100 MB/s over the PCI bus. The PCI bus can operate as fast as 66 MS/s (132 MB/s) and generally is the core technology responsible for high streaming (continuous) data rates in PC-based data acquisition systems.
A scope’s throughput is limited by its relatively slow I/O capabilities and data processing speed. Consequently, the best a high-speed scope can do is to capture successive data bursts. The maximum length currently available is 16 MB. Slower-speed digitizers and data recorders can write data directly to a hard disk, achieving multigigabyte storage capability.
Looking at the data transfer rate in another way, many applications only require intermittent data capture, but the bursts of signal activity may occur close together. According to Andrew Dawson, product manager for board-level products and advanced measurement systems at Gage Applied Sciences, “Rapid transfer of data records is very important for signals with a high pulse repetition frequency (PRF). Examples of high PRF applications include scanning radar, time-resolved ultrasonics, time-of-flight mass spectrometry, and nuclear particle counting.”
References
1. Newton, H., Newton’s Telecom Dictionary, Flatiron Publishing, 1996.
2. “The Dynamic Range Benefits of Large-Scale Dithered Analog-to-Digital Conversion in the HP 89400 Series VSAs,” Product Note 89400-7, Hewlett-Packard Co., 1994.
3. “Tutorial: Selecting A/D Converters, Digitizers, and Oscilloscopes,” Test System and VXI Products Data Book, Hewlett-Packard, 1997, pp. 539-545.
Oscilloscopes and Digitizers
Plug-In DSO Card
The Model 220 PC-based DSO Card provides two channels with D T and D V cursors, up to 50-ns/div sweep speed, a scroll mode sweep rate down to 60 min/div, and 20 MS/s sampling and 32 kwords of memory per channel. The card plugs into any PC expansion slot, works with Windows 95, and has a point-and-click graphics interface. Up to eight cards can be controlled simultaneously. Probes and CD-ROM software are included. $230. HC Protek, (201) 767-7242.
High-Resolution PCI Card
The 12-bit resolution CompuScope 12100 is a PCI bus data acquisition card that provides a 100-MS/s sampling rate on one channel or simultaneous sampling at 50 MS/s on two. A 60-dB signal-to-noise ratio ensures high immunity to digital noise. Other features include autocalibration, up to 4 MS of on-board memory, a high pulse repetition frequency mode that stacks successive acquisitions, and PCI bus mastering to achieve a 100-MB/s data transfer rate. The 12100 runs under GageScope for Windows, and software drivers are available for all popular compilers. 12100-1M 1-MB version: $5,995. Gage Applied Sciences, (800) 567-4243.
Fast Digitizers
The DP210 PCI Plug-In Card and the Model DC240 CompactPCI 6U Module provide 2-GS/s-sampling and have 500-MHz bandwidths. Both digitizers feature oscilloscope-like input signal conditioning with 50-mV to 5-V full scale ranges, 50-W and 1-MW coupling, variable offset, internal calibration, input protection, and fast overload recovery. Dead time is less than 500 ns in the sequential trigger mode, and memory can be expanded up to 4 Mwords from the basic 256 kwords. Single-channel DP210: $8,190; dual-channel DC240: $10,990. Acqiris, (877) 227-4747.
Acquired Data Averager
The Eclipse™ Digital Signal Averager converts analog output from a triggered experiment into an 8-bit resolution digital record from 512 to 262,000 samples long. n successive records, for example, from time-of-flight mass spectrometry, are added in hardware at the full 500-MS/s conversion rate. The average is computed in software by dividing by n, resulting in signal-to-noise improvement if uncorrelated noise was obscuring the signal originally. A 12-bit DAC provides a dithering signal, allowing the 8-bit ADC to yield 12-bit results. $12,950. EG&G Instruments, (800) 251-9750
Digitizer With Analysis
The LSA1000 is a two-channel, 1-GHz bandwidth waveform digitizer/analyzer with an internal 96-MHz PowerPC processor and up to 64 MB of system memory. The maximum sampling rate is 1 GS/s, and memory length ranges from 100 kB to 4 MB per channel. Channels may be combined to double memory and sampling speed. Analysis functions include averaging, extrema, FFT, a high-resolution mode, histogramming and trending, add, subtract, multiply, and divide. Additional waveform analysis packages are available. Remote control and data I/O are via a 10/100Base-T Ethernet port. An ActiveX control provides connectivity and control with Windows 95/NT computers. LSA1000 (100-kB memory): $14,950; LSA1000L (4-MB memory): $26,350. LeCroy, (800) 553-2769.
Flexible-Resolution Scope
The two-channel NI 5112 PC-based, 100-MHz bandwidth oscilloscope for PXI™/CompactPCI systems operates in two modes. For bandwidths above 4 MHz, it has 8-bit resolution and samples at up to 100 MS/s with a 2.5-GHz random interleaved sampling rate. For lower bandwidths, resolution up to 21 bits is available at slower sampling rates. Other features include 16-MB memory depth per channel, ±25-mV to ±25-V full scale input range with up to ±50 VDC offset, and a user-selectable input bandwidth limit. $2,495. National Instruments, (800) 258-7022.
Very-High-Speed DSO
The 3-GHz bandwidth, TDS 694C four-channel DSO provides a 10-GS/s sampling rate and 30-kB memory per channel. In addition to a 7″ color display and a floppy disk drive, the 31-lb instrument features 29 automatic measurements with statistics, a 100-ps/div to 10-s/div time-base range, 8-ps rms trigger jitter, and a ±25-ns channel-to-channel deskew adjustment range. Dual-window zoom, waveform histograms, FFT, and limit testing are standard. Video trigger, 120-kB memory length, an internal hard disk, and several types of passive and active probes are optional. $37,995. Tektronix, (800) 426-2200.
Mixed-Signal Scope
The 500 MS/s DL7100 SignalExplorer four-channel DSO processes four 500-MHz bandwidth analog and 16 digital channels (optionally) simultaneously. A new data stream engine ensures at least a 20-Hz display update rate when presenting up to 1 Mpoints from each of four channels. The memory can be expanded to 4 MB per channel from the basic 1 MB. Combining pairs of channels doubles the memory length and the sampling rate to give up to 8 MB and 1 GS/s. Segmentation creates a history memory for storing previous acquisitions. An automatic history search examines acquisitions for violation of user-specified waveform criteria. 1-Mwords/channel: $12,495; 4-Mwords/channel: $18,495. Yokogawa, (770) 251-8700.
GUI-Controlled Scopes
The Infinium family of two- and four-channel color-display digital scopes features simple, analog-like front-panel controls and a graphical user interface based on Windows 98. The bandwidth ranges from 500 MHz to 1.5 GHz with sampling rates from 1 to 4 GS/s. The memory length is 32-kB per channel, but pairs of channels can be combined to give 64 kB and doubled sampling rate. An anti-aliasing filter and selectable (sin x)/x waveform interpolation are standard. Trigger facilities include glitches as narrow as 500 ps; pattern; state; TV; and runt, setup/hold time, pulse width, and transition violations. $9,995 to $29,995. Agilent Technologies, (800) 452-4844.
SIDEBAR
Signal Conditioning Terms
DSOs and digitizers convert analog input signals into digital data. Signal distortion and noise are added during amplification, filtering, and analog-to-digital conversion. Many technical terms describe the analog and digital signals involved. Rather than present a list of the more important terms in a glossary, related items are grouped together and their interactions discussed.
Bandwidth, Sample Rate, Aliasing, Oversampling, Sampling Scope Operation, Random Sampling, Equivalent Time Sampling
The bandwidth of a system is the frequency range over which the signal amplitude remains within 3 dB of its midband value. Scope calibration is referenced to the signal amplitude at 50 kHz. DC-coupled instruments exhibit only an upper 3-dB point. AC-coupled instruments also have a lower 3-dB point although it’s often less than 100 Hz.
Nyquist’s theorem requires at least two samples for each cycle of the input signal to adequately represent analog waveforms in a sampled data system. Taking more than two samples per cycle is defined as oversampling; less than two is undersampling. An obvious, but overlooked, point is that Nyquist’s theorem refers to the highest frequency present in the input signal.
Failing to sample at or above the Nyquist rate leads to aliasing, an effect caused by samples from successive waveform cycles being mixed with earlier ones. If aliasing has occurred, there is no way to recover the original signal. Input band-limiting filters eliminate higher-frequency signal components that might otherwise alias. The filter will distort the signal, so the filter type and the roll-off rate must be chosen carefully.
Undersampling is useful, in spite of Nyquist, if the signal is repetitive. A classical analog sampling scope captures single points from triggered cycles of the signal, slightly delaying successive sampling times from preceding ones. The acquired points are reassembled into a greatly expanded view of the original waveform.
Digital random sampling, or equivalent time sampling (ETS), captures a number of points related to each trigger event and is an asynchronous process. Waveform reconstruction depends upon accurately measuring the trigger-to-sampling clock delay. ETS allows pretrigger acquisition.
Only classical sampling scopes that operate in an analog random sampling mode can present pre-trigger information. It is important to distinguish between ETS and direct or transient sampling. ETS can only be used with exactly repetitive signals.
Accuracy, Slew Rate and Harmonic Distortion, Linearity, Dynamic Range, Overload Recovery
Accuracy refers to how well something matches a reference, whether a forged dollar bill or a voltmeter measurement. A linear amplifier that does not add noise to the signal produces an output that is an accurately scaled representation of its input.
Slew rate limiting results from having insufficient current available to charge a capacitive load during signal transients. Given a certain amplifier bandwidth and perfect small-signal linearity, slew rate limiting may occur at high signal amplitudes.
In contrast, harmonic distortion can be caused by nonlinearities inherent in the amplifier at any signal level or by slew rate limiting. Output signal amplitude limiting caused by exceeding the amplifier’s input voltage range is an extreme form of inherent nonlinearity.
The linear voltage range at the input or output of an amplifier is commonly termed its dynamic range. The dynamic range of RF equipment refers to the ratio of the largest signal size to the value of the noise floor. The dynamic range of an ADC is the ratio of the largest to smallest digital output signal. Dithering significantly improves dynamic range and can correct nonlinearities.
The input voltage range of a preamplifier determines the point above which clipping and saturation due to large pulses, for example, may occur. Until the preamplifier recovers from overload, its output and the output of the following ADC will be wrong. Recovery time varies from about 1 ns in a modern, fast scope to hundreds of microseconds in a high-resolution digitizer.
Monotonicity, Resolution, Noise, Sampling Aperture Uncertainty,
Effective Number of Bits
ADCs should produce continuously increasing output code values in response to a smoothly increasing input voltage ramp. An ADC that has any decreasing codes is non-monotonic. The number of discrete output codes that an ADC produces is 2n, where n is the number of bits of resolution.
Aperture uncertainty refers to jitter in the precise timing of each sample. The period between actual samples is not exactly regular even if the period of the clock driving the ADC is. Noise can be expressed in a number of ways including total harmonic distortion (THD), signal-to-noise ratio (SNR), and spurious-free dynamic range (SFDR). Because of the statistical nature of random noise, the peak value is about six times the rms value. Flash ADCs will convert the peaks but integrating ADCs will not.
Effective number of bits (ENOB) is a measure of how closely an ADC system matches the performance of an ideal ADC with the same resolution and the same input signal. ADC internal reference divider nonlinearity; noise; and variations among the transient responses, input offsets, and overdrive sensitivities of the internal comparators contribute to lowered ENOB. ENOB typically falls off at high input signal frequency.
Copyright 2000 Nelson Publishing Inc.
January 2000