Accurate, Repeatable Oscilloscope Performance for Digital Design Verification and Debug for High-speed Applications

Trends in computer and digital system architecture and their underlying technology have created requirements for new capabilities in debug and verification tools. For example, serial interfaces are being adopted at an increasing rate, displacing earlier parallel interfaces. The rapidly growing speeds of these serial interfaces continue to accelerate. Table 1 illustrates this trend.

This increase in speed creates new challenges for designers and the tools they use to verify and debug systems. As data speeds escalate, the rise and fall times of the signals become faster and the time window for transmitting each successive bit of information decreases. As a result, the reduced timing margin makes jitter more critical. Therefore, oscilloscopes must be able to measure fast-switching signals accurately, as well as measure jitter and eye openings reliably.

Table 1. Technology trends in serial interfaces

Bandwidth has always been the "banner spec" for oscilloscopes, and for good reason. There are two ways to relate the required bandwidth of an oscilloscope to the technology in use. One is to analyze the harmonics of the signal. The other way is to analyze the accuracy of the rise and fall time measurements for the technology under consideration. Table 2 shows some examples of newer interface technologies and the resultant harmonics.

Table 2. Bit rates of selected technologies

For a given bit rate, an alternating 1-0-1-0 pattern will be a square wave with a frequency equal to ½ the bit rate. This is the fastest square wave on a serial interface with an embedded clock. This is shown in the second column of Table II. Columns 3 and 4 indicate the 3rd and 5th harmonic frequencies of this fundamental square wave. The higher the bandwidth of the oscilloscope, the higher the harmonics that will be captured accurately, and thus the more faithfully the oscilloscope will render the signal.

This analysis could lead to an overly conservative estimate of the required bandwidth or not conservative enough. What really matters are the rise and fall times of the signals, and these are not necessarily indicated just by knowing the bit rate. The time allocated to transmit one bit, known as a "unit interval," sets the upper limit of the rise and fall times. Rise time plus fall time cannot add to more than one unit interval. In many industry standard specifications, the combined rise and fall times can be much less than a half unit interval. In section three, we will explore rise time accuracy and repeatability as an indicator of an oscilloscope's performance.

Jitter is the second critical parameter that must be measured reliably with a real-time oscilloscope. At lower bit rates, the timing margins were generous therefore small amounts of jitter added by the oscilloscope were not critical in the overall jitter measurement, or in eye diagram mask measurements. With many new technologies exhibiting unit intervals as low as 100 ps, jitter dominates system reliability and performance and the margin taken up by the measuring device becomes critical. Sub-ps RMS jitter noise floor in the oscilloscope is required to ensure adequate margin. In section four we will examine how scopes introduce jitter into measurements.

Finally, most measurements start at the probe tip. The probe limits the bandwidth and noise floor of the final measurement. In section five, we will take a look at scope probe performance and how it affects measurements.

In section six, we will look at eye diagram measurements which are a good overall indicator of an oscilloscope's performance, in the same way engineers use eye diagrams as an overall indicator of the performance of high-speed serial links.

Section Two - Noise

Noise is an unavoidable limitation to an oscilloscope's performance. This is especially true at higher bandwidths. Broadband "white" noise is distributed across the frequency spectrum so the broader the bandwidth of an oscilloscope, the more noise energy will be present. Section four will discuss the fact that noise is the dominant determinant of jitter noise floor, therefore noise is especially critical.

Agilent's new DSO80000 series scopes make a significant breakthrough in noise performance. As seen in table three, the total RMS noise of these new oscilloscopes is on the same order as Agilent's 6 GHz and 7 GHz oscilloscopes. Therefore the noise power per hertz in the spectrum is significantly less. In addition, the noise floor of these new scopes is also significantly lower than competing 7 and 8 Ghz real-time scopes.

Table Three. Noise

Without Agilent's breakthroughs in low-noise, wideband amplifier design, it may still have been possible to manufacture an oscilloscope with 13 GHz bandwidth but its usability for measuring small signals would have been significantly compromised. As mentioned earlier, the higher the speed, the smaller the jitter that is acceptable. Noise is the primary contributor to jitter, so it is doubtful if an oscilloscope with a higher noise floor would be capable of making usable jitter measurements on high-speed serial interfaces operating above 3 Gb/s.

Section Three - Rise Time and Bandwidth

Rise time and bandwidth have been the figures of merit for oscilloscopes since the 1930's. The bandwidth and rise time of an oscilloscope must be sufficient to accurately track incoming signals. If a scope doesn't have sufficient bandwidth, other parameters or features are immaterial.

In the early days of analog oscilloscopes, the generally accepted rise time and bandwidth rules were:

• Rule #1: bandwidth = 0.35/(rise time), and rise time = 0.35/bandwidth
• Rule #2: The scope must have <1/3 the rise time of the incoming signal to measure its rise time with an error of 5% or less

These rules were based on the quasi-Gaussian response characteristics of analog oscilloscopes. Fortunately, modern, high-bandwidth digitizing sampling oscilloscopes have a response closer to "brick-wall" than Gaussian, so these those rules have become mute. The new rise time and bandwidth calculations include:

• Rule #1: Bandwidth ~0.43/(rise time)
• Rule #2: The oscilloscope's rise time only needs to be ~0.7 times the rise time of the signal to measure rise time with an accuracy of a few percent.

The probe must support this rise time all the way to the probe tip.

To track signals accurately, it is important that the oscilloscope not only have a wide bandwidth and fast rise time, it must have a flat frequency response all the way to the roll-off frequency, and it must have a linear phase response. Complex digital signals can be analyzed as a sum of a large number of sine waves of various frequencies, magnitudes, and phases. If the individual sine wave components do not pass through the scope with equal gain and delay, the resulting waveform will be distorted and measurements on it will not be accurate.

In digitizing oscilloscopes, the original waveform must be reconstructed from a set of samples to recover all the information in the original waveform. Agilent has developed interpolation algorithms that give consistent, accurate rise time and delay measurements on high-speed signals.

Even more stringent than the average, or mean, error in a rise time (or other time-interval) measurement is the repeatability of the measurement. Oscilloscopes are frequently used in a single-shot mode in which the measurement depends on capturing the waveform once. If the oscilloscope returns a range of different answers for the rise time, users cannot depend on it to accurately measure the rise time on any one occasion, even if the average of many measurements is highly accurate. Once again, low noise is required to get a consistent measurement.

The Agilent DSO81204A, with a 12 GHz bandwidth, is capable of measuring a 40 ps (20-80 percent) rise time with less than 5 percent error, with a standard deviation for repetitive measurements of less than 1 ps.

Figure One. Measurement of a 45 ps fall time on an Agilent DSO81204A

Figure Two. Measurement of a 45 ps fall time on an 8 GHz oscilloscope

Figure Three. Measurement of a 45 ps fall time on an Agilent 86100B DCA oscilloscope with 50 GHz bandwidth, for comparison

Section Four - Jitter

As mentioned earlier, jitter becomes increasingly critical as speeds increase and timing margins decrease. With many new technologies exhibiting unit intervals as low as 100 ps, jitter dominates system reliability and performance and the margin taken up by the measuring device becomes a point of importance.

Agilent's DSO80000 series scopes have typically less than 300 fs RMS trigger jitter. The trigger jitter remains the same if a high-speed pseudorandom binary sequence (PRBS) is applied to the scope, which indicates very low intersymbol interference (ISI) in the oscilloscope's trigger hardware.

Basic trigger jitter is not the only figure of merit for oscilloscopes. A more important figure of merit is jitter noise floor. This establishes the lower limit on jitter that you can measure on a serial data signal. The oscilloscope has internal software that can extract the clock from a serial data signal post-capture and then measure the timing of all edges captured in the memory to determine the jitter. This eliminates trigger jitter as a source of uncertainty from the measurement. In this case, the primary contributor to jitter noise floor is vertical noise.

Signals have a finite slew rate therefore any noise is translated to variation in the times when the signal crosses defined thresholds. Due to the exceptionally low noise of the Agilent DSO80000 series oscilloscopes, the jitter noise floor is typically less than 800 fs RMS for a time interval error measurement. This is the industry's lowest jitter noise floor for a wideband scope. The jitter noise floor for the Agilent 6 GHz 54855A oscilloscope is typically 1.4 ps.

Section Five - Probes

Probes have traditionally been the weakest link in the scope measurement chain. There are four critical factors in a probe that affect measurement accuracy, as well as usability:

• Loading of the circuit under test
• Faithful reproduction of the signal at the probe tip; flat magnitude and phase response across the frequency spectrum
• Noise
• Ease of access to the circuit under test

The award-winning Agilent InfiniiMax architecture, introduced in 2002, has been extended to include a new probe amplifier and three probe heads that provide 12 GHz measurement bandwidth all the way from the probe tip, with excellent fidelity. This new probe defines a new low noise floor performance class in such a wideband probe, contributing only 2.5 mV RMS noise.

With only 220 fF of equivalent capacitive loading at high frequencies, the loading of high-frequency signals also establishes a new benchmark.
The tip of the Agilent InfiniiMax solder-in probe is a mere 0.115 inches across, making it easy to solder it on to high-density electronic circuits.

In addition to the solder-in probe head, a new differential browser probe head and a new dual-SMA differential probe head are available. All of these accessories support measurement bandwidth of 12 GHz to the probe tip.

Section Six - The Eye Diagram

The eye diagram is the acid test of scope performance, just as it is the acid test of high-speed digital circuit performance. Any performance shortfall - noise, intersymbol interference, jitter - will close the eye seen on the oscilloscope. Customers use eye diagrams to evaluate the overall performance of their designs, so it is critical that the scope provide adequate margin in eye diagram testing.

Even at signaling speeds that are well beyond the oscilloscope's ability to accurately track the rise and fall due to its bandwidth, the oscilloscope presents very clean, open eye diagrams due to its exceptionally low noise, jitter, and ISI.

Figure Four. An eye diagram of a 6.6 Gb/s PRBS serial data signal seen on the Agilent DSO81204A 12-GHz oscilloscope.

Conclusion

To keep up with today's technology trends and continue to be the engineer's most valued debug and verification tool, oscilloscopes must meet new requirements. Designers require major leaps in oscilloscope performance to validate and debug new, higher-speed serial interface designs. The Agilent DSO80000 series oscilloscopes meet these challenges by measuring fast rise and fall times with bandwidth up to 13 Ghz, measuring very small amounts of Sub-ps jitter and measuring eye diagrams accurately on high-speed signals.

###

Contacts:

Janet Smith, Agilent
+1 970 679 5397
janet_smith@agilent.com

Related links for more information
	Press Release:	Agilent Technologies surpasses double-digit bandwidth barrier with industry's first 13 GHz real-time oscilloscope and probing measurement system (October 4, 2004)
	> More Press Releases > More Backgrounders