Pulse generators boasting rise times in the sub-10 ps regime have a wide range of important applications in electronics and semiconductor design, manufacturing, and test. While shortest possible rise (or fall) times remain the banner specification, there are many other factors that contribute to the performance of a sub-10 ps pulse generator. Two architectures are commonly employed for these high performance devices. The choice of architecture has a strong impact on the speed, quality, and flexibility of the edge produced. This article will present an overview of sub-10 ps pulse generator architectures, important performance considerations, and conclude with a brief discussion of key applications. Introduction
Until very recently, the need for pulses with sub-10 ps rise times consisted of small niche applications at the periphery of worldwide technology research and development. However, the never ending march toward faster processing and communications technologies has pushed data rates ever higher and the corresponding signal frequencies into the 10s of GHz. To keep up, measurement systems such as the real time oscilloscope have rapidly pushed into record acquisition bandwidths of greater than 60 GHz.
Click on image to enlarge.Figure 1:
7.2 ps fall time (90%-10%) generated using the Agilent N2806A Calibration Pulse Generator and measured at 63 GHz on the Agilent Infiniium 90000 Q-Series real time oscilloscope.[Get a 10% discount on ARM TechCon 2012 conference passes
by using promo code EDIT. Click here to learn about the show and
A whole host of important applications now demand pulse generation of sub-10 ps edge speeds. From calibration and metrology standards, to step response measurements, to characterization of devices using Time Domain Reflectometry and Transmissometry (TDR and TDT), engineering at today’s data rates and bandwidths requires ultra-fast pulses with rise times below 10 ps and frequency content well beyond 60 GHz.
There are two very different architectures used to produce sub-10 ps pulses. The first, and perhaps most intuitive, is the ubiquitous differential amplifier. Simply fabricate an amplifier that can switch fast enough to generate the desired rise time and we’re finished. The challenge is many applications demand edge speeds that push the limits of even the fastest amplifier designs and semiconductor processes in the world. We will return to the differential amplifier, but first our discussion focuses on the second pulse generator architecture, the shock line. Only recently have semiconductor processes enabled differential amplifiers to match, and even surpass, the edge speeds produced by shock lines. For this reason, the shock line has historically been the dominant architecture for applications that demand the absolute fastest edge speeds.The Shock Line
The shock line is simply a nonlinear transmission line, and has been studied for its ability to sharpen electromagnetic pulses since the 1960s , and perhaps even earlier. The shock line has proven to be an elegant and effective architecture for the generation of ultra-fast pulses.
Understanding the performance of shock line based pulse generators requires a basic understanding of how they work. Figure 2
shows an oversimplified diagram of a shock line. As mentioned earlier, a shock line is simply a nonlinear transmission line. There are a variety of ways to achieve the desired nonlinearity, however one of the most common is to replace the capacitors with reverse biased diodes in the semiconductor process. The transmission line becomes nonlinear as the capacitance at any point on the line depends on the instantaneous voltage at that point.
Click on image to enlarge.
Oversimplified shock line diagram; produce a nonlinear transmission line by replacing capacitors with reverse biased diodes. Capacitance varies with instantaneous voltage on the line. In this example we assume the shock line pulse has negative polarity (as is common in sub-10 ps pulse generators), producing the desired reverse bias on the diodes as drawn.
Basic transmission line theory states that propagation delay through a transmission line is given by:
τ = √LC
Thus, an increase in capacitance also increases the delay of the signal through the line. The result is the leading edge of the pulse is delayed with respect to the rest of the pulse and the pulse sharpens considerably as it propagates through the shock line. With proper engineering, the pulse produced by a shock line can be shaped to reliably output the desired edge speed.
Unfortunately, the shock line architecture has inherent drawbacks. It should be apparent that this architecture is polarity dependent. For a given shock line, depending on the voltage polarity and orientation of the diodes, you can output either a fast rising edge or a fast falling edge, but not both. Successful attempts have been made in research literature to accelerate both rising and falling edges using a MOS varacter shock line in Silicon , however achieving close agreement between the rising and falling edge speeds continues to prove challenging. The Differential Amplifier
In this section, we will not discuss how a differential amplifier works. Rather, we will briefly review the technology that has enabled differential amplifiers fast enough to generate sub-10 ps rise and fall times. Common implementations use a wave guide machined into a solid metal chassis to propagate the signal to two custom ASICs; a pre and post amplifier with switching speeds capable of producing the desired edge speed. Often exotic semiconductors are necessary for this application, with InP being the most prevalent. Figure 3
shows an example of a machined aluminum package with InP amplifiers.
Click on image to enlarge.
Inside the remote head of the Agilent N2806A Pulse Generator. The packaging is gold plated, machined aluminum. The InP pre and post amplifiers are visible in the corner nearest the 2 differential RF outputs.
There are only a handful of semiconductor processes in the world capable of producing these edge speeds. The benefit of this architecture, switching speed aside, is the flexibility of a differential amplifier; rising and falling edge speeds are closely matched and square waves can be output up to the bandwidth limitations of the device.Performance Considerations
When evaluating sub-10 ps pulse generators, how do these differences in architecture translate into performance metrics that can be found on a data sheet? While the fall time remains the banner specification, many other metrics are significant factors in the ultimate performance of the pulse generator. Fall Time
– The specification that gets the headlines, the fall time is typically faster than the rise time and thus more prominently displayed. The number itself has significant nuances to be aware of. First, be sure to know whether the spec is for a 90%-10% or 80%-20% measurement. Second, because these edge speeds are so fast, measurement systems such as oscilloscopes do not have enough bandwidth to directly measure the true edge speed accurately. Therefore, some math is required to convolve out the bandwidth limits of the scope to get a more accurate fall time produced by the generator. Unfortunately vendors estimate the fall time of their pulse generators differently leaving considerable room for ambiguity in this critical performance metric. Rise Time
– You may have to dig a little to find specified rise times as they can be orders of magnitude slower than the fall time specification. However, comparing this spec will help you determine the polarity dependence of the generator. Differential Outputs
– This may not be explicitly specified, especially if the generator is not differential. However, a differential generator will have two RF outputs that enable many differential measurements and applications not accessible to a single-ended device. Typically only generators with closely matched rise and fall times will be differential. This is a critical advantage of using a differential amplifier architecture for pulse generation.Maximum Rep Rate
– This spec is also closely tied to the matching of rise and fall times. A generator that can output sub-10 ps rising and falling edges can generally handle extremely high rep rate signals. This enables the generator to output square waves and PRBS signals into the many tens of gigahertz with ultra-fast edge speeds.Step Duration
– Most shock line generators have both minimum and maximum hold times for a step, typically on the order of tens of nanoseconds. Differential amplifiers have the flexibility to follow the input signal and hold a step indefinitely. Deviation from an Ideal Step
– This performance metric is very unlikely to be found on a data sheet as it is difficult to quantify with a single number. However, the measurement and corresponding plot versus frequency is a critical parameter in evaluating the quality of the step generated, not just the raw speed. Many generators have a lot of sharp transitions in their spectral content. This spectral impurity can cause significantly degrade measurement repeatability. Figure 4
shows a plot comparing two different sub-10ps pulse generators: a differential amplifier generator, dark blue (magnitude) and dark red (phase), versus a shock line based generator, bright blue (magnitude) and bright pink (phase). Regardless of your application, a higher quality edge will give you better, more accurate, and more repeatable results.
Click on image to enlarge.Figure 4:
Plot of deviation from the ideal step for Agilent N2806A, dark blue (magnitude) and dark red (phase), versus popular shock line based generator, bright blue (magnitude) and bright pink (phase). Notice the lack of resonances in the spectral content of the Agilent generator, much more closely approximating an ideal step for more accurate and repeatable measurements.Example Applications
There are many important uses for sub-10 ps pulse generators as technologies push into uncharted bandwidths and data rates. Below, we highlight just a couple of these applications.Time Domain Reflectometry and Transmissometry
– The interaction of a fast edge with a circuit element can be used to characterize the behavior of that element. Observing the signal that is reflected back by the circuit element is called Time Domain Reflectometry (TDR). Similarly, observing the signal that is transmitted through the element is called Time Domain Transmissometry (TDT). The combination of these techniques, both of which fall under the umbrella of Network Analysis, can fully characterize an element in a circuit so that its effects can be removed, modeled, or otherwise accounted for.
In the frequency domain, the total bandwidth over which the characterization is accurate depends on the frequency range of the spectral content of the fast edge. In order to characterize circuit elements beyond 40 GHz, a sub-10 ps edge is necessary to provide sufficient high frequency content. In these instances a pulse generator can be employed to accelerate the calibration edge to the speeds necessary to achieve the desired characterization bandwidth. Figures 5 and 6
show example measurements of TDT and TDR respectively using a sub-10 ps pulse generator.
Click on image to enlarge.Figure 5:
Frequency response magnitude plot showing insertion loss correction of 1.85mm cable using Agilent’s PrecisionProbe Advanced and 90000 Q-Series real time oscilloscope. The light blue is the uncorrected response of the cable, the yellow is the correcting filter applied by the scope. The dark blue is the corrected response of the cable, flat out to 62 GHz. This high bandwidth TDT is enabled by the 63 GHz of bandwidth on the oscolloscope and the sub-7 ps edge speeds generated by the Agilent N2806A Calibration Pulse Generator.
Click on image to enlarge.Figure 6:
Incident and reflected step from high bandwidth TDR measurement on a broadband open circuit. The measurement was made using an Agilent DCA86100C Sampling Oscilloscope with 2 modules, 86118A 70 GHz electrical module and 54754A TDR module. The edge out of the TDR module was accelerated using the Agilent N2806A Calibration Pulse Generator.Step Response and Calibration
– Having a known step with an extremely fast edge and a nearly ideal step response in the frequency domain is an invaluable tool in demanding step response and calibration measurements. Many test and measurement systems, as well as custom electrical and optical systems in research and industry, require calibration to very high frequencies using a fast, nearly ideal step. For example, a sub-10 ps pulse generator is an ideal piece of equipment to measure the rise time capabilities and calibrate the response of a high bandwidth oscilloscope. References
GUTZWILLER, M.C., and MIRANKER, W.L.: ‘Nonlinear wave propagation in a transmission line loaded with thin Permalloy films’, IBA4 J., October 1963, pp. 278-287
Grischkowsky, D., et al.: ‘Electromagnetic Shock Waves from Transmission Lines,’ Physical Review Letters, vol. 59, num. 15, pp. 1663-1666, Oct. 1987.
KATAYEV, I.G.: ‘Electromagnetic shock waves’ (Iliffe Books Ltd., London, 1966)
M. G. Case, “Nonlinear Transmission Lines for Picosecond Pulse, Impulse and Millimeter-Wave Harmonic Generation,” Ph.D. dissertation, Univ. of California, Santa Barbara, 1993.
M. J. W. Rodwell, M. Kamegawa, R. Yu, M. Case, E. Carman, and K. Giboney, “GaAs nonlinear transmission lines for picosecond pulse generation and millimeter-wave sampling,” IEEE Trans. Microwave Theory Tech., vol. 39, no. 7, pp. 1194–1204, Jul. 1991.
E. Afshari and A. Hajimiri, “Nonlinear Transmission Lines for Pulse Shaping in Silicon,” IEEE J. of Solid-State Circuits, vol. 40, no. 3, pp. 744-752, March 2005.About the authorDaniel Ruebusch
manages strategic marketing of high performance oscilloscopes at Agilent Technologies. Daniel joined Agilent in 2011. He has past experience in semiconductor device physics and processing and consumer sales and marketing. Ruebusch holds a B.S. from Cornell University in both Electrical Engineering and Materials Science and an M.S. in Electrical Engineering from U.C. Berkeley.