A Tunable Ultra-Wideband Pulse Generator Using a ... - CiteSeerX

Report 0 Downloads 35 Views
A Tunable Ultra-Wideband Pulse Generator Using a Variable Edge-Rate Signal Erick Maxwell and Thomas Weller

Jeffrey Harrow

James A. Haley VA Medical Center Tampa, Florida 33612, USA Email: Jeffrey.Harrow Wva.gov

University of South Florida, Dept. of Electrical Engineering Tampa, Florida 33620, USA Email: {emaxwel2, weller} d,eng.usf.edu Abstract This paper presents a pulse-duration tunable UltraWideband (UWB) generator that is developed using a variable edge-rate signal. Edge-rate variability is introduced by applying a step recovery diode (SRD) to compress the edges of the source and then employing a simple RC network to adjust the edge-rate. Afterwards, the compressed signal is differentiated using microstrip transmission lines in a short circuit stub configuration. The tunable generator resulting from this approach demonstrates Gaussian and monocycle pulses with: good symmetry and low distortion over the tunable range; pulse width variation from 800 to 1150ps over a 1-2OpF capacitance range; and good agreement between simulated and measured results.

I. INTRODUCTION Ultra-Wideband (UWB) microwave systems are finding application in the form of impulse radio, as well as respiratory, cardiovascular and other sensing/monitoring applications [1]. The Federal Communication Commission (FCC) defines UWB as an intentional radiator with an instantaneous 10dB-fractional and total bandwidth of at least 0.2 and 500MHz, respectively [2]. This bandwidth is achieved primarily by radiating short pulses that are derived from a basic Gaussian pulse shape. Tunable pulse generators are useful in UWB radar and radiometric measurement because they provide a platform for optimizing the absorbed and reradiated power of an isolated target. Power optimization may be required for enhancing discrimination and evaluation of electrical characteristics associated with a target [3]. Consequently, a tunable generator can also be used to achieve varying penetration depth, radiation intensity and range resolution by controlling the shape, bandwidth and center frequency of the spectrum this is due to a target's frequency-dependent electrical properties. Tuning the spectrum shape and center frequency may be accomplished with pulse differentiation. However, bandwidth tuning is accomplished by adjusting the pulse-duration, which is a greater challenge and is presented here. Thus, pulse-duration tunable generators provide a valuable tool for research requiring UWB measurement of the electrical properties of materials.

This work was supported in part by the UNCF*MERCK Science Initiative and by the National Science Foundation IGERT Program under Grant DGE 0221681.

vs

Variable Edge Rate Compressor

.'c2 '

Pulse Shaper

RF.Ficrowave Differentiator

Figure 1. Schematic of a tunable UWB generator.

One approach for designing a pulse-duration tunable generator is based upon switching in sequential sections of transmission lines to combine short pulses or varying the circuit impedance. These systems often require a number of discrete components in addition to power biasing for each section. Moreover, they often loose their Gaussian symmetry as more pulses are combined for increasing the width [4], [5]. A pulse-duration tunable generator using a variable edge-rate signal provides a simpler approach to pulseduration tuning. In this paper we report on the development of a new tunable pulse generator that is designed using a variable edge-rate signal, which operates by generating and differentiating a variable edge-rate rectangular pulse [6]. The resulting generator produces Gaussian and monocycle pulses with good symmetry and low distortion over the tunable range, a tuning variation of 800 to 1150ps using a 1-2OpF capacitance, and good agreement between simulated and measured results. II. CIRCUIT DESCRIPTION AND DESIGN The tunable pulse generator in Fig. 1 is implemented with three sub-circuits, including: a variable edge-rate compressor, pulse shaper and RF/microwave differentiator. The variable edge-rate compressor sub-circuit provides a mechanism for producing a tunable pulse width by allowing slew rate control. This sub-circuit is constructed using a Metelics SMMD840-SOT23-OS step recovery diode to rapidly charge up and snap back on the rising (SRI) and falling (SR2) edges of the source. Although a sharp falling edge is not typically used in the construction of a Gaussian

1695

1-4244-0387-1/06/$20.00 (@2006 IEEE

LD

waveform, the corresponding step recovery diode (SR2) contributes to the pulse shape, width and low distortion achieved in this circuit.

Figure 2. Photo of the single-stage tunable UWB pulse generator. From left to right is the variable edge-rate compressor, amplifier, pulse forming network and RF/microwave differentiator

The edge-rate associated with the rectangular pulse that results from the step recovery diode is controlled with a simple RC network, where RN is a 60-ohm chip resistor - a value that was determined by optimizing the pulse shape in ADS. The RC time constant that results from this network provides a means to vary signal rise time by modifying the capacitance. The capacitance may be determined by first recognizing that the series inductance and shunt capacitance forms a first order low-pass filter. The cut-off frequency for this filter is obtained from the reciprocal of the step-function rise time. Consequently, the low-pass filter inductance (L k) and capacitance (C k) are defined by the following equations:

The resulting variable edge-rate signal is then passed to the pulse shaper sub-circuit, which forms a Gaussian pulse. This sub-circuit consists ofthe following: a Picosecond Pulse Labs 5840A-107 amplifier with 21dB of gain and 35dB of isolation over a 8OkHz-9.3GHz bandwidth; an attenuator of 6dB that is used to help meet the OdB input power requirement for the amplifier; and a pulse forming network that is used to differentiate the incoming rectangular pulse. Since, the pulse-forming network includes a differentiator; the initial value for the lengths LI and L2 was set to a quarter wavelength of the maximum frequency [8]. Since, the maximum frequency is determined from the rise time, the length may be expressed by the following equation:

(3)

Len = 1.2 Z10-90% u 4

-J

where £ is the electric permittivity, pt is the magnetic permeability, and the factor of 1.2 is applied to approximate the full rise time from the 10-90% rise time (cl090%). The resulting length was then optimized in ADS to achieve a desirable ripple and overshoot. As a result, the pulseforming network is constructed from a short circuit stub with 020

0.15

0.10 k

L-Ro Lk RL. -

(1) 0.

Ck C'

k

Ct

(2)

where, R0 is the circuit resistance, Ck is the normalized capacitance, Lk is the normalized impedance and CoC is the cut-off frequency [7]. Since, this circuit is based on achieving pulse-duration tuning by modifying the stepfunction rise time, a minimum and maximum rise time is used to constrain the filter elements. The full rise time associated with the 70.Ops transition time for the Metelics step recovery diode is used to limit the minimum stepfunction rise time. However, this time does not account for the diode junction capacitance since it is measured in a test fixture. In addition, the transition time is measured between a 20-80% rise in amplitude. A 10 ns minority carrier lifetime is used to limit the maximum step-function rise time because it affects the operational frequency of the diode. These limits are applied to equations 1 and 2 above, using normalized element values, which correspond to a maximally flat pass-band. This endeavor results in an inductance range of 4.0-0.35nH and capacitance range of 0.17-14.5pF. Consequently, the LD and CN are implemented with a 0.35nH chip inductor and a capacitor trimmer with a 1-2OpF range, respectively (Sprague-Goodman SG2020).

1696

-

0,15

(A)

D.15

D.05

D10 D.C) a 0Q D. 15 o

45 5A 0

Time (ns)

55

60

6

(B Figure 3. Simulated and measured waveforms at 1, 2, 5, 10 and 2OpF capacitance values (from left to right); A: Simulated data at variable edgerate compressor ; B: Measured data. Note: since SRDs are used for the rising and falling edges, the duty cycle follows that of the source. Also, both a sinusoid and square wave produce similar waveforms.

APCCAS 2006

30

35

40

45

the input of the circuit and an HP 54750A digitizing oscilloscope with an HP54715A 20GHz module to capture the output. All measured data obtained from this setup was compared to that simulated using Advanced Design System (ADS) 2003A. The simulation and measurement data were taken on each sub-circuit above. Measurement data was collected using an HP54750A oscilloscope with an HP54753A, 20GHz TDR plug-in. Figure 3 demonstrates the tuning range of the compressor as the capacitance is varied from 12OpF. The amplitude associated with the waveforms, in this figure, is a function of the amount of charge that is available from the step recovery diodes at the time of the snap. Since voltage amplitude may be expressed mathematically as the equivalent charge over the total capacitance, it is expected that the signal amplitude would decrease with an increase in the circuit's capacitance. Consequently, as the capacitance is trimmed the signal rise time decreases while the amplitude increases. Good agreement between simulated and measured data may be observed in this figure. Likewise, the measured and simulated data at the output of the pulse forming sub-circuit (Fig. 4) demonstrates low distortion and good agreement. The waveform increases in width as the capacitance is increased from lpF, where it is 800ps wide, to 2OpF where the waveform is 1150ps wide.

50

Time (ns)

(A)

45

50

Time (ns)

55

60

6!

Figure 4. Simulated and measured Gaussian waveforms at 1, 2, 5, 10 and 2OpF capacitance values (from left to right); A: Simulated Gaussian pulse at the output of the pulse shaper ( 1 st derivative of the edge-rate compressor output); B: Measured pulse over capacitance. Note: There is good amplitude agreement in measured and simulated data. However, the amplitudes associated with 5, 10 and 2OpF show increased attenuation

u~

Q~

a length LI of 103mm and a width of 2.5mm. Since, the resulting Gaussian contains both positive- and negativegoing pulses, an Agilent Technologies Schottky barrier diode package (HSMS2862) is used to clamp the negative going reflections by providing a ground path through DNI as well as blocking through DN2. The resulting Gaussian is then passed to the RF/Microwave differentiator sub-circuit where a monocycle is formed. This sub-circuit contains a DC18GHz 3dB attenuator to minimize circuit reflections. It is implemented using a 100-ohm resistor for matching and a short circuit stub L2 that is 80mm long having a width of 1.25mm. Each sub-circuit has been developed on a separate printed circuit board for the prototype, as shown in Fig 2. III. FABRICATION AND MEASUREMENT The variable edge-rate compressor and pulse forming networks are fabricated on a Rogers Corporation high frequency laminate (RT5870) with a relative dielectric constant of 2.33 and a board thickness of 0.787mm. The RF/Microwave differentiator is fabricated on a 0.787mm FR4 glass epoxy substrate having a relative dielectric constant of 4.2. The setup used to test this generator includes the use of an Agilent 33120A arbitrary waveform generator to produce the 14MHz lOVpp sinusoidal stimulus required at

35

40

45

50

55

60

65

7

50

55

60

65

7

Time (ns)

(A)

35

40

45

Time (ns)

(B)

Figure 5. Simulated and measured monocycle waveforms at 1, 2, 5, 10 and 2OpF capacitance values (from left to right); A: Simulated monocycle pulse resulting from a 2nd derivative ofthe pulse forming sub-circuit; B: Measured monocycle pulse resulting from differentiating the Gaussian.

APCCAS 2006

1697

The shape ofthe waveform remains Gaussian throughout the tunable range and has a distinct peak-amplitude as well as a slope that varies as a function of pulse width. However, the amplitude decreases sharply with an increase in capacitance for the measured results, i.e., the simulated amplitude at 2OpF is about 25mV, whereas the measured amplitude is about 5mV. Consequently, the measured amplitude at 2OpF is very close to the ringing noise in the circuit. However, the Gaussian shape may be restored by increasing the gain of the amplifier at the input ofthe pulse forming sub-circuit. The RF/microwave differentiator sub-circuit produces a monocycle pulse from the Gaussian that results from the pulse forming sub-circuit. Figure 5 shows the simulated and measured waveforms. As demonstrated, a monocycle having a 1.6ns width, results from differentiating the 800ps Gaussian. The differentiator used in this sub-circuit produces pulses that have closely matched amplitudes at the positive and negative levels. 0 E -10 20

=

30 40 50

-6o

q)

X 70 -80 0

1

2

3

4

5

6

7

8

9

The normalized spectrum associated with the Gaussian and monocycle waveforms above are shown in Fig. 6. These figures show that the tuning range and bandwidth decreases upon the application of a second derivative in the RF/Microwave Differentiator sub-circuit. The tuning range measures 300 NMHz and 160 NMHz, for the Gaussian and monocycle, respectively. Likewise, the bandwidth associated with the 1OpF capacitance measures 1.4 GHz for the monocycle spectrum (Fig. 6B) and 1.7 GHz for the Gaussian spectrum (Fig. 6A). A comparison of these responses against the shaped spectrum of the FCC mask for medical imaging demonstrates that pulse duration modifications alone may not provide sufficient shaping to optimize the mask. However, better optimization may be obtained by a combination of methods including pulseduration tuning and pulse differentiation [9]. IV. CONCLUSION We have developed a tunable pulse generator based on a novel mechanism of utilizing step recovery diodes for variable edge-rate compression. This approach simplifies UWB generator design by allowing a focus on generating a smooth slope for the step in a rectangular pulse and then developing RF/microwave differentiators. The waveforms that result from this approach demonstrate good Gaussian symmetry throughout a tuning range of 800ps to 1150ps using a 1-2OpF capacitance trimmer. In addition, these circuits require only an AC input and a DC supply for the amplifier, do not require biasing, and contain only 8 discrete components. This approach should be useful for applications that require a tunable UWB source.

10

REFERENCES

Frequency (GHz)

[1] R. Fontana, "Recent system applications of short-pulse ultrawideband (UWB) Technology," IEEE Trans. Microwave Theory Tech., vol. 52, no. 9, pp. 2087-2104, 2004.

[2] Federal Communications Commission, Code of Federal Regulation, Title 47, ch. 1, part 15, "Radio Frequency Devices," Sub-Part F, Ultra-Wideband Operation, Sec. 503, pp. 767-768, 2003. [3] J. Taylor, Ultra-Widband Radar Technology, Boca Raton: CRC Press,

L

:S. (1)

C-0

0 CL

2001.

C-0

(1) CY

[4] J. Han and C. Nguyen, 'Ultra-Wideband Electronically Tunable Pulse

(1) -o

Generators," IEEE Microwave And Wireless Components Letter, vol. 14, no. 3, March 2004. [5] J. Han and C. Nguyen, 'Microstrip Impulse Generators with Tunable

2 0) m

73

.D m

[6]

07

cn

0

1

2

3

4 5 6 Frequency (GHz)

7

8

9

10

[7]

(B)

[8]

Figure 6. The shaped spectrum for the FCC mask >for medical imaging and the normalized frequency response of wavefor ms generated using a lpF and 1OpF capacitance value; A: Response of Goaussian waveforms; B: Response of Monocycle wavefori'ms.

1698

[9]

Pulse Duration for Ultra-Wideband Applications," 2003 Asia Pacific Microwave Conference, Seoul, Korea, Nov. 2003. C Hsue, T. Cheng and H Chen, "A second-order microwave differentiator," IEEE Microwave And Wireless Components Letter, vol. 13, no. 3, March2003. D. M. Pozar, Microwave Engineering, 2d Ed., New York:John Wiley & Sons, Inc., 1998, pp 422-496. C Hsue, T. Cheng and H Chen,, "Implementation of First-Order and Second-Order Microwave Differentiators," IEEE Transactions on Microwave Theory and Techniques, vol. 52, no. 5, May 2004. I Oppermann, M. Hamalainen, and J. Linatti, UwB theory and applications, West Sussex:John Wiley and Sons, 2004 pp.2-8.

APCCAS 2006