703 bommga #(w, zowof? se/ oa4, email: mutambd@vax. …applications of integrated circuits and...
TRANSCRIPT
Investigation of tapered microstrip transmission
lines for high speed detection circuits
D. Mutambo, D. Mukherjee
Physical Electronics Group, School of Electrical, Electronic and
Information Engineering, South Bank University,
703 BomMgA #(W, ZoWof? SE/ OA4,
Email: mutambd@vax. sbu. ac. uk
Abstract
The analysis of high speed digital interconnects using a generalised time domain scatteringparameters is presented. The scattering parameters can represent any arbitrary N-port passivenetwork under an impedance reference system, and can be obtained from measurements. Atime domain simulation of pulse propagation through a tapered line is performed, the resultsof which are shown and used in the preliminary evaluation of the performance enhancementof a low-noise satellite receiver amplifier. This work forms a general basis for theinvestigation of the exponential tapered line for use in high speed detection circuits.
1 Introduction
The transmission lines and their associated interconnections play an important
role at virtually all aspects of today's communication technology involving
applications of integrated circuits and printed circuit boards. With the design of
fast devices having switching times in the picosecond range, transmitting data
at high megabaud rates has become very commonplace in modern digital
computers and switching networks used for telecommunication. Signal delays
and rise times are more and more limited by interconnection lengths rather than
by device speed and represent a potential obstacle to the ultimate scaling on
VLSI technology. In recent years, modelling of interconnections has become a
major focus of interest in the implementation of digital and microwave circuits.
Shorter rise and fall times as well as higher frequency signals have compelled
most transmission lines to operate within ranges where dispersion is no longer
negligible. Skin effect and losses contribute to signal corruption leading to
waveform attenuation as well as pulse rise and fall time degradation. In wafer-
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
82 Simulation and Design of Microsystems and Microstructures
scale integration, these losses can become very significant and may lead to an
RC type time delays.
The implementation of high-density compatible packaging scheme is
essential for the design of high-speed digital systems such as gallium arsenide
integrated circuits. For microwave or digital applications, printed circuit
boards, chip carriers, and modelling of these networks represent the first step
towards implementing reliable design guidelines.
In this study, time domain approach is used to formulate the propagation
equations on a loss-less line with non-linear behaviour at the termination. A
time-domain flow-graph representation of the solution is also derived. In the
loss-less case, the solution reduces to very simple expressions which greatly
increases computational efficiency.
Recent advances in microwave circuit designs and the wide variety of
emerging applications have pushed for more accurate circuit simulation as a
necessary step towards high performance, low cost, and highly miniaturised
systems. The circuits containing non-uniform transitions are able to improve
the performance of common components like couplers, filters and matching
circuits. In such circuits, the conductor widths, the propagation constant, and
the characteristic impedance are not constant along the direction of
propagation. To simulate these circuits, reliable tools with accuracy and speed
suitable for CAD applications must be developed. Unfortunately, very few
practical modelling and simulation techniques for these circuits are available
today. One is, therefore, often forced to resort to costly fully three dimensional
field simulators for this purpose, the exception being the numerical approach
described in [1] or the analytical solution proposed in [2]. The analysis of these
lines have been a subject of interest for several decades and many authors have
contributed significantly to the study of non-uniform lines [2-8].
This work involves the design and modelling of a high speed signal (1-4
GHz) detection circuit, based on loss-less tapered (non-uniform) transmission
lines. Exponential tapered configurations is being investigated using the
coplanar strip-line (CPS) method. The CPS's inherent property of reducing
line-to-line coupling means more compact layouts can be achieved. For
increased functionality and lower cost it is important to minimise the size of
monolithic microwave integrated circuits (MMIC). Exponential transmission
lines (ETL) transform the characteristic impedance gradually from one value to
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Simulation and Design of Microsystems and Microstructures 83
another for matching purpose. The detection circuit utilising tapered lines
involves the use of components having low resistance, typically in the range of
3 to 10 ohms. The most common type of semiconductor used for these devices
consists of the ffl-V compounds, i.e. GaAs and related compounds.
Semiconductor laser diodes and PIN photodiodes are useful in many circuit
applications in microwave circuits because of their low cost, very small size,
light weight, high speed, time and space coherence. Laser diodes and high
speed photodiodes are normally coupled to 10 ohm circuits. Such
arrangements, beside limiting the bandwidth of components, make it difficult to
handle short electrical pulses in time scale of pico or femto-second, as required
for high speed optoelectronics [3]. ETL is used to match the input resistance of
such components to 50 ohms, allowing considerable improvement of the
temporal response of semiconductor laser diode and high speed photodiodes, as
compared with conventional coupling. The ETL eliminates the need for a
matching network between the transmission line and the detection circuit. Its
gradual change of characteristic impedance along the propagation direction
makes it possible to connect the circuit with low input resistance directly.
There are several mathematical methods of analysing the non-uniform lines.
An important aspect of previous studies [1] is that, except for limited cases [4-
8], they have dealt almost exclusively with non-uniform lines in frequency
(steady state) domain. In other words, the transient behaviour of non-uniform
lines in the time domain has not been studied extensively. The transient
response of a tapered line is a rather complicated function that depends on the
impedances, the line length, the propagation velocities and the time. These are
the most important parameters that determine the efficiency. To the author's
knowledge, no literature had treated the time domain scattering parameters of
an exponential line analytically. This investigation is partly motivated by the
desire to study the interaction of non-uniform lines with non-linear loads. Here
we limit our intentions to the time-domain characteristics of an exponential
line. For analysis of linear circuits, frequency domain scattering parameters are
used to evaluate the circuit performance. However, when transmission lines are
terminated with non-linear loads, the above technique is no longer an adequate
approach. A more appropriate approach is to characterise the transmission line
by a set of time-domain scattering parameters (S-parameters), so that the
interaction between the transmission line and non-linear loads can be expressed
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
84 Simulation and Design of Microsystems and Microstructures
by the corresponding time-domain convolution integral [5]. Therefore, in this
investigation, the transient characteristics of an exponential tapered line will be
analysed by time domain, and based on the results obtained, the formulation of
the tapered lines will be presented.
2 Time-Domain Formulation
Linear and non-linear subsystems can be modelled using time domain
measurement data and the entire system consisting of the individual subsystems
can then be very efficiently analysed. The model of the subsystems use linear or
non-linear time-domain transfer functions, not conventional equivalent circuits.
This approach gives computationally efficient non-linear models which can be
identified from time domain measurement. This results in a much reduced
effort and volume of the measured data compared to frequency domain
measurements. In the frequency domain a large number of measurements have
to be carried out to determine all four scattering parameters, in amplitude and
phase, as functions of frequency and at a number of power levels. In the time
domain the response is a real function and usually much fewer power levels
and time points are required to model non-linear systems. Scattering functions
for these systems are modelled by non-linear functions in the time domain as
shown in fig. 1.
82, (t)
Port 1 - S,, (t) " 822(0 Port2
Vlr(t) Si2(t) V%(t)
Figure 1: Time domain scattering functions model of a linear or non-linear two
ports.
The model in fig. 1 is similar to that used in frequency domain analysis but the
equations describing the model are now written and solved in the time-domain.
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Simulation and Design of Microsystems and Microstructures 85
This is an important difference between frequency domain scattering
parameters which are only functions of frequency and time domain scattering
functions which are functions of instantaneous values of the signal. The
advantage of solving the equations in the time domain is that the scattering
functions can be updated at each time step according to the signal strength.
Using the scattering functions signal flow-graph to represent the
interconnection of the subsystem, the whole system can be simulated in the
time domain explicitly and efficiently without the need to use iteration to solve
non-linear equations.
A non-uniform line can be described as a set of time domain scattering
parameters which relate two reflected waves and two incident waves [6]:
'* a z f , (1)
M * a,(f) , (2)
where t is the time, * denotes a convolution in the time domain, aj(t), b,(t), az
(t), b:(t) are incident and reflected waves for port 1 and 2, respectively, see fig.
2, S-j(t)(iJ = 1,2) are the time domain scattering parameters. The voltages at
ports 1 and 2 are the summation of the incident and reflected waves
(3)
(4)
and the expression for the currents are
(6)
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
x=0
Figure2: Scattering parameters representation of a non-uniform transmission
line.
The non-uniform line extends over a distance / and is terminated with twouniform reference lines at both ends. These are Z^, and Z^ , as the source
and load end reference impedances, which are equal to the characteristicimpedance of the exponential line at the left and right sides, respectively. Z^,
and Z^2 are used as the reference impedance in determining the scattering
parameters 5,-,-M of the non-uniform line.
To evaluate the S -parameters of the ETL, we consider a loss-less, non-
uniform line having a characteristic impedance as follows,
(7)
where Z , and Z %, [9] are the characteristic impedances at the left (source)
and right (load) of the exponential line, respectively, / is the length of the line
and x is the space variable (exponential line extends from x = 0 to x = /). By
using the values listed in section 2.1 in equation (7), Z(%) = 50£2exp (-0.2618%)
2.1 ETL synthesis
The length of the ETL is given as
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
(8)
where F is the maximum reflection coefficient, Z^ is the normalised ( with
respect to source impedance) value of the load impedance to be matched, and/I is the guide wavelength at the lowest frequency of interest. The normalised
impedance profile as a function of length / is give as
/ . (9)
The result obtained from the above equation can the be used to obtain the
characteristic impedance of the ETL by substituting it in (7).
The time domain S parameters of the ETL section were computed with the
use of the method [4] at frequencies of up to 2 GHz. Its length is calculated by
substituting these values, Zs = 50 Q , ZL=20Q , maximum reflection coefficient
of 0.1 at 2 GHz and effective dielectric constant of 9.8, in equations (8) and (9).
Giving us, L = 3.5 cm. The significance of this simulated line length is that, it
is the required length which will give ETL lossless properties for the above
conditions.
2.2 Matched Condition
We assume that both the load and source ends are terminated with matchingresistances, i.e., Z, =Z^, and Z =Z ,, where Z, and Z are the source
and load impedances. Under such conditions, a (t) vanishes and V (t) = b (t).
For a step voltage source u.j(t), as is evident from (2), the output voltage 1/2 (t)
is equal to
The output voltage 1/2 W is proportional to the integral of the scattering
parameter $2i(t), which can be found in [8].
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Simulation and Design of Microsystems and Micro structures
(a) (b)
Figure 3a and b: Step response at the load end for a well matched condition (a)
and an ideal voltage at the input end, (b). A normalised line interval A(f) = 1
represents the propagation delay for the wave travelling across the exponential
line.
Fig. 3 shows the step responses at the load end. The step response starts with a
maximum first arriving wave and the decreases to a steady state value. The
decrease can be approximated by a decay time constant which is dependent on
the physical line length and the characteristic impedances of the exponential
line at both ends [5]. The magnitude of the first arriving wave is proportional to
the square root of the ratio of the load end reference impedance to the source
end reference impedance. The horizontal scale is normalised with respect to the
propagation delay in which the wave travels across the non-uniform line. For
the values, say Z /Z /y = 4 and 9,the output voltage Vi (t) reaches the steady
state value at t ~ 3. If we invert the impedance ratio, i.e., Z /Z y = 0.25 and
0.11, aside from the normalised impedance factor, the outputs appear to be the
same as those for Z«/2 /Zre/i = 4 and 9, respectively
When Zs = 0, we have an ideal voltage source. The time-domain reflection
and transmission coefficients at the source end are -S(i) and S(i), respectively.
S(t) represents the impulse delta function commencing at t = 0. Therefore the
reflected wave bj (t) will be totally re-reflected back to the non-uniform line
and affect the output voltage at the load end. We assume that the load is
terminated with a matching impedance, i.e., #2 (t) = 0. For such a case, we may
use recursive equations [6] to compute the incident waves aj (t) and #2 (t). The
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
SY/?7W/(/f/6V7 fW D 9/g/7 6>/ M/C/YASl ^ ZA f/W M/C/ .S7r ra/r ( 9
reflected waves b\ (t) and i>2 (t) can be obtained by means of convolution
integrals in (1) and (2).
Figure 4: Step response of exponential and uniform lines terminated with a
FET
Figure 5: Pulse signal response of exponential and uniform lines loaded with a
FET.
The performance obtained by using the ETL as shown in fig. 4 compared to
uniform stub matching lines are by far better. The figure shows the response of
the exponential and uniform line to the first arriving pulse. The magnitude of
the pulse is 5 V and the pulse width is 1.2. The output of the exponential line
circuit reach the steady state value 4.3 V, while the output of the uniform line
configuration reach 3.6 V. Furthermore, the exponential line generates smaller
trailing ringing signal than the uniform line. Fig. 5 indicates that the falling
time of the pulse responses is smaller than the rise time of the pulse responses.
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
90 Simulation and Design of Microsystems and Microstructures
and the overall size of the unit reduced, achieving the goal for MMIC size. This
improvement is likely because the exponential line provides maximum first
arriving voltage wave to the load. The first arriving wave improves the rise
time of the output voltage when the circuit incurs a logic low to a logic high
transition. The time variation of b](t) due to the incident wave over time, the
non-uniform line, as compared to a uniform line ,appears to have a continuous
interactions with non-linear load which cause the output voltage to reach the
steady state in a relatively shorter period of time. It should be noted that the
settling down time of step response of an ETL depends on the signal line length
and impedance ratio [8]. Furthermore, to get a good matching result the
capacitive time constant at the load should be comparable to the settling down
time of the exponential line. Touchstone software was used to simulate the S
parameter of the amplifier, fig. 7, shows the response of the amplifier.
DBEGMAX]AMP
.DBCMF]AMP
32.60
38.88
28.88
1.788
1.588
1.3881.287 1.388 FREQ-GHZ 1.313
Figure 7 Touchstone simulation results for the amplifier (1.287 - 1.313 GHz).
3 Conclusion
In this paper, a technique was described for analysing exponential lines as used
for interconnects in high speed digital circuits. First, the theory was developed
and tested, and then it was extended to testing of the low noise amplifier. The S
parameters for the ETL and uniform microstrip lines for the amplifier were
calculated, and comparisons were made with result s by other methods. Good
agreement was found between this technique and others.
A method of analysing digital pulse propagation through the tapered
microstrip lines terminated with non-linear loads was described. Again, the
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Simulation and Design of Microsystems and Microstructures 91
This is due to the different terminal resistances when the output is entering the
logic high from the logic low or the reverse.
~>lfe£
&
W?-Input
Stage
Input — )port
matchin
network
/!\
; / TL
*1
r,
Hi
\ Short\Stub
h-
ed
•*
]
1 1 ,LJn
device +, H, matchi ig ^
Output Stage
r r1 * 2 * LVdd
\ ^ to next
-jHVrET ?
/MRS
{A
RL
Figure 6a Low noise amplifier matching block network with its equivalent
schematic matching diagram.
Void
Rd
to nextstage
D (drain)G (gate)
(from preceeding'L-iS (source)stage) -L-
Figure 6b Circuit arrangements of ETL replacing the stub matching network.
These uniform lines (stub matching) have been used in the design and
building of a two stage FET satellite receiver low noise amplifier. See fig. 6a.
From the above analysis, we can predict that by replacing the uniform line
matching networks with the ETL, fig. 6b, the performance will greatly improve
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
92 Simulation and Design of Microsystems and Micro structure 8
results were compared with previously published results, and agreement was
good. To confirm the validity of the technique, the calculated results were
compared with experimental results. Measurements were taken to determine
the S parameters of the microstrip structures, and calculated values showed
good agreement with measured values. Note, these results were for the low
noise amplifier with stub-matching. As for the ETL, the analysis and evaluation
of the system is underway.
4 References
[1] O. A. Palusinki and A. Lee, "Analysis of Transients in Non-uniform and
Uniform Multiconductor Lines," IEEE Trans. Microwave Theory Tech., Jan.
79% /%;. 727-7J&
[2] C. Nwoke, "An Exact Solution for the Non-uniform Transmission Line
Problem," IEEE Trans. Microwave Theory Tech., July 1990, pp. 944-946.
[3] M. Cristina, R. Carvalho, W. Margulis and J. R. Souza, "A new, small-
sized Transmission Line impedance transformer, with applications in high
speed optoelectronics," IEEE Microwave and Guided Wave Letters, vol. 2, No.
11, Nov. 1992.
[4] J. F. Lee, R. Mittra and J. Joseph, "Time-domain scattering parameters of
an exponential transmission line," IEEE Trans. Microwave Theory Tech., vol.
39, No. 11, pp. 1891-1895, Nov. 1991.
[5] Hsue and Hectman, "Transient response of an exponential transmission line
and its applications," IEEE Trans. Microwave Theory Tech., vol. 42, No. 3,
March 1994, pp. 458-462.
[6] J. E. Schutt-Aine and R. Mittra, "Scattering parameter transient analysis of
transmission lines loaded with non-linear terminations," IEEE Trans.
Microwave Theory Tech., vol. MTT-36, pp. 529-536, arch 1988.
[7] C. W. Hue, " Time-domain Scattering parameters of an exponential
transmission line," IEEE Trans. Microwave Theory Tech., vol. 39, pp. 1891-
1895, Nov. 1991.
[8] C. W. Hue and C.D. Hechtman, "Transient analysis of non-uniform, high
pass transmission line," IEEE Trans. Microwave Theory Tech., vol. 38 pp.
1023-1030, Aug. 1990.
Transactions on the Built Environment vol 31, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509