Professional Documents
Culture Documents
Gyrator-Based Inductance
1.0. INTRODUCTION
The unavailability of inductors characterized by high quality factors, or high Q, is a
shortcoming of monolithic fabrication processes. This shortfall is limiting when circuit design
objectives entail the realization of narrowband radio frequency (RF) amplifiers, high selectivity
bandpass and notch filters, and other circuits for a variety of communication and information
processing applications. Planar spirals of metalization are used commonly, of course, to synthesize on chip inductors with inductance values in the few tens of nanohenries[1]. These
structures, which consume large surface areas, are difficult to parameterize reliably because
their inductance and quality factor values are mathematically intricate functions of geometry
and the electrical dynamics of distributed parasitic energy storage elements implicit to their
underlying bulk silicon[2]-[4]. Moreover, they rarely produce inductors having quality factors
larger than four to seven at signal frequencies of at least the high hundreds of megahertz. To be
sure, anemic inductive Q can be offset by incorporating Qenhancing negative resistance compensating circuitry[5]-[6]. Unfortunately, such compensation increases power dissipation,
degrades circuit noise figure, and limits dynamic range. Moreover, the sensitivity of Q
enhancing subcircuits to parasitic energy storage elements, as well as their outright potential
instability, mandates the incorporation of automatic on chip tuning schemes. In addition to
requiring further increases in standby power, these tuning subcircuits almost unavoidably
degrade circuit frequency response[7].
An alternative to the passive on chip inductor with or without Qenhancing compensation is the active inductor. Although plagued by higher noise and higher power consumption than are counterpart uncompensated passive realizations, active inductors are theoretically
capable of producing relatively high quality factors. Various analog architectures are available
to emulate the electrical characteristics of inductors. These include linear feedback methods[8],
the application of nonminimum phase networks[9] and design approaches predicated on the
gyrator[10]-[11]. The gyrator approach to active inductance synthesis is the most popular of
available techniques[12]. This popularity derives from the fact that gyrators can be configured
straightforwardly with operational transconductors, whose attainable broadband input/output
transfer characteristics have improved in direct proportion to the rapid maturation of deep submicron CMOS device technology[13]. Moreover, gyrators realized with operational transconductors feature transconductances than can be adjusted with applied bias, thereby allowing for
inductors whose values can be adjusted, or tuned, electronically.
The gyrator approach to inductance emulation, like other active synthesis methods, suffers from potentially serious noise, power dissipation, and dynamic range problems. A lesserknown detriment is potential instability in the sense of a propensity to produce underdamped
responses within the network passband. This latter problem, which causes the simulated
inductance characteristics to depart radically from those indigenous to ideal passive inductors,
is examined herewith. The principle cause of underdamping is analytically deduced, and the
propriety of proposed compensation techniques is verified.
Gyrator-Based Inductance
and output currents, I1 and I2, the positive senses of which are delineated in Fig. (1a), the terminal volt-ampere characteristics of an ideal gyrator are given by[15]
I1
0
I
g
2
g V1
,
0 V2
(1)
I1
V1
I2
I1
V2
Gyrator:
Gyration
Ratio = g
I2
V1
gV2
gV1 V2
(a).
(b).
where g is a designable transconductance parameter known as the gyration ratio of the active
network. Equation (1) suggests, as is diagrammed in Fig. (1b), that only two voltage controlled current sources are required in the corresponding two port equivalent circuit.
If the gyrator in Fig. (1a) is terminated at its output port in a capacitance, C, as shown
in Fig. (2a), the electrical model in Fig. (2b) produces a driving point input impedance, Zi(s),
given by
Zi (s)
Zi(s)
V1 s
I1 s
sC
g2
I1
(2)
Zi(s)
I2
V1
Gyrator:
Gyration
Ratio = g
V2
(a).
V1
I1
I2
gV2
gV1 V2
(b).
Figure (2). (a). Ideal Gyrator Terminated At Its Output Port In A Capacitance, C. (b). The Two Port
Equivalent Circuit Of The Terminated Gyrator In (a).
Clearly, the input impedance predicted by (1) is inductive, where the effective inductance, L, is
observed to be
.
(3)
g2
It follows that the ideal gyrator is capable of transforming an ideal capacitance incident with its
output port to an input port inductance whose value is inversely proportional to the square of
the network gyration ratio. Because the gyrator modeled in Fig. (2) is ideal and therefore
lossless, the quality factor of the synthesized inductor is infinitely large.
Operational transconductance amplifiers (OTAs), whose electrical characteristics emulate voltage controlled current sources, provide a logical foundation for the network realization
of a gyrator. Given the electrical symbol and model of the OTA shown in Fig. (3a), a gyrator
based inductor assumes the topological form offered in Fig. (3b). Observe in the latter diagram
T. Bakken & J. Choma
Gyrator-Based Inductance
that the input and output currents, I1 and I2, subscribe to the matric in (1), whence the effective
inductance given by (3) is forged at the input port of the network at hand. A laudable attribute
of the OTAbased approach toward realizing active inductances is that the individual OTA
transconductances, g, can be adjusted through the application of externally applied bias voltage
or current (not shown in the subject diagram). This flexibility effectively allows for electronically tunable active inductances.
I2
V1
I2 V2
Zi(s)
V1
V2
gV1
V1
(a).
I1
I2
gV1 V2
gV2
(b).
Zi(s)
V 1 I1
g
gV2
gV1
I2
V2
(c).
Figure (3). (a). Electrical Model Of An Ideal Operational Transconductor (OTA). (b). The Electrical Equivalent Circuit Of A
Capacitively Terminated Ideal Gyrator (c). The Transconductor Realization Of The Capacitively Terminated Ideal
Gyrator.
Gyrator-Based Inductance
derives largely from gate-source and gate-source overlap capacitances and has a value typically
in the range of tens to the low hundreds of femptofarads. In bipolar OTA structures, Ci derives
from the depletion and diffusion components of base-emitter junction capacitance and from
Miller multiplication of the base-collector junction transition capacitance. The output capacitance, Co, which is usually dominant in common source MOS transistor circuit architectures, is
determined primarily by drain-bulk capacitance and is generally of the order of the tens to a
few hundreds of femptofarads. In bipolar circuits, Co is the collector-substrate capacitance and
depending on the extent of Miller multiplication of the basecollector junction capacitance, it
may not be the dominant energy storage element. The value of Co in either device technology
can also be impacted significantly by requisite common mode bias compensation circuitry.
Finally, Cf represents the net feedback capacitance between the input and output ports of an
OTA. Because an OTA is usually a multistage topology, Cf is generally of the order of at most
a few femptofarads. Nevertheless, it is a troublesome energy storage element because at best,
it establishes an OTA right half plane zero serving to degrade phase margin. At worst, it can
incur significant underdamping or even outright instability in gyrator and other feedback networks. The upshot of the foregoing disclosures is the OTA macromodel abstracted in Fig. (4).
Cf
Ri
Ci V1
I2 V 2
Ro
Co
Figure (4). The High Frequency Macromodel Of The Operational Transconductor Depicted In Fig. (3a).
CO Ci2 Co1 ,
CF C f 1 C f 2
(4)
(5)
Gyrator-Based Inductance
Cf2
Zi(s)
V1
g2V1
g2
Ri2
Ci2
Ro2
Cf1
Co2
V2
g1V2
Ro1
g1
Co1
Ri1
Ci1
(a).
Zi(s)
V1
CF
RI
CI
V2
g2V1
g1V2
RO
CO
(b).
Figure (5). (a). The GyratorBased Active Inductor With Resistive And Capacitive Input Port,
Output Port, And Feedback Elements Incorporated. (b). The Linearized High Frequency Two Port Equivalent Circuit Of The Network In (a).
Note that if the operational transconductors in Fig. (5a) are identical and no additional circuit
elements are appended to either the input or the output ports of each transconductor, CI CO,
and RI RO.
An analysis of the model at hand produces the driving point input impedance function,
Re sLe
Zi ( s )
,
(6)
2
2
s
1
s
n
n
where
RI
1
(7)
1 g1 g 2 RI RO
g1 g 2 RO
is the low frequency resistive component of the input impedance, and
R R C CF C
C CF C
(8)
O
Le I O O
1 g1 g 2 RI RO
g1 g 2
is the low frequency value of the effective input inductance produced by the gyrator. The
approximations in the foregoing two relationships reflect the presumption that the resistances,
Re
Gyrator-Based Inductance
RI, and/or RO, are sufficiently large to ensure g1g2RIRO >> 1. Since RI and RO very large and
CO and CF very small collapse the network in Fig. (5a) to its simplified counterpart in Fig. (3b),
it is not surprising that (8) corroborates with (3) for the case of g1 = g2.
Unfortunately, the electrical synergy between the actual and simplified circuits
degrades with increasing signal frequency since (6) indicates undamped resonance at a frequency, n, which can be shown to be
1
n
.
(9)
CF CO C
Le CI
CO CF C
The last term in the bracketed expression in the denominator on the right hand side of (9) is
smaller than CF, which itself is a small capacitance. Accordingly, (9) suggests that the inductance produced at the input port of the circuit in Fig. (5a) effectively resonates with the net
shunt input capacitance, CI. The damping factor, , (which is the inverse of twice the quality
factor, Q, of the circuit) in (6) derives from
CI CF
g2 g1 CF .
2
1
1
(10)
n Le
Q n Le
RI
RO CO CF C CO CF C
Observe that g2 < g1 and a sufficiently large feedback capacitance, CF, result in a negative
damping factor, which is tantamount to network instability. Fortunately, g2 g1 precludes
right half plane poles and therefore ensures asymptotically stable circuit responses. To the
extent that g2 g1 and RO is very large, (10) delivers an approximate damping factor of
L
n e .
(11)
2RI
As expected, a progressively smaller input resistance, RI, which appears directly in shunt with
the resonant circuit comprised of the generated inductance, Le, and the shunt input capacitance,
CI, serves to increase the damping factor, thereby circumventing potential oscillation problems.
In addition to bracketing the relative stability of the system undergoing investigation,
the damping factor in (10) or (11) sets the degree to which the frequency dependence of the
effective port resistance and port inductance are rendered nominally constant over the signal
passband of interest. To clarify this contention, write (6) in the form,
Re sLe
Zi ( s )
(12)
Re sLe F s ,
2
2
s
1
s
n
n
with F(s) understood to be the second order lowpass function,
1
.
(13)
F s
2
2
s
1
s
n
n
Obviously, frequency invariant resistance and frequency invariant inductance require F(s) = 1,
which is satisfied only for low signal frequencies and approximated only for appropriate
damping factors and suitably large undamped resonant frequencies. Indeed, larger than the
inverse of root two precludes |F(j)| > 1, but constrains the 3dB bandwidth of |F(j)| to a
frequency smaller than n. On the other hand, smaller than the inverse of root two increases
T. Bakken & J. Choma
Gyrator-Based Inductance
function bandwidth at the price of incurring response peaking in excess of one within the passband. The optimum designable constraint appears to be equal to the inverse of root two,
which imposes a maximally flat constraint on |F(j)|; that is, |F(j)| displays no maxima
above unity and is constant to within three decibels of unity for all signal frequencies in the
closed interval, 0 n[16]. It follows from (11) that
L
(14)
RI n e .
2
is a reasonable constraint for ensuring nominally constant resistance and inductance over the
widest possible passband.
f
y
.
(15)
n
fn
Under steady state sinusoidal excitation conditions, (6) therefore becomes
Zi (jy) Re Zi (jy) j Im Zi (jy) ,
(16)
where using (11),
2
1 y 2 Re 2 y 2 n Le
1 y 2 Re 2 y RI
,
Re Zi (jy)
2
2
2 2
2 2
1 y 2 y
1 y 2 y
(17)
and
y 1 y n Le 2 Re
Im Zi (jy)
2
2
1 y 2 2 y
2
n Le y 1
1 y
2 2
y2
RI
.
Re
2 y
(18)
The last expression verifies that the input impedance is not inductive over all signal frequencies since a normalized crossover frequency, say yco, exists for which Im[Zi(j)] 0 for y yco.
This normalized crossover frequency is
f
R
(19)
yco co 1 e 1 ,
fn
RI
where the approximation reflects the fact that from (7), Re << RI. In short, the electrical nature
of the driving point input impedance changes from inductive to capacitive at nominally the
undamped resonant frequency of the gyrator circuit. Since the objective at hand is the realization of circuit inductance, it follows that the utility of the structure proposed in Fig. (5a) is limited to signal frequencies that are smaller than the undamped resonance predicted by (9). On a
positive note, (14) confirms that unconditional circuit stability, in the sense of a positive real
input impedance, is virtually assured as long as the damping factor, , subscribes to the inequality,
Gyrator-Based Inductance
Re
.
(20)
4RI
This inequality is relatively easy to satisfy in view of the fact that Re << RI and is chosen to
be approximately 0.707.
Using (17) and (18), the quality factor, Q(y), of the actively synthesized inductor is
given approximately by
R
n Le y 1 e y 2
RI
.
Q(y)
(21)
2
2
1 y Re 2 y RI
Over the normalized frequency range, y < 1, where the input impedance is inductive, (21) can
be approximated as
n Le y
Q(y)
,
(22)
2
Re 2 y RI
which displays a maximum at
Re RI
f
ym m
.
(23)
fn
n Le
Recalling (7), (23) implies that the frequency at which the inductor quality factor is maximized
is
1
RI
fm
.
(24)
2 Le g1 g 2 RO
The actual value of the maximum quality factor, Qm, derives from substituting (23) into (22):
1
Qm Q(ym )
g1 g 2 RI RO .
(25)
2
Since resistance RI is selected to prescribe a damping factor commensurate with maximally flat
inductance over the frequency passband, it follows that the shunt output resistance, RO, determines the maximum Q of the synthesized inductor. Note therefore that the largest possible
value of maximum Q is essentially dictated by the ability to realize an operational transconductor characterized by very large shunt output resistance. Equivalently, the value of Qm is
limited by the degree to which the output port of a transconductor emulates an ideal current
source.
The ratio, fm/Qm, is a useful design figure of merit for it effectively stipulates the requisite shunt output resistance corresponding to a desired inductance value and given operational
transconductor parameters. This ratio also complements the solution to the problem of determining the engineering practicality of realizing the desired inductance specifications. Note
that
fm
1
,
(26)
Le g1 g 2 RO
Qm
which is independent of the shunt input resistance, RI. Consider, for example, that the operational transconductors in Fig. (5a) have g1 = g2 = 2 millimho and that an inductance, Le, of 15
nanohenries is desired. If this inductance realization is to deliver a peak quality factor (Qm) of
T. Bakken & J. Choma
Gyrator-Based Inductance
100 at a frequency (fm) of 400 MHz, (26) delivers a requisite shunt output resistance, RO, of
about 1.3 megohms. Such an output resistance is unrealistically large within the constraints
imposed by low power transconductor design in state of the art deep submicron CMOS technology. However, it may be attainable in silicongermanium heterostructure bipolar junction
transistor technology. A more reasonable resistance value is RO = 100 kiloohms. Then for Le
= 15 nH and g1 = g2 = 2 mmhos, fm/Qm = 53.05 MHz. For this ratio, Qm = 15, which is a quality factor considerably larger than that afforded by an on chip spiral. It follows that fm = 796
MHz is achievable, provided fm < fn. With Qm = 15 and RO = 100 K, (24) yields RI = 2.25
K, which requires that the input port of the gyratorbased active inductor circuit be shunted
by an appropriate resistance. Unfortunately, it is unlikely that this value of RI guarantees
maximal inductance flatness over the circuit passband. To this end, the damping factor in (11)
corresponding to the required net effective shunt input resistance must be checked to see if
design compromises are mandated.
10
Gyrator-Based Inductance
Fig. (6) is the resultant schematic diagram, while Figs. (7) through (9) display pertinent
OrCad SPICE simulation results. As the following overview confirms, excellent corroboration between simulated and analytically deduced results is obtained.
(1). Fig. (7) displays the dependence of inductance on signal frequency, where the
inductance has been calculated as the frequency derivative of the imaginary component of the driving point input impedance, Zi(j). The simulated low frequency
inductance is 24.91 nH. This inductance is maintained to within 20% for signal
frequencies through 1.42 GHz.
(2). Fig. (8) depicts the frequency responses of both the real and the imaginary
components of the driving point input impedance for the circuit in Fig. (6). The
simulated low frequency resistance is 1.145 , and this resistance remains under
50 for signal frequencies through 1.12 GHz. The real part of the input impedance shows resonance at 5.75 GHz, while its imaginary counterpart is zero at 5.72
GHz.
(3). Fig. (9) offers the frequency response of the inductor quality factor. A maximum
Q of 11.77 is observed at a frequency of 173.8 MHz.
Zi(s)
OTA #2
g2
640
g1
54.4 K
594 fF
OTA #1
Figure (6). Active Inductor Realization For The Design Example Documented In The Text.
Each Transconductor Has An Input Resistance (RI) of 5 M, An Output Resistance (RO) Of 100 K, An Input Capacitance (CI) Of 5 fF, An Output Capacitance
(CO) Of 20 fF, A Feedback Capacitance (CF) Of 3 fF, And A Forward Transconductance (g1, g2) Of 5 mS.
11
Inductance (nH)
Gyrator-Based Inductance
30.00
20.00
10.00
0.00
0.01
0.03
0.10
0.32
1.00
3.16
10.00
-10.00
-20.00
-30.00
Frequency (GHz)
Impedance (Ohms)
Figure (7). The Simulated Frequency Response Of The Active Inductor In Fig. (6). The Inductance Value At Low Signal Frequencies Is 24.91 nH And Deteriorates To 80% Of This
Value At A Signal Frequency Of About 1.42 GHz.
700
500
Real Part
300
Imaginary
Part
100
0.01
-100
0.03
0.10
0.32
1.00
3.16
10.00
-300
Frequency (GHz)
Figure (8). The Simulated Frequency Responses Of The Real And Imaginary Parts Of The Driving
Point Input Impedance Of The Gyrator In Fig. (6). At Low Signal Frequencies, The Real
Part Impedance Is About 1.14 Ohms; It Remains Below 50 Ohms For Signal Frequencies
Under 1.12 GHz.
4.0. CONCLUSIONS
T. Bakken & J. Choma
12
Gyrator-Based Inductance
(2).
It is essential that the two operational transconductors in Fig. (3b) have sufficiently large
transconductances. From (8), large transconductances imply a large output port capacitance, C, for a desired inductance, Le, thereby reducing the inductance sensitivity to parasitic OTA capacitances.
Inductance values larger than the mid tens of nanohenries are potentially counterproductive because the generated inductance effectively resonates with the parasitic capacitance
of the OTA input port. It is essential that this resonant frequency be placed outside the
passband of the system in which the gyratorbased inductance is embedded.
14
12
10
8
6
4
2
0
0.01
-2
0.03
0.10
0.32
1.00
3.16
10.00
Frequency (GHz)
Figure (9). The Simulated Quality Factor Of The Inductance Produced By The Gyrator Circuit In
Fig. (6). The Peak Quality Factor, Which Occurs At A Signal Frequency Of 173.8
MHz, Is 11.77.
(3). From (11) the effective shunt input resistance, RI, of the circuit effectively sets the circuit
damping factor, while (25) and (24) show that the effective output resistance, RO, controls the maximum attainable quality factor and the frequency at which this maximum Q
is achieved. Very large Q demands that the output port of the OTA closely emulate an
ideal current source over the frequency range of interest. Higher Qs can be obtained with
progressively larger RI, but only at the expense of decreased damping factor. In turn,
damping factors less than 0.707 give rise to diminished stability and an effective inductance that does not remain constant throughout the circuit passband. In general, quality
13
(4).
Gyrator-Based Inductance
factors in the low to mid tens are attainable with transconductors realized in deep submicron CMOS technology.
It is essential that the transconductances, g1 and g2, of the two OTA units in Fig. (3b) satisfy the inequality, g2 g1. Failure to satisfy this constraint can result in instability. If g2
is designed to be larger than g1, an increased damping factor for a given quality factor
results, but the resonant frequency correspondingly decreases.
5.0. REFERENCES
[1].
[2].
[3].
[4].
[5].
[6].
[7].
[8].
[9].
[10].
[11].
[12].
[13].
[14].
[15].
[16].
14