Current mode quadrature oscillator using current

differencing transconductance amplifiers (CDTA)

A.Ü. Keskin and D. Biolek

Abstract: A CDTA-based quadrature oscillator circuit is proposed. The circuit employs two
current-mode allpass sections in a loop, and provides high-frequency sinusoidal oscillations in
quadrature at high impedance output terminals of the CDTAs. The circuit has no floating
capacitors, which is advantageous from the integrated circuit manufacturing point of view.
Moreover, the oscillation frequency of this configuration can be made adjustable by using voltage
controlled elements (MOSFETs), since the resistors in the circuit are either grounded or virtually
grounded.

1 Introduction ductance amplifier (CDTA). Based on these canonic

sub-sections, a quadrature oscillator is designed. Simulation
In the last decade, various new current-mode (CM) active results verifying the theoretical analysis are also included.
building blocks have received considerable attention owing
to their larger dynamic range and wider bandwidth with
respect to operational amplifier (opamp)-based circuits. As 2 Current differencing transconductance
a result, current-mode active components have been amplifiers
increasingly used to realise high-speed and high-bandwidth
circuits operating in the current or the voltage mode. The CDTA element [9–11] with its schematic symbol in
The term quadrature oscillator (QO) is used because such Fig. 1 has a pair of low-impedance current inputs p and n,
a circuit provides two sinusoids with 901 phase difference. and an auxiliary terminal z, whose outgoing current is the
Various types of QOs are reported in the literature. In [1], difference of input currents. Here, output terminal currents
an active RC integrator and a passive RC integrator are are equal in magnitude, but they flow in opposite directions,
combined with a negative resistance to yield a QO. Another and the product of transconductance (gm) and the voltage at
QO using active-R circuits is described in [2]. The study
in [3] presents an operational transconductance amplifier In Ix−
(OTA)-based current-mode integrator and an all-pass Vn x−
n
section and their application to design of a dual-mode
QO. Operational transresistance amplifier (OTRA)-based
QOs using virtually grounded passive components have CDTA
been introduced in [4]. Maheshwari [5] reports a QO using
three current-controlled current conveyors of negative type
(CCCII-) and floating passive components, generating Vp P x+
relatively low-amplitude and unequal sinusoidal output z Ix+
Ip
signals. The QO described by Toker et al. [6] uses two
positive second-generation current conveyors, but a voltage
Iz
buffer is also required to avoid the loading problem.
Alternatively, Horng [7] introduces a QO consisting of two
a
current differencing buffered amplifiers (CDBAs), however
this circuit does not exploit the full capacity of the CDBA,
gmVz
since the positive input terminal of one active element is Vz
unconnected. Another earlier work on building a QO using p x+
the CDBAs is reported in [8]. ip z
In this paper, we propose new first-order current-mode
allpass sections using the current differencing transcon-
Z

− gmVz
r The Institution of Engineering and Technology 2006
IEE Proceedings online no. 20050304 n ip-in
x−
doi:10.1049/ip-cds:20050304 in
Paper first received 28th July 2005 and in final revised form 22nd January 2006
A.Ü. Keskin is with the Department of Biomedical Engineering, Yeditepe b
University, Kayisdagi 34755, Istanbul, Turkey
D. Biolek is with the Department of Microelectronics, Brno University of Fig. 1 Symbol and ideal model of CDTA
Technology, Udolni 53, Brno, Czech Republic a Symbol for the CDTA
E-mail: dalibor.biolek@unob.cz b Ideal model of CDTA. Here, Z is externally connected impedance

214 IEE Proc.-Circuits Devices Syst., Vol. 153, No. 3, June 2006
 Vdd M8 M10 M17 M16 M15 M18 M19 M20

IB1

I+ I−
M3 M13 M14
M5 M11
Vp Vn
M1 z
n p
M2
IB3
M6 M12
M4
M21 M22

IB2 M7 M9 M23 M24

Vss

Fig. 2 CMOS-based CDTA

IB1 ¼ IB2 ¼ 85 mA, IB3 ¼ 200 mA, bandwidth ¼ 400 MHz, VDD ¼
VSS ¼ 2.5 V

the z terminal gives their magnitudes. Therefore, this active R1

element can be characterised by the following equations lin p Iout
lin p Iout
CDTA x C2 CDTA x
Vp ¼ Vn ¼ 0; Iz ¼ Ip
In ; Ixþ ¼ gm Vz ; Ix
¼
gm Vz ð1Þ n n
z R2 z
C1
where Vz ¼ Iz  Zz and Zz is the external impedance R4
R3
connected to the z terminal of the CDTA. CDTA can be
thought of as a combination of a current differencing unit a b
followed by a dual-output operational transconductance
amplifier, DO-OTA. Ideally, the OTA is assumed as an Fig. 3 Two variants of CDTA-based current-mode allpass filters
ideal voltage-controlled current source and can be described
by Ix ¼ gm(V+ – V
), where Ix is the output current, and
V+ and V
denote the non-inverting and the inverting
input voltage of the OTA, respectively. Note that gm is a
function of the bias current. When this element is used
in CDTA, one of its input terminals is grounded
p x
(e.g., V
¼ 0 V). With dual output availability, Ix+ ¼
Ix

p x C2 CDTA 2
condition is assumed.
A possible CMOS-based CDTA circuit realisation R1 CDTA 1 n
z x−
suitable for the monolithic IC fabrication is displayed in R2
n Io2
z x−
Fig. 2. In this circuit, transistors from M1 to M12 perform C1
Io1 R4
the current differencing operation while transistors from R3
M13 to M24 convert the voltage at the z-terminal to output
currents at the two outputs of the DO-OTA section.
DO-OTA’s transconductance (gm) is controllable via its bias Fig. 4 Current-mode quadrature oscillator using CDBAs
current IB3. Also, a resistor (Rz) connected at the z-terminal Sinusoidal signals at x terminals of both CDTA elements have 901
can be used to adjust the gain of CDTA, while the voltage phase difference
at the input of DO-OTA is Vz ¼ Iz  Rz.
The gate terminals of the output transistors M11 and
M12 in current-differencing section are connected to bias- The CDTA-based QO is constructed by using the
voltages to provide drain output and high impedance at z AP sections of Fig. 3 cascaded in a loop as shown in
terminal (ideal current controlled current source, CCCS). Fig. 4.
Since an external resistor with relatively low resistance value The current transfer function for the AP filter of
(at the order of few kO) is connected to the z-terminal, the Fig. 3a is
use of diode-connected transistors M11 and M12 becomes 1
advantageous from IC manufacturing point of view. This is s

Iout ðsÞ R1 C1
due to the fact that the need for two additional bias-voltages H ðsÞ ¼ ¼
gm R3 ð2Þ
Iin ðsÞ 1
is eliminated by using diode-connected transistors. sþ
R1 C1
3 CDTA–based quadrature oscillator This circuit provides a phase shift of
Allpass (AP) filters are widely used in analogue signal jðoÞ ¼
2  arctanðoR1 C1 Þ ð3Þ
processing in order to shift the phase while keeping the The second allpass circuit yields a similar form of current
amplitude constant, to produce various types of filter transfer function (2), but with a sign difference. This second
characteristics and to implement high-Q frequency selective AP circuit in Fig. 3b provides a phase shift
circuits. Many AP filter circuits are described in the
jðoÞ ¼ 180
2  arctanðoR2 C2 Þ ð4Þ
literature using various types of CM active elements
[12–23]. However, only few of these AP filter circuits are For the sake of circuit uniformity and ease in practical
suitable for QO realisation. implementation, it is convenient to set R ¼ R1 ¼ R2 and

IEE Proc.-Circuits Devices Syst., Vol. 153, No. 3, June 2006 215
C ¼ C1 ¼ C2 in the quadrature oscillator circuit of Fig. 4. auxiliary circuitry, but the oscillation frequency is now
We then obtain the oscillation frequency of the oscillator as decreased.
1 For simplicity, consider the identical values of
oosc ¼ ð5Þ R3 ¼ R4 ¼ Rz, gm1 ¼ gm2 ¼ gm, and identical parasitic resis-
RC tances Rp ¼ Rn ¼ Rin of both CDTA elements in the
Therefore, the frequency of the proposed circuit is oscillator. Then the analysis leads to the modified amplitude
insensitive to variations in CDTA transconductances and oscillation condition
R3 and R4. Rin
However, the amplitude oscillation condition must be ðgm Rz Þ2 ¼ 1 þ 2 ð9Þ
fulfilled R
In contrast to the ideal case, to maintain steady-state
gm1 gm2 R3 R4 ¼ 1 ð6Þ oscillations, it is necessary to increase current gains gmRz of
where gm1 and gm2 are transconductances of CDTAs 1 and 2. both allpass sections.
Note that this condition can be adjusted without affecting Then the oscillation frequency will be decreased from the
the oscillation frequency. ideal value (5) as follows
During the steady-state oscillation, condition (6) must oosc
o0osc ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð10Þ
be dynamically fulfilled. During the initial transients, the Rin
left-hand term of (6) must be greater than 1 to assure the 1þ2
R
soft-start of the oscillator. There are more possibilities how
to provide it by controlling the left-hand parameters by This effect can be significant for relatively high input
the oscillation amplitude. resistances of CDTAs.
The effect of non-zero input impedances of the p and n
4 Non-ideal effects terminals can be reduced – if possible - by choosing R1 and
1/(ooscC1) much greater than Rp and Rn, or by considering
In a non-ideal case, the CDTA can be characterised by R1 reduced by Rp during the oscillator design stage. A more
general method consists in decreasing the input resistance
Vp ¼ Vn ¼ 0; Iz ¼ ap Ip
an In ; Ixþ ¼ gm Vz ; Ix
¼
gm Vz by negative feedback as shown in Figs. 5a and b. It can be
ð7Þ proved easily that in the ideal case, the current transfer
functions of the filters in Fig. 3 and Fig. 5 are identical.
where ap, an are the parasitic current gains between the p
z, However, there are more benefits of circuits in Fig. 5: We
n
z terminals of the CDTA, respectively, which are save the resistors R3 and R4, and parasitic input resistances
generally deflected from their ideal unity values by the are decreased owing to negative feedback. A drawback is
current-tracking errors, those absolute values being much that we need an additional copy of output current as the
less than one. feedback signal.
In non-ideal case, assuming that the tracking errors of
both CDTA elements of the proposed QO have the same feedback
Iout
values, the frequency of oscillations for the QO can be easily
R1
shown to be independent of the tracking errors of the lin p Iout lin p Iout
CDTA, and still described by (5). CDTA x C2 CDTA x
However, the amplitude oscillation condition is modified n n
z z
C1 R2
to R4
R3 Iout feedback
a2p þ a2n
gm1 gm2 R3 R4 ¼1 ð8Þ a b
2
Note that this condition can also be adjusted without Fig. 5 Modifications of the sections in Fig. 3
affecting the oscillation frequency. Absorption by negative feedback of resistances
a R3 ¼ 1/gm,
The z terminal of CDTA has high impedance, and a
b R4 ¼ 1/gm
small parasitic capacitance exists between this terminal and
ground in the non-ideal CDTA model. Therefore, one may
expect to see a pole at the angular frequency which is Detailed analysis of the oscillator containing such
equivalent to o ¼ 1/(Rz  Cz), where Rz and Cz are the modified allpass sections shows a further essential feature.
internal parasitic resistance and capacitance of CMOS Consider the oscillator of Fig. 4 with the first and the
CDTA at this terminal. Since Rz value is at the order of second sections are modified according to Figs. 5a and b,
megaohms, when a resistor of value RooRz is connected respectively. The analysis will be again performed on the
at this terminal, Rz77RER. assumption that R3 ¼ R4 ¼ Rz and gm1 ¼ gm2 ¼ gm. How-
Another non-ideal effect is caused by the non-zero input ever, let us distinguish different parasitic input resistances
impedances of terminals p and n. If input resistances Rp/Rn Rp and Rn of CDTAs. The analysis leads to some interesting
between the p/n terminal and the ground are taken into conclusions. The amplitude oscillation condition is now
consideration, then ideal transfer functions (2) and (3) of fulfilled automatically owing to the local negative feedback
allpass sections of Figs. 3a and b are modified. For example, in both allpass sections. To ensure reliable soft-start of the
the resistance between the input and the p terminal of the oscillation, some of the classical methods of amplitude
section in Fig. 3a is increased by Rp, and the capacitive stabilisation should be used.
reactance between the input and the n terminal is modified The oscillation frequency is now changed according to
by real part Rn. As a result, both the magnitude and phase the formula
responses are modified. The phase response is shifted oosc
towards the lower frequencies. Similar consequence is also o0osc ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 ð11Þ
valid for the alternative allpass section in Fig. 3b. Note that Rp
Rn
1
R
the amplitude oscillation condition can be again adjusted by
216 IEE Proc.-Circuits Devices Syst., Vol. 153, No. 3, June 2006
As shown in Section 5, the frequency deviation from the
ideal case can be now considerably less than for the classical
topology in Fig. 4. In the case of identical resistances Rp and
Rn, the influence of parasitic input resistances on the
oscillation frequency is totally cancelled.
It should be noted here that, one might combine the
analysis of non-idealities alpha and Rin together, and modify
the study by adding frequency dependence of alpha rather
than assuming it a constant term.
5 Simulation results
The quadrature oscillator configuration presented in this
study is simulated using the CMOS-based CDTA circuit
given in Fig. 2. For this purpose, the parameters of the
0.5 mm MIETEC real transistor model are implemented for
all MOSFETs in the circuit. Transistor aspect ratios are
indicated in Table 1.
Table 1: Transistor W/L ratios used in CDTA circuit
simulations
Transistor
M1–M6
M7–M10
M11–M12
M13–M14
M15–M20
M21–M24
CDTA transconductance is controlled by IB3. SPICE
simulations have verified that for IB3 in the range from 20
to 700 mA, gm is proportional to the logarithm of IB3. For
200 mA, gm ¼ 479 mA/V. As follows from condition (6), for
steady-state oscillation, the corresponding external resis-
tance connected to the z terminal is 2.88 kO.
The input resistances Rp and Rn are rather high for this
topology. Small-signal analysis leads to approximate values
of Rp ¼ 7 kO and Rn ¼ 2 kO. With respect to values of
resistors R1 and R2 mentioned below, one may expect
indispensable influence of input resistances on the oscilla-
tion frequency.
In the first step, a 1 MHz oscillator (Fig. 4) was designed
with the following element parameters: R1 ¼ R2 ¼ 15.9 kO,
C1 ¼ C2 ¼ 10 pF, IB3 ¼ 200 mA. To ensure soft-start oscilla-
tion, R3 and R4 were set to 3.2 kO. PSpice simulation led
to the following results. The steady-state oscillations were
reached within 25 ms after turning the supply sources on.
The amplitudes of output currents Io1 and Io2 are
approximately 130 mA. Owing to the absence of auxiliary
circuitry for loop gain control, the THD of generated
waveforms exceeds 7%.
The oscillation frequency is 779 kHz instead of 1 MHz
owing to the effect described in Section 4. According to (10),
this drop-off would be caused by parasitic input resistance
of 5 kO on the simple assumption that Rp ¼ Rn.
Finally, the principle of feedback current injection
(Fig. 5) was applied. Resistors R3 and R4 were removed.
Currents Ip1 and In2, injected into the p terminal of CDTA1
and the n terminal of CDTA2, were controlled as follows
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
IQ IQ 2 þ I2
Ip1 ¼ Io1 ; In2 ¼
Io2 ; IQ ¼ Io1
100 mA 100 mA
IEE Proc.-Circuits Devices Syst., Vol. 153, No. 3, June 2006
Owing to the quadrature character of currents Io1 and Io2,
IQ is a constant value, equal to the amplitude of generated
waveforms. In (12 a, b), this amplitude is set to be 100 mA.
For concrete values of Rp, Rn, and R, (11) gives an
estimation of oscillation frequency to 1.05 MHz. It can be
concluded that the effect of parasitic input resistances is
practically supressed.
The simulation results are shown in Fig. 6. The steady-
state oscillations are achieved within 4 ms. The oscillation
frequency is 1 MHz and the amplitude 100 mA, which can
be adjusted by a circuitry, operating on the basis of (12).
The THD factor is about 1%. It can be further decreased
by improving the loop gain stabilisation circuitry.
100 µA
0A
−100 µA
W/L (mm) 5 µs 10 µs
I (V × 1) I (V × 2)
time
8/1
5/1 Fig. 6 PSpice simulation
Sinusoidal signals at the x terminals of CDTAs in the oscillator shown
20/2
in Fig. 4 with feedback current injection as indicated in Fig. 5
16/1
6/1
4/1 6 Conclusion
In this study, a CDTA-based canonic quadrature oscillator
circuit is proposed. The proposed configuration is attractive,
because a) it can provide high frequency sinusoidal
oscillations in quadrature and at high impedance output
terminals of the CDTAs, b) the circuit has only virtually
grounded capacitors, which is advantageous from the
integrated circuit manufacturing point of view, and c) the
oscillation frequency of this configuration can be made
adjustable by using voltage-controlled elements (MOS-
FETs) [24–26], since the resistors in the circuit are either
grounded or virtually grounded. A special modification of
allpass sections is proposed which effectively supresses the
influence of parasitic input resistances of CDTAs on the
oscillation frequency.
It is expected that the proposed circuit will be useful
in various current-mode analogue signal processing
applications.
7 Acknowledgments
This work is supported by the Grant Agency of the Czech
Republic under grants No. 102/04/0442 and 102/05/0277,
and by the research programmes of Brno University of
Technology MSM0021630503 and MSM0021630513.
8 References
1 Sedra, A. S., and Smith, K. C.: ‘Microelectronic circuits’ (Oxford
University Press, 1998, 4th edn.)
2 Vidal, E., Poveda, A., and Ismail, M.: ‘Describing functions and
oscillators’, IEEE Circuits Devices Mag., 2001, 17, (6), pp. 7–11
3 Minaei, S., and C- ic-eko$glu, O.: ‘New current-mode integrator and all-
pass section without external passive elements and their application to
design a dual-mode quadrature oscillator’, Frequenz, 2003, 57, (1–2),
o2
pp. 19–24
4 Salama, K.N., and Soliman, A.M.: ‘Novel oscillators using the
operational transresistance amplifier’, Microelectron. J., 2000, 31, (1),
(12a, b, c) pp. 39–47
217
 5 Maheshwari, S.: ‘New voltage and current mode APS using current
controlled conveyor’, Int. J. Electron., 2004, 91, (12), pp. 735–743
.
6 Toker, A., Ozcan, S., Kuntman, H., and C - ic-eko$glu, O.: ‘Supplemen-
tary allpass sections with reduced number of passive elements using a
single current conveyor’, Int. J. Electron., 2001, 88, pp. 969–976
7 Horng, J.-W.: ‘Current differencing buffered amplifiers based single
resistance controlled quadrature oscillator employing grounded
capacitors’, IEICE Trans. Fundam. Electron. Commun. Comput. Sci.,
2002, E85-A, (2), pp. 1416–1419
8 Salama, K. N., and Soliman, A. M.: ‘Novel MOS-C quadrature
oscillator using the differential current voltage conveyor’. Proc.
Midwest Symp. on Circuits and Systems, 1999, Vol. 1, Las Cruces,
New Mexico, pp. 279–282
9 Biolek, D.: ‘CDTA – building block for current- mode analog signal
processing’. Proc. ECCTD’03, 2003, Vol. III, Krakow, Poland,
pp. 397–400
. Biolek, D., Hancioglu, E., and Biolkov!a, V.: ‘Current-
10 Keskin, A.U.,
mode KHN filter employing current differencing transconductance
amplifiers’, AEU- Int. J. Electron. Commun. Accepted for publication,
2005
11 Biolek, D., Gubek, T., and Biolkov!a, V.: ‘Optimization of CDTA-
based Circuits Simulating Ladder Structures’, WSEAS Trans. Math.,
2004, 3, (4), pp. 783–788
12 Salawu, R.I.: ‘Realization of an all-pass transfer function using the
second-generation current conveyor’, Proc. IEEE, 1980, 68, (1),
pp. 183–184
13 Soliman, A.M.: ‘Generation of current conveyor-based all-pass filters
from op amp-based circuits,’, IEEE Trans. Circuits Syst. II, Analog
Digit. Signal Process., 1997, 44, (4), pp. 324–330
14 Cıcekoglu, O., Kuntman, H., and Berk, S.: ‘All-pass filters using a
single current conveyor’, Int. J. Electron., 1999, 86, (8), pp. 947–955
. guz, S., C
15 Toker, A., Ozo$ - ic-eko$glu, O., and Acar, C.: ‘Current mode
allpass filters using CDBA and a new high Q bandpass filter
218
configuration’, IEEE Trans. Circuits Syst. II, Analog Digit. Signal
Process., 2000, 47, (9), pp. 949–954
16 Maundy, B.J., and Aronhime, P.: ‘A novel CMOS first-order all-pass
filter’, Int. J. Electron., 2002, 89, (9), pp. 739–743
17 Toker, A., Ozo$ . guz, S., and C - ic-eko$glu, O.: ‘New realization of
current mode tunable all-pass filters’, Frequenz, 2003, 57, (5–6),
pp. 128–131
18 Toker, A., and Ozo$. guz, S.: ‘Tunable filter for low voltage operation’,
Electron. Lett., 2003, 39, (20), pp. 175–176
19 Metin, B., Toker, A., Terzio$glu, H., and C - ic-eko$glu, O.: ‘A new all-
pass section for high-performance signal processing with a single
CCII-’, Frequenz, 2003, 57, (11-12), pp. 241–243
20 Bodur, S., Kuntman, H., and Çiçekoğlu, O.: ‘Design of
f"yrst-order allpass f"ylters employing single modified third
generation current conveyor’. Proc. ELECO03, Bursa, Turkey, 2003,
pp. 74–78
21 Ibrahim, M.A., Kuntman, H., and C - ic-eko$glu, O.: ‘Canonical
biquadratic all-pass and notch filters employing differential difference
current conveyor’, Frequenz, 2003, 57, (7-8), pp. 162–165
22 Maheshwari, S., and Khan, I.A.: ‘Simple first order translinear-C
current-mode allpass sections’, Int. J. Electron., 2003, 90, pp. 79–85
23 Toker, A., and Ozo$ . guz, S.: ‘Novel all-pass filter section using
differential difference amplifier’, AEU- . Int. J. Electron. Commun.,
2004, 58, pp. 153–155
24 Czarnul, Z.: ‘Novel MOS resistive circuit for synthesis of fully-
integrated continuous-time filters’, IEEE Trans. Circuits Syst., 1986,
CAS-33, pp. 718–721
25 Takagi, S., Czarnul, Z., Iida, T., and Fujii, N.: ‘Generalization of
MRC circuits and its applications,’, IEEE Trans. Circuits Syst, I
Fundam. Theory Appl., 1997, CAS- 44, pp. 777–784
26 Keskin, A.U., . and Hancıo$glu, E.: ‘CDBA-based synthetic floating
inductance circuits with electronic tuning properties’, ETRI J., 2005,
27, (2), pp. 239–242
IEE Proc.-Circuits Devices Syst., Vol. 153, No. 3, June 2006