Download A 4.3-GHz VCO with 2-GHz tuning range and low phase noise

142
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 1, JANUARY 2001
A 4.3-GHz VCO with 2-GHz Tuning Range and Low Phase Noise
Paavo Väänänen, Mikko Metsänvirta, and Nikolay T. Tchamov
Abstract—A voltage-controlled oscillator (VCO) based on
double cross-coupled multivibrator structure with a center frequency of 4.3 GHz and a tuning range of 2 GHz has been designed
and implemented in standard 0.35- m BiCMOS technology. The
measured phase noise is 113 dBc/Hz at 600-kHz offset frequency
from the carrier. The VCO draws 14.6 mA from the 2.5-V supply.
Index Terms—BiCMOS technology, multi-GHz VCO, wide
tuning range.
I. INTRODUCTION
T
HE voltage-controlled oscillator (VCO) is one of the most
challenging parts of the radio transceiver to integrate using
relatively cheap BiCMOS or CMOS processes. One reason for
values of the integrated inductors are low
this is that the
compared to those of the discrete ones. This causes, for instance, lowering of the phase-noise performance of the resonance-tank-based oscillators. Due to the switching-based principle of the relaxation oscillators the of the inductors does not
affect the performance, and low quality factor integrated inductors can be used to increase the switching speed of the transistors
in the circuit.
This paper introduces a new version of a double cross-coupled VCO [1], [2]. Also, in this newer topology, the need for
differential control is avoided, and so the control block is simplified and the tuning sensitivity can also be more easily adjusted.
The VCO controlling principle is presented first. Second, the
effect of using inductive loads is analyzed, the simulation and
measurement results are given, and brief comments are made.
Finally, the measured results are compared to other published
relaxation VCO results. Implementation was carried out using
the ST-Microelectronics 0.35- m BiCMOS process.
Fig. 1.
Schematic of the new double cross-coupled multivibrator ICO.
Transistors
and
direct the current drawn by the conwhich defines the current that bytrollable current sink
passes the timing capacitor . This method requires only one
controllable current sink and avoids the need for two differentially controlled sinks, as in [1], [2].
III. USING
II. VCO CONTROL PRINCIPLE
The schematic of the new double cross-coupled current-controlled oscillator (ICO) is presented in Fig. 1. The current
which is
through the circuit is defined by the current
and
form the emitter-coupled
kept constant. Transistors
and
act as a differential pair directing the
multivibrator,
control current through the timing capacitor [1], [2], and
and
together with the MOS transistors
and
function
as active-pull-down output buffers [3].
THE
RL-LOAD IN DOUBLE CROSS-COUPLED
MULTIVIBRATOR
Bandwidth extension of the amplifier by adding a zero is [4]
known also as shunt peaking. Here it is used to extend the bandwidth of the multivibrator collector load circuit.
is the load resistor,
is the
In the following equations,
and
equivalent capacitance as seen at the collector nodes of
, and is the inductor value. Calculating the impedance of
the collector load circuit when only the resistor is used gives
(1)
Manuscript received June 13, 2000; revised September 8, 2000. This work
was supported by the National Technology Agency of Finland (Tekes) as part
of the project Digital and Analog Techniques in Flexible Receivers.
P. Väänänen and M. Metsänvirta are with Nokia Mobile Phones, 33720
Tampere, Finland (e-mail: [email protected]; mikko.metsanvirta@
nokia.com).
N. T. Tchamov is with the Telecommunications Laboratory, Tampere University of Technology, 33720 Tampere, Finland (e-mail: [email protected]).
Publisher Item Identifier S 0018-9200(01)00444-9.
and
(2)
when the additional inductor is used.
0018–9200/01$10.00 © 2001 IEEE
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 1, JANUARY 2001
143
TABLE I
SUM UP OF THE EXTRACTED SIMULATION RESULTS AND MEASUREMENT RESULTS
Fig. 3.
Differential pulse fronts.
Fig. 2. Load impedances as the function of the frequency.
Using the values
,
nH, and
pF, and
plotting the impedances as a function of the frequency produces
the curves shown in Fig. 2. We can clearly see the bandwidth
extension caused by adding an inductor into the circuit, and so
one zero into the equivalent collector impedance function.
In the time domain, this means that the rise time of the pulse
is decreased. This can be analyzed easily by finding the step
response of the load circuitry. The equations to be solved are
simple first- and second-order differential equations and the step
responses using the values mentioned above are plotted in Fig. 3.
It is clear that the rise time is reduced when the inductor is used
and so the circuit can be tuned into higher frequencies than when
using a simple resistive load.
The phase noise in relaxation oscillators is mainly caused by
the uncertainty in the triggering moment [5]. The voltage over
the timing capacitor that sets the relaxation period is modulated
by the noise contributed by the control block and the oscillator
components. Therefore, the low of the inductor is not an issue
in this instance.
IV. SIMULATION AND MEASUREMENT RESULTS
The simplified schematic of the designed and implemented
circuit is shown in Fig. 4. The circuit was simulated using
CADENCE Spectre RF circuit simulator. The simulated center
frequency is 4.3 GHz with a tuning range of 1.6 GHz. The
simulated current consumption was 15.5 mA from the 2.5-V
supply. The simulated tuning curve together with measured
one is shown in Fig. 5. The measured spectrum at 5.17 GHz
with second and third harmonics is shown in Fig. 6. All the
measurements were carried out using a 2.5-V supply and the
measured direct current drawn from the power supply was
14.6 mA. The phase noise at the control voltage of 2.5 V
is shown in Fig. 7. In Table I, the results from the extracted
simulations are compiled to summarize the achievements.
The chip micrograph is shown in Fig. 8. The size of the chip
mm .
including the pads is
The measured center frequency is 4.4 GHz and the tuning
range is 2 GHz. As can be seen from Fig. 5, the measured VCO
tuning range is slightly greater than the simulated one since the
circuit operated at lower control voltages. It can also be seen
from Fig. 5 that the circuit tuning is somehow stepwise and not
continuous as shown by the simulations. One possible reason
for this is that the buffer in the circuit is in fact part of the oscillator core and so the core is not separated from the load circuitry. This means that all kinds of noise sources, for example
transmission line reflections coming from the connections to the
measurement units, can be seen at the base of the lower cross
transistors, thus modulating the voltage over the timing capacitor. Additional buffers would help alleviate this problem. Another possible reason for the stepwise tuning is that either the
control circuit or the control principle does not function well as
in simulations.
The phase-noise performance was measured using a downconverter and phase-noise measurement system. The simulator
prediction was accurate close to the center frequency, e.g., at
1.4-V control voltage. In the ends of the curve, the performance
144
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 1, JANUARY 2001
Fig. 4. Simplified schematic of the implemented and designed double cross-coupled multivibrator VCO.
Fig. 5. Tuning curves.
was opposite to what was simulated. The difference in the low
end can be explained by problems with the control of the VCO
mentioned above. However, the measured output power at fundamental frequency was behaving as in the simulations, i.e., the
power decreased as a function of frequency. Therefore, there is
no simple explanation for failure of the simulator to predict the
phase-noise performance of this circuit, but it sets a need for
further investigations.
Fig. 6. The measured spectrum at 5.17 GHz. The second and third harmonics
are 18.4 and 20.7 dB below the carrier frequency, respectively.
V. COMPARISON
Table II presents some published relaxation oscillators
[6]–[8] and two LC-resonator-based VCOs [9], [10]. In the
fourth column of the table, the reported phase noise is given
and in the next column, the reported phase noise is normalized
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 1, JANUARY 2001
145
TABLE II
COMPARISON TABLE
[11]. In the class of relaxation oscillators, the phase-noise
performance reported here is the best with low power consumption. Compared with the LC-tank-based VCOs, which
have center frequencies comparable to our design and a broad
tuning range, the VCO designed in this work has the lowest
phase noise and largest tuning range with only slightly higher
current consumption.
VI. CONCLUSION
Fig. 7.
I
Measured phase noise at a control voltage of 2.5 V, V
= 14:6 mA.
= 2 5 V, and
:
A 4.3-GHz double cross-coupled VCO with 15-mA power
consumption from a 2.5 -V power supply and 2-GHz tuning
range was implemented in ST Microelectronic’s 0.35- m
BiCMOS process. The measured phase noise is 113 dBc/Hz
at 600-kHz offset from the carrier, which shows that the double
cross-coupled architecture can achieve low phase noise and a
large tuning range when compared to other published VCOs.
A new way to control a double cross-coupled oscillator was
described and the effect of using an inductive load in the
collectors was analyzed to show its advantages.
REFERENCES
Fig. 8. Chip micrograph. The size is 1:25
2 121 0 mm .
:
to the frequency offset of 600 kHz. We have assumed that the
phase noise decreases with a constant slope of 20 dB/decade
[1] N. Tchamov, A. Popov, and P. Jarske, “VCO with double cross-coupled
high-speed and low-power multivibrator architecture,” in Proc. ISCAS,
vol. 1, 1996, pp. 45–48.
[2] N. Tchamov and P. Jarske, “Emitter-coupled multivibrator circuit,” U.S.
Patent 5 825 256, Oct. 20, 1998.
[3] K. M. Sharaf and M. I. Elmasry, “Active-pull-down nonthreshold logic
BiCMOS circuits for high-speed low-power applications,” IEEE J.
Solid-State Circuits, vol. 30, pp. 691–695, June 1995.
[4] T. H. Lee, The Design of CMOS Radio-Frequency Integrated Circuits. Cambridge, U.K.: Cambridge Univ., 1998, ch. 8.
[5] M. H. Wakayama and A. A. Abidi, “A 30-MHz low-jitter high-linearity
CMOS voltage-controlled oscillator,” IEEE J. Solid-State Circuits, vol.
SC-22, pp. 1074–1081, Dec. 1987.
[6] T. Sowlati and M. H. Shakiba, “Automatic swing control in relaxation
oscillators,” IEEE J. Solid-State Circuits, vol. 33, pp. 1979–1986, Dec.
1998.
[7] M. Soyuer and J. D. Warnock, “Multigigahertz voltage-controlled oscillators in advanced silicon bipolar technology,” IEEE J. Solid-State Circuits, vol. 27, pp. 668–670, Apr. 1992.
146
[8] B. Razavi, “A study of phase noise in CMOS oscillators,” IEEE J. SolidState Circuits, vol. 31, pp. 331–343, Mar. 1996.
[9] T.-P. Liu, “A 6.5-GHz monolithic CMOS voltage-controlled oscillator,”
in Proc. ISSCC, 1999, pp. 404–405.
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 1, JANUARY 2001
[10] C. Lam and B. Razavi, “A 2.6-GHz/5.2-GHz CMOS voltage-controlled
oscillator,” in Proc. ISSCC, 1999, pp. 402–403.
[11] B. D. Leeson, “A simple model of feedback oscillator noise spectrum,”
Proc. IEEE, vol. 54, pp. 329–330, Feb. 1966.