A 39 GHz-80 GHz millimeter-wave frequency doubler with low
power consumption in 65nm CMOS tehnology
Qian Chen, Fazhi An, Guangyao Zhou, Shunli Ma, Fan Ye, Junyan Ren*
State Key Laboratory of ASIC and System, Fudan University, Shanghai 200433, China
* Email: jyren@fudan.edu.cn
Abstract
A wideband injection locking frequency doubler is
proposed for millimeter wave frequency generation in
CMOS. The circuit consists of two push-push pairs along
with a cross-coupled oscillator which share a current
source. The injection locking frequency doubler can
generate frequencies from 39 GHz to 80.6 GHz,
achieving a wide injection locking range of 69.5 % . The
conversion gain of the doubler is more than -15dB in the
whole locking range when driving a 30 fF capacitor.
Designed in TSMC 65nm CMOS, the circuit has an
active area of 340×180 um
2
. The injection locking
frequency doubler consumes 11.8 mW dc power from
1.2V supply, while the output buffer consumes 5.96
mW .
1. Introduction
During the past few years, millimeter-wave (mmW)
CMOS applications are flourishing due to the increasing
speed of CMOS devices in ultra-scaled nodes. Generally
speaking, there are two ways to generate local oscillator
(LO) signals in the millimeter-wave band. The first
method is generating LO signals directly using
fundamental frequency oscillators [1-3]. The second one
is using lower frequency oscillators cascaded with
frequency multipliers to generate the desired frequencies
[4-7]. A major obstacle for the first method is the poor
performance (poor spectral purity, limited tuning range
and large power consumption) of voltage-controlled
oscillators (VCOs) because of the degradation of key
passive components. The alternative method relies on
frequency multiplier circuits seem to be more attractive,
since it is much easier to design high-performance VCOs
at low frequency. The VCOs running at a fraction of the
output frequency reduce the power consumption, extend
the tuning range by the multiplication factor with the
reduced sensitivity to parasitic and process variations.
Moreover, high-frequency dividers which consume large
power are abandoned when using a low-frequency VCO.
There are several ways to realize the frequency
multiplication. Exploiting the nonlinearity of active
devices such as FETs, diodes, varactors can generate
higher harmonics of the input signal. A resonant load is
then used to select the desired harmonic and suppress the
fundamental. However, this method suffers from
low-conversion gain or even loss when design in CMOS.
Another approach is to mix fundamental signal with
itself or its harmonics and up-convert to a higher order
harmonic. But the large DC offset appearing at the output
saturates the mixer and limits the conversion gain.
Sub-harmonic injection locking of an oscillator can also
achieve frequency multiplication. It is widely adopted for
mmW frequency multiplication for its relative high
conversion gain and smaller input capacitance loaded to
the pre-stage circuit. Recently, several CMOS
injection-locked multipliers have been introduced [6-9].
But the performances of frequency multipliers based on
injection-locked oscillator topologies may suffer from
process variations and limited tuning range.
In this paper, a new topology of injection locking
CMOS frequency doubler based on push-push FET is
introduced and analyzed. The proposed circuit leads to a
low power dissipation and provides a differential output
over a broad frequency range. This novel frequency
doubler composed of two push-push pairs and a
cross-coupled oscillator which share a tail current source
is implemented in TSMC 65nm CMOS technology. From
the post-simulation results, the proposed injection
locking frequency doubler shows a very wide 69.5%
locking range with 17.7 mW power, including the output
buffer.
2. Circuit Design
2.1 Injection Locking Oscillators
Synchronization of oscillators by injecting an
appropriate signal has been studied extensively. Once the
injection locking oscillator becomes locked, it behaves
like a first order phase-locked-loop (PLL). But if the
injected signal lies out of the locking range (not very far),
the oscillator may become injection pulled. Terms like
“injection locking” or “injection pulling” may remind
readers versed in the literature the Adler’s 1946 analysis
of injection locking. Adler wrote his celebrated equation
for small injected signal to find the frequency range over
which the injected oscillator becomes locked, that is the
locking range. A. Mirzaei has introduced a new model
for injection locking and expanded the Adler’s equation