Gigahertz CMOS Monolithic Frequency Synthesizer
Rong-Jyi Yang
Department of Electrical Engineering
and Graduate Institute of Electronics
Engineering
National Taiwan University
Taipei, Taiwan 10617, R.O.C
886-2-23635251ext328
r0943049@ee.ntu.edu.tw
Ming-Zhe Liu
Department of Electrical Engineering
and Graduate Institute of Electronics
Engineering
National Taiwan University
Taipei, Taiwan 10617, R.O.C
886-2-23635251ext328
r9921003@ee.ntu.edu.tw
Shen-Iuan Liu
Department of Electrical Engineering
and Graduate Institute of Electronics
Engineering
National Taiwan University
Taipei, Taiwan 10617, R.O.C
886-2-23635251ext239
lsi@cc.ee.ntu.edu.tw
Abstract—In this paper, a gigahertz monolithic
frequency synthesizer was realized in a standard 0.25-
µ
µµ
µ
m
CMOS technology. It consumes 55 mW from a single 2.5-
V power supply, and is fully functional over the complete
frequency range of the interest. The output signal
spectrum of 2.25 GHz with the phase noise –108 dBc/Hz
at 0.5-MHz offset. Its die size is 0.9
✕
✕✕
✕
0.9 mm
2
.
. IntroductionⅠ
ⅠⅠ
Ⅰ
Today’s high-performance RF frequency synthesizers
are often required: (1) to operate over a wide frequency range
which can cover the required range; (2) to have smooth
transition among channel intervals; (3) to have small phase
jitter and frequency variation; and (4) to have an integrated
loop filter on the chip. Up to now, the frequency synthesizer
still is a major concern for monolithic integration.
In general, the PLL as shown in Figure 1 contains five
major building blocks, such as a VCO, a high frequency
divider (also called prescaler), a phase detector, a charge
pump and a low-pass filter. By varying the divide ratio of the
divider, the PLL can synthesize a new frequency based upon
the reference input while retaining the stability, accuracy, and
spectral purity of the original reference. Of course, the
addition of the fractional-N division technique can achieve
the narrow channel frequency spacing resolution. As shown
in Figure 1, the digital
Σ∆
-modulator generates a stream of
integers M(t) which can interpolate a fraction number
corresponding to the control input [1-3].
This divided signal is compared to the reference signal
in the phase detector, which gives output signals, U and D,
equal to the phase deference between its two inputs. Through
the charge pump circuit, the signals are low-pass filtered by
the loop filter, and are converted to be the control input to the
VCO. Under conditions of lock, the two inputs of the phase
detector have a constant phase relationship and thus equal
frequency. The output frequency therefore is (N+M)
⋅
f
REF.
Changing the frequency is done by changing the
divider modulus M(t), which result in an adjustable change of
the VCO control voltage as the loop acquires its steady-state
operating.
In the design of a PLL frequency synthesizer, the
specification of every component must be chosen with
extreme care. The VCO must be of high quality, the
frequency divider must have high speed, the phase detector
must be accurate, and the filter must provide an optimum
compromise in switching speed, frequency settling, output
noise, etc…
. Circuit DescriptionⅡ
ⅡⅡ
Ⅱ
2.1 Phase frequency detector
A common drawback for some phase detectors is a
dead zone in the phase characteristic at the equilibrium point.
The dead zone generates phase jitter since the control system
does not change the control voltage when the phase error is
within the dead zone [4].
A dynamic phase detector as shown in Figure 2(a) is
proposed in [5]. The proposed dynamic CMOS PFD is shown
in Figure 2(b), where the two PMOSFETs in the first stage
and the two NMOSFETs in the last stage have been
exchanged. Thereby the data line transistor is moved to the
middle, while the clock line transistor is moved to the power
rails.
The phase characteristics of the conventional PFDs and
dynamic PFDs are depicted in Figure 3. Unlike the
conventional PFD, there is less dead zone in the
characteristics of the dynamic PFDs. A magnification of the
characteristics at a zero phase is shown in Figure 3. All data
are based on simulation with 0.25-
µ
m CMOS model when
supply voltage of 2.5V and operating frequency of 10MHz.
Furthermore, the dynamic CMOS PFD has a better
property in dead zone effect than the conventional PFD. It
can be explained in the following. When the PLL is near
locked, if the signal R lags the signal V, there is a very small
output signal to denote their phase difference. Just after the
output of D is active, the input signal R changes to low and
make MN
1
off and MP
3
on as shown in Figure 2.
Simultaneously, the voltages of C
1A
, C
2A
and C
1B
, C
2B
are
equal to supply value. Then the reset path will reactivate
through the NAND gate since U has changed to high; it can
turn on MP
1A
and MP
1B
, then indirectly turn on MN
3A
and
MN
3B
. However, when MN
3A
is turned on, it should
discharge the charge stored on C
2A
and C
1A
since MP
3A
had
been on, resulting in the reset speed slow. In the right hand of