CLOCK DIVIDER DESIGN
3.1 Introduction
In a PLL synthesizer, of all the PLL components, circuit design of the VCO and
the frequency divider is very critical. The reason partly is these components operate at
the highest frequencies within the PLL and also consume most of the power as
compared to the other components. As part of this thesis, a digital frequency divider
based on True Single Phase Clocking (TSPC) has been designed. It was more an attempt
to learn the different aspects of digital design at high frequencies. This work is mostly
based on the concepts provided in [20]. A divider with dual modulus of 256/288 has
been designed to divide the band of frequencies specified for the 802.11a WLAN
standard which is from 5.14 GHz to 5.72 GHz to a frequency of 20 MHz which forms
the reference frequency to the frequency synthesizer. This work does not include a
swallow counter needed for dynamic programmability
3.2 Circuit Configuration and Working
The frequency divider designed here consists of three main blocks, a divide-by2,
a divide-by-8/9 and a divide-by-16 block, as shown in figure 3.1. The divide-by-8/9
prescaler whose structure is shown in figure 3.2 proves to be a rather critical block. The
input frequency is first halved by a divide-by-2 circuit so that the speed constraints of
this prescaler can be relaxed. The prescaler is usually followed by a program counter
and a swallow counter which allows to dynamically control the overall division ratio
between 256 and 288 and to sweep the output frequency between 5.14 and 5.74 GHz in
1
20-MHz steps. Hence, the synthesized signal is suitable to operate both in the
HiperLAN II bands [18] and in the 802.11a lower and middle bands [19], assuming
direct-conversion or low-IF receiver architectures.
Figure 3.1 Block diagram of frequency divider
Generally, a TSPC logic flip-flop (FF) is used to synchronize the programmable
divider output to the first divide-by-2 circuit output. This FF can cancel out the
accumulated time jitter due to the divide-by-8/9 prescaler and the divide-by-16 program
counter [12]. Two or three cascaded CMOS logic inverters are needed after the first
divide-by-2 to drive the divide-by-8/9 prescaler and the synchronizing FF.
Figure 3.2 Divide-by-8/9 prescaler topology
3.2.1 Divide-by-2 Circuit
The divide-by-2 circuit realized in the TSPC logic is shown in figure 3.3. The
salient feature of the TSPC clocking technique is that there is only one clock signal
needed to trigger the flip-flops and no extra clock phase is required whatsoever. This
technique is mainly used in dynamic CMOS circuits and helps to simplify the design.
2
V
Figure 3.3 Circuit schematic of a TSPC divide-by-2
The circuit consists of three parts. The first part is a gated inverter that consists
of MP1, MP4 and MN1, which passes the divider output to the following stage when In
goes low. The second part is a latch stage that consists of MP2, MP3, MN2, MN3, MN4
and MN5. This circuit will be activated and store the output of the gated inverter when
In is high. The PMOS transistors MP1 and MP2 are used to pre-charge the internal
nodes to increase the speed of the circuit. The output of the flip-flop is directly
connected back to the D-input to obtain the divide-by-2 function because the TSPC
circuit can completely isolate the sense and latch stage at different phases of the clock
signal. The static power of the circuit is zero because no direct path from supply to
ground exists and it only consumes dynamic power. One of the advantages of the TSPC
3
divider is its simplicity. The circuit consists of only nine transistors. In some cases
where the inverted output is required an inverter is included which adds two more
transistors. The circuit requires an input signal of large amplitude, and it is very
sensitive to the slope of the signal. Therefore, a high-frequency input buffer may
sometimes be needed in front of the divider to drive it. The speed of the circuit greatly
depends on the voltage supply. The circuit will be slow if a low-voltage supply is used.
In order to operate at higher frequency, larger sizes of the transistors are needed to
increase the gm and thus make it operate faster. However, increasing the size also
increases loading for the previous stage and thus the trade-off should be considered
during design. Moreover, using larger transistor sizes will increase the degree of charge
leakage and charge sharing at the output nodes and thus will affect the minimum
operation frequency of the circuit.
3.2.2 Divide-by-8/9 Circuit
The divide-by-8/9 prescaler shown in figure 3.2 above, is comprised of a
divideby-2/3 synchronous divider and two asynchronous divide-by-2 circuits. While the
latter are implemented again in TSPC logic, the divide-by-2/3 circuit has been realized
using the extended true-single-phase-clock (E-TSPC) logic proposed in [21]. A block
diagram representation and the schematic of the divide-by-2/3 circuit are shown in
figure 3.4 and figure 3.5 respectively. Compared to the classical TSPC logic, the E-
TSPC avoids the stacked MOS structure that slows the switching speed, and all the
transistors are free from the body effect. For these reasons, E-TSPC logic allows higher
operating frequencies, although it features static power dissipation. However, this
4
causes only a small increase in power dissipation, since at the frequencies of interest the
dynamic power consumption is dominant over the standby current [21].
Figure 3.4 Block level representation of divide-by-2/3 circuit
Figure 3.5 Schematic of divide-by-2/3 circuit
This logic also allows embedding complicated logic functions within the latches. This
results in very compact circuits and in a reduced number of transistors. The concept is
applied in the design of the divide-by-2/3 circuit, where the AND and the OR gates are
realized by adding only one transistor each as evident from figure 3.4.
5
3.2.3 Divide-by-16 Circuit
The divide-by-16 program counter can be viewed as an extension of the
divideby-2 circuit. It is implemented by simply cascading four divide-by-2 circuits as
shown in figure 3.6. The output of this circuit is fed into the frequency synthesizer
where the PFD compares it with the reference frequency.
Figure 3.6 Block diagram of divide-by-16 circuit
Figure 3.3 Circuit schematic of divide-by-2
3.4 Simulation Results and Inferences
6
A frequency divider has been designed to operate at a frequency of 5 GHz. The
circuit was simulated using the Mentor Graphics Tool. The transient response of the
circuit to both sine wave and square wave input is shown in the plots that follow.
(a)
7
(b)
Figure 3.9 Transient response of divide-by-2 circuit to (a) sine wave input and (b)
square wave input
Figure 3.10 Schematic of divide-by-2 circuit
8
The divide-by-2/3 circuit has a control input MC. By applying the right voltage
at the control input the divider can be configured to work as either a divide-by-2 or
divide-by-3. If divide-by-2 operation is required MC should be set to a binary value of 1
(2.5 V) and if divide-by-3 operation is required MC should be set to a binary value of 0
(0 V). The waveforms are shown in figures 3.11 and 3.12.
(a)
9
(b)
Figure 3.11 Transient response of divide-by-2/3 circuit with control input MC = 0, to (a)
sine wave input and (b) square wave input
(a)
10
(b)
Figure 3.12 Transient response of divide-by-2/3 circuit with control input MC = 1, to (a)
sine wave input and (b) square wave input
11
Figure 3.13 Schematic of divide-by-2/3 circuit (refer figure 3.5)
12
(a)
(b)
Figure 3.14 Transient response of divide-by-8/9 circuit with control input MC = 1, to (a)
sine wave input and (b) square wave input
13
(a)
(b)
Figure 3.15 Transient response of divide-by-8/9 circuit with control input MC = 0, to (a)
sine wave input and (b) square wave input
14
Figure 3.16 Waveforms at intermediate nodes in circuit
Figure 3.17 Waveform of complete circuit
15
Layout
A full custom layout of the divider shown in Figures 3.18 and 3.19 was designed
to optimize for area and reduce interconnect parasitics. This helps to achieve greater
speed of operation.
Figure 3.18 Layout of divide-by-2 circuit
Figure 3.19 Layout of divide-by-2/3 circuit
16
The widths of the transistors used are shown in the table below. The length of all
the transistors is kept at the minimum the technology allows which is 0.25 µm.
Table 3.2 Transistor Widths of Divide-by-2 Circuit
Transistor Width (W) in µm
MP1 1.5
MP2 1.9
MP3 2.4
MP4 1.5
MN1 0.5
MN2 1
MN3 1.5
MN4 1.7
MN5 1.7
Table 3.3 Transistor Widths of Divide-by-2/3 Circuit
17
PMOS Transistor Width (W) in µm NMOS Transistor Width (W) in µm
Mp1 3 Mn1 0.5
Mp2 2 Mn2 1.3
Mp3 3 Mn3 2
Mp3 1.5 Mn4 2
Mp5 3 Mn5 0.5
Mp6 3 Mn6 0.5
Mp7 2 Mn7 1.3
Mp8 3 Mn8 3
18