Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02238093 1998-0~-20
DIRECT DIGITAL FREQUENCY SYNTHESIZER USING PULSE GAP SHIFTING
TECHNIQUE
S Field of the Invention
The present invention relates to digital frequency synth~?si7ers and is
particularly concerned with directly digitally synthesizing a desired frequency using
pulse gap ~hifting
LO Back~round to the Invention
Systems for transmitting digital signals often use a phase lock loop (PLL) in
clocking related functions, such as recovering timing information from a received
signal to provide a clock signal for system timing. In a PLL, a frequency synthesizer is
used to output a signal having a frequency which is dependent upon an error signal.
LS The error signal l~resent~ the difference between the phase of a reference signal
which is supplied to the PLL and an output signal.
Frequency synthesizers are known to be susceptible to extraneous conditions,
such as temperature variations or input noise. These conditions affect the frequency of
their output signals. When a frequency synthesizer is used in a PLL, these conditions
20 manifest themselves as jitter in the phase of the PLL output signal.
The effect of timing error in a digital signal is to reduce the duration in which a
zero or one state can be detected in a given time interval. The given time interval is
inversely proportional to the tr~n~mi~sion rate of the digital signal. As a result, an
CA 02238093 1998-0~-20
increase in the tr~n~mi~sion rate of the signal decreases the given time interval,
subsequently re4ui~ g a decrease in the allowable amount of timing error in the
slgnal.
With increasing tr~n~mi~sion rates, specifications for maxhllulll allowable
5 tirning error in digital signals are becoming increasingly clçm~n-lin~ System clock
~ign~lc, which are often provided by PLLs, must meet m~Yimum allowable timing
error specifications required by the tr~n~mi~sion rate of the signals. As these
specifications are reduced, the need for clock sources with improved accuracy and
controllability arises.
Summary of the Invention
An object of the present invention is to provide an improved direct digital
frequency synth~ei7er (DDFS).
In accordallce with an aspect of the present invention there is provided a direct
digital frequency synth~ .r for deriving a synth~ci7.ed clock signal from a reference
clock signal having a sllccçs~ n of cyclic pulses comprising: inputs for the reference
clock signal and a digital control word; an output for the synthesized clock signal; a
phase accllmnl~tor coupled to the inputs for the digital control word and the reference
20 clock signal and having an output fnr a phase shifter control signal, and means for
ge~ lillg an a;liv~Llg event in the phase shifter control signal in depen-l.once upon
the digital control word and a first number of pulses of the l~,felcllce clock signal; and
a phase shifter having inputs for the reference clock signal and for the phase shifter
CA 02238093 1998-0~-20
control signal, an output for an intermediate frequency signal which is coupled to the
synthesized clock signal, means for passing a second number of pulses of the reference
clock signal to the intermediate frequency signal and then inhibiting a pulse from
passing thereto, and means for passing, responsive to the activating event, a third
5 number of pulses of the reference clock signal to the intermediate frequency signal and
then inhibiting a pulse from passing thereto.
An advantage of this aspect of the invention is that it provides the ability to
adjust the frequency of the synthesized clock signal according to the value of the
10 digital control word, whereby the number of pulses of the reference clock signal which
are passed to the synth~si7e~ clock signal are dependent upon the activating event,
which is in turn dependent upon the value of the digital control word. Thus, the
frequency of the synth~si~e~l clock signal is dependent upon the value of the digital
control word and the frequency of the reference clock signal.
A further advantage of this aspect of the invention is that it provides a means to
adjust the phase of the synthesized clock signal by an amount equal to the period of the
reference clock signal divided by the sum of the second number of pulses plus one.
Conveniently, the digital control word is a signed digital control word capable
of having a positive or a negative polarity. Furthermore, the phase accllmlll~tor
comprises: outputs for a fo~ signal and a back signal; means for generating, in the
fol ~r~ signal, a first activating event in dependence upon the first number of pulses
CA 02238093 1998-0~-20
-4-
and for one polarity of the digital control word; means for passing, responsive to the
first activating event, the third number of pulses of the reference clock signal to the
intermediate frequency signal and then inhibiting a pulse from passing thereto; means
for generating, in the back signal, a second activating event in dependP.nce upon the
S first number of pulses and for the other polarity of the digital control word; and means
for passing, re~ol~ive to the second activating event, a fourth number of pulses in the
reference clock signal to the intermediate frequency signal and then inhibiting a pulse
from passing thereto.
Conveniently, the third number of pulses is less than the second number of
pulses and the fourth number of pulses is greater than the second number of pulses.
Advantageously, this provides a means to increase or decrease the frequency of the
synthesi7ed clock signal in depen-lPn~e upon the polarity of the digital control word.
Further, the phase of the synthesi7ed clock signal can be likewise advanced or delayed.
In accordance with another aspect of the present invention there is provided a
direct digital frequency synthesizer comprising inputs for a reference clock signal and
an N-bit signed control word, and an output for a synthesized clock signal; a phase
accumulator having inputs coupled to the N-bit signed control word and the reference
20 clock signal, and having outputs for a ro~ signal and a back signal; a phase shifter
having inputs coupled to the reference clock signal and the fol~va~.l and back signals
with an output for an intermediate frequency signal; and a divider having an input
CA 02238093 1998-0~-20
coupled to the intermediate frequency signal with an output coupled to the synthesized
clock sign.al.
Conveniently, the phase accllm~ tQr comprises an N-bit signed adder having
5 inputs coupled to the N-bit signed control word and another N-bit signed word and
outputs for an overflow signal, an underflow signal and an N-bit signed result word,
whereby an event on the overflow signal or underflow signals is responsive to the
result of adding the N-bit signed control word and the another N-bit signed word; a
divider having an input coupled to the reference clock signal and an output for a
10 divided reference clock signal; and a latch having a clock input coupled to the divided
reference clock signal, and inputs for the overflow signal, the underflow signal and the
N-bit signed result word with their l~e~ tive latched outputs provided on outputs for
the forwar.d signal, the back signal and for the another N-bit signed word.
Conveniently, the phase shifter comprises a counting means for counting a
number of pulses of the reference clock signal in dependence upon a controlling
means; a pulse inhibiting means for inhibiting the next pulse after the number of pulses
has been counted; an output means for oul~ulLh~g all but inhibited pulses of the
reference clock signal to the phase shifter output; a controlling means for controlling
20 the nurnber of pulses to be counted in dependence upon the forward and back signals,
whereby the number of pulses to be counted after the next inhibited pulse is decreased
by one when the ro~ signal has the event, and after counting the number of pulses
the next pulse is inhibited, after which, the number of pulses to be counted is increased
CA 02238093 1998-0~-20
- 6 -
by one, and whereby the number of pulses to be counted after the next inhibited pulse
is increased by one when the back signal has the event, and after counting the number
of pulses the next pulse is inhibited, after which, the nDber of pulses to be counted is
decreased by one. Advantageously, this provides a means to increase or decrease the
5 frequency of the synthesized clock signal in dependence upon the polarity and value of
the digital control word. Further, the phase of the synthesi~ecl clock signal can be
likewise advanced or delayed by an amount equal to the period of the reference clock
signal divided by the sum of the number of pulses to be counted plus one.
Another advantage of the DDFS is that it provides linear control of the phase
and frequency of the synth~ ed clock signal irrespective of extraneous conditions,
such as tc,,,lpc;l~lule~ that fall within its normal operating range. Increments of
controlled variation in the phase of the synthesi7ed clock signal are a fraction of the
period of the reference clock signal XCLK. Further, since the DDFS responds to a
15 digital control word it lends itself more easily to techniques for colllpe~ g for other
components. For example, a look-up table could be used to compensate for drift in the
left;lel~e clock signal due to temperature, whereby a compensation factor from the
table is added to the control word before the control word is applied to the DDFS.
Finally, since the DDFS is digital it allows for larger scale integration of digital PLL
20 circuits, possibly with an entire digital PLL being on one integrated circuit, thus
leading to reduced m~nuf~ctllring costs and greater reliability of the digital PLL.
CA 02238093 1998-OS-20
Brief Description of the Drawings
The present invention will be further understood from the following detailed
description with reference to the drawings in which:
Fig. 1 illustrates in a block diagram a direct digital frequency synthP-ei~er in
5 accordance with an embodiment of the present invention;
Fig. 2a illustrates in a block diagram the DDFS of Fig. 1 in greater detail; and
Fig. 2b is a state diagram of the modes of operation of the DDFS of Fig. 2a.
In the drawings, like rereLe.lce characters in the different figures l~les~lll the
sarne features.
Detailed Description
Referring to Fig. 1 there is illustrated in a block diagram a direct digital
frequency synthesizer in accordance with an embodiment of t_e present invention.
The DDFS 10 includes an input 12 for a reference clock signal XCLK, an input 14 for
15 a control word Y and an output 16 for a synthesized clock signal SCLK.
In operation the DDFS 10 takes the reference clock signal XCLK, which has a
frequency fx of 84.24 MHz, and the control word Y, which is a 21-bit signed integer,
as inputs~ and outputs the synthesized clock signal SCLK, which has a frequency fs of
2C1 12.96 MHz. The frequency fs is adjusted by writing the control word Y, of an
~propliate value, to the DDFS 10.
CA 02238093 1998-0~-20
Referring to Fig. 2a, there is illustrated in a block diagrarn the direct digital
frequency synthesizer of Fig. 1 in greater detail. The direct digital frequency
synthesizer 10 includes a phase accumulator 20 and a phase shifter 22. The phase
accllm~ ror 20 includes an N-bit signed adder 24, a divide by M divider 26 and a latch
5 28. The N-bit signed adder 24 has inputs X and Y and an output Z, as well as outputs
for overflow and underflow signals. The divide by M divider 26 has an input for the
reference clock signal XCLK and an output for a divided clock signal. The divided
clock signal is applied to the clock input of the latch 28. The latch 28 provides a
latched output of the adder output Z to the input X of the adder 24 and provides a
10 latched output of the overflow and underflow signals to the phase shifter 22 as forward
and back signals, respectively. The phase shifter 22 includes a pulse inhibit logic 30
that has inputs for receiving the foJ ~vard signal, the back signal and the reference clock
signal XCLK, as well as an output for an interrnediate frequency signal. The
interrnediate frequency signal, having a frequency fO of 77.76 MHz, is divided by six
15 by divider 34 to which results in the synthesi7ed clock signal SCLK having a
frequency fs of 12.96 MHz.
With reference to Fig. 2a, in operation the result of the addition of the N-bit
signed words provided to the X and Y inputs of the N-bit signed adder 24 is provided
20 at its Z output as a result word Z. When the result of an addition operation is greater
than 2N-l the overflow signal wil] transition to its high voltage level ca~l~ing it to have
a rising edge, and the least significant N bits of the result word Z will be provided at
CA 02238093 1998-0~-20
the output Z. When an addition operation results in a number less than _2N 1, which
can occur when the control word Y is negative, the underflow signal will transition to
its high level call~ing it to have a rising edge and the output Z will provide the least
significant N bits of the result word Z.
The divided clock signal, which is applied to the clock input of the latch 28,
controls l~tchin~ of the signals applied to the inputs of the latch 28. The divided clock
signal has a frequency fa equal to the frequency fx Of the reference clock signal XCLK
divided by M. As a result, the overflow and underflow signals and the result word Z
10 are latched by the latch 28 once every M pulses of the reference clock signal XCLK.
The latched overflow and underflow signals are the forward and back signals
provided at outputs of the latch 28. They are applied to inputs of the pulse inhibit
logic 30 and are used to control its mode of operation. The pulse inhibit logic 30 has
15 three modes of operation as lep-esented by 32 in FIG. 2b. The modes are neutral, up
and down. In these modes of operation an integer variable I represents a number of
pulses. The difference between the modes of operation is the number of pulses that are
applied to the pulse inhibit logic 30 for every pulse that is inhibited. This number can
be I, I-l or I+l depending on the mode of operation. The operation of the pulse inhibit
2Ci logic 30 in each mode, and the effect of the fol~v~d and back signals on it, is
explained below with reference to FIG. 2b.
CA 02238093 1998-0~-20
- 10-
Neutral Mode: count to I-l and inhibit the next pulse. Thus, in the
neutral mode, I pulses are applied for each inhibited pulse. If there is a rising
edge on the rolw~d signal then go to the up mode after the next inhibited pulse
(A). If there is a rising edge on the back signal then go to the down mode afterthe next inhibit pulse (D).
Up Mode: count to I-2 and inhibit the next pulse, then return to the
neutral mode (B). Thus, in the up mode, I-l pulses are applied for each
inhibited pulse.
Down Mode: count to I and inhibit the next pulse, then return to the
neutral mode (C). Thus, in the down mode, I+l pulses are applied for each
inhibited pulse.
The variable I is chosen to achieve a desired value of the frequency fO of the
interrnediate frequency signal. This value is related to the frequency fx of the reference
15 clock signal XCLK as follows; in lhe absence of a rising edge on either the forward or
back signals the pulse inhibit logic 30 will remain in the neutral mode, inhibiting a
pulse every I pulses. Therefore, in the neutral mode, for every I pulses applied to the
pulse inhibit logic 30, by the reference clock signal XCLK, I-l pulses will be output to
the intermediate frequency signal. It follows that the frequency fO when the pulse
20 inhibit logic 30 is in neutral mode (denoted by fO n ) is related to the frequency fx and
the number of pulses I by the following equation:
CA 02238093 1998-0~-20
EQ. 1 for~ fl( I )
In this embodiment I=13 and fX=84.24 MHz and therefore fO, n=77.76 MHz.
5 The frequency fs Of the syntllç~i~ecl clock signal SCLK is equal to the frequency fO
divided by six, by the divider 34. I'he frequency fO n is 77.76 MHz and therefore the
frequency fs is 12.96 MHz when the pulse inhibit logic 30 is in the neutral mode.
Retllrning now to the effect of the fo~ d and back signals on the pulse inhibit
10 logic 30, their effect is to shift the timing of the inhibited pulse by advancing or
delaying its cyclic timing by one pulse, depending on whether the rol~v~d or back
signal has a rising edge. The timing of the inhibited pulse occurs one pulse period
earlier as a result of a rising edge in the forward signal, and one pulse later as a result
of a rising edge in back signal.
Returning to the phase accllm~ tor 20, the result word Z is latched by latch 28
once in every M pulses of the reference clock signal XCLK. A new value of the result
word Z is present at the output Z of the adder 24 shortly thereafter. The time delay
after which the new value present at the output Z is valid is dependent on the adder 24.
20 This delay should be less than the period of the frequency fa Of the divided clock
CA 02238093 1998-0~-20
- 12-
signal. In this way, the result word Z will be valid at the latch 28 input when the
divided clock signal activates the latch 28 to latch its inputs.
The process of adding control word Y to the result word Z, which is applied to
S the X input, continues nor~nally until the result is greater than the magnitude of the
adder, which is 2N- 1 . When this condition occurs the voltage level of the overflow
signal transitions to high level, after which it is latched by the latch 28 on the next
rising edge of the divided clock signal. Thus, the overflow signal voltage level
transitions to a high level once in every 2N l/y pulses of the divided clock signal,
10 which in turn pulses once every M pulses of the reference clock signal XCLK. This
relationship is expressed in EQ. 2, below:
EQ. 2 A 2 M
1~
where A r~lesents the nurnber of pulses of the reference clock signal XCLK
15 for each rising edge of the overflow, or underflow, signal.
For a negative value of the control word Y, the operation of the adder 24 is the
same as clescribed above with one difference. The difference is that the underflow
signal transitions to a high voltage level when the result of an addition operation is less
2CI than minus 2N 1. The number of pulses of the reference clock signal XCLK for every
rising edge of the underflow signal is also expressed by EQ. 2.
CA 02238093 1998-0~-20
As mentioned earlier, the effect of the forward and back signals is to shift the
time at which a pulse is inhibited by one period of the reference clock signal XCLK.
However, the pulse inhibit logic 30 is only responsive to the forward and back signals
5 when it is in the neutral mode, as described earlier. In this mode, a rising edge on the
fo~ d signal will cause the pulse inhibit logic 3() to go into the up mode after the
next inhibited pulse, and return to the neutral mode after a pulse is inhibited from the
up mode. Therefore, the pulse inhibit logic 30 can be in the up mode half of the time at
most. This limits the range of frequencies that the synth~si7ed clock signal SCLK can
10 be set to for a given value of I. Similarly, the pulse inhibit logic can only be in the
down mode half of the time at most.
In the up mode, a pulse is inhibited by the pulse inhibit logic 30 for every I-1
pulses of the reference clock signa] XCLK. As a result, there are I-2 pulses in the
intermediate frequency signal, for e very I-1 pulses of the reference clock signal
1 5 XCLK.
The frequency fO is det~nnined by averaging the number of pulses per second
in the intermediate frequency signal over a relatively long period of time. The
relationship between the frequency fO and the frequency fx is detçrrnined by the
20 percentage of time that the pulse inhibit logic 30 spends in the up mode in relation to
the amount of time it spends in the neutral mode. This percentage is dependent on the
control word Y, the divisor M of the divider 26, and the magnitude of the N-bit signed
CA 02238093 1998-0~-20
- 14-
adder 24. The number of pulses of the reference clock signal XCLK that occur for each
rising edge of the forward signal is given by the result, A, of EQ. 2. This rising edge
causes the pulse inhibit logic 30 to go into up mode once for every A pulses of the
reference clock signal XCLK, providing this is not more than half of the time. The
S rem~in~ler of the time the pulse inhibit logic 30 will be in the neutral mode.
Accordingly, the number of pulses of the reference clock signal XCLK that will be
applied to the pulse inhibit logic 30 for each rising edge of the forward signal equals
the number applied in the neutral mode plus the number applied in the up mode. This
relationship is expressed by EQ. 3, below:
EQ. 3 Pulses appliedfor each rising edge of the forward signal = bI + (I -1)
where b denotes the number of times that the pulse inhibit logic 30 cycles
through the neutral mode (with I pulses applied each time) and I-l represents the
15 number of pulses applied during one cycle through the up mode.
It should be evident that the nurnber of pulses of the reference clock signal
XCLK for each rising edge of the overflow (or underflow) signal given by EQ. 2
equals the same number of pulses applied in Up and Neutral modes for each rising
edge of the fol vv~d signal given by EQ. 3. This should be evident because the forward
20 signal is the overflow signal after it has been latched by the latch 28. Likewise with the
underflow and back signals.
CA 02238093 1998-0~-20
- 15 -
Equating EQ. 2 and EQ. 3 allows one to solve for b. The result is given in EQ.
4, below. This equation will be useful later in detçnnining a range of useful values of
the control word Y given the values of N, M and I.
5 EQ. 4 b= 2 M ~
The frequency fO of the intermediate frequency signal output by the pulse
inhibit logic 30 equals the number of pulses ll~slllilled divided by the number applied
times the number of pulses applied per second. This relationship is expressed in EQ. 5,
10 below.
Pulses 'rransmitted
Q f o f ~ Pulses Applied
The above equation will be considered in f~ther detail for a positive value of
15 the control word Y, that is, when the pulse inhibit logic 30 operates in the up and
neutral modes. According to the previous description of the operating modes of the
pulses inhibit logic 30, in the neutral mode I-l pulses are transmitted for every I pulses
that are applied. In the up mode, I-2 pulses are transmitted for every I-l pulses that are
applied. The pulse inhibit logic 30 cycles through the neutral mode b times for every
20 one cycle through the up mode. Given the above information, EQ. S can be ~l,lGssed
in terms of b and I as shown below, where fO (n U) r~l~selll~ the frequency fO of the
CA 02238093 1998-05-20
-16-
intermediate frequency signal when the pulse inhibit logic 30 operates in the up and
neutral modes.
EQ 6 f = f (I -2)+b(I -1)
' o,(n,u) r (I - 1) + b(I)
s
The change in the frequency fO which results from the rising edges of the
fol ~v~d signal is given by subtracting EQ. 1 from EQ. 6. The result is expressed in
EQ. 7, below.
10 EQ. 7 ~ f f ~ --1
Recognizing that the denominator of the term in the brackets is equal to the
right hand side of EQ. 3 and rec~llin~ that EQ. 3 equals the right hand side of EQ. 2, a
substitution can be performed to obtain an equation for the change in the frequency fO
that results from a rising edge ofthe folvv~d signal. This change in the frequency fO,
15 in terms of the Y, I, N and M, is given by EQ. 8, below.
fr Y
EQ. 8 ~fO(nu) I 2N-IM
The change in phase of the frequency fO for each rising edge of the fo. ~d
signal is equal to the change in frequency times the amount of time over which the
CA 02238093 1998-0~-20
-17-
change occurs. This amount of time is the number of pulses for each rising edge of the
forward signal, given by EQ. 3, times the period of the reference clock signal XCLK.
The result is that the change in phase is inversely proportional to the number of pulses
I applied in the neutral mode. This is shown by EQ. 9, below.
Q ~ (~)o,(n,u) =--I
Where the change in phase of the frequency fO given by EQ. 9 is in unit
intervals of the frequency fO.
The DDFS 10 synthesi7.es frequency offsets as a series of phase steps. The
10 phase steps are a fraction of the period of the reference clock signal XCLK, which is
the highest rate signal input to the DDFS 10. Using EQ. 8 and EQ. 1, an ~pression, in
terms of Y, I, N and M, for the frequency fO when the pulse inhibit logic operates in
the up and neutral modes can be derived by recognizing that the resultant frequency
equals the initial frequency plus the change in frequency. This ~"ession is given by
1 '; EQ. 10, below.
EQ. 10 f o,(n,u) f x[ I 2N-' MI~
There are three different cases for a positive value of the control word Y that
20 will be considered next. The first case is when the control word Y equals zero. When
the control word Y equals zero the result of the adder 24 does not increase or decrease
CA 02238093 1998-0~-20
-18-
and hence a rising edge never occurs on the overflow or underflow signals. In absence
of a rising edge on these signals the pulse inhibit logic 30 will remain in the neutral
mode. In this mode the frequency i-'o is given by EQ. 1. The second case is when the
control word Y has a value such that the variable b, in EQ. 4, is less than or equal to
'i one. Such a condition indicates that the pulse inhibit logic 30 would be in the neutral
mode less than half of the time. However, as described earlier, the pulse inhibit logic
30 can not be in the neutral mode less than half of the time. Therefore, the minim~
value for b in operation is one. For the case when EQ. 4 results in a value of b<l, the
frequency fO is given by EQ. 6 with b equal to one, the solution of which provides a
10 lower frequency limit of the frequency fO.
Values of the control word Y that result in a solution for EQ. 4 of b<l will, in
operation, set the frequency fO to the lower limit described above. The values of the
control word Y that will cause this condition are supplied by EQ. 11, below.
N-l M
l'i EQ. 11 2I-l
l'he third case, for a positive value of the control word Y, is when b, given by
EQ. 4, is greater than one. In this case, substituting b into EQ. 6 provides the frequency
21~ fO of the intermediate frequency signal, alternatively, EQ. 10 can be used.
Furthermore, solving EQ. 4 for b > 1 produces a requirement on the divisor M. This
CA 02238093 1998-0~-20
- 19-
requirement is expressed by EQ. 12, below. For values of I and N, EQ. 12 specifies
the relationship between the divisor M and the control word Y that must be m~int~ined
to keep the frequency fO above its lower limit.
5 EQ. 12 2N-I
In order to allow the greatest amount of control over the frequency fO, the
largest range of values of the control word Y that have an effect on the frequency fO
should be used. This range of effective values will be such that only one value of the
lC control word Y will cause the frequency fO to be set to its lower limit. All other values
of the control word Y will result iIl the frequency fO to be set above its lower limit. In
terms of the variable b, this range of values should provide the largest range over
which the control word Y can be varied while keeping b > 1.
1 ~i The control word Y can take on values from zero to 2N 1. By substituting the
largest value of control word Y into EQ. 12 a relationship between the divisor M of the
divider 26 and the variable I is obtained, as expressed in EQ. 13, below. ~ t~ g
this relationship between M and I will provide the largest range of effective values of
the control word Y.
2()
CA 02238093 1998-0~-20
- 20-
EQ. 13 M 2 2I - 1
In this embodiment I=13 and therefore EQ. 13 indicates that the divisor M
should be greater than or equal to 25 when the control word Y is a positive number.
Turning now to the case of a negative value of the control word Y, a negative
value will cause the pulse inhibit logic 30 to operate in the down and neutral modes. In
the down mode, a pulse is inhibited for every I+l pulses of the reference clock signal
XCLK that are applied to the pulse inhibit logic 30. The neutral mode operates as
10 described earlier, with one pulse being inhibited for every I pulses applied. As a result,
the number of pulses of the reference clock signal XCLK for each rising edge of the
back signal equals the number of pulses applied in the neutral mode plus the number
applied in the down mode. This relationship is expressed in EQ. 14, below:
15 EQ. 14 Pulses appliedfor each rising edge of the back signal = cI + (I + 1)
where c denotes the number of times that the pulse inhibit logic 30 cycles
through the neutral mode (with I pulses applied each time) and I+l represents the
number of pulses applied during one cycle through the down mode.
2~)
Recalling that the number of pulses applied for each rising edge of the underflow
signal equals the number applied for each rising edge of the back signal, EQ. 2 can be
equated to EQ. 14 to solve for c resulting in EQ. 15, below.
CA 02238093 1998-0~-20
-21-
N-IM (I+l)
EQ. 15
As stated earlier, the frequency fO equals the number of pulses per second
'; applied to the pulse inhibit logic 30, multiplied by the number of pulses tr~n.cmitted by
it for the number of pulses applied to it. For a negative value of the control word Y this
relationship is expressed by EQ. 16, below.
I +C(I -l)
EQ. 16 fo f~ (I+l)+c(I)
1)
In EQ. 16, the terms c(I) and c(I-l) represent the number pulses applied and
tr~n.cmilte~l in the neutral mode, respectively. The terms (I+l) and (I) l~lesell~ the
number of pulses applied and tla~ ed in the down mode, respectively.
IJsing EQ. 1 6, equations for the change in the frequency fO for each rising edge
of the back signal, and the frequency fO for a negative value of the control word Y can
be derived in terms of Y, I, M and N. These equations are the same as EQ. 8 and EQ.
10, respectively, derived for positive control word Y. Furthermore, an equation for the
change in phase of the frequency fO for each rising edge of the back signal can be
CA 02238093 1998-0~-20
derived. I he resultant equation is E~Q. 9, for a positive value of the control word Y,
times minus one to reflect an increase in ph~e for each rising edge of the back signal.
There are two c~es for a negative value of the control word Y that will be
5 considered next. The first is when the control word Y h~ a value such that the variable
c, given by EQ. 15, is less than or equal to one. Such a condition indicates that the
pulse inhibit logic 30 would be in lhe neutral mode less than half of the time.
However, ~ described earlier, the pulse inhibit logic 30 can not be in the neutral mode
less than half of the time. Therefore, the ",i~ ,lll, l value for c in operation is one. For
10 the case when EQ. 15 results in a value of c<l, the frequency fO is given by EQ. 16
with c equal to one, the solution of which provides an upper frequency limit of the
frequency fO.
Values of the control word Y that result in a solution for EQ. 15 of c<l will, in
15 operation, set the frequency fO to the upper limit described above. The values of the
control word Y that will cause this condition are supplied by EQ. 17, below.
N-IM
EQ. 17 IYl 2I + 1
The second c~e, for a negative value of the control word Y, is when EQ. 15
results in a value of c that is greater than one. Solving EQ. 15 for c > 1 yields a
CA 02238093 1998-OS-20
-23-
requirement on the divisor M that specifies the relationship between the divisor M and
the control word Y that must be m~int~ined to keep the frequency fO below its upper
limit. This requirement is expressed by EQ. 18, below.
S EQ.18 M (2I+?¦~
When values of M, N, I and Y result in EQ. 18 being hue, sub~liLuling c into EQ. 16
provides the frequency fO of the intermediate frequency signal, ~ltPln~tively, it can be
calculated using EQ. 10.
In order to allow the greatest antlount of control over the frequency fO, the
10~ largest range of values of the conh ol word Y that have an effect on the frequency fO
should be used. This range of effective values will be such that only one value of the
control word Y will cause the frequency fO to be set to its upper limit. All other values
of the conhrol word Y will result in the frequency fO to be set below its upper limit. In
terms of the variable c, this range of values should provide the largest range over
15 which the conhrol word Y can be varied while keeping c > 1.
The conhrol word Y can take on values from zero to minus 2N 1. By
subslilulillg the largest negative value of conhrol word Y into EQ. 18 a relationship
between the divisor M and the variable I is obtained, as expressed in EQ. 19, below.
CA 02238093 1998-0~-20
- 24 -
Maintaining this relationship between the divisor :M and the variable I will provide the
largest range of effective values of the control word Y.
EQ. 19 M 2 2I + 1
In this embodiment I=13, and therefore EQ. 19 indicates that the divisor
M should be greater than or equal to 27 when the control word Y is a negative number.
Th.e results for positive and negative control word Y define the value of the
10 divisor M as greater than or equal to 25 for a positive control word Y, and greater than
or equal to 27 for a negative control word Y. In order to satisfy both of these
requirements the divisor M should be greater than or equal to 27. With a value of the
divisor M that meets these conditions the DDFS 10 will operate with values of the
control word Y in the range of +/- 2N 1 without being unduly limited by the upper
and lower frequency limits. For I=13 and fx =84.24 MHz, EQ. 16 with c=l provides
an upper irequency limit for the frequency fO of 78 MHz. Under the same conditions
EQ. 6 with b=l yields a lower frequency l.imit of 77.5008 MHz. The collc~ollding
upper and lower frequency limits ofthe synth~i7e 1 clock signal SCLK are obtained
by dividing the latter two limits by six, repres~nting the division pel~lmed by the
20 divider 34. This calculation produces upper and lower frequency limits of the
synthlo.si7e~ clock signal SCLK of 13 MHz and 12.9168 MHz, respectively. Therefore,
the frequency fs can be varied from 12.9168 MHz to 13 MHz for values of the control
CA 02238093 1998-0~-20
word Y from 22~ to -22~, respectively. When the control word Y equals zero, the
frequency fs will be 12.96 MHz.
In addition to controlling the firequency fs ofthe syntll~i7ed clock signal
5 SCLK, the control word Y can also be used to control its phase. Referring to EQ. 2,
and ~sl-min~ that the values of the control word Y are in its largest effective range,
this equal:ion can be used to calculate the number of pulses of the reference clock
signal XCLK, for each shift in phase of the intermediate frequency signal. For
exarnple, if the control word Y initially equals zero, then by setting it to a new value
1~ of 2l6, the phase of the synthesized clock signal SCLK will be delayed after a certain
number of pulses. The amount of delay will be equal to one period of the intermediate
frequency signal divided by I. The delay will occur after 432 periods of the reference
clock signal XCLK firom the time that the control word Y was set to its new value. The
value of 432 periods is calculated firom EQ. 2 using M=27, N=21 and Y=216. If the
firequency fO ofthe intermediate fiequency signal equals 77.76 MHz and I=13, the
amount of phase delay in terms of time is 989 picoseconds. This delay occurs after
5.13 (=432/ fx ) microseconds ha ~e elapsed. After the phase delay has occurred the
control word Y can be reset to zero to return the frequency fs to 12.96 MHz. This
resetting would have to be done w ithin 5.13 microseconds after the phase delay, in
2() order to avoid another phase delay from occurring.
CA 02238093 1998-0~-20
-26-
The capability to control the phase of the synthesized clock signal SCLK could
be used in a phase locked loop circuit (PLL) . In a PLL circuit, there is a need to
control the phase of the output signal relative to the phase of an input signal. This is
5 accomplished by providing feedback of the output signal to a phase colllp~alor. The
phase CO1L11J~alOr output is a phase difference signal, which is used for controlling the
phase of the output signal. To use the DDFS 10 in this application, the phase
difference slgnal would need to collespolld to an applul,l;ate value of the control word
Y. This value could be used to shift the phase of the synthesized clock signal SCLK,
10 by the required amount, according to the method described in the previous example.
Various other modifications of the invention may be made without departing
from the principle thereof. For example, the pulse inhibit logic 30 could be
impl~mented with only two modes of operation. These could be up and neutral, or
15 neutral aIId down. In either case the pulse inhibit logic 30 would only be responsive to
a rising edge on one of the forward or back signals. The other signal would be non-
functional, and therefore not required. Furthermore, the DDFS 10 would only be
responsive to either positive or negative values of the control word Y, depending on
the modes of operation that were used. However, a value of the control word Y in the
20 middle of the effective range of values could be selected as a base value, and the
control word Y set to values lower or greater than it, in order to set the frequency fs to
a desired value. With these modifications, the frequency fO could still be calculated
CA 02238093 1998-OS-20
from EQ. 1, EQ. 6 and EQ. 16, depending on the modes of operation that were
implemented. This modification could simplify the pulse inhibit logic 30, however, at
the expense of control over the frequency fo.
S Another modification to the pulse inhibit logic 30 would be to enable the mode
of operation to remain in the up or down mode if another rising edge occurred on the
fc~l~val~d or back signals, re~e~iliv~ly. The result would be different from the disclosed
embodiment, wherein the rol~va~d and back modes are always followed by the neutral
mode. The effect of this modification would be to change the upper and lower
l 0 frequency limits of the frequency fO. This is because the pulse inhibit logic 30 could be
in the neutral mode less than half of the time. Therefore, the modification would
remove the limitation that b and c must be greater than or equal to one in operation.
The new upper and lower frequency limits could be calculated from EQ. 16 and EQ. 6,
respectively, with c and b equal to zero. This modification would increase the range
1 'i between the upper and lower frequency limits at the expense of reduced sensitivity to
the frequency fO to the control word Y.
A modification to the phase accumulator 20 would be to remove the divider 26.
The effect of this modification on the effective range of values of the control word Y
2() can be ~letennined by setting M equal to one in EQ. 11 and EQ. 17. In addition, the
adder 24 would have to be fast enough to be able to provide valid values of the result
word Z and the overflow and underflow signals before the latch 28 latches its inputs.
CA 02238093 1998-0~-20
-28-
This modification reduces the number of components in the DDFS 10 but requires a
modification of the pulse inhibit logic 30 and requires that the adder 24 be capable of
running at a faster speed of operation than in the described embodiment.
Another modification to the DDFS 10 would be to remove the divider 34. In
this case, the frequency fs Of the synthesi7ed clock signal SCLK would equal the
frequency fO of the intermediate frequency signal This modification would reduce the
number of components in the DDF'S 10 but would increase the amount jitter of the
10 synthesized clock signal SCLK.
Yet another modification would be to rep]ace the N-bit adder 24 with an
loadable N-bit up-down counter. In this case, the counter would be loaded with the
absolute value of the control word Y. The counter would either count up or down,
15 depending on the whether the control word was negative or positive. The underflow
and overIlow signals could be generated when the counter counts past zero or 2N,
respectively. After either of these conditions occurred, the counter could be reloaded
with the control word Y to repeat l:he same counting process. This modification would
change the relationship bclween the number of pulses of the reference clock signal
20 XCLK and the occurrence of a rising edge on the ~o,vv~d and back signals. The values
of control word Y would not necessarily correspond to the same value of the frequency
CA 02238093 1998-0~-20
- 29 -
fO as with the disclosed embodiment, however, this difference would not have a
substantial effect on the operation of the DDFS 1().
Stïll yet another modification would be to change the sequence of the pulse
5 inhibit operation from a regular sequence of one pulse inhibited in I pulses transmitted,
to another sequence. For example, the sequence could be pseudo random, in that the
number of pulses passed to the intermediate frequency signal output from the reference
clock signal input, before a pulse was inhibited in the neutral mode, could be a pseudo
random n~lmber instead of a constant number. The pseudo random numbers would
10 repeat in a sequence over a long period of time and hence the frequency of the
intermediate frequency signal would be constant on average taken over a relatively
long interval. By .chi~ing the sequence of inhibited pulses in the same manner as
previously described, that is, by advancing or delaying the timing of the next inhibited
pulse, yet still m~ t~ g the sequence after the shift, the phase and thus the
15 frequency of the intermediate frequency signal could be controlled.
Finally, the values of N, I, M and the divisor of divider 34 could be varied
from those in the disclosed embodiment. Depending on the values used, changes in the
sensitivity of control over the frequency fs~ and the upper and lower frequency limits
20 could be affected.
CA 02238093 1998-0~-20
-30-
A direct digital frequency synthesizer has been disclosed according to an
embodiment of the present invention. Numerous ~modifications, variations, and
adaptatio:ns may be made to the particular embodiment of the invention described
above without departing from the scope of the invention that is defined in the claims .
