BACKGROUND OF THE INVENTION Frequency synthesizers have traditionally made use of loop dividers which are designed to operate on an output frequency signal of a voltage controlled oscillator (VCO). Integer frequency synthesizers generally use loop dividers which divide an input frequency to a lower output frequency using a complete input frequency cycle period, or two pi (27r) radians, as the minimum unit of resolution. Fractional frequency synthesizers also employ frequency loop dividers which use 2z radians as the minimum unit of resolution. Typically, the loop dividers used in fractional frequency synthesizers obtain a fractional frequency divide value by periodically changing the integer value of loop divider so that, over several cycles of the output waveform, the average divide inclues a fractional component. While the method of averaging a loop divider value does produce an output signal at the desired divided down frequency, lower frequency subharmonics are also present in the output. The subharmonic frequency components are not desirable in frequency synthesizers since they introduce unwanted noise energy in the output spectrum. Often, the subharmonic frequency components must be attenuated by low pass filtering within the frequency synthesizer, which consequently increases the lock time, and decreases VCO sideband noise attenuation. The practice of dividing, and fractionalizing a divided value in units of complete input clock cycles results in low frequency, and relatively large amplitude subharmonic spurs, which require such filtering. It would be beneficial to develop a method by which a fractional frequency synthesizer would produce subharmonic spurs that occur at higher frequencies and at smaller amplitudes. This would reduce filtering requirements and improve synthesizer performance.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a fractional synthesizer, in accordance with the present invention.

FIG. 2 is a waveform diagram of a fractionally divided signal generated using a conventional fractional synthesizer.

FIG. 3 is a frequency spectrum diagram the signal shown in FIG.

2.

FIG. 4 is a waveform of a fractionally divided signal using a fractionalization technique, in accordance with the present invention.

FIG. 5 is a frequency spectrum diagram corresponding to the fractionally divided signal of FIG. 4, in accordance with the present invention.

FIG. 6 depicts a prior art fractional loop divider.

FIG. 7 depicts a fractional loop divider, in accordance with the present invention.

FIG. 8 is a waveform diagram illustrating a first example of double resolution operation, in accordance with the present invention.

FIG. 9 is a waveform diagram illustrating a second example of double resolution operation, in accordance with the present invention.

FIG. 10 is a state diagram describing the algorithm embodied in the state machine portion of the prescaler modulus controller, in accordance with the present invention.

FIG. 11 depicts a prior art dual modulus prescaler.

FIG. 12 is a waveform diagram depicting selected signals of the prescaler of FIG. 11.

FIG. 13 depicts a first embodiment of a symmetrical dual modulus prescaler, in accordance with the present invention.

FIG. 14 is a waveform diagram depicting selected signals of the prescaler of FIG. 13.

FIG. 15 depicts a second embodiment of a symmetrical dual modulus prescaler, in accordance with the present invention.

FIG. 16 is a waveform diagram depicting selected signals of the prescaler of FIG. 15.

FIG. 17 is a block diagram of a radio incorporating a synthesizer, in accordance with the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT FIG. 1 illustrates a preferred embodiment of a double resolution fractional N frequency synthesizer 100. A reference frequency 120 is coupled to one input of a biphase detector 110, and a divided frequency 135 is coupled to another input of the biphase detector 110. The reference frequency 120 incorporates a waveform that is symmetrical. The biphase detector 110 compares the phase difference of the reference frequency 120 and the divided frequency 135, and outputs an error signal 115 which represents the difference in phase between the two signals 120,135. The biphase detector 110 is distinguished from a conventional tri-state phase detector in which only the positive or negative phase transitions of the reference frequency 120 and the divided frequency are compare. The biphase detector compares the phase on both the negative and positive transitions of the reference frequency 120. The biphase detector 110 mazez be a mixer type phase detector, an exclusive-OR gate, or a biphase tri- state detector as described below.

The error signal 115 is passed through a low pass filter 118 to produce a filtered error signal 125 that is inputted to a voltage control oscillator (VCO) 138. The low pass filter 118 is used to attenuate unwanted spurious signals. The VCO 138 produces an output frequency signal 130 which corresponds to the voltage level of the filtered error signal 125. The output frequency signal 130 is coupled to a symmetrical divider 165 which produces the divided frequency 135. Preferably, the divided frequency 135 has a symmetrical waveform, or fifty percent (50CXc) duty cycle, so that when compare to the symmetrical reference frequency 120 (by the biphase detector 110), no spurious errors signals result due to mismatches in the shapes of the two waveforms.

The double resolution fractional N divider 180 inclues a symmetrical divider 165 having an additional divided frequency output 140 that is twice (2x) the divided frequency 135. The 2x divided frequency

140 is synchronous with the divided frequency 135 such that the positive edge transitions of the 2x divided frequency 140 are coincident with either the positive or negative edge transitions of the divided frequency 135. The 2x divided frequency 140 is used to drive resolution specification circuitry 170,175 that comprises a prescaler modulus controller 170, and a fractionalization circuit 175. The fractionalization circuit 175 inclues a provision for programming a numerator 160, and a denominator 162 to determine the fractional portion of the divide value. The divide modulus 145 may be varied over each half cycle of the divided frequency 135 in such a manner as to allow frequency division along half cycle units, or 7r radians, of the output frequency signal 130. The fractionalization circuit provides a fractionalization control signal 150 as an input to the prescaler modulus controller 170. The prescaler modulus controller 170 uses the fractionalization control signal 150 in combination with an integer value control signal 155, the divided frequency 135, and the 2x divided frequency 140 to generate a divide modulus or divisor 145. The integer value 155 is coupled to the prescaler modulus controller 170.

FIG. 2 illustrates a fractionally divided waveform as conventionally generated. Here, three cycles of six nanoseconds (6 ns) are followed by one cycle of seven nanoseconds (7 ns) which, when taken together, are intended to generate a frequency of 160 megahertz. FIG. 3 is a frequency spectrum showing the resulting frequencies generated.

Note that in addition to the intended frequency of 160 MHz, an additional frequency of 40 Mhz is generated at a particular amplitude (A). The 40 Mhz frequency represents an undesirable subharmonic frequency.

FIG. 4 illustrates a fractionally divided waveform used to generate a 160 Mhz frequency, in accordance with the present invention. Here, the divide modulus is specified for each half cycle and is varied as needed to produce the desired result. The resultant divided frequency, in this example, has in succession three half cycles of 3 ns; one half cycle of 3.5 ns; three half cycles of 3 ns; and one half cycle of 3.5 ns. FIG. 5 is a frequency spectrum diagram showing the generated frequencies. Note that a subharmonic frequency of 80 Mhz is generated instead of 40 Mhz,

and at half the amplitude (A/2). Further, only two cycles (four half cycles) are required to accomplis the fractional frequency averaging.

The increased subharmonic frequency and lower amplitude allows the performance of a fractional N frequency generator to be significantly improved, by allowing for increased loop bandwidth, particularly when the loop bandwidth is limited by subharmonic spurs.

FIG. 6 is a schematic diagram which illustrates a conventional fractional N loop divider 600 that provides an output signal 638. The fractional divider 600 incorporates an integer divider 602 which inclues a dual modulus prescaler 610, having possible divide values of either 4 or 5. The prescaler output signal 625 is coupled to a fixed divider 620, which is embodied as a divide-by-16 circuit. A feedback network 622 combines the various states of the fixed divider 620, and a P8 programming bit 665, a P4 programming bit 660, a P2 programming bit 655, and a PlA programming bit 650, to generate a prescaler modulus control signal 640.

The divide value is determined by multiplying the low modulus value of the dual modulus prescaler 610, which in this embodiment is 4, by the value of the fixed divider stage 620, which in this embodiment is 16.

Consequently, the base divider value is 64. The P8, P4, P2, Plia, and P1B programming bits 665,660,655,650,636 add to the base divider value, when enabled, by periodically controlling the dual modulus prescaler 610 to divide by 5 instead of by 4. The P8 programming bit 665, when set to logic high, adds 8 to the base divider value of 64, for a programme divide value of 72. The P4 programming bit 660 adds 4 to the base divider value, when set to logic high, and the P2 programming bit 655 and the P1A programming bit 650 add 2 and 1, respectively, to the base divider value when set to logic high.

The fractionalization circuit 175 is implemented using a single accumulator. The accumulator circuit 175 is coupled to the divider output 638 of the divider 600. The accumulator circuit 175 has a programme denominator 162 which determines its capacity, and a programme numerator 160 which determines the number by with the value of the accumulator circuit 175 increments at each positive

transition of the divider output 638. The accumulator circuit 175 operates such that the numerator 160 value is added to the contents of the accumulator circuit 175 at each positive transition of the divider output 638. When the contents of the accumulator circuit 175 equal or exceed the value of the denominator 162, the carry out signal 636 is set to logic high, and any difference between the accumulator contents and the denominator 162 value, is left in the accumulator contents. The carry out signal 636 is set to logic high at a rate which reflets the fractional value of the numerator 160 divided by the denominator 162. The carry out signal 636 is coupled to the feedback network 622 such that one (1,) ils added to the base divider value whenever the carry out signal 636 is set to logic high.

FIG. 7 is a schematic diagram of the fractional N loop divider 180, in accordance with the present invention. Here, significant details and interoperation of the symmetrical divider 165, the prescaler modulus controller 170, and the fractionalization circuit, are highlighted. The symmetrical divider 165 comprises a symmetric dual modulus prescaler 715, a fixed divider stage 705, and a feedback network 710. The prescaler modulus controller 170 comprises a state machine 732, and a set of data latches 734. The state machine 732 has five output signals 754,756,758, 760, and 762 which are coupled to the feedback network 710 through the set of data latches 734. The set of data latches 734 is used to reduce the possibility of data setup errors due to long propagation delays. The feedback network 710 and the fixed divider stage 705 are selected such that the divider output 135 has a symmetrical waveform for all possible integer divide values, which facilitates double resolution frequency synthesis. The fixed divider stage 705 inclues a set of D-flip-flops 722, 724,726,728, labeled aA","B","C", and"D,"respectively. A variety of divider ratios may be used to provide other configurations.

The control signals 764,766,768,770, and 772, which are output by the set of data latches 734, propagate through a portion of the feedback network 710 to form the MC8 control signal 750, the MC4 control signal 748, the MC2 control signal 746, the MC1B control signal 742, and the

MC1A control signal 744, respectively, which are OR'd together to form the prescaler modulus control signal 752. The PROG [3: 0] bus 736 has a 4- bit width. The four signals which comprise the PROG [3: 0] bus 736 determine the integer divide value of the divider, in a similar fashion as the programming bits 665, F60, 655, and 650 in FIV : 6. For the divider 165, however, the programming bits embodied in the PROG [3: 01 bus 736, are first combine and manipulated in the state machine 732, when necessary, in order to accommodate the fractional signal 150. The fractional signal 150 comprises a number from the fractional circuit, or accumulator 730. This number has a range of values depending on the order, or type of fractional circuit used, and is periodically added to the programme divider moulus. For a two-accumulator fractional circuit, the fractional signal 150, would be a number from the set: (-1,0,1,2).

For a single accumulator fractional circuit, the fractional signal 150, would be a number from the set: (0,1). In general, higher order fractional circuits produce fractional signals which have a greater range of possible values to be added to the programme divider moulus.

FIG. 8 is a waveform diagram illustrating a sample operation of the double resolution divide circuit 180, in accordance with the present invention. Referring to FIG. 7 and FIG. 8, the circuit 180 adds a multiple of 7r radians (0.5 cycles) of the output frequency signal 130 to a half cycle of f he divided frequency 135 waveform, as directe by the fractional circuit 175, or more specifically, by the accumulator 730. The four signals which comprise the PROG [3: 01 bus 736 are set to a logical 1,0,1, and 1 respectively, so that the MC8 control signal 750, the MC2 control signal 746, and the MC1A control signal 744 is normally enabled. The PROG [3: 0] bus 736 is therefore programme to obtain the odd integer divide value of the base divide value: 4 x 16 = 64, plus the offset provided by the programming bits PROG [3: 0] which is 8 + 2 + 1 = 11. Thus, the total integer divide value corresponds to a divide by 75. In FIG. 8, the divide frequency 135 waveform period is designated by"OUT", and the waveforms associated with the flip-flops 722,724,726,728 are designated as"A","B","C", and"D,"respectively. The divided frequency 135

waveform is the output of flip-flop 726, designated as"C"on the waveform diagram, and shown having two half cycles 845,850. The prescaler output signal 774 is designated as"PCLK,"and the MC8 control signal 750 is designated as"MC8", with the other control signals having a similar corresponding notation.

The duty cycle of"PCLK"774 is shown for the divide by 4 mode 810 and the divide by 5 mode 815 of the prescaler 715. Note the symmetry of "PCLK"signal 774. In divide by 4 mode 810, each half cycle of"PCLK"is labeled with the value"2", which represents two cycles of the output frequency signal 130. In divide by 5 mode 815, each half cycle of"PCLK" is labeled with the value"2.5", which represents 2.5 cycles of the output frequency signal 130. The fractional circuit 175 adds or radians (0.5 cycles) ouf the output frequency signal 130 waveform to the half cycle of the divided frequency 135 designated by 850, where the divided frequency 135 waveform is logic low. During the logic low portion of the divided frequency 135 waveform, the state machine 732 swaps the MU LA control signal 744 from being active, and instead activate the MC1B control signal 742. It can be seen that if the two control signals 742, and 744 are swapped, then the MU LA control signal 744, and the MC1B control signal <BR> <BR> <BR> <BR> 742 each adds Z radians of the output frequency signal 130 waveform to the logic low portion of the divided frequencv 135 waveform. Normally, only Z radians is added to the logic low portion 850 of the divided frequency 135 waveform by the MU LA control signal 744. In order to resume the normal programme divide value, the MU LA control signal 744 remains deactivated, and the MC1B control signal 742 remains activated following the logic low portion 850 of the divided frequency 135 waveform. This condition is referred to as the"SWAP"state 1020 in the state diagram described below. If the fractional circuit 175 later attempts to add 7r radians to a half cycle of the divided frequency 135 waveform, the MCLA control signal 744 and MC1B control signal 742 may be swapped again, in accordance with the logic level of the divided frequency 135 at that time. It is also possible to subtract or radians from a particular half

cycle of the divided frequency 135 waveform by swapping the MU LA control signal 744 and MC1B control signal 742.

The timing of the MCIA control signal 744, and the MC1B control signal 742 is significant in the operation of the double resolution divider 180, in that they occur during the transitions of the divided frequency 135 waveform. For instance, because the MU 1A control signal 744 selectively occurs at the negative transition of the divided frequency 135 waveform, a corresponding prescaler output 774 waveform occurs such that one half cycle of the prescaler output 774 waveform is on either side of the divided frequency 135 waveform negative transition. Since the prescaler output 774 waveform is always symmetrical, 7r radians of the output frequency signal 130 waveform may be selectively added to both sides of the divided frequency 135 waveform negative transition. In short, transitions of the divided frequency 135 waveform occur dring the middle of a cycle of "PCLK". This enables a half cycle (0.5) of the output frequency signal 130 waveform to be selectively added or subtracted to half cycles 845,850 of the divided frequency 135 waveform by varying the value of the prescaler modulus 820,830 that occurs during the transition of the divided frequency 135 waveform.

When the symmetrical divider 165 is programme for an even valued integer divide, such as divide by 74, the MC1A control signal 744 and the MC1B control signal 742, will not normally be activated. FIG. 9 is a timing diagram illustrating this operation. As in FIG. 8, the addition of ir radians of the output frequency signal 130 waveform is made to a half cycle 945 (logic high) of the divided frequency 135 waveform. The MCLA control signal 744 and MC1B control signal 742 are both activated by the state machine 732, regardless of the polarity 945, 950 of the divided frequency 135 waveform. Afterwards, in order to resume the programme integer divide value, the state machine 732 decrements the programme value of the MC8 control signal 750, MC4 control signal 748, and the MC2 control signal 746. This condition is referred to as the"SUB1"state 1030 in the state diagram described below.

A"SUBO"state 1010 occurs when no subtraction or swapping of control

signals occurs, due to an odd valued number from the fractional circuit 175.

FIG. 10 shows a state diagram 1000 that describes the algorithm embodied in the state machine 732 shown in FIG. 7. The state diagram 1000 illustrates the general method by which a symmetrical divider having a dual modulus symmetrical prescaler adds or subtracts any number of half cycles of an input frequency signal to or from any half cycle of a divided output frequency waveform. The state diagram 1000 indicates seven possible transitions labeled"A","B","C","D","E","F', and"G,"between three states 1010,1020,1030. For each transition, several different input conditions may trigger the transition, and consequently several different output conditions may result from a particular set of inputs. The inputs to a state machine embodying the state diagram 1000 are: 1) the programme integer divide value PROG [3: 0] bus 736 ; 2) the signed number that the fractional circuit 175 adds to the programme integer divide value during a half cycle of the divided frequency 135 waveform; and 3) the current logic level of the divided frequency 135 waveform. The outputs of the state machine are: 1) the control signals P2 P4 and P8 which are not timed to correspond to the transitions of the divided frequency 135 waveform; 2) the PlA control signal 760 which is timed to correspond to the negative transition of the divided frequency 135 waveform; and 3) the P1B control signal 762 which is timed to correspond to the positive transition of the divided frequency 135 waveform. The inputs and outputs of the state machine 732 and the corresponding state diagram 1000 for any transition are summarized in the following table: TRANS. INPUTS OUTPUTS (1) (2) (3) (1) (2) (3) A Odd-No Hi (PN~P2)-Int (No/2) 0 1 A Odd-No Lo (PN-P2)-Int (No/2)- 0 1 1 TRANS. INPUTS OUTPUTS A Odd No Hi (PN#P2) +Int (No/2) 0 1 +1 <BR> <BR> <BR> A Odd No Lo (PN#P2) +Int (No/2) 0 1<BR> <BR> <BR> <BR> <BR> <BR> <BR> <BR> <BR> B Odd-Ne * (PN#P2)-(Ne/2) 0 1<BR> <BR> <BR> <BR> B Odd Zero * (PN#P2) 0 1<BR> <BR> <BR> <BR> B Odd Ne * (PN#P2) + (Ne/2) 0 1<BR> <BR> <BR> <BR> <BR> <BR> <BR> <BR> <BR> C Odd-No Hi (PN#P2)-Int(No/2)- 1 0 1 C Odd-No Lo (PN~P2)-Int (No/2) 1 0 C Odd No Hi (PN~P2) +Int (No/2) 1 0 C Odd No Lo (PN#P2) +Int (No/2) 1 0 +1 *****CEven D Even Zero * (PN#P2) 0 0 <BR> <BR> <BR> D Odd Zero * (PN#P2) 1 0<BR> <BR> <BR> <BR> <BR> -Ne*(PN#P2)-(Ne/2)00D* D Ne * (PN#P2) + (Ne/2) 1 0 E Even No 11(PN#P2)+Int(No/2) E Even-No * (PN#P2)-Int(No/2)- 1 1 1 F Even No * (PNP2) +Int (No/2) 0 0 F Even-No * (PN#P2)-Int(No/2)- 0 0 1 F Odd

TRANS. INPUTS OUTPUTS G Even Ne * (PN~P2) + (Ne/2)-1 1 1 G Even Zero * (PN~P2)-1 1 1 G Even-Ne * (PN#P2)-(Ne/2)-1 1 1 where: PN~P2: is an integer from 2 to N added to the base divide value; Odd: is an odd valued programme integer divide value 736; Even: is an even valued programme integer divide value 736; No: is an odd valued number from the fractional circuit 175; Ne: is an even valued number from the fractional circuit 175; Int (No/2): is the integer result of No/2; Hi: is logic high level of the divided frequency 135 waveform; Lo: is the logic low level of the divided frequency 135 waveform; is a don't care value; 0: is a deactivated control signal; 1: is an activated control signal; FIG. 11 illustrates a conventional dual modulus prescaler configuration. The input clock 1110 is coupled to first, second and third flip-flops 1140,1150, and 1160 respectively. The first, second and third flip-flops 1140,1150,1160 are also serially coupled and provided with a feedback path 1180 in order to provide a frequency divide value of 4 or 5.

As shown, the modulus control signal 1120 selectively couples the third flip-flop 1160 through the AND gate 1155, so that when the modulus control signal 1120 is low, a frequency divide value of 4 results, and when the modulus control signal 1120 is high, a frequency divide value of 5 results. As shown in FIG. 12, when the modulus control signal 1120 is logic low, the output signal 1130 is symmetrical, and when the modulus control signal 1120 is logic high, the output signal 1130 is non- symmetrical.

FIG. 13 illustrates a first embodiment of a dual modulus prescaler 1300 which produces a symmetrical waveform for both even an odd valued frequency divides, in accordance with the present invention.

Here, first, second, and third flip-flops 1340,1350,1360, are coupled similar to the circuit in FIG. 11. An additional flip-flop 1395 is selectively coupled to the second flip-flop 1350 through the modulus control signal 1320. The additional flip-flop 1395 is clocked on the opposite clock edge of input signal 1310, as compare to first, second and third flip-flops 1340, 1350,1360. Also, the output signal 1330 is the logical OR of the output 1375 from the additional flip-flop 1395, and the output from the second flip-flop 1350. FIG. 14 shows a timing diagram illustrating that the output signal 1330 is symmetrical for a divide by 4 and a divide by 5.

FIG. 15 illustrates a second embodiment of a dual modulus prescaler 1500 that produces a symmetrical waveform for both even an odd valued frequency divides, in accordance with the present invention.

In response to a clock signal 1510, the prescaler 1500 provides for a low divide value which is odd, and a high divide value which is even. The modulus control signal 1520 when high, enables the divider to provide a frequency divide of 6. When the modulus control signal 1520 is low, the divider provides a frequency divide of 5. A flip-flop 1595 provides a signal 1575 which is logically OR'd with a signal from flip-flop 1550 to generate an output signal 1530. The waveform diagram in FIG. 16 illustrates that the output signal 1530 is symmetrical for both the divide by 5 and the divide by 6 operation.

FIG. 17 shows a block diagram of a radio communication device 1700, in accordance with the present invention. The communication device 1700 is preferably a two-way radio or radio telephone that is operable to provide telephone interconnect, dispatch, paging, private call and data services. In the radio 1700, a controller 1710 is coupled to a memory 1780, to a transmitter 1740, and to a receiver 1750, to provide communication circuitry. The transmitter 1740 and receiver 1750 are coupled to a fractional synthesizer 1745 formed in accordance with the present invention. The transmitter 1740 and the receiver are also

coupled via an antenna switch 1760 to an antenna 1770. For transmit operations, the controller 1710 configures the antenna switch to couple the transmitter 1740 to the antenna 1770. Similarly, for receive operations, the controller 1710 couples the antenna 1770 via the antenna switch 1760 to the receiver 1750. Receive and transmit operations are conducted under instructions stored in the memory 1780. The radio 1700 also inclues a display 1720, a keypad 1730, and a speaker 1790, that together provide a user interface for accessing radio functions.

A method and apparats has been disclosed which enables a frequency synthesizer loop to fractionally synthesize an output frequency from a reference frequency based on half cycle periods. or Jr radians. The technique provides for subharmonic spurs which are twice in frequency and half in amplitude when compare to subharmonic spurs which would occur in a frequency synthesizer using a resolution based on 2 Jr radian periods.

Numerus modifications, changes, variations, substitutions and equivalents will occur to those skilled in the art without departing from the spirit and scope of the present invention as defined by the appende claims.