Signal generator, and method

[0002] The present invention generally relates to an arrangement that generates periodic electrical signals, and more particularly, to a signal generator for a microprocessor, and to a method.

[0003] Many applications, such as microprocessors, require a clock having a high frequency. While quartz oscillators operate at a lower frequency, conversion is often required.

[0004] FIG. 1 illustrates a simplified block diagram of a phase-locked loop (PLL) circuit. The PLL circuit has a phase comparator (PC), a low pass filter (LPF) with capacitor, and a voltage controlled oscillator (VCO). The PLL circuit provides that the frequency F_OUTof the output signal Y remains locked to the frequency F_REFof the input signal X (or to a multiple of F_REF). The phase comparator detects phase differences between the input signal X and the output signal Y and acts on the VCO to minimize the difference.

[0005] It is desired to integrate the complete PLL circuit and the processor into a monolithic integrated chip. For many “on-chip” applications, the input signal X has a low reference frequency (e.g., F_REF≈32 kHz) and the output signal Y has a very high output frequency (e.g., F_OUT≈100 . . . 200 MHz). Frequency multiplication is required by a factor F_OUT/F_REFof about 3000 . . . 6000. Since noise influence increases frequency variations, the PLL circuit would require a large capacitor (e.g., 10 nF) external to the chip and connected by an unwanted extra pin. Integrating the capacitor would be difficult too; for example, the silicon area of such a capacitor would be larger than the area of the PLL circuit itself.

[0006] There in an ongoing need to provide an improved PLL circuit as well as a method that mitigate some or all of these and other disadvantages and limitations of the prior art.

[0007] FIG. 1 illustrates a simplified block diagram of a phase-locked loop (PLL) circuit;

[0008] FIG. 2 illustrates a simplified block diagram of a signal generator according to the present invention;

[0009] FIG. 3 illustrates a simplified block diagram of a multiplier in the generator of FIG. 2;

[0010] FIG. 4 illustrates a simplified block diagram of a detector in the multiplier of FIG. 3;

[0011] FIG. 5 illustrates a simplified block diagram of a digital filter in the multiplier of FIG. 3;

[0012] FIG. 6 illustrates a simplified block diagram of a digital fractional divider in the multiplier of FIG. 3;

[0013] FIG. 7 illustrates a simplified spectrum of an intermediate signal in the generator of FIG. 2, as well as illustrates a transfer function;

[0014] FIG. 8 illustrates a simplified signal diagram of an input signal as well as intermediate signals in the multiplier of FIG. 3;

[0015] FIG. 9 illustrates a simplified signal diagram of an intermediate signal in the multiplier of FIG. 3;

[0016] FIG. 10 illustrates a simplified signal diagram of an intermediate signal in the multiplier of FIG. 3;

[0017] FIG. 11 illustrates a simplified signal diagram of an intermediate signal provided by the multiplier of FIG. 3; and

[0018] FIG. 12 illustrates a simplified flow chart diagram of a method to operate the multiplier of FIG. 3.

[0019] The following description of a preferred embodiment follows the order (a) short introduction into the signals, (b) explanation of components, (c) detailed explanation of the signals, and (d) method of operation.

[0020] (a)

[0021] When referring to signals, terms like “intermediate”, “clock”, “input”, “output” are introduced once but can conveniently be left out later. For convenience, same symbols (e.g., symbol “dK”) are used for the signals (e.g., “signal dK”) and any numeric value that the signal represent (e.g., “integer dK”).

[0022] For convenience and simplicity, signals X₁, X₂, X₃, Z are considered to be digital single-bit signals of rectangular shape, cf. signal X₂in FIG. 2 as representative example. First logical state “1” and second logical states “0” are defined, for example, by high value 118 and low value 117, respectively, of a physical quantity, such as a voltage or a current. The duration of both logical states is conveniently assumed to be substantially equal. Neglected is the transition time between the states, i.e., from “1” to “0” and vice versa. For convenience, a period T is conveniently defined between consecutive transitions from “1” to “0” (“falling edges”). The period T is the reciprocal of a frequency of the signal, that is

[0023] T=1/F  (2)

[0024] Signals such as dK and K are considered to be multi-bit signals that represent a positive or negative numerical value. Periods are conveniently defined between time points that allow changes in the numerical value.

[0025] (b)

[0026] FIG. 2 illustrates a simplified block diagram of signal generator 100 according to the present invention. Signal generator 100 comprises digital frequency multiplier 110 (“arrangement”, “multiplier”) and PLL circuit 180. It is an advantage of the present invention that generator 100 is implemented on a single monolithic chip.

[0027] Generator 100 receives external input clock signal X₁having frequency F₁(e.g., F₁≈32 kHz) at input 121. An example for X₁is explained in connection with FIG. 8. The period T₁of signal is defined as

[0028] T₁=1/F₁  (4)

[0029] T₁=31.25 μs (example)  (6)

[0030] Since signal X₁is, preferably, provided by a quartz oscillator (not shown), frequency variations of F₁are neglected.

[0031] Generator 100 provides output clock signal Y having frequency F_Yat output 185. F_Yis larger than F₁(e.g., F_Y≈100 MHz). The frequency jitter of signal Y (variations of F_Yover the time) are, preferably, in the range of plus/minus one per cent (minimum 99 MHz, maximum 101 MHz).

[0032] Multiplier 110 uses intermediate clock signal Z having frequency F_Z. The period T_Zof signal Z is defined as:

[0033] T_Z=1/F_Z  (8)

[0034] Details for signal Z are explained in connection with FIG. 8.

[0035] Multiplier 110 receives signal X₁and provides intermediate clock signal X₂having frequency F₂at output 145. Details for signal X₂are explained in connection with FIG. 11. The period T₂(n) of signal X₂is defined as the time interval two consecutive falling edges 119.

[0036] Multiplier 110 selectively provides transitions of signal X₂(e.g., falling edges 119) at consecutive time intervals T₂(n); the duration of T₂(n) is determined by a variable number (e.g., A+P(n)) of periods T_Zof signal Z. Details for multiplier 110 are explained in connection with FIG. 3.

[0037] The time average F′₂of frequency F₂is substantially constant, in other words, measuring F₂at consecutive time points and calculating the arithmetic average of the measurements substantially results to one and the same F′₂. Average F′₂is higher than F₁but lower than F_Y, that is:

[0038] F′₂>F₁  (10)

[0039] F′₂<F_Y  (12)

[0040] Average F′₂is related to F₁by multiplication factor M₁, that is:

[0041] F′₂=M₁*F₁  (14)

[0042] Conveniently, factor M₁is an integer; preferred values for factor M₁are in the range between 128 and 1024 (powers to 2); in the example, M₁=512. Preferably, factor M₁is programmable, so that generator 100 operates with either value.

[0043] To simplify calculations and further explanation, M₁is assumed to be M₁=500 (instead of 512), thus resulting in average F′₂=16 MHz and an average period T′₂=1/F′₂=0,0625 ns.

[0044] PLL circuit 180 receives signal X₂(at input 181) as reference to lock and provides signal Y at output 185; frequency F_Yis related to average F′₂by multiplication factor M₂, that is:

[0045] F_Y=M₂*F′₂  (16)

[0046] F_Y=M₁*M₂*F₁  (18)

[0047] Conveniently, factor M₂is a real number.

[0048] In the illustrative example with F_Y=100 MHz and F′₂=16 MHz, factor M₂is 6.25.

[0049] PLL circuit 180 low-pass filters out frequency fluctuations of signal X₂(transfer function, cf. FIG. 7). The frequency jitter of signal Y is less than the jitter of signal X₂.

[0050] FIG. 3 illustrates a simplified block diagram of multiplier 110 in the generator of FIG. 2. Multiplier 110 looks similar to a conventional PLL with digital components, however, its operation and its signals are different.

[0051] In a core portion, multiplier 110 comprises digital phase detector 120 (“detector”), digital filter 130 (“filter”, details in FIG. 5), fractional divider 140 (“divider”, details in FIG. 6), local oscillator 150 (“oscillator”); and feedback unit 160.

[0052] Filter 130 and divider 140 has registers or other memory means (no shown in FIG. 3) to store numeric values needed for operation, such as, for example, filter coefficients, and look-up tables. Persons of skill in the art can implement the registers without the need of further explanation herein. Oscillator 150 has digital-to-analog converter 151 (“DAC”). Signal lines with in the core portion are illustrated by plain lines.

[0053] Preferably, to assist operation, multiplier 110 further comprises calibration unit 163, and initialization unit 164. For convenience, signal lines to and from units 163, 164 are illustrated by dashed lines.

[0054] In the core portion (120, 130, 140, 150, 160), signals are distributed as follows: X₁(cf. FIG. 2, input 121 of detector 120), dK (output 124 of detector 120 to input 137 of filter 130), K (output 138 of filter 130 to input 142 of divider 140), X₂(output 145 of divider 140 to input 161 of feedback unit 160), X₃(output 162 of feedback unit 160 to input 122 of detector 120), correction signal for frequency divider noise correction (FDNC, output 143 of divider 140 to input 123 of detector 120), Z (output 155 of oscillator 150 to input 146 of divider 140 and to input 125 of detector 120).

[0055] In a first control loop (detector 120, filter 130, divider 140, feedback unit 160), signal X₂is derived from signal X₁, wherein signal Z serves as an auxiliary signal. Preferably, this first control loop has two signal feedback paths:

[0056] (i) the path via feedback unit 160 to alter K in a coarse manner; and—optionally—

[0057] (ii) the direct path from divider 140 to detector 120 for fine adjusting the operation of detector 120 using FDNC (real number, details later).

[0058] In a second control loop, calibration unit 163 compensates the influence of the environment temperature on oscillator 150.

[0059] Oscillator 150 provides signal Z (mentioned above). There is no need to synchronize signal Z with signal X₁. Preferably, F_Zis higher than F₁, and F_Zis also higher than the average F′₂of F₂, that is:

[0060] F_Z>F₁  (20)

[0061] F_Z>F′₂  (22)

[0062] A preferred value for F_Zis, for example, the rated frequency F_{Z RATED}=160 MHz (T_Z=6.25 ns). Due to tolerances in manufacturing generator 100, F_Zis allowed to vary in a predetermined operating range (i.e. F_{Z MIN}. . . F_{Z MAX}) that is larger than the frequency range of signal Y (i.e., F_{Y MIN}. . . F_{Y MAX}), but that is smaller than the frequency F_Y, that is:

[0063] |F_{Z MAX}−F_{Z MIN}|>|F_{Y MAX}−F_{Y MIN}|  (24)

[0064] Preferably, F_Zvaries by plus/minus 10 percent (e.g., 144 MHz . . . 176 MHz). However, other ranges can also be used (e.g., 85 MHz . . . 250 MHz).

[0065] Oscillator 150 is, preferably, implemented on the chip without the use of frequency stabilizing quartz components. Oscillator 150 is preferably, implemented as a frequency controlled oscillator VCO that receives a control voltage V_Cfrom DAC 151. In other words, oscillator 150 is a digitally controlled oscillator.

[0066] Oscillator 150 exhibits a frequency drift that is a function of the temperature and control voltage V_C.

[0067] Alternatively, oscillator 150 is implemented as a delay ring oscillator with a loop of delay elements (example explained in U.S. Pat. No. 5,832,048).

[0068] Feedback unit 160 converts signal X₂into feedback clock signal X₃, for example, by division with factor M₁. Signal X₃has a frequency that is about frequency F₁.

[0069] Division can be accomplished, for example, by changing the logical state (causing transitions) of X₃every M₁periods of X₂. Preferably, feedback unit 160 is implemented by a counter to M₁.

[0070] Decimation is well known in the art, and described, for example, in “Crochiere, R. E. and Rabiner, L. R.: “‘Interpolation and Decimation of Digital Signals—A Tutorial Review’, Proceedings of the IEEE, vol. 69, No.3, March 1981”.

[0071] Detector 120 calculates the phase difference between signal X₁(frequency F₁) and signal X₃(also about frequency F₁, cf. feedback loop via unit 160), preferably, with an accuracy of a single period T_Zand provides numeric representation “dK” of this phase difference. Preferably, detector 120 is implemented as an up-and-down counter. Details are explained in connection with FIGS. 4 and 8.

[0072] Filter 130 samples the numeric representation “dK” with a sampling frequency F_Sthat is, preferably, higher than F₁and thereby low-pass filters “dK” to further numeric representation “K”. Details for the operation of filter 130 are also explained in connection with FIGS. 9-10.

[0073] Filter 130 provides stability in the first control loop (120, 130, 140, 160) by preventing positive feedback.

[0074] Divider 140 provides signals X₂. As explained with more detail below, the period T₂(n) of X₂is variable and depends on K and on Z (cf. T₂(n) in FIG. 11).

[0075] FIG. 4 illustrates a simplified block diagram of detector 120 in multiplier 110 of FIG. 3. Detector 120 comprises state-machine 126, counter 127, and—optionally—adder 128. State machine 126 receives signal X₁at input 121, signal X₃at input 122, as well as signal Z at input 125, and forwards control signals (collectively labeled “CONTROL”) to counter 127. Counter 127 provides signal dK at output 124, hence dK is an integer.

[0076] Optionally, as indicated by the dashed frame, adder adds signal FDNC (cf. FIG. 3) at input 123 to dK, resulting in dK# at output 124# With FDNC being a real number, dK# is also a real number.

[0077] The following explanation considers dK as integer; persons of skill in the art can apply the mentioned fine adjusting by using dK# instead (output 124# coupled to input 123 of filter 130, second control loop).

[0078] In state machine 126, circles illustrate the following states: WAIT, COUNTER RESET (A), COUNTER RESET (B), COUNT DOWN, and COUNT UP. Unidirectional arrows between the states indicate state transitions that are triggered, as in the example of FIG. 4, by falling edges (symbols ↓) of signals X₁, X₃and Z that occur at any time point. The WAIT state has 3 departing arrows that correspond to each signal X₁, X₃and Z. The other states have 2 departing arrows only so that they remain unchanged even for the falling edge of the “non-arrow” signal.

[0079] If a falling edge triggers a state transition, i.e. when an arrow goes to a circle, state machine 126 causes counter 127 either to reset dK to zero (“RESET”); to increment dK by 1 (“COUNT UP”); to decrement dK by 1 (“COUNT DOWN”); or to output dK to adder 128 (that substantially simultaneously forwards dK to output 124).

[0080] Preferably, when dK has reached its maximum value (i.e. dK=dK_MAX) and state machine 126 issues COUNT UP, then dK remains unchanged at dK_MAX. Similarly, when dK has reached its minimum value (i.e. dK=dK_MIN) and state machine 126 issues COUNT DOWN, then dK remains unchanged at dK_MIN.

[0081] Without departing from the invention, persons of skill in the art can modify state machine 126, for example, by making it respond to rising edges.

[0082] FIG. 5 illustrates a simplified block diagram of digital filter 130 in multiplier 110 of FIG. 3. Digital filter 130 comprises multiply unit 132, adder units 133 and 135, register 131, and divide unit 136. Filter 130 receives dK at input 137 (cf. FIG. 3) and provides K at output 138 (cf. FIG. 3).

[0083] Multiply unit 132-1 digitally multiplies representation dK to dK′, for example, by a multiplication factor of 1.5. Adder unit 133 digitally adds dK from input 137 and dK(iii) from register 131 to dK(i); register 131 delays dK(ii), preferably, by a single period 1/F₂to dK(iii); divide unit 136 divides dK(iii) to dK(iv), preferably, by dividing with M₁; and adder unit 135 adds dK(i) and dK(iv) to dK(v); and multiply unit 132-2 digitally multiplies dK(v), preferably, by 1/M₁, to K at output 138. Digitally multiplying signals is well known in the art.

[0084] Since the multiplication factor in units 132-1 and 132-2 is a real number, dK(i), dK(iv), dK(v) as well as K are real numbers too, each having an integer portion and a fractional portion. The integer portion represents a positive or negative numeric integer value; and the fractional portion represents a numeric value between 0 and +1. Preferably, K is positive.

[0085] For simplicity, the figures represent the signal flow by single lines; persons of skill in the art can implement the connections, for example, in parallel by separate bit lines. Conveniently, the real portions are transmitted by bit lines for 2⁰, 2¹, 2²and so on; and similarly, the fractional portions are transmitted by bit lines for 2⁻¹, 2⁻²and so on. For convenience of further explanation, the term “most significant bits” (MSB) is used for the integer portion; and the term “least significant bits” (LSB) stands for the fractional portion.

[0086] FIG. 6 illustrates a simplified block diagram of digital fractional divider 140 in multiplier 110 of FIG. 3. Divider 140 receives representation K at input 142 from filter 130, receives signal Z at input 146 from oscillator 150, provides signal X₂at output 145 (cf. FIG. 3), and—optionally—provides FDNC at output 143. Divider 140 comprises delta-sigma modulator 140-1, controlled frequency divider 140-2, integrator 140-3, and adders 140-4 and 140-5. Coming from input 142, the bits of K split into most significant bits (KMSB) and least significant bits (KLSB), cf. the explanation of FIG. 5 above. Modulator 140-1 modulates KLSB to KLSB′; adder 140-4 forwards the sum of KLSB40 and KMSB to divider 140-2; under control of Z, divider 140-2 provides X₂; X₂is fed back to modulator 140-1; and—optionally—adder 140-5 forwards the difference between KLSB and KLSB′ to integrator 140-3 for further integration to FDNC.

[0087] Since some components (like modulator 140-1) introduce a delay, additional delay stages are convenient to ensure simultaneous signal arrival at adders 140-4 and 140-5, has well as at output 143. Illustrated are delay stage 140-6 (between input 142 and adder 140-4), delay stage 140-7 (between input 142 and adder 140-5), and delay stage 140-8 (between integrator 140-3 and output 143).

[0088] (c)

[0089] FIG. 7 illustrates a simplified spectrum of signal X₂in generator 100 of FIG. 2 and the transfer function of PLL circuit 180. The abscissa axis shows F₂in MHz. Graph 1 shows the power spectrum density, and graph 2 shows the transfer function. The transfer function is substantially inverse to the spectrum, so that low frequency components of X₂are prioritized over high frequency components. In other words, while PLL circuit 180 passes the “low” frequency components at F₂, PLL circuit 180 filters out the unwanted “high” frequency components.

[0090] Assuming sampling with 16 MHz, the spectrum is shown up to 8 MHz only (Nyquist criterium), wherein higher frequency components are not considered here.

[0091] FIG. 8 illustrates a simplified signal diagram of signals X₁, X₃and Z in multiplier 110 of FIG. 3. Also, FIG. 8 indicates how detector 120 derives signal dK. As mentioned above, detector 120 calculates the phase difference between signal X₁and signal X₃with an accuracy of plus/minus a single period T_Zof signal Z.

[0092] At output 124, detector 120 provides dK as a numeric representation of the time T_dKbetween consecutive falling edges of X₁and X₃counted by periods T_Z, that is:

[0093] dK=T_dK/T_Z  (28)

[0094] Preferably, the sign of dK also indicates which signal leads: dK is positive when the falling edge of X₁comes earlier than the falling edge of X₃as in FIG. 8, and dK is negative otherwise. Values of dK are, preferably, the following range:

[0095] dK_MIN≦dK≦dK_MAX  (30)

[0096] −T₂*T_Z*0.5≦dK≦+T₂*T_Z*0.5  (32)

[0097] −250≦dK≦+250 (example)  (34)

[0098] However, in most cases, the falling edges occur at short intervals, so that dK is in the range ±32.

[0099] Since detector 120 determines dK based on a period T_Z, the value of dK is linearly related to frequency F_Z. In other words, suppose Td remains constant, dK increases if oscillator 150 speeds up and dK decreases if oscillator 150 slows down. In other words, dK is proportional to the frequency ratio between F_Zof signal Z and F₁of signal X₁, and proportional to the period ratio between T₁and T_Z, that is:

[0100] dK=factor*F_Z/F₁  (36)

[0101] dK=factor*T₁/T_Z  (38)

[0102] with “factor” being a proportional factor introduced for convenience of explanation.

[0103] FIG. 9 illustrates a simplified signal diagram of signal dK in multiplier 110 of FIG. 3. Values of dK are shown be vertical lines. Since falling edges in each signal X₁and X₃occur every T_1≈T3, a new value of dK becomes available every T₁and remains constant during that interval (horizontal lines).

[0104] FIG. 10 illustrates a simplified signal diagram of intermediate signal K provided by filter 130 in multiplier 110 of FIG. 4. K represents a real number (floating point). Compared to dK, the signal K appears less jumping and more smoothed.

[0105] Compared to dK, K changes at the above mentioned shorter sampling periods, that is:

[0106] T_S<T₁  (40)

[0107] This is conveniently achieved by sampling dK by filter 130 with signal Z or with signal X₂.

[0108] T_S<T_Z  (42)

[0109] T_S<T₂  (44)

[0110] Similar as dK, K is proportional to the frequency and period ratios mentioned above. In other words, K gives also a qualitative representation of the frequency of signal Z.

[0111] FIG. 11 illustrates a simplified signal diagram of signal X₂provided by multiplier 110 of FIG. 4. Signal X₂is illustrated with consecutive periods between falling edges 119:

[0112] T₂(1), T₂(2), T₂(n) . . . T₂(N)  (46)

[0113] As mentioned above, period T₂(n) varies in duration. Each period T₂(n) is allowed to differ from its predecessor, that is:

[0114] T₂(n−1)≠T₂(n)  (48)

[0115] In other words, the actual period T₂(n) differs from the average period T′₂(i.e., 1/F′₂) by a predetermined offset ΔT₂(n)

[0116] T₂(n)=T′₂+ΔT₂(n)  (50)

[0117] also called “jitter”, indicated below for each falling edge 119; ΔT₂(n) is either zero (T₂(n) substantially equal to T′₂), positive (T₂(n) longer than T′₂), or negative (T₂(n) shorter than T′₂).

[0118] The above mentioned average T′₂of T₂(n) is calculated over N periods as follows:

[0119] $\begin{array}{cc}{T}_2^{\prime }=\frac{1}{N}\sum _{n=1}^N{T}_2\left(n\right)& \left(52\right)\end{array}$

[0120] Preferably, N is larger than the number of periods T₂(n) that fit into a single period T₁. In other words, N is larger than M₁. As mentioned above, T₂(n) is a multiple of T_Z, that is:

[0121] T₂(n)=T_Z*[A+P(n)]  (54)

[0122] Both A and P(n) are integers; A is a positive constant, and P(n) is a positive or negative offset variable, that is:

[0123] P(n)=0, ±1, ±2 . . . ±P_MAX  (56)

[0124] Preferably, the absolute value of P_MAXis smaller than or equal to the absolute value of A, that is:

[0125] |P_MAX|≦|A|  (58)

[0126] In the example, T₂(1) is longer than T′₂(i.e., ΔT₂(1)=+T_Z); T₂(2) is substantially equal to T′₂(i.e., ΔT₂(2)=0); T₂(3) is shorter than T′₂(i.e., ΔT₂(1)=−T_Z), thus giving an offset sequence starting with P(1)=+1, P(2)=0, and P(3)=−1.

[0127] Using the offset sequence P(1) to P(N), the previous equations are re-written as follows: $\begin{array}{cc}{T}_2^{\prime }=T_z*\left[A+\frac{1}{N}\sum _{n=1}^{N}P\left(n\right)\right]& \left(60\right)\end{array}$  T′₂=T_Z*[A+P′]  (62)

[0128] Preferably, the offset sequence P(1) to P(N) is predetermined and depends on K. The average P′ of the sequence as well as T_Zdetermine T′. Changes of T_Zare allowed but the influence of such changes is compensated by P′.

[0129] In operation, divider 140 provides signal X₂, for example, by repeating the following event sequence: (i) at the beginning of each period T₂(n), divider 140 outputs falling edge 119 of X₂(as mentioned above, falling edges can also be used); (ii) also at the beginning of T₂(n), divider 140 starts to count A+P(n) periods of T_Z; (iii) divider 140 output a rising edge of X₂(e.g., after half of A+P(n) periods); and (iv) after A+P(n) periods T_Z, divider 140 outputs falling edge 119 again (starting T₂(n+1)). Persons of skill in the art are able to provide the P(n) sequence, for example, by a look-up table; providing a counter is well known too, so that table and counter are not illustrated. Using a sigma-delta modulator for providing the P(n) sequence is also convenient. Such a sigma-delta modulator can relate KLSB to P′.

[0130] In the example, to achieve the above mentioned F₂=16 MHz and T₂=62.5 ns for X₂, for various frequencies F_Zof signal Z, A is selected as A=10; the following values for P′ are convenient: P′=−1 for F_Z=144 MHz, P′=−0.5 for F_Z=152 MHz, P′=0 for F_Z=160 MHz, P′=+0.5 for F_Z=168 MHz, and P′=+1 for F_Z=176 MHz.

[0131] While P′ depends on K; the sequence of P(n) is distributed by quasi random, provided that the average P′ reaches the required value. If K changes, P′ changes too; if K remains unchanged, and P′ remains unchanged. This random distribution avoids noise at predetermined frequencies in the spectrum of X₂and hence also in Y. It is an advantage of the present invention that generator 100 is substantially stable after about 10 periods T₁of “low frequency” signal X₁. Compared to the prior art (cf. FIG. 1), that is an improvement by the factor 100.

[0132] (d)

[0133] FIG. 12 illustrates a simplified flow chart diagram of method 300 to operate multiplier 110 of FIG. 2. At start 301, multiplier 110 is enabled by providing power signal X₁to generator 100. As mentioned above, the stabilized time base of period T₁of signal X₁is available to multiplier 110 at substantially all the time.

[0134] The following steps 310, 320 and 330 are considered as calibration step 305 (dashed frame) of oscillator 150, preferably, executed by calibration unit 163.

[0135] In a first time frame, i.e. in step starting oscillator 310, oscillator 150 provides signal Z with an increasing frequency F_Zuntil F_Zreaches a predetermined minimum frequency F_{Z MIN}that allows detector 120 to operate properly. For example, F_{Z MIN}is half the rated frequency F_{Z RATED,}that is:

[0136] F_{Z MIN}=F_{Z RATED}/2  (64)

[0137] F_{Z MIN}=80 MHz (example)  (66)

[0138] In a second time frame, i.e. in step counting periods 320, calibration unit 163 counts periods T_Z, for example, within a single period T₁of signal X₁and thereby measures the frequency F_Z. This is possible, because period T₁is known; counting over more than one periods is also possible. During the operation of generator 100, frequency F_Zcan be measured at substantially any time, it is convenient to measure from time to time only.

[0139] In a third time frame, i.e., adjusting frequency 330, F_Zis altered such that F_Zreaches a predetermined rated value in the predetermined operation range F_{Z MIN}. . . F_{Z MAX}. Conveniently, calibration unit 163 instructs DAC 151 to alter control voltage V_Cuntil signal Z reaches the desired frequency. Preferably, calibration unit 163 uses a translation table having separate values for V_Cassigned to separate frequency values.

[0140] Performing steps 310, 320 and 330 as explained here is convenient, but not essential for the present invention. Persons of skill in the art are able to start and calibrate oscillator 150 otherwise, such as, for example, as explained in U.S. Pat. No. 5,774,006.

[0141] Having signal Z stabilized is the condition for the operation of divider 140. Leaving calibration 305, in the following, step 340 providing intermediate signal X₂, divider 140 provides signal X₂as described above.

[0142] During operation of divider 140 (substantially simultaneously with step 340), feedback unit 160, detector 110 and filter 130 provides signal K that represents a predetermined relation between signal X₂, intermediate clock signal Z and input reference clock signal X₁, Further, during operation, the filter coefficients of filter 130 can be adjusted (feedback loop).

[0143] In other words, method 300 to derive an output clock signal Y from input reference clock signal X₁comprises the following steps:

[0144] generating 305 intermediate clock signal Z;

[0145] providing 340 intermediate clock signal X₂by counting periods T_Zof intermediate clock signal z wherein the number A+P(n) of periods of intermediate clock signal Z in intermediate clock signal X₂varies and wherein the spectrum of intermediate signal X₂is being shaped;

[0146] dividing 350 intermediate clock signal X₂to feedback clock signal X₃at substantially the same frequency as input reference clock signal X₁;

[0147] calculating 360 numeric representation K of the phase difference between input reference clock signal X₁and feedback clock signal X₃by counting periods of intermediate clock signal Z;

[0148] using numeric representation K to adjust 370 the number of periods of intermediate clock signal Z in intermediate clock signal X₂; and

[0149] filtering 380 intermediate clock signal X₂to output signal Y with a transfer function that is inverse to the spectrum.

[0150] While the invention has been described in terms of particular structures, devices and methods, those of skill in the art will understand based on the description herein that it is not limited merely to such examples and that the full scope of the invention is properly determined by the claims that follow.