Avoiding very low duty cycles in a divided clock generated by a frequency divider

The instant patent application is related to and claims priority from the provisional India patent application entitled “Clocking Scheme for Delta Sigma Modulator Working with a Fractional Divider Of Variable Length”, Serial No.: 201841022886, Filed: 19 Jun. 2018, which is incorporated in its entirety herewith to the extent not inconsistent with the description herein.

Embodiments of the present disclosure relate generally to frequency dividers, and more specifically to a technique for avoiding very low duty cycles in a divided clock generated by a frequency divider.

Frequency dividers are well-known in the relevant arts, and refer to electronic circuits that divide the frequency of an input clock to generate an output clock. The factor by which a frequency divider divides the frequency of the input clock may be an integer (e.g., 100) or a fraction (e.g., 100.6).

Duty cycle refers to the ratio of the logic high duration to the period of a periodic waveform. It may be desirable at least in some application environments to avoid very low duty cycles of a clock output generated by a frequency divider. One reason to avoid such low duty cycles is that circuits that receive the clock output of a frequency divider may not operate reliably when the duty cycle of the clock output is very low.

Several aspects of the present disclosure are directed to avoiding low duty cycles in a clock output by a frequency divider.

Example embodiments of the present disclosure will be described with reference to the accompanying drawings briefly described below.

FIG. 1 is a block diagram of an example device in which several aspects of the present disclosure can be implemented.

FIG. 2A is a diagram of a prior frequency-dividing unit.

FIG. 2B is a timing diagram illustrating the operation of a frequency-dividing unit in an embodiment of the present disclosure.

FIG. 3A is a block diagram of a prior frequency divider.

FIG. 3B is a timing diagram illustrating the operation of a prior frequency divider.

FIG. 4 is diagram of a frequency-dividing unit in an embodiment of the present disclosure.

FIG. 5 is a block diagram of a frequency divider in an embodiment of the present disclosure.

FIG. 6A is a diagram illustrating some portions of a frequency divider in an embodiment of the present disclosure.

FIG. 6B is a block diagram of a block that generates divider-select signals in an embodiment of the present disclosure.

FIG. 7 is a timing diagram illustrating the waveforms at various nodes of a frequency divider in an embodiment of the present disclosure.

FIG. 8A is a block diagram of a one-shot used in a frequency divider in an embodiment of the present disclosure.

FIG. 8B is a timing diagram illustrating the operation of a one-shot in an embodiment of the present disclosure.

In the drawings, like reference numbers generally indicate identical, functionally similar, and/or structurally similar elements. The drawing in which an element first appears is indicated by the leftmost digit(s) in the corresponding reference number.

1. Overview

A frequency divider includes a set of frequency-dividing units coupled in series in a sequential order, with the sequence of frequency-dividing units including a lowest unit and a highest unit, with the remaining units being disposed in series between the lowest unit and the highest unit. The lowest unit is coupled to receive an input clock whose frequency is to be divided and provided as an output clock. Each frequency-dividing unit in the set is coupled to receive a corresponding first clock as an input and is operable to generate a corresponding second clock as an output. The frequency divider includes a logic block to generate a first set of edges of the output clock synchronous with the input clock. The logic block is designed to generate a second set of edges of the output clock synchronous with the output clock of a highest operative frequency-dividing unit in the set.

Several aspects of the present disclosure are described below with reference to examples for illustration. However, one skilled in the relevant art will recognize that the disclosure can be practiced without one or more of the specific details or with other methods, components, materials and so forth. In other instances, well-known structures, materials, or operations are not shown in detail to avoid obscuring the features of the disclosure. Furthermore, the features/aspects described can be practiced in various combinations, though only some of the combinations are described herein for conciseness.

2. Example Device

FIG. 1 is a block diagram of an example device in which a frequency divider implemented according to several aspects of the present disclosure can be used. Phase locked loop (PLL) 100 (which can be used as a frequency synthesizer) of FIG. 1 is shown containing phase frequency detector (PFD) 110 , charge pump 120 , low-pass filter (LPF) 130 , voltage controlled oscillator (VCO) 140 , frequency divider 150 , frequency divider 160 , delta-sigma modulator (DSM) 170 and logic block 180 . PLL 100 may be implemented as an integer-only PLL or a fractional PLL (as noted below), and may be implemented in integrated circuit (IC) form. Several aspects of the present disclosure are described below with respect to frequency divider 160 presented in the context of a PLL. However, it is to be understood that such aspects can be realized in other contexts such as a stand-alone frequency divider, as well as in frequency divider 150 .

Frequency divider 160 receives an input clock 169 (Clk-in) from an external device (not shown), divides the frequency of Clk-in according to either signal 176 received from DSM 170 (when fractional division is desired) or signal 186 received from logic block 180 (when only integer division is desired), and generates an output clock Fref ( 161 ).

Logic block 180 receives a user input (e.g., from a user or a processing block, not shown) on path 181 , with the user input representing the number (divide ratio) by which Clk-in is to be divided to generate Fref. The term “divide ratio” is used herein to refer generally to the ratio of the divider's input clock's frequency and the divider's output clock's frequency. When the divide ratio to be applied by frequency divider 160 is a fraction (e.g., 100.23), logic block 170 forwards the fractional portion (e.g., 0.23) of the user input (on path 181 ) to DSM 170 on path 187 , and the integer portion of the number (e.g., 100) to frequency divider 160 on path 186 . When the divide ratio is an integer, logic block forwards the integer value to frequency divider 160 on path 186 .

DSM 170 generates (in one of several known ways) a sequence of divide values corresponding to (or representing) the fractional part of the divide ratio received on path 187 . DSM 170 forwards the numbers in the sequence successively (one number per cycle of Fref, with the sequence repeating after the last number in the sequence is forwarded) to frequency divider 160 on path 176 . More specifically, DSM 170 forwards on path 176 the next number in the sequence noted above at the end of the immediately previous cycle of Fref, which is indicated by the falling edge of Fref (which is also provided to DSM 170 as an input) to enable DSM 170 to forward the next number in the sequence to frequency divider 160 .

Frequency divider 160 determines the divide ratio per cycle (i.e., the divide ratio to be obtained in each cycle) of Fref by adding the inputs received on paths 187 and 176 corresponding to the cycle. Alternatively, such addition may be performed in a separate block, not shown, which would then provide the sum to frequency divider 160 . Thus, when fractional division of Clk-in is desired, frequency divider 160 successively divides Clk-in by values in a sequence, such that the effective average frequency of Fref equals the desired fraction of Clk-in. When integer-only division is desired, DSM 170 is not implemented, and logic block 170 forwards the integer number received on path 181 to frequency divider 160 on path 186 , the integer number itself representing the divide ratio. In an alternative embodiment, the input received on path 181 represents a desired output frequency (for Fref), and logic block 170 computes the corresponding divide ratio based on the input.

PFD 110 receives as inputs, Fref on path 161 and a feedback clock Ffb on path 151 , and operates to generate error signals UP and DOWN on respective paths 112 U and 112 D. The ON (active) durations of error signals UP and DOWN are respectively proportional to the amount of phase by which Fref leads or lags Ffb.

Charge pump 120 converts the UP and DOWN outputs of PFD 110 to a corresponding charge on path 123 . While signals UP and DOWN have been noted as being applied to charge pump 120 , signals derived from UP and/DOWN (e.g., logical inverse of the signals) may instead be applied to charge pump 120 depending on the specific design of charge pump 120 . LPF 130 is a low-pass filter and rejects frequency variations at node 123 above a certain cut-off limit. LPF 130 converts the low-pass-filtered charge to a voltage 134 .

VCO 140 generates an output signal Fvco on path 145 , with the frequency of Fvco being determined by the (instantaneous) magnitude of voltage received on path 134 . Fvco is typically a square wave or sinusoidal wave, and may be used by other systems (not shown) as a clock signal (after appropriate processing or conditioning, if so desired).

Frequency divider 150 receives Fvco as an input, divides the frequency of Fvco by a divide value (which may be different from the divide ratio used by frequency divider 160 ), and provides the frequency-divided signal as a feedback clock Ffb on path 151 . The divide value may be programmed in frequency divider 150 by a user from an external device, not shown.

Frequency divider 160 may be required to support a very wide range of divide ratios. As an example, frequency divider 160 may need to bring a frequency range of 2.1 gigahertz (GHz) to 8 Kilohertz (KHz) to a range 8 megahertz (MHz) to 8 KHz, which translates to a range of divide ratios of about 1 to 250000. Such a requirement may be due to the manner in which components of PLL 100 (e.g., PFD 110 ) are implemented, in that some of such components can operate reliably only over the smaller frequency range.

Prior frequency dividers are associated with one or more drawbacks at least when used to support the very wide range of divide ratios noted above. One drawback is that when the frequency of CLK-in is very large (e.g., of the order of Megahertz) and the specific divide ratio is small (e.g., of the order of 8 or 16), the prior frequency dividers would generate the output clock (Fref) with a duty cycle that is very small. As noted above, DSM 170 needs to forward a next number in the sequence representing the divide ratio to frequency divider 160 at the end of an immediately previous cycle of Fref, as indicated for example, by the falling edge of Fref. When the duty cycle of Fref is very small, DSM 170 may not be able to reliably forward the next number at the correct time instant to frequency divider 160 , thereby resulting in errors/imperfections in Fref.

Several aspects of the present disclosure are directed to a frequency divider which overcomes the drawback(s) noted above, as described in detail below, and will be clearer in comparison with a prior approach. Accordingly, the prior approach and the building blocks used to implement a frequency divider in such a prior approach are first described next.

3. Prior Frequency-Dividing Unit

FIG. 2A is a block diagram of a frequency-dividing unit that is used in building a programmable frequency divider according to a prior approach. Two or more of units as unit 200 of FIG. 2A may be cascaded (connected in series) to obtain a frequency divider which supports a desired range of divide ratios. In FIG. 2A , frequency-dividing unit 200 is a divider which can divide an input signal by 2 or 3, as controlled by corresponding signals, described below. Only a brief description of unit 200 (with reference also to the timing diagram of FIG. 2B ) is provided herein, as the design of such unit is well-known in the relevant arts. For more details of unit 200 , the reader is referred to the lecture notes available from MIT OpenCourseWare at http://ocw.mit.edu/cours es/electrical-engineering-and-computer-science/6-976-high-speed-communication-circuits-and-systems-spring-2003/lecture-notes/lec14.pdf.

Frequency-dividing unit 200 receives an input clock Clk-in ( 201 ), and generates an output clock Clk-out ( 299 ) whose frequency is either half or one-third of the frequency of Clk-in 201 . Other signals shown in FIG. 2A and/or FIG. 2B include nc ( 235 ) and Mod-in ( 209 ), Mod-out ( 211 ), n 12 ( 234 ), div 3 b ( 267 ), n 22 ( 289 ) and clkob ( 291 ). Nc is a control signal (which is a binary signal) provided as input to frequency-dividing unit 200 , and which determines if frequency-dividing unit 200 is to be allowed to divide by 3 at all (or divide by 2 only). Mod-in 209 is a mode control signal received from another unit (not shown, but typically the next higher unit when multiple ones of unit 200 are cascaded), and determines when (i.e., at what time instant) frequency-dividing unit 200 is to divide by 3. Mod-out ( 211 ) is a mode control signal generated by frequency-dividing unit 200 , and may be connected as an input to a previous stage in a cascade (not shown). The other signals noted above are signals internal to frequency-dividing unit 200 .

FIG. 2B is a timing diagram illustrating the operation of frequency-divider unit 200 of FIG. 2A . As may be observed from FIG. 2B , in the interval t 21 to t 23 , frequency-dividing unit 200 operates as a divide-by-2 unit, while in the interval t 23 to t 25 frequency-dividing unit 200 operates as a divide-by-3 unit. In interval t 22 -t 23 , signals nc and Mod-in are high, and cause frequency-dividing unit 200 to divide by 3 in the interval between t 23 and t 25 . Mod-in transitions to logic low at t 23 , and thereafter, nc transitions to logic low at t 24 and remains at logic low. In response, frequency-dividing unit 200 operates as a divide-by-2 unit from t 25 .

It may be observed from FIG. 2B that in intervals in which nc is at logic low, unit 200 divides by 2 irrespective of the value of Mod-in. To cause unit 200 to divide by 3, both nc and Mod-in should be held at logic high, and the system consequently performs a divide by 3 by skipping an input clock (CLk-in) cycle as shown in FIG. 2B . By appropriate control of nc and Mod-in, frequency-dividing unit 200 can be made to operate as a divide-by-2 or divide-by-3 unit.

By connecting multiple ones of units such as a frequency-dividing unit 200 in series (i.e., in a cascade), corresponding ranges of divide ratios between an input clock and an output clock are achieved. As an example, a cascade of three units such as frequency-dividing unit 200 forming a frequency divider 300 is shown in FIG. 3A , and the timing diagram of 3 B shows the corresponding waveforms when a divide ratio of 15 is desired. In general, frequency divider 160 adds the integer portion received on path 186 and the DSM output received on path 176 to obtain the bit values of Nc- 2 , Nc- 1 and Nc- 0 (of FIG. 3A ). Frequency divider 160 internally contains the logic circuits (not shown) for generating Nc- 2 , Nc- 1 and Nc- 0 (each of which is a binary signal) based on the values received on paths 186 and 187 , and such logic may be implemented in a known way. The values of signals Nc- 2 , Nc- 1 and Nc- 0 determine the specific divide ratio to be used by divider 300 . Input terminals 307 , 317 and 327 together represent the “divide input” of divider 300 . Unit 200 - 1 is the ‘lowest’ (first) unit, while 200 - 3 is the ‘highest’ unit.

In FIG. 3A , three units, each identical to frequency-dividing unit 200 , are used in a cascade. Clk- 0 ( 309 ) is the input signal, and is provided as input to the first unit 200 - 1 . Unit 200 - 1 receives as input a mode control signal 308 (Mod- 1 ) from the next (higher) stage (or unit) 200 - 2 , and a control signal 307 (Nc- 0 ). Depending on the value of Nc- 0 , unit 200 - 1 performs only divide-by-2 operation, or divide-by-3 once in a divide cycle and divide-by-2 the rest of the divide cycle. A divide cycle refers to one period of the output clock (e.g., Out-Clk 333 in FIG. 3B . Unit 200 - 1 provides a divided clock Clk- 1 ( 312 ) and a mode control signal 311 (Mod- 0 ) as outputs. 2 k

Unit 200 - 2 receives Clk- 1 ( 312 ) as an input clock. Unit 200 - 2 receives as input a mode control signal 318 (Mod- 2 ) from the next (higher) stage 200 - 3 , and a control signal 317 (Nc- 1 ). Depending on the value of Nc- 1 , unit 200 - 2 performs only divide-by-2 division or divide-by-3 once in a divide cycle and divide-by-2 the rest of the divide cycle. Unit 200 - 2 provides a divided clock Clk- 2 ( 322 ) and mode control signal 308 (Mod- 1 ) as outputs.

Unit 200 - 3 receives Clk- 2 ( 322 ) as an input clock. Unit 200 - 3 , being the ‘highest’ unit, the mode control signal 328 (Mod- 3 ) provided as input to unit 200 - 3 is tied to logic high. Unit 200 - 3 receives a control signal 327 (Nc- 2 ). Depending on the value of Nc- 2 , unit 200 - 3 performs only divide-by-2 division, or divide-by-3 once in a divide cycle and divide-by-2 the rest of the divide cycle. Unit 200 - 3 generates a divided clock Clk- 3 ( 332 ) and mode control signal 318 (Mod- 2 ) as outputs.

In the example of FIG. 3A , it is assumed that each of Nc- 0 , Nc- 1 and Nc- 2 is set to logic high. Therefore, each of stages/units 200 - 1 , 200 - 2 and 200 - 3 divides-by-3 once in a divide cycle. Unit 200 - 1 divides-by-3 in interval t 31 -t 32 , unit 200 - 2 divides-by-3 in interval t 31 -t 33 , and unit 200 - 3 always divides by 3. The values for the waveforms shown in FIG. 3B repeat after t 34 . Clk- 3 has a frequency that is 1/15 of that of Clk- 1 . The general expression that specifies the divide ratios that can be obtained by divider 300 (with 3 stages) is: [2 3 +(Nc- 2 ) 2 2 +(Nc- 1 )2 1 +(Nc- 0 )2 0 ]. By appropriate selection of Nc- 2 , Nc- 1 and Nc- 0 , a divide ratio in the range 8 to 15 (both inclusive) can be obtained. Nc- 0 , Nc- 1 and Nc- 2 are each generated by a logic unit (not shown), but contained within frequency divider 300 based on the divide ratio (here 15) to be applied. In general, with a cascade of ‘k’ units/stages, a range of [2 k , 2 k+1 −1] can be obtained. In the example of FIG. 3A , for a divide ratio of 15, each of Nc- 2 , Nc- 1 and Nc- 0 is a logic high.

Referring again to FIG. 3B , although Clk- 3 has the desired frequency ( 1/15 of that of Clk- 0 ), in a prior implementation of divider 300 , instead of Clk- 3 an output clock derived from Mod- 0 , Clk- 0 and the highest Mode signal that is guaranteed to be operational (here Mod- 2 ) is provided as the output clock 333 (Out-Clk). One reason for not using Clk- 3 as the final output is that the jitter associated with clock outputs of higher stages may be unacceptable. In FIG. 3B , Out-Clk ( 333 ) represents the output clock provided by divider 300 , with the falling edges of Out-Clk being generated by the combination of Mod- 0 and Clk- 0 , and the rising edges of Out-Clk being generated by the rising edge of Mod- 2 , or equivalently the falling edge of Clk- 3 .

The highest mode signal that is guaranteed to be operational (and not a fixed binary value) depends on the number of stages in the cascade of units of the frequency divider that are required to support a particular divide range and therefore that would be operational. When divider 300 is used to generate a divide range [8, 15], Mod- 2 is the highest mode signal that is operational (i., active, not fixed). However, when divider 300 is used to generate a divide range of [4, 5] for example, only stages 200 - 1 and 200 - 2 would be operational, and Mod- 2 would be tied to logic high (and thus fixed at one logic level) and therefore cannot be used to generate the rising edge of the output clock. Hence, Mod- 1 , which is the highest mode signal that is guaranteed to be operational (not fixed at any logic state), would have to be used for generating the rising edge of the output clock even when all three stages/units 200 - 1 through 200 - 3 are operational, as in the example of a divide ratio of 15 of FIG. 3A . It may be appreciated that the duty cycle of the resulting output clock would be relatively smaller. In general, the dynamic use of the highest Mode signal guaranteed to be operational in the prior approach to generate the rising edges of the output clock forces the use of a lower mode signal (e.g., Mod- 1 rather than Mod- 2 ) even when a higher mode signal is available for use. Consequently, the prior approach may suffer from the drawback that when the divide ratio to be obtained is very low, the duty cycle of the output clock may be very small.

One consequence of low duty cycle of the output clock is when such clock is used to indicate a required action to another circuit component. Referring to FIG. 1 , it may be observed that output clock 161 (Fref) is provided as an input to DSM 170 to indicate to DSM 170 when to start a next divide cycle. If the duty cycle of Fref is very small, the logic high duration of Fref is also small, and consequently Fref may not be able to cause DSM to reliably trigger the next divide cycle.

Several aspects of the present disclosure overcome the problem noted above by avoiding low duty cycles in the output clock, as described next with examples.

3. Divide-by ⅔ Unit

FIG. 4 is a schematic diagram illustrating the implementation details of a frequency dividing unit implemented according to several aspects of the present disclosure. Frequency-divider unit 400 is shown containing AND gates 410 , 420 , 440 and 470 , latches 430 , 450 , 480 and 490 , inverters 460 and 495 , and multiplexer (MUX) 480 . AND gates 410 , 420 , 440 and 470 , latches 430 , 450 , 480 and 490 , and inverters 460 and 495 are respectively identical to AND gates 210 , 220 , 240 and 270 , latches 230 , 250 , 280 and 290 , and inverters 260 and 295 of FIG. 2A , and together operate similar to the circuit of FIG. 2A and the timing diagram of FIG. 2B , and the description is not repeated here in the interest of conciseness.

Frequency-dividing unit 400 receives an input clock Clkin ( 401 ), and generates an output clock Clkout ( 499 ) whose frequency is either half or one-third of the frequency of Clkin 401 . Other signals shown in FIG. 4 include nc ( 435 ), Modin ( 409 ), Modout ( 411 ), n 12 ( 434 ), div 3 b ( 467 ), n 22 ( 489 ) and clkob ( 491 ). Nc is a control signal (which is a binary signal) provided as input to frequency-dividing unit 400 , and which determines if frequency-dividing unit 400 is to be allowed to divide by 3 at all (or divide by 2 only). Modin 409 is a mode control signal received from another unit (not shown, but typically the next higher unit when multiple ones of unit 400 are cascaded), and determines when (i.e., at what time instant) frequency-dividing unit 400 is to divide by 3. Modout ( 411 ) is a mode control signal generated by frequency-dividing unit 400 , and may be connected as an input to a previous stage in a cascade (not shown). The other signals noted above are signals internal to frequency-dividing unit 400 . When unit 400 is the rightmost (highest) unit in a cascade, Clkout ( 499 ) is connected to fb-rst-in ( 498 ), and Modin ( 409 ) is tied to logic high.

MUX 480 receives a select signal (divsel 482 ), and forwards on path 481 (as signal fb-rst-out) the corresponding one of signals Clkin 401 and fb-rst-in 498 based on the value of select signal divsel. In the example of FIG. 4 , MUX 480 forwards Clkin as output fb-rst-out when divsel 482 is a logic 0, and forwards fb-rst-in as output fb-rst-out when divsel is a logic 1. Several of units such as frequency-divider unit 400 may be cascaded to form a divider providing a desired range of divide ratios. A divider with three cascaded units, each implemented similar to unit 400 , is shown in FIG. 5 . In general, with a cascade of ‘k’ units/stages, a range of [2k, 2k+1 −1] can be obtained.

The description below is provided in the context of a frequency divider with three frequency-dividing units merely as an illustration, and frequency divider 500 can have more or fewer number of frequency-dividing units 400 . The specific number of units 400 used is generally determined by the overall range of divide ratios that need to be supported by frequency divider 500 . FIG. 5 is a block diagram of a frequency divider 500 built using three units/stages of unit 400 . Unit 500 - 1 is the ‘lowest’ (first) unit, while 500 - 3 is the ‘highest’ unit. The input clock, output clock, input mode control signal, output mode control signal, feedback input and feedback output signals of each of units 500 - 1 , 500 - 2 and 500 - 3 are respectively received/transmitted on the corresponding terminals Clkin ( 401 ), Clkout ( 499 ), Modin ( 409 ), Modout ( 411 ), fb-rst-in ( 498 ) and fb-rst-out ( 481 ) of the respective divider units 400 .

Ck 0 ( 509 ) is the input clock, and is provided as input to the first (leftmost or lowest) unit 500 - 1 . Unit 500 - 1 receives as inputs a mode control signal 508 (Md 1 ) and feedback signal 591 (Fb 1 ) from the next (higher) stage/unit 500 - 2 , a control signal 507 (Nc 0 ) and a ‘divider-select’ (divsel) signal divsel 0 . Unit 500 - 1 is operational when divsel 0 is a logic high (logic one), and not-operational when divsel 0 is a logic low (logic zero). Depending on the value of Nc 0 , unit 500 - 1 performs only divide-by-2 division, or divide-by-3 once in a divide cycle and divide-by-2 the rest of the divide cycle. A divide cycle refers to one period of the output clock (e.g., Out-Ck 671 in FIG. 7 ) of feedback divider 500 , and is also equal to the period of the clock output (e.g., Ck 3 ) of the rightmost frequency-divider unit. Unit 500 - 1 provides a divided clock Ck 1 ( 512 ), a mode control signal 511 (Md 0 ) and feedback signal 590 (Fb 0 ) as outputs. Signal Fb 1 is received on the fb-rst-in ( 498 in FIG. 4 ) terminal of unit 500 - 1 and Fb 0 is transmitted on the fb-rst-out ( 481 in FIG. 4 ) terminal of unit 500 - 1 .

Unit 500 - 2 receives Ck 1 ( 512 ) as an input clock. Unit 500 - 2 receives as inputs a mode control signal 518 (Md 2 ) and a feedback signal 592 (Fb 2 ) from the next (higher) stage 500 - 3 , a control signal 517 (Nc 1 ) and divsel 1 . Unit 500 - 2 is operational when divsel 1 is a logic high (logic one), and not-operational when divsel 1 is a logic low (logic zero). Depending on the value of Nc 1 , unit 500 - 2 performs only divide-by-2 division or divide-by-3 once in a divide cycle and divide-by-2 the rest of the divide cycle. Unit 500 - 2 provides a divided clock Ck 2 ( 522 ), mode control signal 508 (Md 1 ) and feedback signal 591 (Fb 1 ) as outputs. Signal Fb 2 is received on the fb-rst-in ( 498 in FIG. 4 ) terminal of unit 500 - 2 and Fb 1 is transmitted on the fb-rst-out ( 481 in FIG. 4 ) terminal of unit 500 - 2 .

Unit 500 - 3 receives Ck 2 ( 522 ) as an input clock. Unit 500 - 3 , being the ‘highest’ unit, the mode control signal 528 (Mod- 3 ) provided as input to unit 500 - 3 is tied to logic high. Unit 200 - 3 receives a control signal 527 (Nc 2 ), a feedback signal 593 (Fb 3 ) and divsel 2 as inputs. Since unit 500 - 3 is the highest unit, Ck 3 is connected to Fb 3 . Unit 500 - 3 is operational when divsel 2 is a logic high (logic one), and not-operational when divsel 1 is a logic low (logic zero). Depending on the value of Nc 2 , unit 500 - 3 performs only divide-by-2 division, or divide-by-3 once in a divide cycle and divide-by-2 the rest of the divide cycle. Unit 500 - 3 generates a divided clock Ck 3 ( 532 ), mode control signal 518 (Md 2 ) and feedback signal 592 (Fb 2 ) as outputs.

The general expression that specifies the divide ratios that can be obtained by divider 500 (with 3 stages) is: [2 3 +(Nc 2 )2 2 +(Nc 1 )2 1 +(Nc 0 )2 0 ]. By appropriate selection of Nc 2 , Nc 1 and Nc 0 , a divide ratio in the range 8 to 15 (both inclusive) can be obtained.

Control signals Nc 0 , Nc 2 and Nc 2 are each generated as a logic high by a logic unit (not shown, but contained within frequency divider 500 ) based on the divide ratio (here 15) to be applied. In the example of FIG. 5 , it is assumed that each of Nc 0 , Nc 1 and Nc 2 is set to logic high. Therefore, each of stages/units 500 - 1 , 500 - 2 and 500 - 3 divides-by-3 once in a divide cycle.

Divider-select signals divsel 0 , divsel 1 and divsel 2 are generated by divsel generator 690 (shown in FIG. 6B , and contained within divider 500 ) based on Nc 0 , Nc 1 and Nc 2 . The logic implemented in divsel generator 690 to generate the divider-select signals employs the following logic:

If divsel<x>=1, then divsel<x−1> equals 1; divsel<x>=1 if (divsel<x+1>=1) OR nc<x+1>=1; wherein x represents values 0, 1, 2, 3, etc., and ‘1’ represents a logic high, and ‘OR’ represents a logic OR operation.

To illustrate for the case of divide ratio of 15, with each of Nc 0 , Nc 1 and Nc 2 being 1, each of divsel 0 , divsel 1 and divsel 2 is also 1 according to the logic noted above. With each of divsel 0 , divsel 1 and divsel 2 being a 1, Ck 3 is forwarded as Fb 0 , as may be verified from an inspection of FIG. 4 and FIG. 5 .

It is noted here that even though clock Ck 3 has the desired frequency ( 1/15 of Ck 0 ), Ck 3 is not used as the output clock of divider 500 for the reason that Ck 3 may exhibit unacceptable jitter. Instead, the output clock is generated based on Ck 0 , Md 0 and Fb 0 . FIG. 6A is a block diagram of a circuit portion of frequency divider 500 that is employed to generate output clock Out-Ck ( 671 ) based on signals Fb 0 , Ck 0 and Md 0 . The implementation details of FIG. 6A merely represent one example manner of implementation. Other circuits and/or connections can also be used to achieve the same effect as the circuit portion of 6 A, as would be apparent to one skilled in the relevant arts. FIG. 6A is shown containing one-shot 610 , NMOS (N-channel metal oxide semiconductor) transistors 620 and 630 , inverters 640 , 650 , 660 and 670 , and AND gate 680 . One-shot 610 receives Fb 0 and generates Fb-rst ( 611 ). AND gate 680 generates the result Ck 0 .Md 0 ( 681 ) of a logical AND operation on CK 0 and Md 0 . Inverters 660 and 670 are used for delaying signal 646 to generate the output clock Out-Ck ( 671 ). The combination of inverters 640 , 650 transistors 620 and 630 represents a latch, and can be replaced by corresponding latch circuits, well known in the relevant arts. Nodes 611 and 681 may be viewed respectively as the set and reset inputs of the latch, while nodes 645 and 646 may be viewed respectively as Q and Q-bar outputs of the latch.

FIG. 8A is a block diagram illustrating the implementation details of one-shot 610 in an embodiment of the present disclosure, and is shown containing inverters 820 and 840 , delay unit 810 and AND gate 830 . Inverter 840 generates the logical inverse of Fb 0 on path 841 . Delay unit 810 provides a fixed delay to signal 841 to generate delayed signal 812 . Inverter 820 generates the logical inverse of signal 812 on path 823 . AND gate 830 generates the result ( 611 ) of a logical AND operation on signals 823 and 841 . FIG. 8B is a timing diagram illustrating the operation of one-shot 610 . The manner in which a positive pulse is generated on Fb-rst based on Fb 0 is illustrated there.

FIG. 7 shows the waveforms of various nodes of frequency divider 500 of FIG. 5 and the circuits of FIG. 6A . Unit 500 - 1 divides-by-3 in interval t 71 -t 72 , unit 500 - 2 divides-by-3 in interval t 71 -t 73 , and unit 500 - 3 always divides by 3. The values for the waveforms shown in FIG. 5 repeat after t 76 . Ck 3 has a frequency that is 1/15 of that of Ck 0 . Since each of divsel 0 , divsel 1 and divsel 2 is a logic high for a divide ratio of 15, Ck 3 propagates through units 500 - 3 , 500 - 2 and 500 - 1 , and is provided on terminal Fb 0 as the feedback signal output of unit 500 - 1 .

With combined reference now to FIG. 6A and FIG. 7 , the falling edge of Fb 0 (also the same as Ck 3 in the example) results in a positive pulse being generated by one-shot 610 at t 34 . Therefore, at t 34 , transistor 620 is switched ON for the duration of the pulse width of Fb-rst ( 611 ), thereby pulling node 645 ( set ) to ground, and node 646 (reset) to logic high. Thus, at t 34 , a rising edge of Out-Ck is generated. The logical AND of CK 0 and Md 0 generates a positive pulse starting at t 75 , with the pulse causing transistor 630 to be turned ON for the duration of the pulse width of signal 681 . As a result, at t 75 node 646 (reset) is pulled to ground and node 645 (set) is set to logic high, and a falling edge of Out-Ck is generated at t 75 . Time instance t 76 represents the end of a divide cycle. The waveforms of FIG. 7 repeat for subsequent divide cycles (each equal to the duration t 71 -t 76 ), and Out-Ck is generated with a frequency equal to 1/15 of that of Ck 0 .

The techniques described above enable the generation of an output clock (Out-Ck) whose falling edges are always generated based on Ck 0 and Md 0 , and therefore synchronous with the input clock Ck 0 . The term ‘synchronous’ implies that the falling edges of Out-Ck occur at an edge of Ck 0 . However, the term ‘synchronous’ also covers instances when the falling edges of Out-Ck is triggered by Ck 0 and occur within a timing window of one cycle of Ck 0 . Due to being generated synchronous with Ck 0 (which has the least jitter of all the clocks in FIG. 5 ), the falling edges therefore exhibit very low jitter (than for example if the highest clock Ck 3 had been directly used as the output clock).

The rising edges of the output clock (Out-Ck) are generated based on (and synchronous with) the highest operational clock signal (here Ck 3 ), i.e., based on the clock output of the highest operational unit (here unit 500 - 3 ). Again, the term ‘synchronous’ also covers instances when the falling edges of Out-Ck is triggered by Ck 0 and occur within a timing window of 1 cycle of the highest operational clock signal (Ck 3 in the example). It may be observed from the timing diagram of FIG. 7 that the use of Ck 3 in the example (of a divide ratio being 15) renders the positive (logic high) durations of the output clock (Out-Ck) to have an acceptably long duration than, for example, if the output (e.g., Md 1 ) of a lower unit been used.

In general, the use of frequency divider unit 400 in the manner illustrated with respect to divider 500 of FIG. 5 enables an output clock to be generated with acceptable duty cycles due to the use of the highest active/operational clock signal to generate the rising edges of the output clock. In particular, the use of MUX 480 in each frequency-divider unit, along with the corresponding connections between the units as illustrated in FIG. 5 , ensures that the highest available clock signal (rather than a fixed/hardwired signal such as that in the prior technique) is always used to generate the rising edges.

Depending on the specific divide ratio used in divider 500 , the highest operational clock (clock generated by the rightmost frequency-divider unit) among clocks Ck 1 , Ck 2 and Ck 3 would be forwarded as Fb 0 . For example, if the divide ratio is such that only units 500 - 1 and 500 - 2 are operational (for example for a divide ratio of 4), then divsel 2 would be a logic zero, while divsel 0 and divsel 1 would be a logic high. Hence, Ck 2 would be forwarded as Fb 2 by unit 500 - 3 (which is now not used except to forward Ck 2 on node Fb 2 ). Since both divsel 0 and divsel 1 are logic high, Ck 2 would be forwarded as Fb 0 . As a result, the rising edges of the output clock Out-Ck would now be generated at falling edges of Ck 2 (whose period now would also equal the divide cycle).

When used in place of frequency divider 160 of FIG. 1 , a frequency divider implemented according to various aspects of the present disclosure, as described in detail above, may provide an output clock (as Fref of FIG. 1 ) with an acceptably long duty cycle so as not to affect the operation of DSM 170 in reliably forwarding the next number at the correct time instant to frequency divider 160 .

4. Conclusion

References throughout this specification to “one embodiment”, “an embodiment”, or similar language means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, appearances of the phrases “in one embodiment”, “in an embodiment” and similar language throughout this specification may, but do not necessarily, all refer to the same embodiment.

While in the illustrations of FIGS. 1, 4, 5, 6A, 6B and 8A , although terminals/nodes are shown with direct connections to (i.e., “connected to”) various other terminals, it should be appreciated that additional components (as suited for the specific environment) may also be present in the path, and accordingly the connections may be viewed as being “electrically coupled” to the same connected terminals. In the instant application, power and ground terminals are referred to as constant reference potentials.

While various embodiments of the present disclosure have been described above, it should be understood that they have been presented by way of example only, and not limitation. Thus, the breadth and scope of the present disclosure should not be limited by any of the above-described embodiments, but should be defined only in accordance with the following claims and their equivalents.