Dual Frequency Clock Recovery Using Common Multitap Line

This invention relates in general to telecommunication circuits, and more particularly to a circuit for recovering a clock signal in a telecommunication circuit.

In area of telecommunications, signals are sent between devices for long distances over communication cables and/or fibers. These signals are not accompanied by a clock signal; hence, the device receiving the data signal must generate its own clock (known in the art as "clock recovery"). This clock is used to retime the incoming signal.

Typically, data is transmitted between devices as return-to-zero ("RZ") data. With RZ data, a logic "1" data pulse returns to zero prior to the end of a clock period, as shown in FIG. 1a. In the telecommunications industry, RZ data typically returns to zero halfway through a clock period. Additionally, for telecommunications, RZ data is typically transmitted in a bipolar form, shown in FIG. 1b, where a logic "1" may be of positive or negative polarity. Usually, the bipolar signal is split into positive and negative data, shown in FIGS. 1c and 1d, at the receiving portion of the system. Clock recovery can be performed before or after the positive and negative data is recombined. Clock recovery is typically performed prior to operations such as B3ZS decoding, bipolar violations, and similar operations, such that these operations can be performed synchronously.

A second data form is non-return-to-zero (NRZ) data. With NRZ data, a logic "1" data pulse does not return to zero at any time during the period, as shown in FIG. 1e. NRZ data types are most commonly found internal to the device. Because NRZ data is not suited for long distance transmission, it is rarely transmitted between devices, with the exception of a fiber-optic transmission system. In most cases, NRZ data contains all of its information in a single data stream; however, there may be cases in which NRZ data may be split into more than one data stream.

One prior art clock recovery circuit uses a resonant (ringing tank) clock circuit to generate a clock signal in response to received data. Ringing tank circuits present several problems. First, in order to maintain the circuit in a "ringing" condition, a minimum repetition rate of a data pulse is necessary. Further, ringing tank circuits are subject to temperature variations which may result in a drift in frequency. In order to maintain the circuits, a considerable amount of testing and adjusting is necessary to keep the ringing tank tuned to the appropriate frequency.

A second clock recovery circuit uses an analog phase-locked loop (PLL). These devices present the usual problems of imprecision of a analog circuit relative to digital techniques. Accordingly, a number of digital PLL circuits have been proposed for clock recovery. One of these is shown in U.S. Pat. No. 4,819,251 to Nelson. In this patent, a digital clock recovery circuit is presented which uses a delay line to produce a plurality of delayed sample signals, which are then used to sample incoming data in a phase detector. The resultant sample data is then resampled by the tentatively correct apparatus clock output signal. The resampled data provides a direct indication of the phase difference between the data and the clock and the value can be obtained using a summing circuit. If the summed amount is outside an allowable range of values, a phase altering signal is applied to an oscillator to change the phase of the apparatus clock output signal. While circuits such as these overcome some of the problems associated with analog clock recovery circuits, such as precision tuning of the VCO or VCXO, phase adjustments, high cost and the need to use precision analog parts, these digital clock recovery circuits are not suited to clock recovery from RZ data and/or require high frequency clock sources in order to operate.

In some situations, it is highly desirable to recover a clock from signals of two different frequencies. DS3 and STS-1 signals, for example, are almost identical in their transmitted characteristics, each being RZ bipolar signals which return to zero halfway through the period. The DS3 has a nominal frequency of 44.736 MHz, while the STS-1 has a nominal frequency of 51.84 MHz. Since both of these signals occupy equivalent levels in the telecommunications hierarchy, it will be common for a particular piece of equipment to accommodate both types of data. Consequently, a clock recovery circuit which can recover a clock from both types of signals would be a great benefit to avoid duplicate circuitry for each type of signal. Heretofore, prior art clock recovery devices do not easily switch between signals of different frequencies.

Therefore a need has arisen in the industry for an accurate clock recovery system, which is capable of handling multiple frequency input signals, has a large retime margin, and is capable of handling both RZ and NRZ data inputs.

In one aspect of the present invention, a single clock recovery circuit generates a recovered clock signal from a data stream which may be transmitted in one of a plurality of communications types such as DS3, STS-1 or NRZ. Reference clock circuitry generates a reference clock signal responsive to a control signal indicating the communications type. Delay line circuitry generates a plurality of phased clock signals from the refernece clock of similar frequency to the data type reference clock. Based on the communications type, a set of said plurality of phased clocks sample the data stream to determine which of the phased clocks will first detect a predetermined logic level on the data stream. The recovered clock is generated by dynamically selecting one of said set of phased clocks responsive to the sampling.

The present invention provides significant advantages over the prior art. It is often desirable to use the same equipment to operate on multiple communications types, such as DS3 and STS-1, which are similar in their transmitted characteristics and which occupy equivalent .levels in the communications hierarchy. For efficiency, it is desirable to minimize the circuitry needed to recover the clock signal from the incoming data and to retime the data with the recovered clock. The present invention allows a single circuit to recover a clock from more than one type of data by selecting a set of phased clocks, the set depending upon the particular communications type, with which to compare the incoming data to obtain phase information. Further, the same circuit may be used elsewhere in a system to recover a clock from an NRZ data stream.

For a more complete understanding of the present invention, and the advantages thereof, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:

FIGS. 1a-e illustrate timing diagrams showing examples of RZ, bipolar RZ, RZ--P, RZ--N, and NRZ signals;

FIG. 2a illustrates a first embodiment of a circuit for recovering a clock signal from received data and retiming the received data;

FIG. 2b illustrates a multitap delay line for use in the circuit of FIG. 2a;

FIG. 3 illustrates a second embodiment of a circuit for recovering a clock signal from received data and retiming the data;

FIG. 4 illustrates a block diagram of the clock recovery circuitry used in the circuits of FIGS. 2a and 3;

FIG. 5 illustrates a schematic representation of the P.sub.-- DET circuitry used in the clock recovery circuit of FIG. 4;

FIG. 6 illustrates a timing diagram illustrating operation of the P.sub.-- DET circuitry of FIG. 5;

FIG. 7 illustrates a schematic representation of the N.sub.-- DET circuitry used in the clock recovery circuit of FIG. 4;

FIG. 8 illustrates a schematic representation of the hold circuitry used in the clock recovery circuit of FIG. 4;

FIGS. 9a-b illustrate the compare circuitry used in the clock recovery circuitry of FIG. 4;

FIG. 10 illustrates a schematic representation of counter circuitry used in the clock recovery circuitry of FIG. 4;

FIG. 11 illustrates a schematic representation of the PCS circuity used in the clock recovery circuitry of FIG. 4;

FIG. 12 illustrates a schematic representation of the data retime circuitry used in clock recovery circuitry of FIG. 4;

FIGS. 13a-b illustrate an alternative embodiment for the N--DET circuitry of FIG. 7 and

FIG. 14 illustrates an alternative embodiment for the data retime circuitry of FIG. 12.

FIGS. 2a-b and 3 illustrate first and second embodiments of a circuit for recovering a clock signal from incoming data and for retiming the incoming data with the recovered clock. In the first embodiment, shown in FIG. 2a, a circuit 10 comprises a clock recovery circuit 12 coupled to data lines 14 and a plurality of phased clocks on lines 16. The clocks (CLK1-CLK8) are generated by delay line circuitry 18, which is coupled to a clock (shown in FIG. 1 as either a DS3 or STS-1 clock) selected by control signal DS3.sub.-- STSF through multiplexer 20. The control signal DS3.sub.-- STSF is also coupled to the clock recovery circuitry 12. The clock recovery circuitry 12 outputs the recovered clock (RCLK) and the recovered data (RDATA).

Generation of the clocks corresponding to DS3 and STS-1 data types is well known in the art. Other clock frequencies could be provided as well, where appropriate.

In this embodiment, phased clocks are generated using an 8-tap fixed delay line 18, shown in greater detail in FIG. 2b. At each tap 22 of the delay line 24, the reference-clock (the output of multiplexer 20) is delayed an additional 2.75 nanoseconds (for the embodiment shown using DS3 or STS-1 data). All eight clock lines (CLK1-CLK8) are input to the clock recovery circuit 12 via lines 16. As described hereinbelow, the clock recovery circuit 12 uses the clocks output from delay line 18 to generate the RDATA and RCLK signals. When recovering a clock from DS3 data, a DS3 clock reference (i.e., a 44.736 MHz signal) is input to the delay line 24 and all eight phased clocks are used in the clock recovery circuit 12. On the other hand, when recovering a clock from STS-1 data, an STS-1 clock reference (i.e., a 51.84 MHz clock signal) is input into the delay line 18 and seven of the eight phased clocks are used by the clock recovery circuit 12. The decision whether to use seven or eight phase clocks is controlled by the DS3.sub.-- STSF control signal).

FIG. 3 illustrates a block diagram of a circuit 26 which comprises a second embodiment of the circuit for clock recovery and data retime. This embodiment is identical to the embodiment shown in FIG. 2a, with the exception that a phase-locked loop 28 is used to lock onto the output of multiplexer 20. In the preferred embodiment, there are two types of phase-locked loop circuits which may be used to implement PLL 28. The first type of phase-locked loop circuit using a VCO frequency that is approximately eight times the frequency of the recovered clock. The output of the phase-locked loop is then divided by eight using a Johnson counter. The Johnson counter provides the eight phased clocks which are input to the clock recovery circuit 12. One known disadvantage of this type of circuit is the possible undesirable effects due to the high frequency of the PLL 28, if the PLL 28 is located on the printed circuit board. The PLL 28 could be located, however, within the same device that contains the clock recovery circuit 12. In this situation, the negative effects of the high frequency VCO can be contained within the device.

A second type of PLL would use a VCO frequency which is about the same frequency as the data. This type of VCO would use a voltage controlled ring oscillator that is tapped at eight equidistant points around the ring. The eight taps would serve as the eight phased clocks to be input to the clock recovery circuit 12.

FIG. 4 illustrates a block diagram of the clock recovery circuit 12. Data is received into the clock recovery circuit 12 as PDT (positive data) and NDT (negative data). The PDT signals would be, for example, the positive RZ signals shown in FIG. 1c, while the NDT signals would be the negative RZ signals shown in FIG. 1d after inversion. It should be noted that the PDT (and NDT) data signals could be NRZ data as well. The PDT signals are input to positive level detection circuitry 30 (hereinafter P.sub.-- DET circuitry 30). The NDT signals are input to negative detection circuitry 32 (hereinafter N.sub.-- DET circuitry 32). The DS3.sub.-- STSF control signal is input to P.sub.-- DET circuitry 30 and N.sub.-- DET circuitry 32, as are eight clocks generated by the delay line 18 or PLL 28. The P.sub.-- DET circuitry 30 outputs positive retime data (inverted) PRTDT(1:8)F to data retime circuitry 34. Similarly, N.sub.-- DET circuitry 32 outputs negative retime data (inverted) NRTDT(1:8)F to data retime circuitry 34. NRTDT(1:8)F is also input to hold circuitry 36, along with the first clock (CLK1) from delay line circuitry 18 or PLL circuitry 28. The output of the hold circuitry 36 is phase data PH(1:8) and PH(1:8)F (the inverted PH(1:8) data), which is received by compare circuitry 38, along with the CLK1 signal. The compare circuitry 38 outputs a control signal, COUNT.sub.-- RSTF, to counter circuitry 40 and SELECT(1:8) signals to phase controlled selector circuitry 42 (hereinafter PCS 42). PCS 42 also receives a control signal LOAD from counter 40 and the eight phased clock signals CLK(1:8). The output of PCS 42 is the recovered clock signal RCLK which is input to the data retime circuit 34. The data retime circuit 34 outputs signals PRDATA and NRDATA, which are the retimed PDT and NDT signals, respectively.

The P.sub.-- DET circuitry 30 is shown in greater detail in connection with FIG. 5. The PDT signal is input to eight flip-flops 44.sub.1 -44.sub.8. Respective clock signals, CLK1-CLK8, are coupled to the clock input of the flip-flops 44.sub.1 -44.sub.8. The non-inverted output (Q) of each flip-flop 44.sub.1 -44.sub.8 is coupled to the input to a second stage flip-flop 46.sub.1 -46.sub.8. Clocks CLK1-CLK8 are input to the clock input of respective flip-flops 46.sub.1 -46.sub.8. A plurality of NOR gates 48.sub.1 -48.sub.8 have their outputs connected to the clear (CL) input of the respective flip-flops 46.sub.1 -46.sub.8. Each NOR gate 48.sub.1 -48.sub.8 has as its inputs the non-inverted outputs of each flip-flop 46.sub.1 -46.sub.8, except the flip-flop to which its outputs is coupled. Hence, NOR gate 48.sub.1 has as its inputs the non-inverted outputs of flip-flops 46.sub.2 -46.sub.8. Similarly, NOR gate 48.sub.2 has as its inputs the non-inverted outputs of flip-flops 46.sub.1 and 46.sub.3 -46.sub.8. The inverted output of each flip-flop 46.sub.1 -46.sub.8 is the PRTDT(n)F signal. Hence, the inverted output of flip-flop 46.sub.1 is the PRTDT1F signal and the inverted output of flip-flop 46.sub.2 is the PRTDT2F signal. The DS3.sub.-- STSF control signal is coupled to the clear input of flip-flop 44.sub.8.

In operation, the P.sub.-- DET circuitry 32 has two functions. The first function is that of a phase detector which identifies the first clock (CLK1-CLK8) to detect a logic high signal. The second function is to convert positive RZ data (assuming the input is RZ data) into positive NRZ data.

The operation of the P.sub.-- DET circuitry 30 is best understood in relation to FIG. 6, which shows a sample data stream PDT and the resultant output signals PRDT(1:8)F. The flip-flops 44.sub.1 -44.sub.8 latch their input on the rising edge of the respective clock signal. Hence, at point 50, (shown on the CLK1 timing diagram) the PDT signal is low and, hence, a zero is latched at the non-inverting output of flip-flop 44.sub.1. On the next low-to-high transition of CLK1, the zero is propagated to the non-inverting output of flip-flop 46.sub.1, as shown at point 52 of the PRDT1F signal. The two flip-flops 44.sub.1 and 46.sub.1, in series, provide conditioning against metastable anomalies that invariably occur in an asynchronous apparatus.

Each pair of flip-flops 44.sub.n and 46.sub.n thus sample the PDT data (twice for metastability considerations). The first pair of flip-flops to detect logic "1" in the data stream disables all of the other flip-flop pairs by activating the clear (reset) pins of the second flip-flop (46.sub.n) of the pair. Referring again to FIG. 6, at point 54, CLK3 detects a logical high value on the PDT signal. Hence, at point 56, the PRDT3F signal goes low. When the PRDT3F goes low, NOR gates 48.sub.1 -48.sub.2 and 48.sub.4 -48.sub.8 each output a logic "0" which resets the respective flip-flops 46.sub.1 -46.sub.2 and 46.sub.4 -46.sub.8. The flip-flops remain in the reset state until PRDT3F transitions to a logical high at point 56 (responsive to detection of a low value on the PDT signal at point 58). At this point, all flip-flops 46.sub.1 -46.sub.8 resume the sampling function. As shown at points 60 and 62, the CLK3 signal also detects a logical high on the PDT signal at point 60, resulting in a high-to-low transition of the PRDT3F signal at point 62, which resets the other flip-flops 46.sub.1 -46.sub.2 and 46.sub.4 -46.sub.8.

When the DS3.sub.-- STSF input is in the logic "1" state, the P.sub.-- DET circuitry 30 operates in conjunction with all eight clocks, CLK(1:8). When the DS3.sub.-- STSF control signal is in the logic "0" the P.sub.-- DET circuitry 30 uses only seven of the eight phases of the clock, i.e., CLK(1:7). This allows for clock recovery of two different frequencies, in this case, a DS3 signal and a STS-1 signal. In order to use the seven phase operation of the P.sub.-- DET, the phase clocks should be generated using a fixed, eight tap delay line which is external to the device. For the given example, each tap of the delay line should have a delay of approximately 2.75 nanoseconds. Thus, when a 44.736 MHz clock is input into the delay line, all eight phased clocks are used in the P.sub.-- DET circuitry 30, a clock can be recovered from DS3 data. Likewise, when a 51.84 MHz clock is input into the delay line and only seven phases are used in the P.sub.-- DET circuitry 30, the clock can be recovered from STS-1 data. In either case, the phased clocks are substantially equally spaced apart in terms of phase shift. It should be noted that other alternatives are possible. For example, if an 8.times. clock is divided by a Johnson divide-by-eight counter, eight equidistant phases are created. In this example, the full eight phases are used to perform clock recovery. Different clock frequencies can be recovered by changing the frequency of the 8.times. clock.

The N.sub.-- DET circuitry 32 is illustrated in FIG. 7. The NDT signal is input to eight flip-flops 64.sub.1 -66.sub.8. Respective clock signals, CLK1-CLK8, are coupled to the clock input of the flip-flops 64.sub.1 -66.sub.8. The noninverted output of each flip-flop 64.sub.1 -66.sub.8 is coupled to the input to a second stage flip-flop 66.sub.1 -66.sub.8. Clocks CLK1-CLK8 are input to the clock input of respective flip-flops 66.sub.1 -66.sub.8. A plurality of NOR gates 68.sub.1 -68.sub.8 have their outputs connected to the clear (CL) input of the respective flip-flops 66.sub.1 -66.sub.8. Each NOR gate 68.sub.1 -68.sub.8 has as its inputs the noninverted outputs of each flip-flop 66.sub.1 -66.sub.8, except the flip-flop to which its output is coupled. Hence, NOR gate 68.sub.1 has as its inputs the noninverted outputs of flip-flops 66.sub.2 -66.sub.8. Similarly, NOR gate 68.sub.2 has as its input the noninverted output of flip-flops 66.sub.1 and 66.sub.3 -66.sub.8. The inverted output of each flip-flop 66.sub.1 -66.sub.8 is the NRTDT(n)F signal. Hence, the inverted output of flip-flop 66.sub.1 is the NRTDT1F signal and the inverted output of flip-flop 66.sub.2 is the NRTDT2F signal.

The purpose of the N.sub.-- DET circuitry 32 is similar to that illustrated in connection with FIGS. 5-6 for the P.sub.-- DET circuitry 30. The N.sub.-- DET circuitry 32 detects the presence of a logic "1" in the NDT data stream and finds the phase relationship between the incoming data and the phased clocks CLK1-CLK8 (or CLK1-CLK7 if DS3.sub.-- STSF="0"). N.sub.-- DET circuitry 32 also provides conditioning against metastable anomalies. As with the P.sub.-- DET circuitry 30, the N.sub.-- DET circuitry 32 determines which phase clock CLK(1:8) retimes a logic "1" first from the NDT data stream. This is accomplished by allowing each phase clock to sample the data (twice for metastability considerations). The first sample operation to detect a logic "1" in the data stream disables all of the other sampling operations by activating their reset pins. The reset pins are held in the active state until a logic "0" is detected in the NDT data stream. This operation is repeated every time there is a low-to-high transition in the NDT data stream.

When the DS3.sub.-- STSF input is in the logic "1" state, the N.sub.-- DET circuitry 32 operates with all eight phase clocks. When the DS3.sub.-- STSF control signal is in the logic "0" state, the N.sub.-- DET circuitry 32 uses only seven of the eight phases of the clock. This allows for clock recovery of two different frequencies, as discussed in greater detail in connection with the P.sub.-- DET circuitry 30.

The hold circuitry 36 is illustrated in FIG. 8. The hold circuitry receives the NRTDT(1:8)F signals from the N.sub.-- DET circuitry 32. A first bank of flip-flops 70.sub.1 -70.sub.8 receive respective NRTDT(N)F signals. Hence, flip-flop 70.sub.1 receives the NRTDT1F signal and flip-flop 70.sub.2 receives the NRTDT2F signal. Each flip-flop 70.sub.1 -70.sub.8 is clocked using the CLK1 phased clock. The outputs of the flip-flops 70.sub.1 -70.sub.8 are connected to the inputs of a NAND gate 72. The output of NAND gate 72 is coupled to the select input of multiplexers 74.sub.1 -74.sub.8. The inverting outputs of multiplexers 74.sub.1 -74.sub.8 are coupled to respective inputs of flip-flops 76.sub.1 -76.sub.8. The inverted outputs of flip-flops 76.sub.1 -76.sub.8 are coupled to the A inputs of respective multiplexers 74.sub.1 -74.sub.8. The B input of multiplexers 74.sub.1 -74.sub.8 are connected to respective NRTDT(1:8)F signals. Hence, the B input of multiplexer 74.sub.1 is coupled to NRTDT1F through flip-flop 70.sub.1 and the B input of multiplexer 74.sub.2 is connected to NRTDT2F through flip-flop 70.sub.2. Flip-flops 76.sub.1 -76.sub.8 are clocked by the CLK1 signal. The non-inverted output of the flip-flop 76.sub.1 is the PH(n) signal and the inverted output of flip-flops 76.sub.1 -76.sub.8 is the PH(n)F signal. For example, flip-flop 76.sub.1 outputs the PH1 and PH1F signals.

In operation, the hold circuitry 36 stores new phase information during the interval between logic "1" bits on the NDT signal. The hold circuitry 36 aligns the incoming phase information with the leading phase clock (CLK1) and holds the new phase information during the time between logic "1" bits. By retiming the incoming phase information from the N.sub.-- DET circuitry 32 using the leading the phase clock (CLK1), the phase information is aligned so that other operations can be conducted synchronously. The retiming function is performed by flip-flops 70.sub.1 -70.sub.8. After the phase information is aligned, the hold circuitry determines whether a logic "1" or a logic "0" has been detected. If a logic "0" has been detected, the non-inverted output of all flip-flops 70.sub.1 -70.sub.8 will be "1", resulting in an output of "0" from the NAND gate 72. When the NAND gate 72 outputs a logic "0" the multiplexer passes the previous phase information from the outputs of flip-flops 76.sub.1 -76.sub.8. If a logic "1" is detected, the multiplexers pass the output of the flip-flops 70.sub.1 -70.sub.8 to flip-flops 76.sub.1 -76.sub.8, respectively, thereby updating the phase information.

FIG. 9 illustrates a schematic diagram of the compare circuitry 38. The compare circuitry receives the PH(1:8) and PH(1:8)F Signals from the hold circuitry 36. NAND gates 78.sub.1 -78.sub.8 are coupled to the PH(n) and PH(n)F signals such that the current phase from the hold circuitry 36 can be determined. A table of possible phase information signals is given below in TABLE I.

For a given phase, one of the NAND gates 78.sub.1 -78.sub.8 will output a "0" while the remaining gates will output a "1". Corresponding NAND gates 80.sub.1 -80.sub.8 are coupled to the non-inverted and inverted outputs of flip-flops 82.sub.1 -82.sub.8, which are coupled to the PH(1:8) signals and clocked by the CLK1 signal. Hence, the flip-flops 82.sub.1 -82.sub.8 store the previous phase information. NAND gates 80.sub.1 -80.sub.8 are coupled to the non-inverting and inverting signals of the flip-flops 82.sub.1 -82.sub.8 in the same manner that NAND gates 78.sub.1 -78.sub.8 are coupled to the PH(1:8) and PH(1:8)F signals. The non-inverted outputs of the flip-flops 82.sub.1 -82.sub.8 comprise the SELECT(1:8) signals.

The outputs of respective NAND gates 78.sub.1 -78.sub.8 and 80.sub.1 -80.sub.8 are coupled to OR gates 84.sub.1 -84.sub.8. Hence, OR gate 84.sub.1 has the output of NAND gate 78.sub.1 and the output of NAND gate 80.sub.1 as its inputs and OR gate 84.sub.2 has the outputs of NAND gates 78.sub.8 and 80.sub.2 as its inputs. The outputs of OR gates 84.sub.1 -84.sub.8 are coupled to the inputs of NAND gate 86. The output of NAND gate 86 comprises the COUNT.sub.-- RSTF signal. The output of COUNT.sub.-- RSTF signal indicates whether the phase information output from the hold circuitry 36 on a current cycle is the same as the phase information output from the hold circuitry 36 on the previous clock cycle (using CLK1 as the reference). If the phase information is the same, then the inputs to one of the OR gates 84.sub.1 -84.sub.8 will both be a logic "0", thereby causing the COUNT.sub.-- RSTF to be a logic "1". On the other hand, if the phase information has changed from one clock period to the next, none of the OR gates 84.sub.1 -84.sub.8 will have both inputs set to a logic "0" and therefore COUNT.sub.-- RSTF will transition to a logic "0".

The purpose of the compare circuitry 38 is to insure that the phase information from the hold circuitry is valid and not just a transient which has resulted from a metastable anomaly. Hence, the COUNT.sub.-- RSTF signal indicates when there is a change in phase. This information is received by the counter circuitry 40. The last phase information (SELECT(1:8)) is sent to the PCS circuitry 42.

FIG. 10 illustrates a schematic diagram of the counter circuitry 40. The counter circuitry 40 receives the CLK1 signal and the COUNT.sub.-- RSTF from the compare circuitry 38 on FIG. 9. The counter circuitry 40 outputs a LOAD signal which is the inverted CNT.sub.-- HOLDF signal. In the embodiment shown in FIG. 10, the counter is a linear feedback shift register counter, although other types of counters would be equally suitable. The counter circuitry 40 comprises 7 pairs of respective multiplexer 88.sub.1 -88.sub.7 and flip flops 90.sub.1 -90.sub.7. Each multiplexer 88.sub.1 -88.sub.7 receives the LOAD1S control signal as an input to one of its select inputs. The LOAD1S signal is the output of a NAND gate 92 having the COUNT.sub.-- RSTF and ALLZEROS control signals as inputs. The ALLZEROS control signal is the output of NAND gate 94. The second select input of the multiplexers 88.sub.1 -88.sub.7 is coupled to the CNT.sub.-- HOLDF signal, which is the output of NAND gate 96 (the HOLD signal is the inverted CNT.sub.-- HOLDF signal). Each multiplexer has a first input coupled to the output of its associated flip-flop (i.e., multiplexer 88.sub.1 receives the output of flip-flop 90.sub.1 as its first input and multiplexer 88.sub.2 receives the output of flip-flop 90.sub.2 as its first input). The second input to multiplexor 88.sub.1 is coupled to the X6X7FEEDBK control signal, which is the output of exclusive-OR Gate 98. The remaining multiplexers 88.sub.2 -88.sub.7 receive the non-inverted output of the preceding flip-flop as their second input. In other words, multiplexer 88.sub.2 receives the non-inverted output of flip-flop 90.sub.1 and multiplexer 88.sub.3 receives the non-inverted output of flip-flop 90.sub.2 as its second input. The third and fourth inputs to the multiplexers 88.sub.1 -88.sub.7 are coupled to a logical high signal. Each flip-flop 90.sub.1 -90.sub.7 is clocked by the CLK1 signal and receives as its input the output of its associated multiplexer. Hence, flip-flop 90.sub.1 receives the output of multiplexer 88.sub.1. NAND gate 94 receives the inverted output of each flip-flop 90.sub.1 -90.sub.7. The eighth input to NAND gate 94 is coupled to electrical high signal. NAND gate 96 is coupled to either the inverted or non-inverted outputs of the flip-flops 90.sub.1 -90.sub.7. The configuration of inputs to NAND gate 96 determines the value up to which the counter circuitry 40 will count. Exclusive OR gate 98 is coupled to the non-inverted outputs of flip-flops 90.sub.6 and 90.sub.7.

In operation, the counter 40 has 128 unique states, with the state "0000000" being an illegal state. If counter assumes the illegal state, the ALLZEROS control signal will transition to a logical low, thereby driving the LOAD1S signal high, which will reset the counter to the state "1111111". Similarly, if the COUNT.sub.-- RSTF signal is a logical "0" the LOAD1S control signal will transition to a logical "1" which will again reset the state of the counter to "1111111". Otherwise, while COUNT.sub.-- RSTF is a logical "0" the counter will continue to count responsive to the CLK1 signal until the inputs to NAND gate 96 cause the CNT.sub.-- HOLDF signal to transition to a logical "0". When CNT.sub.-- HOLDF is a logical "0" the non-inverting output of the flip-flops 90.sub.1 -90.sub.7 will pass through the respective multiplexers 88.sub.1 -88.sub.7 until the COUNT.sub.-- RSTF transitions to a logical "0" indicating a new phase.

The purpose of the counter circuitry 40 is to count the number of consecutive phase matches and output a LOAD indication when a predetermined number of matches have occurred. This insures that the current phase information is not just a transient which has resulted from a metastable anomaly. In general, the counter circuitry 40 increments every time COUNT.sub.-- RSTF is a logical "1". If COUNT.sub.-- RSTF is a logical "0", the counter is reset to the initial condition. If n matches occur, where n is a predetermined number of matches chosen to indicate the current phase information is not transient, the LOAD output of the counter circuitry 40 is activated to a logic "1". The output will remain at logic "1" as long as the counter continues to receive the same phased information. The value of n is based on the expected frequency offset between the data and phased clocks.

FIG. 11 illustrates a schematic diagram of the PCS circuitry 42. The purpose of the PCS circuitry 42 is to use the phase information from the compare block 38 in order to select one of the phased clocks (CLK1-CLK8) as the recovered clock without the generation of glitches during transitions between phased clocks.

The select signals (SELECT1-SELECT8) are received from the compare circuitry 38 and are coupled to one input of respected multiplexers 100.sub.1 -100.sub.8. The LOAD signal from the counter circuitry 40 is coupled to the selector input to each of the multiplexers. The outputs of multiplexers 100.sub.1 -100.sub.8 are coupled to the input of respective flip-flops 102.sub.1 -102.sub.8. The non-inverted outputs of flip-flops 102.sub.1 -102.sub.8 are coupled to the other input of respective multiplexers 100.sub.1 -100.sub.8. Flip-flops 102.sub.1 -102.sub.8 are clocked by the CLK1 signal. The non-inverted outputs of flip-flops 102.sub.1 -102.sub.8 are also coupled to the input of a respective flip-flop 104.sub.1 -104.sub.8, although there is no respective flip-flop 104.sub.3 corresponding to flip-flop 102.sub.3). Flip-flops 104.sub.1 and 104.sub.2 are clocked using the CLK5 signal, flip-flop 104.sub.4 is clocked using the CLK2 signal, flip-flop 104.sub.5 is clocked using the CLK3 signal, flip-flop 104.sub.6 is-clocked using the CLK4 signal, flip-flop 104.sub.7 is clocked using the CLK5 signal and flip-flop 104.sub.8 is clocked using the CLK6 signal. The non-inverted output of flip-flop 104.sub.1 is coupled to OR gate 106.sub.1 which has as its other input the non-inverted output of flip-flop 102.sub.1. Similarly, OR gate 106.sub.2 is coupled to the non-inverted output of flip-flops 102.sub.2 and the non-inverted output of flip-flop 104.sub.2. NAND gate 108.sub.1 receives as its input the output of OR gate 106.sub.1 and the CLK1 signal. NAND gate 108.sub.2 receives the output of OR gate 106.sub.2 and the CLK2 signal. NAND gate 108.sub.3 receives as its inputs the non-inverted output of flip-flop 102.sub.3 and the CLK3 signal. NAND gate 108.sub.4 receives as its inputs the non-inverted output of flip-flop 104.sub.4 and the CLK4 signal. NAND gate 108.sub.5 receives its input the output of flip-flop 104.sub.5 and the CLK5 Signal. NAND gate 108.sub.6 receives its inputs the non-inverted output of flip-flop 104.sub.6 and the CLK6 signal. NAND gate 108.sub.7 receives its inputs the non-inverted output of flip-flop 104.sub.7 and the CLK7 signal. NAND gate 108.sub.8 receives its inputs the non-inverted output of flip-flop 104.sub.8 and the CLK8 signal. The outputs of NAND gates 108.sub.1 -108.sub.8 are coupled to the inputs of NAND gate 110. The output of NAND gate 110 is coupled to inverter 112. The output of inverter 112 is the RCLK signal.

In operation, of the PCS circuitry 42 accepts and holds new phase information, i.e., the SELECT1-8 signals when the LOAD signal is a logic "1". The phase information, stored in flip-flops 102.sub.1 -102.sub.8 is used to select one of the phased clocks as the recovered clock. The stored phase information is skewed using an associated phase clock, such that the phase information is retimed with a phased clock that rises during the time when the phase clock to be selected as the recovered clock is low. For example, the SEL4 signal, which is the non-inverted output of flip-flop 102.sub.4, is skewed using CLK2. The CLK2 signal will transition high at a point where the CLK4 signal is low. This prevents the creation of any small pulses, or glitches, which could give false readings when used to retime the data.

This concept of skewing the phase of information works for all but the first two phase positions, i.e., CLK1 and CLK2. These exceptions occur because the phase information that cannot be retimed when the phased clock (CLK1 or CLK2) to be selected as a recovered clock is low. In these cases, the respective OR gate 106.sub.1 or 106.sub.2 allows the low-to-high transition of the phase information output from the respective flip-flop 102.sub.1 or 102.sub.2 to have an immediate effect while the high to low transition of the phase information is delayed by half a bit period. The net result of the skewing operation is that one of the phase clocks is glitchlessly selected as the output from the PCS circuitry 42 as the recovered clock.

The data retime circuitry 34 is shown in FIG. 12. The retime circuitry 34 receives the PRTDT(1:8)F data from the P.sub.-- DET circuitry 30 and the NRTDT(1:8)F data from N.sub.-- DET circuitry 32. This data is retimed using the recovered clock (RCLK) signal from the PCS circuitry 42. The PRTDT1F-PRTDT8F signals are all input to NAND gate 114. Referring then to FIG. 6, only one of the PRTDT(1:8)F signals may be low at any one time; hence the output of NAND gate 114 is the retimed PDT signal in NRZ format. Similarly, the eight NRTDT(1:8)F signals are input to NAND gate 116 resulting in an NRZ output of the retimed NDT signal. The output of NAND gates 114 and 116 are retimed with the RCLK signal from PCS circuitry 42, using flip-flops 118 and 120, respectively. The output of flip-flop 118 is the PRDATA signal, which is the retimed NRZ equivalent of the PDT signal. The output of flip-flop 120 is the NRDATA signal, which is the retimed NRZ equivalent of the NDT signal. The PRDATA and NRDATA signals may be combined for a unitary signal or maintained separately.

FIGS. 13 and 14 illustrate alternative embodiments which can be used for increased transient and under/over-shoot problems, respectively. FIG. 13 illustrates circuit for the N.sub.-- DET circuit 32 of FIG. 7, which requires that a pulse be present at a logical high level for at least two successive phased clocks in order to enable the NRTDT signal. In this embodiment, pairs of flip-flops 64, which are clocked by successive clock signals, are coupled to the inputs of a respective AND gate. Hence, the outputs of flip-flops 64.sub.1 and 64.sub.2 are coupled to AND gate 122.sub.1, the outputs of flip-flops 64.sub.2 and 64.sub.3 are coupled to the inputs of AND gate 122.sub.2 and the outputs of flip-flops 64.sub.3 and 64.sub.4 are coupled to the inputs of AND gate 122.sub.3. The output of AND gates 122.sub.1 -122.sub.8 are coupled to the input of respective flip-flops 124.sub.1 -124.sub.8. Each flip-flop 124 is coupled to a respective clock signal CLK1-CLK8. The outputs of flip-flops 124.sub.1 -124.sub.8 are coupled to the inputs of respective flip-flops 66.sub.1 -66.sub.8. A multiplexer 126 receives the outputs of flip-flops 64.sub.7 and 64.sub.8 and passes one of the inputs to AND gate 122.sub.7 responsive to the DS3.sub.-- STSF signal. The remainder of the circuitry is the same as that shown in FIG. 7.

In operation, an AND gate 122 is enabled only if a logical high voltage is detected by two successive flip-flops 64. Thus, to enable AND gate 122.sub.1, the noninverted outputs of flip-flops 64.sub.1 and 64.sub.2 must both detect a logical high on the NDT signal in order for AND gate 122.sub.1 to output a logical high signal to flip-flop 124.sub.1. Since flip-flops 64.sub.1 and 64.sub.2 are clocked on CLK1 and CLK2, respectively, the-logical high signal on NDT must be at a logical high for at least two successive phased clocks in order for a logical high to be output from flip-flop 66.sub.1. Consequently, a transient with a width less than the difference between two successive clocks will not register as a logical high signal on NDT.

Multiplexer 126 switches the input to AND gate 122.sub.7 such that the output of flip-flop 64.sub.1 is coupled to the second input of AND gate 122.sub.7 when seven phased clocks are used and the output of flip-flop 64.sub.8 is coupled to AND gate 122.sub.7 when eight phased clocks are used.

While the above transient-protection circuitry is described in relation to the N.sub.-- DET circuitry 32, similar circuitry could be used for the P.sub.-- DET circuitry 30 to prevent transients from effecting the proper detection of signals.

FIG. 14 illustrates an alternative embodiment of circuitry to prevent under/over-shoots on the PDT and NDT signals which result as false data at the output of the data retime circuit 34. The circuit of FIG. 14 is similar to the circuit of FIG. 12, with the exception that the outputs of NAND gates 114 and 116 are coupled to respective AND gates 130 and 132, respectively. Further, the output of NAND gate 114 is coupled to the input of inverter 134, the output of which is coupled to the second input of AND gate 132. The output of NAND gate 116 is coupled to the input of inverter 136, the output of which is coupled to the second input of AND gate 130. The output of AND gate 130 is coupled to the input of flip-flop 118 and the output of AND gate 132 is coupled to the input of flip-flop 120. The effect of the AND gates 130 and 132 and inverters 134 and 136 is to prevent the PR data output and the NR data output to be simultaneously in a high logic state. When the output of NAND gate 114 is high, the output of inverter 134 is driven low, thereby driving the output of AND gate 132 low. Similarly, when the output of AND gate 116 is logical high, the output of inverter 136 is logical low, thereby driving AND gate 130 low.

The present invention provides advantages over the prior art. Importantly, the clock recovery system has the ability to adjust to phase and frequency differences whereas most clock recovery schemes that recover a clock from a data stream have only phase detection capabilities. The circuitry can recover a clock from a data stream whose frequency varies widely relative to the frequency of the-phased clocks and can be adapted to almost any frequency by changing the choice of delays and reference frequencies. The clock can be recovered from NRZ data and RZ data and can convert positive and negative RZ data into positive and negative NRZ data. Further, the circuitry can recover a clock from two different data frequencies, such as DS3 and STS-1 using a single multitap delay line.

Although the Detailed Description of the Preferred Embodiment has been directed to certain exemplary embodiments, various modifications of the embodiments, as well as alternative embodiments would be known to those skilled in the art. For example, different logic configurations could be used to perform the same functions as the logic described throughout. The invention encompasses any modifications or alternative embodiments that fall within the scope of the claims.