Patent Grant
7196562
The present invention relates to integrated circuit devices and, more particularly, to integrated circuit devices having clock generators therein and methods of generating clock signals.
Phase-locked loop (PLL) and delay-locked loop (DLL) integrated circuits are frequently used to generate highly accurate internal clock signals on integrated circuit substrates (e.g., chips). In some cases, such PLL and DLL integrated circuits may be used in multi-output clock drivers, which may provide clock distribution to a system integrator in order to optimize the timing of high-performance computer and communication systems. One conventional PLL-based clock buffer that provides only limited phase adjustments is the RoboClock™ CY7B994V, which is manufactured by Cypress Semiconductor Corp. This clock buffer is described more fully in Cypress' publicly available datasheet (Doc. No. 38-07127), which may be obtained at Cypress' website.
Unfortunately, such clock buffers may provide only partial-period clock skew control to those clock signal outputs having frequencies equal to divide-by-two or lower (i.e., divide-by-N (Div/N), where N is a positive integer greater than one), relative to an on-chip divide-by-one synchronizing clock signal. As illustrated by the timing diagram of
Clock signal generators according to the present invention support generation of clock signals with full-period programmable skew characteristic. According to a first embodiment of the present invention, an integrated circuit chip is provided with a clock driver therein. This clock driver supports generation of a plurality of output clock signals having different frequencies in a range between 1 and 1/N times a frequency of an internal clock signal, where N is a positive integer greater than one. These output clock signals also have full-period programmable skew characteristics. The internal clock signal is generated by an internal clock signal generator, which may be a locked loop integrated circuit. The locked loop integrated circuit may be a phase-locked loop (PLL) integrated circuit having a voltage-controlled oscillator (VCO) therein. This voltage controlled oscillator may be programmably set to operate at a frequency within a frequency range (e.g., 200 MHz–800 MHz) that is supported by the clock driver. Alternatively, the locked loop integrated circuit may be a delay-locked loop (DLL) integrated circuit, which receives an externally generated clock signal that sets the frequency of the internal clock signal.
In particular, the clock driver is configured to support generation of a divide-by-N clock signal having a full-period programmable skew characteristic that is stepped in N×M time units having a duration equal to 1/M times a period of the internal clock signal, where M is a positive integer greater than eight. In some embodiments, M=C×F, where C and F are positive integers, which represent a number of coarse time units (ctu) of skew adjustment and fine time units (ftu) within each coarse time unit, respectively. To achieve a sufficient degree of skew control, the values of C and F may equal six (6) and four (4), respectively. Thus, a divide-by-8 clock signal may be set to one of 192 skew values (192=8×6×4), which span a full period. In other embodiments, the internal clock signal generator comprises a differential voltage-controlled oscillator (DVCO) having at least ½C stages therein.
These embodiments further include circuitry that enables full-period clock skew control in a preferred manner. This circuitry includes a divide-by-N clock generator that is responsive to a first skew signal (which may be selected in response to coarse and fine skew select signals) and a synchronization unit. The synchronization unit is electrically coupled to an output of the divide-by-N clock generator and is responsive to the first skew signal. The divide-by-N clock generator is configured to generate N separate divide-by-N clock signals that have the same frequency but are phase shifted relative to each other in equal phase increments. A one-of-N select circuit is also provided within the divide-by-N clock generator. This select circuit is configured to select one of the N divide-by-N clock signals in response to a time unit position signal. The synchronization unit is configured to synchronize the selected divide-by-N clock signal to the first skew signal.
The clock driver may also include a delay chain and phase interpolator circuit that is configured to generate the first skew signal in response to a fine skew select signal. This delay chain and phase interpolator circuit is electrically coupled to an output of a multiplexer, which may be configured to receive a plurality of skew signals from the voltage-controlled oscillator within a PLL. The multiplexer may also be responsive to a coarse skew select signal.
According to other embodiments of the present invention, the clock driver includes a locked loop integrated circuit that is configured to generate a plurality of internal clock signals that are skewed in time relative to each but have the same first frequency (e.g., VCO frequency). A skew signal select circuit is also provided. This skew signal select circuit is configured to generate a selected skew signal from at least one of the plurality of internal clock signals. A divide-by-N clock generator is also provided, which is responsive to the selected skew signal. This divide-by-N clock generator is configured to generate a plurality of divide-by-N clock signals that have the same frequency but are phase shifted relative to each other in increments of 360°/N, where N is a positive integer greater than one. This divide-by-N clock generator includes a one-of-N select circuit that is configured to select one of the plurality of divide-by-N clock signals, in response to a time unit position signal. A synchronization unit is electrically coupled to an output of the one-of-N select circuit. This synchronization unit is configured to synchronize a selected one of the plurality of divide-by-N clock signals to the selected skew signal. The synchronized signal is then passed to an output buffer. The output buffer is provided to drive an off-chip load with a divide-by-N output clock signal having a full-period programmable skew characteristic, in response to a synchronized divide-by-N clock signal that is derived directly or indirectly from the synchronization unit.
In some of these embodiments, the skew signal select circuit includes a multiplexer that is configured to receive the plurality of internal clock signals and a “fine” delay chain and phase interpolator circuit that is electrically coupled to an output of the multiplexer. Alternatively, the skew signal select circuit may include a multiplexer that is configured to receive the plurality of internal clock signals and a phase interpolator circuit that is electrically coupled to a pair of outputs of the multiplexer. In this latter case, the skew signal select circuit may not require the use of a “fine” delay chain.
Further embodiments of the present invention include methods of generating a clock signal having a full-period programmable skew characteristic, by generating a plurality of internal clock signals that have equivalent frequencies, but are phase shifted relative to each other in equal time increments. At least one of the plurality of internal clock signals is then selected, in response to a coarse skew select signal. A selected skew signal is then generated from the at least one selected plurality of internal clock signals. A plurality of divide-by-N clock signals are generated in response to the selected skew signal. The plurality of divide-by-N clock signals, which are of equal frequency, are phase shifted relative to each other in increments of 360°/N, where N is a positive integer greater than one. One of these plurality of divide-by-N clock signals is selected in response to a time unit position signal. The selected divide-by-N clock signal is then synchronized to the selected skew signal. In still further embodiments, a pair of differential divide-by-N output clock signals are generated from the synchronized divide-by-N clock signal.
Still further embodiments of the present invention include a packaged integrated circuit device having a nonvolatile memory device and a programmable clock driver circuit therein. The nonvolatile memory device may be provided as an electrically erasable programmable read-only memory (EEPROM) device that is disposed on a first integrated circuit substrate (e.g., semiconductor chip) and the programmable clock driver circuit may be disposed on a second integrated circuit substrate. The first and second integrated circuit substrates may be packaged side-by-side in an integrated circuit package. In alternative embodiments, the nonvolatile memory device and programmable clock driver circuit may be provided as side-by-side circuits on a single integrated circuit chip.
The programmable clock driver circuit includes a control circuit and at least one clock generator therein. According to preferred aspects of these embodiments, the control circuit is configured to detect an error(s) in configuration data that is used by the programmable clock driver circuit. This configuration data is read from the nonvolatile memory and stored in volatile program registers during program restore operations. The control circuit is further configured to automatically idle the at least one clock generator in response to detecting the error in the configuration data. This automatic idling of the clock generator prevents the clock generator from becoming configured in a manner that could potentially damage or alter the performance of downstream devices receiving clock signals generated by the programmable clock driver circuit. This automatic idling may include operations to set the clock generator at a default setting (e.g., minimum frequency), which applies to all output banks of the driver circuit. In alternative embodiments, a detection of an error may result in a “local” disabling of only a single output bank to which the error in the configuration data applies. The clock generator typically includes a phase-locked loop (PLL) integrated circuit that synchronizes the timing of the clock signals generated at the output banks.
The control circuit performs the error detection operations by reading the configuration data from the volatile program registers and generating a first error check word from the configuration data. These operations are performed in response to a program save command (ProgSave) issued to the integrated circuit device. This first error check word may be generated using a cyclic redundancy check (CRC) algorithm to generate a binary number. The control circuit completes the operations associated with the program save command by programming the nonvolatile memory with the configuration data and the first error check word. Thereafter, in response to a program restore command (ProgRestore) issued to the driver circuit, the control circuit reads the configuration data and the first error check word from the nonvolatile memory device and also generates a second error check word from this configuration data. This second error check word is then compared against the first error check word to determine whether a difference exists. The existence of a difference indicates that an error is present in the configuration data that has been read from the nonvolatile memory device.
In some cases, the program restore command may result in a writing of erroneous data into the volatile program registers, however, this erroneous data will be blocked from influencing performance of the programmable clock driver circuit. Instead, the detection of an error in the configuration data will automatically cause the driver circuit to be held in a default mode of operation that precludes generation of an active lock signal at an output pin of the packaged device. This default mode of operation may include holding the PLL integrated circuit at a minimum clock frequency and precluding phase lock by the PLL integrated circuit. Precluding phase lock may operate to inform another “master” device that the driver circuit (e.g., “slave” device) needs to be reprogrammed to correct the detected error. Such reprogramming operations may be performed in response to issuance of a program write command and new configuration data by the master device. Alternatively, the master device may issue a program restore command to the driver circuit, which causes another internal downloading of the configuration data from the nonvolatile memory to the volatile registers. Still further operations performed by the control circuit may include automatically performing error correction operations on the configuration data once an error has been detected during a program restore operation. In such cases, the error check word may be generated using a Hamming code that supports at least single bit error correction.
The present invention now will be described more fully herein with reference to the accompanying drawings, in which preferred embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Like reference numerals refer to like elements throughout and signal lines and signals thereon may be referred to by the same reference characters. Signals may also be synchronized and/or undergo minor boolean operations (e.g., inversion) without being considered different signals. The suffix B (or prefix symbol “/”) to a signal name may also denote a complementary data or information signal or an active low control signal, for example.
Referring now to
The VCO 18 is illustrated as having a plurality of differential stages, which are each responsible for generating a pair of skew signals. Each pair of skew signals includes a true skew signal and a complementary skew signal, which is 180° out-of-phase relative to the corresponding true skew signal. Thus, three pairs of differential skew signals result in six skew signals (6 SKEWS) that are “coarsely” separated from each other by 60°. As illustrated, these skew signals are provided to a skew signal select circuit, which is configured to generate a plurality of selected skew signals (SEL-SKEW1 to SEL-SKEW6). The skew signal select circuit is illustrated as including an analog multiplexer 20, a coarse skew select circuit 22, a fine skew select circuit 26 and a plurality of delay chain and phase interpolator circuits 24a–24f. Other embodiments of a skew signal select circuit are also possible.
The coarse and fine skew select circuits 22 and 26 may have programmable features that support field programming of the clock driver 100 by an end user. For example, non-volatile memory (not shown) may be provided within a packaged clock driver chip to support automatic updating of skew configuration data during each power-down/up or other reset event. The coarse skew select circuit 22 is configured to generate a plurality of coarse skew select signals, which are received by the multiplexer 20. In response to the coarse skew select signals, the multiplexer 20 routes one or more of the skew signals to a plurality of outputs of the multiplexer 20. In some applications, a single “coarsely” selected skew signal may be routed (i.e., broadcast) to a plurality of outputs of the multiplexer 20.
Each of the coarsely selected skew signals is provided to one or more of the delay chain and phase interpolator circuits 24a–24f. The delay chain and phase interpolator circuits 24a–24f are responsive to respective fine skew select signals, which are generated by the fine skew select circuit 26. The delay chain and phase interpolator circuits may be of conventional design and need not be described further herein. In the illustrated embodiment, four time units of fine skew control are provided by each of the delay chain and phase interpolator circuits 24a–24f and six time units of coarse skew control are provided by the coarse skew select circuit 22 and multiplexer 20. Accordingly, each of the selected skew signals SEL-SKEW1 to SEL-SKEW6 may be skewed in fine time increments of Tvco/24, where Tvco represents the period of the voltage-controlled oscillator 18. These increments are more fully illustrated by
The selected skew signals of
As illustrated by the timing diagram of
Accordingly, the signals generated by the coarse and fine skew select circuits 22 and 26 and the time unit position signal TU-POSITIONn can be programmed to collectively provide a selected divide-by-N clock signal having a desired skew characteristic. Because the time unit position signal TU-POSITIONn is freely programmable over the N possible values, a programmable full period skew characteristic can be achieved at the output of each divide-by-N clock generator 30a–30f. In
This full period skew characteristic is more fully illustrated by TABLE 1, where the voltage-controlled oscillator (VCO) frequency is illustrated as ranging from 200 MHz to 800 MHz and the frequency of clock signals generated at outputs of the clock driver 100 are illustrated as ranging from 16.67 MHz to 800 MHz. The duration of each coarse time unit (ctu) equals one-sixth the period of the voltage-controlled oscillator and the duration of each fine time unit (ftu) equals one-fourth the duration of the coarse time unit. Thus, in the illustrated embodiment, the total number of fine time units across a full period of skew control equals 24×N. Other embodiments having a different number of time units across a full period are also possible.
The full-period shew characteristics associated with selected divide-by-2 (Div/2) and divide-by-3 (Div/3) clock signals are further illustrated by the timing diagrams of
The range of full-period skew characteristics associated with the following clock signals: Div/2, Div/3, Div/4, Div/5, Div/6, Div/8, Div/10 and Div/12, are more fully illustrated by TABLE 2.
Referring again to
Referring now to
Notwithstanding this disclosure of preferred techniques to generate clock signals having full-period skew characteristics, less preferred clock) drivers may be configured to generate clock signals having less than full-period skew characteristics. For example, the divide-by-N clock generators 30a–30f may be configured to generate only N−1 phase-shifted divide-by-N clock signals for those cases where N is greater than two. Thus, in the event N=12, the divide-by-N clock generator 30a may be configured to generate 11 divide-by-N clock signals so that a corresponding output clock signal having a 0.916 times full-period clock skew may be achieved using a one-of-(N−1) select circuit, for N greater than two. Alternatively, in the event N=3, the divide-by-N clock generator 30a may be configured to generate 2 divide-by-N clock signals so that a corresponding output clock signal having a 0.667 times full-period clock skew may be achieved using the one-of-(N−1) select circuit.
Further embodiments of the present invention include packaged clock driver circuits that utilize nonvolatile memory devices to support automatic reprogramming operations during power-up/reset events and enable restoration of a prior configuration profile(s). As illustrated by
The programmable clock driver circuit 210 includes a control circuit and at least one clock generator therein. The control circuit is illustrated in
The control circuit is configured to detect an error(s) in configuration data that is used by the programmable clock driver circuit 210. This configuration data is read from the nonvolatile memory 202 and stored in the volatile program registers 206 during program restore operations. Pages 7–8 of the IDT datasheet illustrate 12 bytes (e.g., 96 bits) of configuration data that may be utilized within a device configured to support six banks of clock driver outputs, including one bank that is responsible for generating a pair of differential feedback clock signals (e.g., QFB,/QFB). The control circuit is further configured to automatically idle the at least one clock generator within the PLL clock driver core 208 in response to detecting the error in the configuration data. This automatic idling of the clock generator prevents the clock generator from becoming configured in a manner that could potentially damage or alter the performance of downstream devices receiving clock signals (i.e., clock driver outputs) generated by the programmable clock driver circuit 210. This automatic idling may include operations to set the clock generator at a default setting (e.g., minimum frequency), which applies to all output banks of the driver core circuit 208. In alternative embodiments, a detection of an error may result in a “local” disabling of only a single output bank to which the error in the configuration data applies. The clock generator typically includes a phase-locked loop (PLL) integrated circuit that synchronizes the timing of the clock signals generated at the output banks.
The control circuit performs error detection operations by reading the configuration data from the volatile program registers into the nonvolatile memory controller 204. In response to this reading operation, a first error check word is generated from the configuration data. This first error check word may constitute a cyclic redundancy check (CRC) byte (e.g., 8 bits) that is generated within the CRC byte check/generate logic 207. These operations may be performed in response to a program save command (ProgSave) issued to the integrated circuit device 200 (see, e.g., page 12 of the IDT attached datasheet). This first error check word may be generated using a cyclic redundancy check (CRC) algorithm that generates a binary number. The control circuit completes the operations associated with the program save command by programming the nonvolatile memory 202 with the configuration data and the first error check word. Thereafter, in response to a program restore command (ProgRestore) issued to the driver circuit 210, the control circuit reads the configuration data and the first error check word from the nonvolatile memory 202 and also generates a second error check word from this configuration data. This second error check word is then compared against the first error check word to determine whether a difference exists. The existence of a difference indicates that an error is present in the configuration data that has been read from the nonvolatile memory device 202.
In some cases, the program restore command may result in a writing of erroneous data into the volatile program registers, however, this erroneous data will be blocked from influencing performance of the programmable clock driver circuit 210. Instead, the detection of an error in the configuration data will automatically cause the driver core 208 to be held in a default mode of operation that precludes generation of an active lock signal (LOCK) at an output pin of the packaged device (see, e.g., pages 6 and 11 of the IDT attached datasheet). This default mode of operation may include holding the PLL within the driver core 208 at a minimum clock frequency and precluding phase lock by the PLL. Precluding phase lock may operate to inform another “master” device (not shown) that the driver circuit 210 (e.g., “slave” device) needs to be reprogrammed to correct the detected error. Such reprogramming operations may be performed in response to issuance of a program write command and new configuration data by the master device. Alternatively, the master device may issue a program restore command to the driver circuit 210, which causes another internal downloading of the configuration data from the nonvolatile memory 202 to the volatile registers within the programming registers and control logic 206. Still further operations performed by the control circuit may include automatically performing error correction operations on the configuration data once an error has been detected during a program restore operation. In such cases, the error check word may be generated using a conventional Hamming code that supports at least single bit error correction.
In some cases, the program restore command may result in a writing of erroneous data into the volatile program registers, however, this erroneous data will be blocked from influencing performance of the programmable clock driver circuit 210. Instead, the detection of an error in the configuration data will automatically cause the driver core 208 to be held in a default mode of operation that precludes generation of an active lock signal (LOCK) at an output pin of the packaged device (see, e.g., pages 6 and 11 of the attached datasheet). This default mode of operation may include holding the PLL within the driver core 208 at a minimum clock frequency and precluding phase lock by the PLL. Precluding phase lock may operate to inform another “master” device (not shown) that the driver circuit 210 (e.g., “slave” device) needs to be reprogrammed to correct the detected error. Such reprogramming operations may be performed in response to issuance of a program write command and new configuration data by the master device. Alternatively, the master device may issue a program restore command to the driver circuit 210, which causes another internal downloading of the configuration data from the nonvolatile memory 202 to the volatile registers within the programming registers and control logic 206. Still further operations performed by the control circuit may include automatically performing error correction operations on the configuration data once an error has been detected during a program restore operation. In such cases, the error check word may be generated using a conventional Hamming code that supports at least single bit error correction.
The above described operations performed by the packaged integrated circuit device 200 are illustrated by the flow diagram of
As illustrated by Blocks 712, 716, 720 and 724, after a power-up/reset event, the clock driver circuit 210 is responsive to a plurality of commands issued by a master device (not shown). The receipt of a write command issued by the master device is accompanied by new configuration data that is written into the volatile registers within the clock driver circuit 210, Block 714. Likewise, the receipt of a read command issued by the master device results in a reading of the configuration data from the volatile registers to an interface (e.g., JTAG/I2C interface/bus) of the packaged integrated circuit device 200, Block 718. The receipt of a program save command (ProgSave) from the master device results in a copying of the configuration data (within the volatile registers) to the nonvolatile memory 202 (e.g., EEPROM). These program save operations include writing the nonvolatile memory 202 with the configuration data and an internally generated error check word (e.g., CRC byte) that is computed from the configuration data, Block 722. In some embodiments, the configuration data may represent 12 bytes of data (i.e., 96 bits of data) and error check word may be defined by a single 8-bit byte.
The issuance of a program restore command by the master device causes the packaged integrated circuit device 200 to undergo operations similar to the automatic program restore operations described above with respect to Blocks 706, 708 and 710. As illustrated by Block 726, the configuration data is transferred from the nonvolatile memory 202 to the volatile registers. An error check operation (e.g., CRC error check) is also performed in the nonvolatile memory controller 204. This operation includes generating a new error check word and comparing the new error check word against the one read from the nonvolatile memory. If an error is detected, Block 728, the PLL clock driver core 208 is set to a default operating level and phase lock is prevented, Block 730. Otherwise, the clock driver circuit 210 is operated in accordance with the configuration data and a new phase lock condition is established, Block 732.
In the drawings and specification, there have been disclosed typical preferred embodiments of the invention and, although specific terms are employed, they are used in a generic and descriptive sense only and not for purposes of limitation, the scope of the invention being set forth in the following claims.
This application is a continuation-in-part of U.S. application Ser. No. 10/648,090, filed Aug. 26, 2003 now U.S. Pat. No. 6,977,539, and claims priority to U.S. Provisional Application Ser. No. 60/551,095, filed Mar. 8, 2004, the disclosures of which are hereby incorporated herein by reference.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5485490 | Leung et al. | Jan 1996 | A |
| 5614855 | Lee et al. | Mar 1997 | A |
| 5684434 | Mann et al. | Nov 1997 | A |
| 5877656 | Mann et al. | Mar 1999 | A |
| 6111445 | Zerbe et al. | Aug 2000 | A |
| 6125157 | Donnelly et al. | Sep 2000 | A |
| 6215689 | Chhor et al. | Apr 2001 | B1 |
| 6271702 | Stansell | Aug 2001 | B1 |
| 6329859 | Wu | Dec 2001 | B1 |
| 6330679 | Conary et al. | Dec 2001 | B1 |
| 6359486 | Chen | Mar 2002 | B1 |
| 6384653 | Broome | May 2002 | B1 |
| 6388478 | Mann | May 2002 | B1 |
| 6433645 | Mann et al. | Aug 2002 | B1 |
| 6466098 | Pickering | Oct 2002 | B2 |
| 6492852 | Fiscus | Dec 2002 | B2 |
| 6509773 | Buchwald et al. | Jan 2003 | B2 |
| 6525584 | Seo et al. | Feb 2003 | B2 |
| 6539072 | Donnelly et al. | Mar 2003 | B1 |
| 6597212 | Wang et al. | Jul 2003 | B1 |
| 6678832 | Gotanda | Jan 2004 | B1 |
| 6691215 | Mirov et al. | Feb 2004 | B1 |
| 6768358 | Birk et al. | Jul 2004 | B2 |
| 20050242853 | Kwack et al. | Nov 2005 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 60551095 | Mar 2004 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10648090 | Aug 2003 | US |
| Child | 10979979 | US |
