The present invention relates to digital memory systems, and more specifically, to synchronous memory systems.
As the operational frequencies of digital computing systems continue to increase, it has become increasingly necessary to use synchronous memory systems instead of the slower asynchronous memory systems. In synchronous memory systems, data is sent between a master device and one or more memory devices in the form of data packets which travel in parallel with, and must maintain precise timing relationships with, a system clock signal.
Because synchronous memory systems impose tight timing relationships between the clock and data signals, the memory interface circuits in the memory devices of the synchronous memory system generally require clock recovery and alignment circuits such as phase locked loops (PLLs) or delay locked loops (DLLs). One drawback of these clock recovery and alignment circuits, however, is that they typically operate effectively only over a limited range of frequencies. For example, a PLL may not be able to lock to the system's clock frequency if the frequency is either too low or too high. Additionally, the performance of these clock recovery and alignment circuits is degraded due to conditions such as temperature, supply voltage, speed binning codes, process, dimensions (i.e. length) of the memory bus, etc.
It is an object of this invention to provide for an adjustable synchronous memory system.
It is a further object of this invention to provide for a synchronous memory system that uses frequency information to improve the performance of the circuits at the system clock frequency.
It is a further object of this invention to provide for a synchronous memory system that uses system parameters to improve the performance of the circuits at the system clock frequency.
The present invention is a method for adjusting the performance of a synchronous memory system. A memory system comprises a master device and a slave device. A memory channel couples the master device to the slave device such that the slave device receives the system operating information from the master device via the memory channel. The slave device further includes means for tuning circuitry within the slave device such that the performance of the memory system is improved.
Other objects, features, and advantages of the present invention will be apparent from the accompanying drawings and from the detailed description that follows below.
The present invention is illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which like reference numerals refer to similar elements and in which:
A method and apparatus for adjusting the performance of a memory system is described. A synchronous memory system wherein the master device has information about its operating frequency and transmits this frequency information to the memory devices is described. The memory devices then use this frequency information to adjust their clock recovery and alignment circuits to improve their performance at the system clock frequency. The master device may send the memory devices information that specifies the exact frequency of operation, or alternatively, the master device may send the memory devices information that specifies a predetermined range of frequencies which includes the system's clock frequency. For one embodiment, the frequency information is sent as a digital code that is received, stored, and decoded by the memory devices to produce a control code for adjusting the performance of the clock recovery and alignment circuits.
The synchronous memory system operates by sending and receiving data in packets which are synchronized with respect to a system clock. In order to do this properly, the memory master as well as all the slave devices must include circuitry that ensures that the data is read/written synchronously with the system clock. This circuitry is placed within the interface circuits of the memory master and slave devices. The key function of this circuitry is to produce internal clock signals within each device that maintain the proper phase relative to that of the external system clock such that data read or written to the channel by each of the devices is done so at the correct time, thereby preserving synchronization in the memory system. Because the memory devices may be used in different systems which use different operating clock frequencies, this circuitry should function effectively over a large range of possible system clock frequencies.
The master device 110 can be a memory controller, a microprocessor, a 3-D firmware chip, or any other microchip that accesses the synchronous memory. The master device 110 includes a memory interface circuit 115 for transmitting and receiving data from the memory bus 180. The master device 110 generates requests to store data into or recover data from the memory devices 120, 130 via the memory bus 180.
The memory bus 180 is a data communications channel. For one embodiment, the memory bus 180 is a collection of wires or transmission lines. For one embodiment, the memory bus 180 comprises matched-impedance printed circuit board traces.
For one embodiment, the memory devices 120, 130 are dynamic random access memories (DRAMs). Alternatively, the memory devices 120, 130 are static random access memories (SRAMs) or other memory devices. Each memory device 120, 130 includes a memory interface circuit 125, 135, respectively, for transmitting and receiving data from the memory bus 180. For one embodiment, the memory devices 120, 130 cannot generate requests for data but instead only respond to requests generated by the master device 110.
The clock source 150 provides the synchronizing clock signal for the memory system at a system clock frequency. In
The terminator 140 provides a matched-impedance termination for the transmission lines of the memory bus 180. For one embodiment, all signals transmitted on the memory bus 180 eventually terminate at the terminator 140. Although it is included in the synchronous memory system of
The master device 110 further includes information circuitry 290. The information circuitry holds information about the system's clock frequency and other system-level information. For one embodiment, the information circuitry 290 holds system clock frequency information. The information circuitry 290 may detect and/or store other information which affects circuit functioning. For one embodiment, the information circuitry 290 may detect and/or store information about the system temperature, or temperature ranges. The information circuitry 290 may detect and/or store information about the supply voltage, or voltage range. The information circuitry 290 may further detect and/or store information about the length of the memory bus 180, speed binning codes, process, and other factors that may affect the operation of the memory system. For one embodiment, the information circuitry 290 includes a PVTR detector. Detecting and storing this type of information is known in the art, as is the influence of the various factors on system operation. For one embodiment, the data in the information circuitry is also used to tune the performance of the CRA circuit 210 in the master device's memory interface 115.
For one embodiment, the same information about frequency, voltage, temperature, etc. that is sent to the memory devices to tune the performance of their CRA circuits is also made available to a CRA circuit inside the master device to tune its performance. Chip-specific information about the master device, such as the master device's process condition may be used along with the information that is sent to the memory devices to tune the performance master device's CRA circuit.
The master device 110 uses the memory bus 180 to access data and control the memory device 120. The master device 110 improves the performance of the CRA circuits in the memory device 120 by sending frequency information though the memory bus 180 to the memory device 120. For one embodiment, the frequency information is sent as a digital code to the memory device 120. Alternatively, for more accuracy, the frequency information may be sent as an analog signal.
For one embodiment, the memory bus 180 includes a plurality of high-speed data lines 230, 270 which transmit data information between the master device 110 and the memory device 120 in parallel with either the CTM 160 or CFM 170 clock signals. The memory bus 180 also includes a plurality of high-speed control signal lines 240, 260 for transmitting address, request, acknowledge, and other control signals. Finally, the memory bus 180 includes lower-frequency “sideband” lines 280 for communicating information at lower speed between the master device 110 and the memory device 120.
For one embodiment, the memory channel includes slow speed lines and high speed lines. For one embodiment, the slow speed lines are used for system control such as nap, and the high speed lines are used for data and addressing. The controller is aware of its operating frequency and communicates this information to the slave devices. For one embodiment, the master device communicates this information to the slave devices via the slow speed lines. The slave devices receive, decode, and use this frequency information to adjust the circuits in their CRA circuits to improve their performance at the system clock frequency and other operating conditions. In other words, frequency control information comes down the slow speed lines to adjust/improve the performance of the high speed lines. For an alternative embodiment, there is only one channel that operates first at low speed to send frequency control information to adjust the CRA circuits and then operates at high speed after adjustment. For another alternative embodiment, there is only one channel that always operates at high speed, but until the frequency control information has been sent, it operates with lower initial margin.
For one embodiment, the master device 110 sends the memory device 120 information that specifies the exact frequency of operation. For another embodiment, the master device 110 sends the memory device 120 information that specifies a predetermined range of frequencies which includes the system's clock frequency.
For one embodiment, the master device further includes a PVTR detector, and the information circuit 290 further sends information from this PVTR circuit to the slave devices. In one embodiment, the master has a PVTR detector or other detector for detecting system operating parameters. The master sends this information to the slave devices so that they can adjust their performance. This data can be sent via any of the 3 ways described above, i.e. over a separate slow speed channel, over a temporarily slow speed channel, or over an initially low margin high speed channel.
In an alternative embodiment, each individual slave device has a PVTR detector to control the performance of its own CRA circuits.
In yet another embodiment, the master sends frequency information to the slaves, but each slave also has its own PVTR detector. The frequency data is combined with the PVTR data to properly adjust the CRA circuits to account for both of these two types of operating information.
For one embodiment, such information is sent to the memory device 120 periodically during operation of the memory system. For another embodiment, the information is sent only once, during initialization of the memory system.
The master device 110 (not shown) sends the information about the system's operating frequency to the memory device 120. The frequency information is encoded onto n bits as described below. This frequency information is then received by the transceiver circuitry 330 in each memory device 120. The use of transceiver circuitry 330 for receiving data from the memory bus 180 is well-known in the art. Upon receiving this information about the system's clock frequency, the transceiver circuitry 330 stores it into the n-bit register circuit 340. For one embodiment, the frequency information is stored in the register circuit 340 during normal system operation or at the initialization of the system.
The n-bit register circuit 340 presents this frequency information to a decoder circuitry 350. The decoder circuitry 350 translates this frequency information into a m-bit control code for adjusting the performance of the CRA circuitry 310. For one embodiment, the control code is used to adjust one or more portions of the CRA circuitry 310 such that the circuitry operates effectively at the system's clock frequency. For another embodiment, the control code adjusts the CRA circuitry 310 to optimize for external factors, such as temperature, memory bus 180 length, supply voltage, etc. There are several ways that the frequency information can be encoded into n bits and then sent to and stored in the memory device 120. For one embodiment, a binary word indicates the time period of the system clock in pico-seconds (ps), where period=1/frequency. For example, using a 16-bit register, a 16-bit digital word can be sent to the memory device 120 that indicates with 1 ps precision that the system's clock period is anywhere from 0 ps to 65,535 ps (216-1). Alternatively, a more compact digital code that indicates one of a predetermined range of frequencies which includes the system's clock frequency may be used. This scheme requires the storage of fewer bits than the first scheme, but only specifies a range of frequencies instead of an exact frequency. One example of a compact code that could be used, and its corresponding range of frequencies, is shown in the table below:
As can be seen, this scheme uses only 2 bits instead of 16 bits.
In one embodiment, a subset of the n-bit register is implemented. The number of bits used may be restricted to the minimum number needed for the applicable frequency range. For example, the full range of frequencies may use 16-bits, specifying a range of frequencies from 0 ps to 65,535 ps. One embodiment may implement 12-bits, limiting the range of frequencies from 0 ps to 4069 ps, if that is sufficient for the system application in question. For one embodiment, a similar method can be implemented for an encoded frequency information. Reducing the number of bits used reduces the circuitry needed to store and decoder the information, and thus reduces the cost of the device.
Once the frequency information has been stored in the n-bit register circuit 340, the decoder circuitry 350 evaluates this n-bit data to produce the required m-bit control code for adjusting the CRA circuitry 310 for optimal operation. For one embodiment, the control code that is decoded from the frequency information specifies a range of operating frequencies. This is simple if the register circuit 340 is given a compact code that specifies a range of frequencies which includes the system's clock frequency. The decoder circuitry 350 is more complex if the register circuit 340 holds the period of the system's clock signal. For one embodiment, the decoder circuitry 350 includes a simple look-up table for frequency ranges corresponding to control codes. For one embodiment, these tables are hard wired. For one embodiment, these tables may be altered by a user. The decoded control code is then sent to the CRA circuit 310.
The CRA circuit 310 adjusts the phase of internal clock signals so that the receive and transmit circuitry of the memory device 120 will be synchronized with the system clock signals CTM & CFM. The CRA circuit 310 may include variable delay elements, phase interpolator (mixer) circuits, and slew rate control circuits. By receiving and responding to these control codes, these circuits enable the synchronous memory system to operate effectively over a larger range of system clock frequencies than would be possible without the control codes. For one embodiment, the CRA circuit 310 is a phase locked loop (PLL) circuit. For another embodiment, the CRA circuit 310 is a delay-locked loop (DLL) circuit.
The control codes are used to adjust the locking frequency range of the clock recovery and alignment circuits to include the operating clock frequency of the system. The control codes are also used to reduce the jitter of the signals on the high-speed lines, and to improve the timing margin of the signals on the high-speed lines.
An input signal 410 is an input to the phase detector 420. For one embodiment, the input signal 410 is a system clock signal such as CTM. The output of phase detector 420 is an input to integrator/filter 430. The output of integrator/filter 430 is input to a voltage controlled oscillator (VCO) 440. The output of the VCO 440 is the output of the phase locked loop 400. The output of the VCO 440 is the reference signal 460, which is input to the phase detector 420.
An example of how the phase locked loop works is as follows. At the beginning the loop is in balance, i.e. the loop error is equal to zero. Assume that the frequency of the incoming signal increases slightly. This means that the phase of that signal changes a little faster—phase is the integral of frequency. Accordingly, the loop error becomes positive because the phase of the reference signal cannot change at once due to inherent delays in the PLL 400. The frequency generated by the VCO 440 follows the changes in the error signal so that it also increases. The final consequence is that an increase in the incoming signal's frequency causes an increase in the frequency of the reference signal. Thus, the reference signal 410 and incoming signal 410 converge on the same frequency. The elements of the PLL 440 are known in the art.
Although this disclosure has stressed the use of frequency information to tune the clock recovery and alignment circuit of the memory device 120, other relevant information could be transmitted to the memory device 120 and held by its register circuit 340 to tune the performance of clock recovery and alignment circuits. Examples of other types of information that could be sent from the master to the memory device 120 are temperature, supply voltage, speed binning codes, dimensions of the memory bus 180, etc. Any one or more of these types of information could be sent from the master device 110 to the memory device 120 to tune the memory device 120 circuits to operate more effectively under the system's operating conditions.
In the foregoing specification, the invention has been described with reference to specific exemplary embodiments. It will, however, be evident that various modifications and changes may be made without departing from the broader spirit and scope of the invention as set forth in the claims. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.
This application is a continuation of U.S. patent application Ser. No. 11/609,135 filed on Dec. 11, 2006, now U.S. Pat. No. 7,337,294, which is a continuation of U.S. patent application Ser. No. 10/386,210 filed on Mar. 10, 2003, now U.S. Pat. No. 7,149,856, which is a continuation of U.S. patent application Ser. No. 10/051,957 filed on Jan. 18, 2002, now U.S. Pat. No. 6,553,452, which is a continuation of U.S. patent application Ser. No. 08/948,774 filed on Oct. 10, 1997, now U.S. Pat. No. 6,513,103.
Number | Name | Date | Kind |
---|---|---|---|
4338569 | Petrich | Jul 1982 | A |
4691124 | Ledzius et al. | Sep 1987 | A |
4884041 | Walker | Nov 1989 | A |
4920489 | Hubelbank | Apr 1990 | A |
5036300 | Nicolai | Jul 1991 | A |
5180994 | Martin et al. | Jan 1993 | A |
5303191 | Eagan et al. | Apr 1994 | A |
5329491 | Brown et al. | Jul 1994 | A |
5347227 | Bortolini | Sep 1994 | A |
5399995 | Kardontchik et al. | Mar 1995 | A |
5422806 | Chen et al. | Jun 1995 | A |
5446696 | Ware | Aug 1995 | A |
5451894 | Guo | Sep 1995 | A |
5485490 | Leung | Jan 1996 | A |
5490059 | Mahalingaiah et al. | Feb 1996 | A |
5506815 | Hsieh et al. | Apr 1996 | A |
5513327 | Farmwald | Apr 1996 | A |
5532633 | Kawai | Jul 1996 | A |
5534805 | Miyazaki et al. | Jul 1996 | A |
5550783 | Stephens et al. | Aug 1996 | A |
5554945 | Lee | Sep 1996 | A |
5563928 | Rostoker et al. | Oct 1996 | A |
5570054 | Takla | Oct 1996 | A |
5614855 | Lee | Mar 1997 | A |
5673295 | Read et al. | Sep 1997 | A |
5712883 | Miller et al. | Jan 1998 | A |
5712884 | Jeong | Jan 1998 | A |
5721837 | Kikinis et al. | Feb 1998 | A |
5742798 | Goldrian | Apr 1998 | A |
5745792 | Jost | Apr 1998 | A |
5752011 | Thomas et al. | May 1998 | A |
5764092 | Wada et al. | Jun 1998 | A |
5768189 | Takahashi | Jun 1998 | A |
5784328 | Irrinki et al. | Jul 1998 | A |
5799051 | Leung | Aug 1998 | A |
5801985 | Roohparvar et al. | Sep 1998 | A |
5838578 | Pippin | Nov 1998 | A |
5870614 | Ang | Feb 1999 | A |
5886564 | Sato et al. | Mar 1999 | A |
5890014 | Long | Mar 1999 | A |
5890100 | Crayford | Mar 1999 | A |
5943206 | Crayford | Aug 1999 | A |
5956289 | Norman et al. | Sep 1999 | A |
5959481 | Donnelly et al. | Sep 1999 | A |
5978926 | Ries et al. | Nov 1999 | A |
6002627 | Chevallier | Dec 1999 | A |
6131073 | Honda et al. | Oct 2000 | A |
6160755 | Norman et al. | Dec 2000 | A |
6233190 | Cooper et al. | May 2001 | B1 |
6249155 | Hartman et al. | Jun 2001 | B1 |
6337589 | Ooishi | Jan 2002 | B1 |
6438057 | Ruckerbauer | Aug 2002 | B1 |
7587262 | Pippin | Sep 2009 | B1 |
20020051396 | Higashiho et al. | May 2002 | A1 |
20030156483 | Feurle et al. | Aug 2003 | A1 |
20030212474 | Pippin | Nov 2003 | A1 |
20040047099 | Pippin | Mar 2004 | A1 |
20040181354 | Gauthier et al. | Sep 2004 | A1 |
20040227553 | Muranishi | Nov 2004 | A1 |
20080162759 | Garlepp et al. | Jul 2008 | A1 |
20090046761 | Pan | Feb 2009 | A1 |
20090129438 | Pan | May 2009 | A1 |
Number | Date | Country |
---|---|---|
0200003583 | Feb 2000 | JP |
WO 9210032 | Jul 1992 | WO |
Number | Date | Country | |
---|---|---|---|
20080162759 A1 | Jul 2008 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 11609135 | Dec 2006 | US |
Child | 12036598 | US | |
Parent | 10386210 | Mar 2003 | US |
Child | 11609135 | US | |
Parent | 10051957 | Jan 2002 | US |
Child | 10386210 | US | |
Parent | 08948774 | Oct 1997 | US |
Child | 10051957 | US |