Electrical computers and digital processing systems: support – Synchronization of clock or timing signals – data – or pulses
Reexamination Certificate
1998-12-14
2001-05-15
Heckler, Thomas M. (Department: 2182)
Electrical computers and digital processing systems: support
Synchronization of clock or timing signals, data, or pulses
C711S167000
Reexamination Certificate
active
06233694
ABSTRACT:
FIELD OF THE INVENTION
The invention relates to processing circuits utilizing synchronous dynamic random-access memories (SDRAMs) and, more particularly, synchronization devices for such SDRAM memories.
BACKGROUND OF THE INVENTION
SDRAM memories were developed to adapt to the ever increasing operating speed of processing circuits of the microprocessor type. While standard DRAM memories have a transfer frequency limited to a few tens of megahertz, architectures with SDRAM memories can achieve operating frequencies of 100 MHz or more.
SDRAM memories utilize a synchronous interface circuit where all the signals are sampled on the same clock edge. These signals are, for example, the address and data signals, the control and instruction signals such as “cs”, “ras”, “cas”, “we”, “dqm” whose definition and functions are explained in any manufacturer documentation concerning SDRAM memories and which will therefore be dealt with only in brief here.
Read operation
The time diagram relating to a read operation for an SDRAM is given in FIG.
1
. This read operation depends on two programmable parameters referred to as “cas latency” and “burst length”.
In the example in
FIG. 1
, the “cas latency” instruction corresponding to the latency time in read mode is equal to 3. This means that the data dq are read three cycles after the read instruction has been received. The “burst length” instruction is equal to 4, which means that four consecutive data are extracted from the same received address information. The dotted line
1
corresponds to the activation of the address (ad) as a “row” address by the ras signal (row address strobe) and the dotted line
2
corresponds to the activation of the address (ad) as a “column” address by the cas signal (column address strobe) with read instruction. The latency time corresponds to the time between the application of the “column” address (ad) and the data output. The dotted line
3
corresponds to the reading of the valid data dq, 3 clock cycles after the cas signal.
Write operation
The time diagram relating to a write operation for an SDRAM is given in FIG.
2
. This operation depends on the “burst length” signal which represents the number of consecutive data which are written from the same address information. The dotted line
4
corresponds to the row addressing (ras). The dotted line
5
corresponds to the column addressing (cas) with the activation of the write instruction (we).
Again with reference to FIG.
1
and the read operations, the time Tac is the access time for the data after the clock rising edge corresponding to the last cycle as defined by the latency time, and Toh is the holding time for these data after the next clock rising edge (therefore corresponding to the latency time). When the operating frequency is high, the data dq which have been read remain valid for a short period of time, as represented by the unhatched “valid data” signal segment through which the dotted line
3
passes in FIG.
1
. For example, for 100 MHz clock frequency and for times Tac and Toh of respectively 8.5 ns and 3 ns, the time for which the data which has been read remains valid and can therefore be acquired is:
10 ns−8.5 ns+3 ns=4.5 ns
Acquisition of these data within this time interval requires that the clock of the circuit managing the SDRAM memory, which circuit will be referred to below as the controller, and the SDRAM memory's own clock be in phase. If this is not so, then it will not be possible for the data to be dealt with reliably.
If the controller is an integrated circuit, or part of an integrated circuit, the control, address and data signals sent to the SDRAM memory by this circuit, which are synchronized with the internal clock derived from the external clock signal received on the clock input of the circuit, will not generally be synchronized with the external clock signal directly received by the SDRAM memory, even though the signals sent to these two clock inputs are the same, as explained below.
FIG. 3
represents an SDRAM memory controller
6
which is connected to an SDRAM memory
7
by a link
8
for exchanging the control, address and data signals, according to the prior art.
The same clock H is sent to the input of the memory controller circuit
6
and to the input of the SDRAM memory
7
. The controller receives this clock on its clock input and distributes it to the various circuits which need to be synchronized with this clock. In order for this to be done, the clock signal which is received needs to be “buffered” several times for a tree distribution, before being delivered to the circuits which exchange signals on the link with the SDRAM memory, as indicated in FIG.
3
. Each buffer circuit creates a delay of a few nanoseconds, and an overall delay of the order of 5 to 10 nanoseconds can make this direct link
8
between the controller and the memory unreliable, because of the desynchronization between the clock H and the signals exchanged on this link.
Of course, the controller may use PLL circuits (phase locked loops). But such circuits increase the complexity and cost of producing ASICs. Delay lines may also be utilized, but there are known problems inherent with components of this type, their calibration needs to be perfect and fine adjustment is expensive.
SUMMARY OF THE INVENTION
The object of the invention is to overcome the drawbacks mentioned above. To this end, the invention relates to a synchronization device for synchronous dynamic random-access memory, which device is connected to at least one memory by an address bus (AD) and a data bus (DO), the memory having a clock input, the device comprising a first buffer (r
5
) which receives data (Di_w) to be sent to the memory on the data bus, a second buffer (r
4
) which receives addresses (ADi) to be sent to the memory on the address bus, and a third memory (r
1
) which receives the data which come from the memory and are sent on the data bus, characterized in that, on a clock output, it delivers a clock signal (clki_b) to be sent to the clock input of the memory, in that the clock signal (clki) utilized for buffering the addresses and the data is the same, with the possible exception of having undergone signal inversion, as the one (clki_b) delivered on the clock output, and in that the same signal (clki_b) is utilized by the third buffer (r
1
) to buffer the data sent by the dynamic memory on the data bus.
REFERENCES:
patent: 5448715 (1995-09-01), Lelm et al.
patent: 5630096 (1997-05-01), Zuravleff et al.
patent: 5802596 (1998-09-01), Shinozaki
patent: 5896347 (1999-04-01), Tomita et al.
patent: 6014339 (2000-01-01), Kobayashi et al.
patent: 01261325 (1991-05-01), None
Allard Jean-Marc
Plissonneau Frédéric
Sorin Alain
Fogelson Joel M.
Heckler Thomas M.
Kurdyla Ronald H.
Thomson Licensing S.A.
Tripoli Joseph S.
LandOfFree
Synchronization device for synchronous dynamic random-access... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Synchronization device for synchronous dynamic random-access..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Synchronization device for synchronous dynamic random-access... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2561120