Integrated memory having a memory cell array containing a plurality of memory banks, and circuit configuration having an integrated memory

BACKGROUND OF THE INVENTION

[0001] 1. Field of the Invention

[0002] The present invention relates to an integrated memory having a memory cell array which contains a plurality of memory banks with respective memory cells. The invention further pertains to a circuit configuration having such an integrated memory.

[0003] Integrated memory devices, for example in the form of so-called DRAMs (Dynamic Random Access Memories), are generally operated in a system environment containing, by way of example, a processor or a controller which accesses the memory for the purpose of reading stored data or for the purpose of writing data which are to be stored. An integrated memory, particularly in the form of a DRAM, generally contains a memory cell array which is normally divided into a plurality of separate memory banks containing respective memory cells. In each of the memory banks, a respective row decoder and column decoder can be used for separately accessing the memory cells, organized in matrix form, for the purpose of reading data or for the purpose of writing data. In memory systems containing a plurality of processors or controllers, it has been customary up to now for each processor or controller to access a dedicated memory connected to it. This can mean, firstly, a comparatively high space requirement and secondly a comparatively high cost involvement, since a respective dedicated memory chip needs to be provided for a plurality of processors or controllers.

SUMMARY OF THE INVENTION

[0004] It is accordingly an object of the invention to provide an integrated memory and a circuit configuration with an integrated memory, which overcome the above-mentioned disadvantages of the heretofore-known devices and methods of this general type and which provides for an advantageous integrated memory that, when used in a system environment containing a plurality of processors or controllers, requires only a comparatively low amount of space. Furthermore, it is also an object of the present invention to specify a circuit configuration containing an integrated memory and a plurality of processors or controllers which allows a comparatively low space requirement.

[0005] With the foregoing and other objects in view there is provided, in accordance with the invention, an integrated memory, comprising:

[0006] a memory cell array containing a plurality of memory banks each with respective memory cells;

[0007] at least two independently operated connection panels each connected to a respective command and address bus and to a respective data bus for enabling external communication with the memory cell array; wherein each the data bus is connectible to each of the memory banks; and

[0008] a control circuit for external access to the memory banks, the control circuit being configured to:

[0009] produce a number of first control signals corresponding to a number of the memory banks, the first control signals being associated with a respective one of the memory banks, and one of the first control signals respectively indicating if an associated the memory bank is being accessed; and

[0010] produce a number of second control signals corresponding to a number of the connection panels, wherein one of the second control signals is produced if an access collision occurs between access to one of the memory banks via one of the connection panels and access to the one memory bank via another of the connection panels.

[0011] In line with the invention, an integrated memory of the type mentioned in the introduction also has two connection panels, which can be operated independently of one another, for external communication by the memory which are connected to a respective command and address bus and to a respective data bus. The respective data bus can be connected to each of the memory banks. In addition, a control circuit is provided which produces a number of first control signals and a number of second control signals for the purpose of external tap-off. The number of first control signals corresponds to the number of memory banks, with the first control signals being associated with a respective one of the memory banks. One of the first control signals respectively indicates if the associated memory bank is being accessed. The number of second control signals corresponds to the number of connection panels. One of the second control signals is produced if an access collision occurs between access to one of the memory banks via one of the connection panels and access to the same memory bank via another of the connection panels.

[0012] The invention thus provides a memory which allows two processors or controllers to access the memory independently of one another. Accordingly, the invention provides that, in a circuit configuration, at least two processor units, for example each in the form of a processor or controller, are respectively connected to a different one of the connection panels in the inventive integrated memory. The processor units respectively receive at least one of the first and second control signals and access the memory independently of one another on the basis of the first and second control signals. The invention thus allows a multiprocessor environment to contain just one integrated memory which can be accessed by a plurality of processor units independently of one another. This permits a comparatively space-saving circuit configuration.

[0013] In one embodiment of the invention, the processor units access different memory banks in the memory in parallel. This allows the processors to access the memory independently of one another at a high access speed without any waiting times, “wait states”. Accordingly, there is no drawback as compared with memory systems which contain processors with respectively dedicated associated memories.

[0014] In accordance with an added feature of the invention, the control circuit is configured to produce one of the second control signals if an access collision occurs between access to one of the memory banks and a refresh mode on the same memory bank.

[0015] In accordance with an additional feature of the invention, the control circuit drives the memory cell array such that a given memory bank is connected to one of the connection panels if an access command on the command bus is applied to the one connection panel for accessing the given memory bank.

[0016] In accordance with another feature of the invention, upon access to one of the memory banks, the control circuit sets a control signal of the first control signals associated with the one memory bank and resets the control signal before access has ended. In a preferred embodiment, a control signal of the second control signals produced by the control circuit in the event of an access collision indicates which of the connection panels should be ignored for access to the memory banks.

[0017] In accordance with a further feature of the invention, there is provided an externally programmable register circuit for storing an information item regarding which of the connection panels is prioritized for access in an event of an access collision, and for outputting the stored information item to the control circuit.

[0018] With the above and other objects in view there is also provided, in accordance with the invention, a circuit configuration, comprising an integrated memory device as outlined above, and two or more processor units respectively connected to a different one of the connection panels. The processor units receive at least one of the first and second control signals, and access the memory independently of one another based on the first and second control signals.

[0019] In accordance with again an added feature of the invention, the two or more processor units access different memory banks concurrently, simultaneously, and in parallel.

[0020] In accordance with again an additional feature of the invention, an externally programmable register circuit of the memory is programmable by one of the processor units with an information item that defines which of the connection panels has priority for access in the event of an access collision.

[0021] In accordance with yet an added feature of the invention, the processor units and the memory interact such that a refresh mode for each of the memory banks can be controlled by one of the processor units. Alternatively, the memory is configured to enable a refresh mode to be controlled for each of the memory banks without involvement of the processor units.

[0022] In other words, the invention allows a “refresh mode” for refreshing the memory cell content in two ways. First, the processor units and the memory interact such that a refresh mode for each of the memory banks can be controlled by one of the processor units. Alternatively, the memory is designed such that a refresh mode for each of the memory banks can be controlled without the involvement of the processor units. Accordingly, the control circuit in the memory produces one of the second control signals if an access collision occurs between access to one of the memory banks and a refresh mode on the same memory bank.

[0023] In accordance with yet an additional feature of the invention, the processor units are configured to check, in each case before access to the memory, whether one of the first control signals has been set, and, if one of the first control signals has been set, not to access the memory bank that is associated with that first control signal.

[0024] In accordance with a concomitant feature of the invention, the processor units check, in each case before access to the memory, whether one of the second control signals has been set, and, if one of the second control signals has been set, not to access or to interrupt access via one of the connection panels.

[0025] Other features which are considered as characteristic for the invention are set forth in the appended claims.

[0026] Although the invention is illustrated and described herein as embodied in an integrated memory having a memory cell array containing a plurality of memory banks, and circuit configuration having an integrated memory, it is nevertheless not intended to be limited to the details shown, since various modifications and structural changes may be made therein without departing from the spirit of the invention and within the scope and range of equivalents of the claims.

[0027] The construction and method of operation of the invention, however, together with additional objects and advantages thereof will be best understood from the following description of specific embodiments when read in connection with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

[0028]FIG. 1 is a block diagram of an embodiment of a circuit configuration according to the invention;

[0029]FIG. 2 is a more detailed block diagram of an embodiment of an integrated memory and of a circuit configuration according to the invention; and

[0030] FIGS. <b>3</b> to <b>8</b> are respective signal diagrams to explain the operation of the memory and of the circuit configuration shown in FIG. 2.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0031] Referring now to the figures of the drawing in detail and first, particularly, to FIG. 1 thereof, there is shown a circuit configuration containing an integrated memory <b>3</b> which has a plurality of memory banks B<b>0</b> to B<b>3</b> containing respective memory cells. The memory cells are diagrammatically indicated as a small memory field <b>4</b> of columns and lines. The memory <b>3</b> also has two connection panels P<b>1</b> and P<b>2</b> which can be operated independently of one another and which are connected to a respective communication bus <b>10</b> and <b>20</b> for the purpose of external communication by the memory. Each of the communication buses <b>10</b> and <b>20</b> can be connected to each of the memory banks B<b>0</b> to B<b>3</b>, shown symbolically by respective switch symbols that can be actuated by a control circuit <b>30</b>.

[0032]FIG. 2 shows a more detailed embodiment of the memory <b>3</b> and of the circuit configuration shown in FIG. 1. Each of the independently operable connection panels P<b>1</b> and P<b>2</b> is connected to a respective command and address bus <b>11</b>, <b>21</b> and to a respective data bus <b>12</b>, <b>22</b>. The command and address bus <b>11</b> is used for transmitting command signals CMD#<b>1</b> and address signals ADD#<b>1</b> and is connected to the connection panel P<b>1</b>. The data bus <b>12</b> is used for transmitting data signals DQ#<b>1</b> and is likewise connected to the connection panel P<b>1</b>. The command and address bus <b>21</b> is used for transmitting the command signals CMD#<b>2</b> and address signals ADD#<b>2</b> and is connected to the connection panel P<b>2</b>. The data bus <b>22</b> is used for transmitting the data signals DQ#<b>2</b> and is likewise connected to the connection panel P<b>2</b>. The command and address bus <b>11</b> and the data bus <b>12</b> are connected to the processor unit <b>1</b>, and the command and address bus <b>21</b> and the data bus <b>22</b> are connected to the processor unit <b>2</b>. The respective data bus <b>12</b>, <b>22</b> can be connected to each of the memory banks B<b>0</b> to B<b>3</b>, likewise shown symbolically by switch symbols in FIG. 2.

[0033] The memory <b>3</b> contains a control circuit <b>30</b> which controls changeover of the data buses <b>12</b>, <b>22</b> for the purpose of connection to the respective memory banks B<b>0</b> to B<b>3</b> and also controls access to the memory cells in the respective memory banks B<b>0</b> to B<b>3</b>. The control circuit <b>30</b> is connected to the respective command and address bus <b>11</b>, <b>21</b>. The control circuit <b>30</b> also produces a number of first control signals BSY#<b>0</b> to BSY#<b>3</b> which corresponds to the number of memory banks B<b>0</b> to B<b>3</b>. The control signals BSY#<b>0</b> to BSY#<b>3</b> are respectively associated with one of the memory banks and respectively indicate if the associated memory bank is being accessed. In addition, the control circuit <b>30</b> produces a number of second control signals CNT#<b>1</b>, CNT#<b>2</b> which corresponds to the number of connection panels P<b>1</b>, P<b>2</b>. One of the control signals CNT#<b>1</b>, CNT#<b>2</b> is produced if an access collision occurs between access to one of the memory banks via one of the connection panels P<b>1</b>, P<b>2</b> and access to the same memory bank via the respective other one of the connection panels P<b>1</b>, P<b>2</b>. The control signals BSY#<b>0</b> to BSY#<b>3</b> and CNT#<b>1</b>, CNT#<b>2</b> are made available on external connections of the memory <b>3</b> and are tapped off by the processor units <b>1</b> and <b>2</b>. These respectively receive the control signals and access the memory <b>3</b> independently of one another on the basis of these control signals.

[0034] In addition, the memory <b>3</b> has a register circuit <b>31</b> which is in the form of a “mode register set” (MRS), in particular. The register circuit <b>31</b> is used, inter alia, for storing an information item regarding which of the connection panels P<b>1</b>, P<b>2</b> has priority for access in the event of an access collision. In one advantageous embodiment, the externally programmable register circuit <b>31</b> is programmed by one of the processor units <b>1</b>, <b>2</b> (“master processor”) with an information item regarding which of the connection panels P<b>1</b>, P<b>2</b> has priority for access in the event of an access collision. The information item stored in the register circuit <b>31</b> is forwarded to the control circuit <b>30</b>.

[0035] The text below explains the operation of the memory configuration shown in FIG. 2 in more detail with reference to the signal diagrams shown in FIGS. <b>3</b> to <b>8</b>.

[0036] The access commands ACT transmitted with the command signals CMD#<b>1</b>, CMD#<b>2</b> signify that a memory bank needs to be activated for access. The commands RDA and WRA signify read access with “autoprecharge” and write access with autoprecharge, respectively. The command REF signifies initiation of a refresh cycle for implementing a refresh mode. The signals transmitted with the address signals ADD#<b>1</b>, ADD#<b>2</b> signify access to the respective memory bank B<b>0</b> to B<b>3</b> in the memory <b>3</b>. The control signal BSY#<b>0</b> is associated with the memory bank B<b>0</b>, the control signal BSY#<b>1</b> is associated with the memory bank B<b>1</b>, and so on. In the present example, the memory <b>3</b> is in the form of a synchronous dynamic memory (control clock CLK), an “SDRAM”. In the present example, it is operated on the basis of a double data rate architecture (DDR DRAM), but the mode of operation shown can be transferred analogously to a single data rate architecture.

[0037] According to FIG. 3, the memory bank B<b>1</b> is accessed using the access command ACT. In this case, access is controlled by the control circuit <b>30</b> in line with the memory shown in FIG. 2 by virtue of the memory bank B<b>1</b> being connected to the connection panel P<b>1</b> when the access command ACT on the command bus <b>11</b> is applied to the connection panel P<b>1</b> for the purpose of accessing the memory bank B<b>1</b>. As soon as the access command ACT is received by the connection panel P<b>1</b>, also referred to as a port, the control signal BSY#<b>1</b> is set by the memory. When the “row cycle time” tRC (cycle time required in an access cycle for the purpose of activating a word line) minus two clock periods of the clock signal CLK has elapsed, the control signal BSY#<b>1</b> is reset by the control circuit <b>30</b> before access has ended. The processor units <b>1</b>, <b>2</b> check, in each case before access to the memory, whether one of the control signals BSY has been set. If the respective control signal BSY has not been set, the memory bank in question can be accessed again in the next clock cycle. In the case shown in FIG. 3, when tRC has elapsed, the processor unit <b>1</b> checks whether the control signal BSY#<b>1</b> has been set. This signal has not been set at time (<b>2</b>), which means that the memory bank B<b>1</b> can be accessed again. In line with the read command RDA, the previous access to the memory bank B<b>1</b> involves data in this memory bank being output on the data bus <b>12</b>.

[0038] In another case, in line with FIG. 4, two access commands ACT for the memory bank B<b>1</b> and for the memory bank B<b>2</b> are output. Accordingly, the control signals BSY#<b>1</b> and BSY#<b>2</b> are set in succession. As a result of the read commands RDA for memory banks B<b>1</b> and B<b>2</b>, the data in the memory banks B<b>1</b> and B<b>2</b> are output in succession via the port P<b>1</b>. When the control signal BSY#<b>1</b> has been reset, the memory bank B<b>1</b> can be accessed further.

[0039] The signal diagram shown in FIG. 5 is used to show parallel access by the processor units <b>1</b> and <b>2</b> to the memory banks B<b>0</b> to B<b>3</b>. Port P<b>1</b> is used to access the memory banks B<b>0</b> and B<b>1</b> using the access commands ACT and read commands RDA in succession. Port P<b>2</b> is used to access the memory banks B<b>2</b> and B<b>3</b> using the access commands ACT and read commands RDA in succession. Any access to the respective memory bank involves the corresponding control signal BSY associated with this memory bank being set. At the respective times, as explained with reference to FIG. 3, the respective control signal BSY is reset. The data in the memory banks B<b>0</b> and B<b>1</b> are output in succession via port P<b>1</b>, and in parallel with this the data in the memory banks B<b>2</b> and B<b>3</b> are output in succession via port P<b>2</b>.

[0040] The signal diagram shown in FIG. 6 is used to show three further cases of different access operations. In case <b>1</b>, memory bank B<b>1</b> is accessed simultaneously both via port P<b>1</b> and via port P<b>2</b>. At time (<b>1</b>), each of the processor units <b>1</b>, <b>2</b> checks the state of the control signal BSY#<b>1</b> associated with the memory bank B<b>1</b>. Since this signal has not been set, a respective access command ACT for memory bank B<b>1</b> is produced at time (<b>2</b>). At time (<b>3</b>), both processor units <b>1</b>, <b>2</b> note that the control signal BSY#<b>1</b> has been set. In this case, an access collision occurs between access to the memory bank B<b>1</b> via port P<b>1</b> and access to the same memory bank B<b>1</b> via port P<b>2</b>. In this case, it is necessary to decide which of the ports P<b>1</b>, P<b>2</b> has priority for access in the event of an access collision. In case <b>1</b>, shown in FIG. 6, access via port P<b>2</b> has priority. The control signal CNT#<b>1</b> is set. This means that access via port P<b>1</b> is ignored for access by the memory. Access to memory bank B<b>1</b> via P<b>1</b> is stopped by the processor unit <b>1</b>, and access via port P<b>2</b> is continued.

[0041] In case <b>2</b>, as shown in FIG. 6, the processor unit <b>2</b> checks at time (<b>3</b>), precisely at the same time as it sends the access command ACT, the state of the control signal BSY#<b>1</b>. In this case too, an access collision arises between a plurality of access operations to memory bank B<b>1</b>. In this case, the control signal CNT#<b>2</b> is set at time (<b>4</b>). The access to port P<b>1</b> which has been started is continued, and the access to port P<b>2</b> is ignored by the memory (“first come, first served”).

[0042] In case <b>3</b>, processor unit <b>2</b> checks the control signal BSY#<b>1</b> before an access command ACT is sent. In the present case, however, in a departure from the specification, an access command ACT for memory bank B<b>1</b> is nevertheless sent. In this case, the control signal CNT#<b>2</b> is set at time (<b>5</b>), and access via port P<b>2</b> is ignored by the memory, while access via port P<b>1</b> is continued. As can be seen from the cases shown in FIG. 6, the control signal CNT#<b>1</b> or CNT#<b>2</b> is necessary only in case <b>1</b>; in the other two cases, the control signal BSY#<b>1</b> is sufficient for access control, in principle.

[0043] The signal diagram shown in FIG. 7 is used to show another four cases of different access operations to the memory <b>3</b>. In case <b>1</b>, the processor unit <b>1</b> had no time to check the signal BSY#<b>1</b>. In this case, an access collision occurs between a refresh mode on memory bank B<b>1</b> and access via port P<b>1</b> likewise to memory bank B<b>1</b>. The control signal CNT#<b>1</b> is set and the access command ACT for port P<b>1</b> is ignored. The processor unit <b>1</b> is informed about the set signal BSY#<b>1</b> at time (<b>2</b>). The signal CNT#<b>1</b> is set at time (<b>3</b>).

[0044] In case <b>2</b>, a similar situation arises to that in case <b>1</b>, but at the same time (<b>1</b>). The access command ACT at port P<b>1</b> is ignored, and the processor unit <b>1</b> in question is informed about CNT#<b>1</b> at time (<b>2</b>).

[0045] In case <b>3</b>, the access command ACT for accessing memory bank B<b>1</b> has already been sent. After the time tRC, the memory bank B<b>1</b> is automatically closed and the refresh mode is implemented on memory bank B<b>1</b>, with the control signal BSY#<b>1</b> being set for a further time rTC.

[0046] In case <b>4</b>, a read command RDA for memory bank B<b>1</b> is initiated. Using the control signal CNT#<b>1</b>, the memory indicates to the processor unit <b>1</b> that no further read command can be accepted. Next, a precharge operation is automatically performed. After that, the refresh command is executed, with the control signal BSY#<b>1</b> being set for a further time period tRC.

[0047] The processor units <b>1</b> and <b>2</b> and the memory <b>3</b> can, in principle, interact such that a refresh mode for each of the memory banks B<b>0</b> to B<b>3</b> can be controlled by one of the processor units <b>1</b>, <b>2</b>. When the control signals BSY and CNT are provided, however, it is also possible for the memory to control the refresh mode for each of the memory banks without the involvement of the processor units. The functionality is similar to in the cases described in more detail above. One difference, however, is that the refresh command REF is produced internally in the memory.

[0048] The signal diagram shown in FIG. 8 is used to show further cases of different access operations. In case <b>1</b>, an access command ACT for memory bank B<b>1</b> has been initiated via port P<b>2</b> after an internal refresh mode on the same memory bank has been initiated at this point. In this case, the signal CNT#<b>2</b> is set. The access command for memory bank B<b>1</b> is ignored by the memory. Memory bank B<b>3</b> is accessed in parallel via port P<b>1</b>.

[0049] In case <b>2</b>, a refresh mode is implemented on memory bank B<b>3</b>. At the same time, an access command ACT for memory bank B<b>3</b> is initiated via port P<b>1</b>. The control signal CNT#<b>1</b> is produced in order to indicate to the processor unit <b>1</b> that access to memory bank B<b>3</b> on account of the access command is not being implemented by the memory. At the same time, the memory bank B<b>0</b> is accessed via port P<b>2</b>. Had another access command for memory bank B<b>3</b> been initiated via port P<b>2</b> at the same time instead of the memory bank B<b>0</b> being activated, a double access collision would occur. In this case, both signals CNT#<b>1</b> and CNT#<b>2</b> would be set and none of the access commands via port P<b>1</b> and port P<b>2</b> would be accepted by the memory.

[0050] In case <b>3</b>, a refresh mode is implemented on memory bank B<b>0</b> after an access command ACT for memory bank B<b>0</b> has been initiated via port P<b>1</b>. The control signal CNT#<b>1</b> is set in order to indicate to the processor unit <b>1</b> that access to memory bank B<b>0</b> is being terminated by the memory. No read or write command for memory bank B<b>0</b> is accepted while the control signal BSY#<b>0</b> has been set. In parallel with this, an access command for memory bank B<b>2</b> is executed via port P<b>2</b>. One clock cycle later, the processor unit <b>1</b> initiates an access command for memory bank B<b>2</b> via port P<b>1</b>. However, the processor unit <b>1</b> cannot use the signal CNT#<b>1</b> for checking an access collision, because this signal has already been used in order to indicate the access collision for memory bank B<b>0</b>. In this case, the processor unit <b>1</b> needs to check the control signal BSY#<b>2</b> in order to establish that the access command for memory bank B<b>2</b> via port P<b>1</b> is being ignored by the memory.

[0051] In case <b>4</b>, a read operation is executed via port P<b>2</b> until the signal CNT#<b>2</b> is set. After that, the memory bank is precharged and a refresh mode on this memory bank is implemented.

[0052] In the case of the exemplary embodiments explained, the refresh mode has been implemented bank by bank. If the refresh mode for a memory bank is controlled by one of the processor units, this means that the respective memory bank address needs to be transmitted. One alternative is to implement a refresh mode for all the memory banks at the same time. In this case, instead of one memory bank, all the memory banks need to be precharged. The number of refresh commands is reduced in this case (by a factor of 4 in the case of the exemplary embodiments cited), but this entails a certain degree of restriction on the flexibility for controlling the memory's refresh mode.