Patent Application
20080151675
This invention relates in general to integrated systems of large integration scale comprising distinct static random access storage resources.
In modern integrated systems, that is in complete functional systems integrated with large scale integration techniques (VLSI) on a single monolithic silicon chip, and generally referred to with the acronym SoC (System on Chip), static memories may typically have the greatest static power consumption
Static power consumption is due to leakage currents of the basic devices and may be strongly dependent on the size of the devices, on parameters of the fabrication process, particularly on the threshold voltage Vth, on dielectric thicknesses and also on working temperature. In memory arrays, the leakage is particularly relevant because of the very large number of cells that may be smaller than the minimum size allowed by the used fabrication technology. Moreover, the number of static memories deployed in modern SoCs increases with each step forward of the fabrication technology.
In mobile applications such as for portable phones and alike and in the automotive field, it may be desirable to limit as much as possible power consumption due to leakage without compromising the overall performances of the integrated system.
Besides the traditional integrated circuit technique known by the acronym RBB (Reverse Body Bias) and besides devices with multiple supplies (of different voltages), the most effective way of limiting the current absorbed in a stand-by state and under certain working conditions of the integrated system may be to keep enabled only a small part of the static memory resources and to disable the remaining part.
So far, the possibility of turning on and off the power supply to different static random access memory resources in a monolithically integrated system is achieved by using distinct memory blocks, each of which is integrated in areas or sections of the device to which it is possible to independently interrupt the supply, as contemplated in the execution program of a particular application by the SoC. Such an organization of known SoCs is schematically depicted in
There is a waste of silicon area. Moreover, designers may desirably take into account the electrical effects caused by the turning off of certain logic circuitry used by a memory block to be disabled.
There is an evident need or usefulness for an architecture of the above mentioned capabilities that occupies a reduced integration area, and that simplifies the realization of complex integrated electronic systems having reduced current consumption during stand-by phases and/or during the execution of applications that do not contemplate access to data stored in any or in some of the distinct random access memories that are part of the integrated system.
These desirable objectives may be attained by the new architecture found by the applicants.
Basically, the gist of this invention that is outstandingly effective consists in realizing a single array or matrix of memory cells of total storage capacity comparable to the sum of the single storage capacities of two or more distinct random access memories for which a distinct management of turn off or turn on phases is contemplated in order to reduce static current absorption during particular functioning phases of the integrated system. Moreover, the peripheral circuitries, i.e. the logic circuits for decoding addresses of circuits for reading and writing data in and from addressed memory locations of the two or more sectors in which the array of cells is electrically subdivided, are unified.
The distinct sectors of the array of memory cells are individually supplied through respective supply lines connected to as many appropriate voltage supply sources through dedicated pins of the integrated device in case of controlled sources outside the device or to distinct controlled supply voltage sources present on the device. This approach may allow a reduction of about 5 to 15% of the global area, depending on the size of the distinctly powerable sectors of random access memory. This is primarily due to reductions of the overhead due to the peripheral circuitry that is substantially unified and to the simplification of electrical paths (reduction of the number of internal vias).
a and 1b are respective schematic block diagrams of the known architecture of a monolithically integrated electronic system (SoC) according to the prior art and an architecture according to the present invention.
According to an embodiment, some or even all the distinct static random access memory resources that are defined in the design phase and that may be independently powered or disconnected as specifically commanded by the software during the execution of a certain application, by the integrated electronic system, includes an SKAM architecture in which the array of cells is subdivided into two or more regions, in the depicted example MEM1 and MEM2, each of which is independently supplied through distinct dedicated supply lines Vdd1 and Vdd2, as schematically illustrated in
By comparing the architectures of
In the illustrated example, the sector 1 or primary sector may be a memory resource that is always supplied for providing storage of data even when the data are not being used, depending on the particular application, executed by the integrated system. Sector 2 or secondary sector, like any other sector in which the single matrix of cells may be subdivided, represents a portion of the memory that on the contrary can be turned off during certain functioning phases of the device.
According to a possible variation of the embodiments of the integrated device, the memory cells belonging to distinct sectors that may be independently powered or turned off, can be realized with threshold values that differ from one sector to the other. For example, the sector 1, if it is to be powered without interruptions, can include cells with a relatively high threshold, while sector 2 can include low threshold cells in order to speed up data processing.
The peripheral circuitry comprises address decoders, the circuitry for controlling and managing the various operations to be carried out in the distinctly allowable memory sectors. The input/output circuitry of data to be stored or read may have optionally and preferably, as illustrated, a separate supply line even if, as an alternative, such a supply line may be in common with the supply of also an eventual primary sector 1 to be supplied without interruptions for storing essential data for carrying out the application.
In the example of
The following table lists the possible configurations or functioning modes and the relative reduction effects on the stand-by current absorption.
The modes 1, 2 and 3 allow a reduction of the stand-by current without compressing the voltage levels of the memory cells. Current absorption may be further reduced in modes 4 and 5 at the cost of reducing noise rejection. In case the application being executed so permits, users may select modes 2 and 5 in which the unified peripheral circuitry of the distinct memory sectors is turned off.
The architecture of this embodiment allows also a higher degree of flexibility in defining the size of the primary memory sector and of secondary memory sectors or regions through a program operation.
In the tested sample embodiment, the monolithically integrated electronic system had first storage resources of 3 kB and secondary storage resources of 5 kB SPAM. In a low-leakage functioning mode of the device, the 5 kB SRAM could be turned off while the 3 kB SRAN kept uncorrupted data for the application stored therein.
With the architecture depicted in
In order to keep the possibility of carrying out ECC controls and to keep permanently driven the output buffers of the memory device, the peripheral circuitry may be kept turned on. This choice depends on the particular application to be carried out and on the amount of leakage that may be attributed to the peripheral circuitry.
In case the number and storage capacity of different secondary memory sectors, that may be turned on or off independently by the application program, are programmable, the array of memory cells may comprise a certain relatively large number of sectors that can be electrically supplied through distinct supply lines, such as to determine a high “granularity” in terms of such numerous distinct memory portions. By programming, the most appropriate memory space subdivision for the particular application to be executed is established by the realization of the desired number of secondary sectors of appropriate capacity each including a sufficient number of the granular portions that may be powered as a group of lesser capacity. The turning on and off of the distinct secondary sectors of appropriate storage capacity may in this case be operated on the grouped memory portions forming the independently powerable secondary sector, such as for minimizing power consumption during different functioning phases of the device.
| Number | Date | Country | Kind |
|---|---|---|---|
| VA2006A000081 | Dec 2006 | IT | national |
