Background of the Invention

This invention relates to field programmable gate arrays and more particularly to methods of programming such arrays.

In a field programmable gate array, CMOS static random access memory (SRAM) devices ("cells") are commonly used for controlling the on/off (relatively low resistance/relatively high resistance) conditions of high-current paths of interconnection ("pass") transistors, typically n-channel MOS transistors. Each of these memory cells is addressable (accessible) through a high-current path of at least one separate access transistor--typically the source-drain path of an n-channel MOS transistor. The high-current path (typically, n-channel MOS transistor source-drain path) of each pass transistor controls a connection between a pair of controlled ("application") circuitry devices, such as logic gates, buffers, latches, look-up tables, or a combination of them. These application circuitry devices, together with the SRAM cells (including their access transistors), are typically integrated in an integrated circuit semiconductor silicon chip. The devices thus form a "field programmable gate array", reflecting the fact that they thus form an array which can be interconnected (configured; "programmed") or re-interconnected (reconfigured; "reprogrammed") at will in the field (i.e., long after the chip has been fabricated) in accordance with various desired interconnection patterns, by writing or re-writing the SRAM cells by means of applying suitable binary digital data signal voltages to the cells through their respective access transistors. These data signal voltages are selected so as to configure or reconfigure the states of the respective SRAM cells (i.e., to write or re-write the cells) in accordance with the desired resulting on/off (relatively low resistance/relatively high resistance) conditions of the respective pass transistors. In this way, the application circuitry can be programmed or reprogrammed in the field.

In CMOS technology, each of the SRAM cells is typically formed by a pair of cross-coupled inverters, each inverter formed by p-channel MOS transistor whose source-drain path is connected in series with an n-channel MOS transistor. Thus each SRAM cell contains a total of four transistors in the inverters, plus one or more access transistors. The semiconductor chip area consumed by all the SRAM cells, together with their pass transistors, ordinarily consumes a substantial fraction of the total chip area.

In U.S. Patent No. 4,821,233, each SRAM cell included a pair of cross-connected CMOS inverters. Each pair of inverters was accessed through a single n-channel MOS access transistor, instead of the more usual balanced configuration using two access transistors, to save semiconductor chip area. Each inverter contained a p-channel MOS transistor in series with an n-channel MOS transistor. The p-channel MOS transistor in one of the inverters in each SRAM cell had a different threshold voltage than that of the p-channel MOS transistor in the other. This use of two different p-channel MOS transistor thresholds was required to ensure that--immediately after power was applied to the chip and hence to the inverters, but before every cell had been written with its correct memory state (appropriate to the desired, properly configured application circuitry)--all the not-yet-properly-written cells would maintain their pass transistors in their off conditions, in order to avoid short-circuits in the application circuitry. This is to say, every cell that had not yet been written in a one-memory-cell-row-at-a-time writing procedure would deliver a low voltage level to the gate of its n-channel MOS pass transistor, whereby such a pass transistor would be in its off condition and hence would not enable a short circuit in the application circuitry insofar as the high-current path of such a pass transistor was concerned. Otherwise, owing to the random initial memory states of the memory cells, and hence the random on/off conditions of the pass transistors, undesirably many short-circuits would likely arise in the application circuitry. The use of these different p-channel MOS transistor thresholds, however, requires an extra photolithographic step ("level") for fabricating the cells, whereby processing yields undesirably are reduced and hence manufacturing costs are increased.

One way to avoid this need for the two different p-channel MOS thresholds is a methodology involving re-writing all the SRAM cells with proper information prior to applying power supply voltage to the application circuitry. However, such a methodology requires either an extra (on-chip) electronic switching device or a separate (off-chip) mechanical switching device for controlling the power supply voltage applied to the application circuitry--the former alternative undesirably dissipating excess power and undesirably introducing an unwanted voltage drop across the on-chip switching device, and the latter alternative being unfriendly to the (human) user. Moreover, in order to change the memory state of a cell from that in which its input terminal (i.e., the terminal connected to the access transistor's high-current path) is originally at a low voltage level to that in which its input terminal is at a high voltage level (that is to say, to change the memory cell's state from "low" to "high", i.e., from "0" to "1"), such a methodology, as well as that of the above-mentioned patent, requires that the channel width-to-length ratio of the access transistor be larger than that of at least one of the n-channel MOS transistors of the SRAM cell, whereby extra semiconductor chip area is undesirably needed for the access transistor.

Therefore, it would be desirable to have a methodology that does not suffer from the foregoing shortcomings.

Summary of the Invention

In accordance with the invention, a method is provided as set forth in claim 1. It is advantageous, in addition, that the method is as set forth in claim 2. It is also advantageous, in addition, that the method is as set forth in claim 3. It is further advantageous that the method is also in accordance with claim 4. It is further advantageous that the method is also in accordance with claim 5. It is further advantageous that the method is also in accordance with claim 6. It is further advantageous that the method is also in accordance with claim 7. It is further advantageous that the method is also in accordance with claim 8. It is further advantageous that the method is also in accordance with claim 9. It is further advantageous that the method is also in accordance with claim 10. It is further advantageous that the method is also in accordance with claim 11. It is further advantageous that the method is also in accordance with claim 12.

Brief Description of the Drawing(s)

• FIG. 1 is a circuit schematic useful for explaining an embodiment of the invention; and
• FIG. 2 is a timing diagram of voltages to be applied to the circuit schematic of FIG. 1, in accordance with a specific embodiment of the invention.

It should be understood that the schematic depicted in FIG. 1 shows only one (exemplary) SRAM cell, but that in practice there would be many such cells, typically arranged in a row-column matrix, all binary data signal input terminals D of all access transistors N3 located in each column being connected together (to form a "bit line"), and all (row) select terminals S located in each row being connected together (to form a "word line"). Each such SRAM cell controls the on/off condition of its pass transistor N4, which, in turn, controls the resistance between an exemplary pair of points A1 and A2 located in application circuitry C1, C2.

Detailed Description

Referring now to FIG. 1, an SRAM cell ("SRAM") is formed by a pair of cross-coupled inverters INV₁ and INV₂. Inverter INV₁ is composed ofp-channel MOS transistor P1 and n-channel MOS transistor N1 whose source-drain paths are connected in series between memory cell power supply voltage terminal PV_DD and ground, and whose control (gate) terminals are connected together to an input terminal I1 of this inverter INV₁. Similarly, inverter INV₂ is composed of p-channel MOS transistor P2 and n-channel MOS transistor N2 whose source-drain paths are connected in series between the memory cell power supply terminal PV_DD and ground, and whose control (gate) terminals are connected together to an input terminal 12 of this inverter INV₂. A node located between the source-drain (current) paths of P1 and N1 is connected to an output terminal O1 of the inverter INV₁, and a node between the source-drain (current) paths of P2 and N2 is connected to an output terminal O2 of the inverter INV₂. The output terminal O1 is connected to the input terminal I2, and the output terminal O2 is connected to the input terminal I1, thus completing the SRAM cell. This SRAM cell has an input terminal I3 connected to a node located between O2 and I1; it has an output terminal O3 connected to a node located between O2 and I1. Operating voltages (denoted PV_DD) are delivered to the SRAM cell at the respective memory cell power supply voltage terminals PV_DD, as more fully described below.

The SRAM is accessed (for changing its memory state) through the source-drain path of the access n-channel MOS transistor N3 having its source-drain path connected between the SRAM cell's input terminal 13 and the binary data signal input terminal D, and having its gate terminal connected to a (row) select input terminal S. The output terminal O3 of the SRAM is connected to the control (gate) terminal of an n-channel MOS pass transistor N4. Depending upon its on/off condition, this pass transistor N4 connects or disconnects the pair of points A1 and A2 in application circuitry C1, C2. Typically, the devices of application circuitry C1, C2 comprise logic gates, buffers, and latches, each of which typically is a CMOS or NMOS device. These devices are supplied with power during operation of the application circuitry by means of a d-c voltage, AV_dd, typically equal to 5v, applied to power supply terminals V_dd of the application circuitry.

The on/off condition of the pass transistor N4 at any moment of time is determined by the memory state of the SRAM: a high ("1") state of the SRAM (a high voltage level being present at the output terminal O3 of the SRAM) causes N4 to be on; a low ("0") state of the SRAM (a low voltage level being present at the output terminal O3) causes N4 to be off. More specifically, when a row-select voltage pulse (S) pulse is applied to the gate terminal S of the access transistor N3, the pulse having a height that is high enough to turn on this access transistor N3, and the voltage being applied to the cell at its power terminal PV_DD is sufficient latch the cell--i.e., sufficient to preserve the cell's memory state after the row-select pulse (S) terminates--then the cell will be written in accordance with a binary digital data signal (D0/D1) being applied to the data signal input terminal D of the access transistor (N3) at the moment of time at which the row-select pulse (S) terminates.

Referring now to FIG. 2, the abscissa represents time, and the ordinate represents voltage. Time interval 10 represents an initialization procedure, to prevent short-circuits that would otherwise be caused by random writing of the SRAM cells during "powering up" of the semiconductor chip--i.e., when power supply voltage is initially applied to the SRAM cells; interval 20 represents an exemplary programming (writing) interval for an exemplary row of the SRAM cell matrix; interval 30 represents a transition interval, viz., from the final programming interval to operation of the application circuitry (C1, C2); interval 40 represents an exemplary reprogramming (re-writing) interval for an exemplary row of the cell matrix, and interval 50 represents a transition interval, viz., from the final reprogramming interval to further operarion of the application circuitry.

It should be understood that reading operations (not shown) are performed between intervals 30 and 40, as well as after interval 50.

During the time interval 10, the application circuitry (C1, C2) is "powered up" at time t_o--that is, all application circuitry devices receive an application voltage supply AV_dd (typically equal to approximately 5v) applied to terminal V_dd (FIG. 1). They continue to receive this voltage AV_dd throughout intervals 10, 20, 30, 40, and 50, (unless otherwise desired at the end of intervals 30 or 50, as in cases where it is desired to cease operation of the application circuitry). Next, at time t₁, (the leading edge of) a row-select voltage pulse (S) is simultaneously applied to the control (gate) terminals S of all access transistors (N3)--i.e., of all access transistors located on all rows. This row-select pulse (S) continues from t₁ to t₃.

Commencing preferably at some time prior to time t₂ (t₂ falling between t₁ and t₃) and persisting for some time after t₃, a binary digital data signal having its low binary level (D0) is applied simultaneously to the data signal input terminals D of all access transistors (N3)--i.e., of all access transistors located on all columns. Also, at time t₂, a memory cell supply voltage (PV_DD) haring an intermediate level, typically of about 3v, is simultaneously applied to the power terminals PV_DD of all SRAM cells in the row-column matrix. At all times prior to time t₂ the memory supply voltage (PV_DD) is kept low enough, typically ground, so as to maintain the pass transistor N4 in its off condition and thereby prevent any short-circuits in the application circuitry C1, C2. Next, at time t₃ the row-select pulse (S) is terminated, whereby every SRAM cell in the matrix is written with a binary digital "0"--that is, with its output terminal O3 delivering a low voltage level to its pass transistor N4, whereby every pass transistor in the row-column matrix is off, as is desired after initialization.

Advantageously, the voltage level of the memory cell supply voltage (PV_DD) during the initialization interval 10 is maintained at an intermediate level (typically about 3v) that is lower than the height of the row-select pulse (S) (typically about 5v), and is higher than the low binary level of the data signal (typically about 0v).

After initialization interval 10 has thus been completed, the programming interval 20 is implemented, one row after another. As depicted in FIG. 2, programming of only one row is indicated by the programming interval 20, but it should be understood that such a programming interval is to be applied to all other rows, one row at a time (one row after another).

During all programming intervals 20 (for each row of the matrix), the memory cell power supply voltage (PV_DD) for all cells is maintained at the same (intermediate) level as it was during the initialization interval 10. A row-select voltage pulse (S), typically having the same height as it did during the initialization interval 10, is applied during the rime interval from T₁ to T₂ to the row-select input terminal S (hence gate terminal) of the access transistor N3 of every cell on the first row that thus is being addressed and hence will be written first During the rime interval from T₁ to T₂, a separate binary digital data signal (D0/D1) is applied simultaneously to the data signal input terminal D of each SRAM cell on the row that is being written, each such signal being in accordance with the desired memory state of the respective cell, i.e., in accordance with the desired programming of all cells on the row. The high binary digital level (D1) is typically equal to 5v; the low value (D0) is typically equal to 0v. This data signal (D0/D1) signal should not cease to be "valid" until a time subsequent to T₂--i.e., it should persist until after T₂--and preferably it commences to be "valid" before time T₁.

In this way, even in the presence of the voltage drop across the high-current path of the access transistor N3, because the resulting input signal (typically about 4v) being delivered from data signal input terminal D to input terminal 13 of the SRAM cell in case the binary level of the input signal is high (D1), the cells can be more easily written into their binary "1" state, regardless of their previous states, and hence without the need for tailoring the width-to-length ratios of the n-channel MOS transistors. At the same time, this initialization procedure avoids the need of different p-channel MOS transistor thresholds, because at the end of the procedure all pass transistors N4 are in their off conditions. Thus, every SRAM cell on the first row is written with a memory state that depends upon the binary level of the data signal (D0/D1) that was delivered to its data signal input terminal D during the programming interval 20.

Then, one row after another row of cells is similarly written by being subjected to a similar programming interval 20, one row at a time.

After all the cells located on all the rows have thus been written, i.e., the SRAM matrix has thus been programmed, the memory cell power supply voltages (PV_DD) being delivered to the power terminals PV_DD of all SRAM cells in the matrix are increased simultaneously at time T₃ (during a single interval 300) to a level that is advantageously higher than that of the application circuitry voltage supply AV_dd that is being applied to terminals V_dd of the application circuitry C₁,C₂. Typically, the memory cell power supply voltage (PV_DD) thus is increased to about 6v, that is, about 1v (i.e., at least the threshold of the pass transistor N4) higher than the application circuitry power supply voltage AV_dd. In this way, the conductances of those pass transistors N4 that are in their condition of being on (low resistance) are desirably high, that is to say, the resistance between exemplary points A1 and A2 is desirably low. After the memory cell supply voltage (PV_DD) has thus been increased, the application circuitry C1, C2 is ready for operation.

If and when it is desired to reprogram the circuitry, the SRAMs can either be subjected to another initialization interval 10, followed by another sequence of row-by-row writing intervals 20, and transition interval 30. Alternatively, especially in cases where only relatively few rows need to be rewritten, the reprogramming interval 40 can be applied to those rows, and only those rows, that require rewriting, whereby reprogramming time can be saved at the expense of possible temporary short-circuits in the application circuitry. The number of these short-circuits, however, generally tend to be relatively small as compared with the number of potential short-circuits that can occur when the memory cells are randomly written. Also, during the reprogramming interval 40 the memory power supply voltage (PV_DD) is low enough to maintain the resistance of the pass transistor N4 at a fairly high value (even when the memory state of its SRAM cell is "1" and thus tends to turn on its pass transistor N4) whereby the (V²/R) power losses due to the temporary short-circuits are relatively unimportantly low.

If it is thus desired to avoid the task of initialization and rewriting all rows, then at time T₄ of the reprogramming interval 40, the memory cell power supply voltage (PV_DD) for all cells in the matrix is reduced, typically to 3v, while the application circuitry supply voltage (AV_dd) can be maintained, typically at about 5v. Next, beginning at time T₅ a row-select pulse (S), having a height of typically also about 5v, is applied to the control terminals S of all access transistors N3 located on the row being written. Before this row-select pulse (S) terminates and preferably before it commences, binary digital data signals (D0/D1) are applied to the data signal input terminals of all access transistors N3 located on that row, each such data signal having a voltage level in accordance with the desired reprogramming (rewriting) of the respective cell. The row-select pulse then terminates at time T₆, whereby all SRAM cells on the row are rewritten, and the reprogramming interval 40 has been completed for that row. Thereafter, the other rows that are to be rewritten are subjected to a similar reprogramming interval 40. No other rows need be rewritten. Finally, after all rows that require rewriting have thus been rewritten, say at time T₇, the transition interval 50 commences, and the memory power supply voltage (PV_DD ) being delivered to all SRAM cells in the matrix at their power supply terminals PV_DD is increased simultaneously to a level again that is advantageously higher, by a threshold voltage of the pass transistor N4, than that of the application circuitry power supply voltage (AV_dd)--i.e., typically to about 6v, i.e., about one volt higher than that of the application circuitry power supply voltage.

Although the invention has been described with respect to a specific embodiment, various modifications can be made without departing from the scope of the invention. For example, the memory cells can be realized in bipolar technology, as can be the access transistors, the pass transistors, and the application circuitry devices.