Combination field programmable gate array allowing dynamic reprogrammability

0001The present invention relates to field programmable gate arrays (FPGA), and more particularly, to a FPGA that allows for both dynamic reprogrammability by refreshing of charge and non-volatile one-time programmability based upon the breakdown of the gate oxide of a transistor.

0002FPGA's are finding increasing application as logic and/or processing elements. One type of FPGA utilizes SRAM cells, which requires six transistors. The SRAM is used to provide configurable or programmable switches. The programming code is stored as a bitmap typically off chip in non-volatile memory. The SRAM based FPGA is programmed by first programming the non-volatile memory. Then, the bitstram from the non-volatile memory is loaded from the non-volatile memory to the SRAM. The SRAM then is used to control the FPGA. This multi-chip solution results in a large form factor and relatively high cost.

0003Another type of FPGA is based upon anti-fuse technology. Although widely accepted, anti-fuse technology requires specialized fuse manufacturing process. Further, a FPGA based upon anti-fuse technology can only be programmed once.

0004Yet another type of FGPA is based upon flash memory technology. However, flash memory technology requires a relatively more complex semiconductor manufacturing process, thereby increasing cost. Further, flash memory technology is typically one to two generations behind advanced CMOS logic processes.

0005<patcit id="pcit0001" dnum="US6650143B"><text>US6,650,143</text></patcit> discloses a field programmable gate array based upon transistor gate oxide breakdown.

0006Aspects of the present invention are set out in the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

0007<ul id="ul0001" list-style="none" compact="compact"><li><figref idref="f0001">FIG. 1</figref> is a schematic circuit diagram of a portion of a FGPA in relation to the present invention.</li><li><figref idref="f0002">FIG. 2</figref> is a partial layout diagram of a portion of the FPGA represented by <figref idref="f0001">FIG. 1</figref>.</li><li><figref idref="f0003">FIG. 3</figref> is a table of voltages showing the operation of the FPGA cell of <figref idref="f0001 f0002 f0003">FIGS. 1-3</figref>.</li><li><figref idref="f0003">FIG. 4</figref> is a table of voltages showing the operation of an alternative embodiment of a FPGA cell.</li><li><figref idref="f0004">FIG. 5</figref> is a schematic circuit diagram of a portion of a FPGA in relation to the present invention.</li><li><figref idref="f0004">FIG. 6</figref> is a table of voltages showing the operation of the FPGA cell of FIG. 5-C.</li><li><figref idref="f0005">FIG. 7</figref> is a timing diagram showing the storage of data onto the cell of <figref idref="f0004">FIG. 5</figref>-C.</li><li><figref idref="f0006">FIG. 8</figref> is an alternative embodiment of a portion of a FPGA in accordance with an aspect of the present invention.</li><li><figref idref="f0007">FIG. 9</figref> is a top layout view of the circuit of <figref idref="f0007">FIG. 9</figref>.</li><li><figref idref="f0008">FIG. 10</figref> is a schematic diagram of a portion of a dynamic memory array.</li><li><figref idref="f0009">FIG. 11</figref> is a top layout view of the circuit of <figref idref="f0008">FIG. 10</figref>.</li><li><figref idref="f0010">FIG. 12A</figref> is a schematic diagram of a dynamic memory array.</li><li><figref idref="f0011">FIG. 12B</figref> is a top layout view of the circuit of <figref idref="f0010">FIG. 12A</figref>.</li><li><figref idref="f0012">FIGS. 13A-13C</figref> show the use of a floating point node to control a NMOS, PMOS, and inverter, respectively.</li><li><figref idref="f0013">FIGS. 14A-14C</figref> show the use of a floating point node and capacitor to control a NMOS, PMOS, and inverter, respectively.</li><li><figref idref="f0014">FIG. 15A</figref> is a schematic diagram of a dynamic memory array.</li><li><figref idref="f0015">FIG. 15B</figref> is a top layout view of the circuit of <figref idref="f0014">FIG. 15A</figref>.</li><li><figref idref="f0016">FIG. 16</figref> is a schematic diagram of a dynamic memory array.</li><li><figref idref="f0017">FIG. 17A</figref> is a schematic circuit diagram of a portion of a FPGA.</li><li><figref idref="f0017">FIG. 17B</figref> is the layout diagram of the FPGA represented by <figref idref="f0017">FIG. 17A</figref>.</li></ul>

DETAILED DESCRIPTION

0008A FPGA based on transistors having an ultra-thin dielectric that can be stressed into breakdown (soft or hard breakdown) to set a leakage current level is disclosed. A suitable ultra-thin dielectric is the high quality gate oxide of about 50 Å thickness or less used in a transistor, as is commonly available from presently available advanced CMOS logic processes. Such oxides are commonly formed by deposition, by oxide growth from a silicon active region, or by some combination thereof. Other suitable dielectrics include oxide-nitride-oxide composites, compound oxides, and so forth.

0009In the following description, numerous specific details are provided to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention can be practiced without one or more of the specific details, or with other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the invention.

0010Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

0011The present invention is related to gate oxide breakdown based flash memory designs developed by the present inventor and assigned to the same assignee as the present invention. Examples are shown in <patcit id="pcit0002" dnum="WO03025944A1"><text>WO 03/025944 A1</text></patcit>, <patcit id="pcit0003" dnum="WO03034331A1"><text>WO 03/034 331 A1</text></patcit>, and <patcit id="pcit0004" dnum="WO03060917A2"><text>WO 03060917 A2</text></patcit>.

0012<figref idref="f0001">FIG. 1</figref> shows an example of a FPGA array 100 formed in respect of the present invention. The array 100 is a two column by two row array, however, it can be appreciated that the array can be any arbitrary size. The array 100 includes four memory cells 102, each of which comprises a select transistor 104, a capacitor 106, and a switch 108.

0013Examining the memory cell 102 at, for example, the crosspoint of the first row R₁ and the first column C₁ (also referred to herein as a "bitline" or "column bitline"), the memory cell 102 has the gate of its select transistor 104 connected to a read bitline (BLR) line. The drain of the select transistor 104 is connected to a row line R₁ (also referred to herein as a "word line" or "row wordline"). The source of the select transistor 104 is connected to one terminal of the capacitor 106. The other terminal of the capacitor is connected to the column bitline (BL).

0014The gate of the switch 108 is also connected to the source of the select transistor 104, thusly, between the source and one terminal of the capacitor 106. This connection point is referred to as a switch control node. The source and drain of the switch 108 are connected in a "daisy chain" manner with the other switches common to that column.

0015As will be seen below, during the programming step, a relatively large voltage is across the capacitor 106 of the selected column and selected row to break down the gate oxide of the capacitor 106. The other memory cells 102 shown in <figref idref="f0001">FIG. 1</figref> are also formed from, in one embodiment, identical cells 102 at the crosspoints of the column bitlines C_x and row wordlines R_y, where y = 1 to N and N is the total number of rows and where x = I to M and M is the total number of columns.

0016The use of memory cells 102 as programmed elements in the FPGA 100 of <figref idref="f0001">FIG. 1</figref> is advantageous because the components can be fabricated using many conventional CMOS processes using only a single polysilicon deposition step, without adding any mask steps to them. This is in contrast to "floating gate" type FPGA flash memory, which requires at least two polysilicon layers. This is also in contrast to the anti-fuse type FPGA, which requires special process steps. Further, with modern technological advances, the size of a capacitor and transistor can be made very small. For example, current 0.18 micron, 0.13 micron, and smaller line width processes would greatly increase the density of the FPGA.

0017Although only a two by two FPGA array 100 is shown, in practice such FPGA arrays may contain tens of thousands cells, or even millions of cells, when fabricated using, for example, an advanced 0.13 µm CMOS logic process. Even larger arrays will be realized as CMOS logic processes improve further.

0018<figref idref="f0002">FIG. 2</figref> shows a partial layout diagram 200 for a portion of the FPGA array 100. The layout diagram of <figref idref="f0002">FIG. 2</figref> is suitable for an advanced CMOS logic process. The term MOS is commonly understood to pertain to any gate material, including doped polysilicon and other good conductors, as well as to various different types of gate dielectrics not limited to silicon dioxide, and the term is so used herein. For example, the dielectric may be any type of dielectric, such as an oxide or nitride, which undergoes a hard or soft breakdown upon the application of a voltage for a period of time. In one embodiment, a thermally grown gate silicon oxide of about (5nm for 0.25µm, 3nm for 0.18µm and 2nm for 0.13µm process) thick is used.

0019The FPGA array 100 preferably is laid out in a grid in which the column lines such as C₁ and C₂, are orthogonal to the row lines such as R₁ and R₂. <figref idref="f0002">Figure 2</figref> shows two cells 102, arrayed as one column by two rows. As seen in <figref idref="f0002">Figure 2</figref>, the metal one (M1) line is used to connect the gate of the switch (SW) with one terminal of the capacitor 106. Further, the drain of the select transistor (ST and 104) is connected to the word line through a n+ diffusion contact, metal one, via one, and metal two (M2). Further, it can be seen that all of the devices (select transistor 104, capacitor 106, and switch 108) are formed from a polysilicon layer over a low voltage (LV) oxide.

0020The operation of the FPGA array 100 is now explained with reference to the illustrative voltages shown in <figref idref="f0003">FIG. 3</figref>. It will be appreciated that the voltages are illustrative, and that different voltages are likely to be used in different applications or when different process technologies are used. During programming, the various cells in the FPGA array 100 are exposed to one of four possible programming voltage combinations, which are shown on lines 301, 303, 305, and 307 of <figref idref="f0003">FIG. 3</figref>. Read voltages are shown on lines 309, 311, 313, and 315. Assume that a FPGA cell 102 (note that the cell 102 in <figref idref="f0002">Figure 2</figref> is a generic reference to all of the cells in the FPGA 100) is selected for programming and is located at the crosspoint of R₁ and C₁. The selected memory cell 102 is referred to as at the selected row and selected column ("SR/SC"). As shown on line 301, the voltage on the selected wordline R₁ (designated as V_wl or "voltage on the wordline") is 0 volts and the voltage on the bitline C₁ (designated as V_bl or "voltage on the bitline") is 8 volts. Further, the voltage on the selected read bitline (designated as V_blr or "voltage on the read bitline") is 3.3 volts.

0021This set of voltages results in the select transistor 104 being "on", which places 0 volts from the word line onto one terminal of the capactior 106. The other terminal of the capacitor 106 is connected to the bitline (V_bl) which is at 8 volts. Thus, the voltage across the capacitor 106 is 8 volts. The gate oxide of the capacitor 106 is designed to break down at this potential difference, which programs the FPGA cell as will be described further below. When the oxide of the capacitor 106 breaks down, this converts the capacitor 106 into a resistor.

0022It can be appreciated that the precise magnitude of voltages applied is dependent upon the thickness of the gate oxide and other factors. Thus, for example, for a 0.13 µm CMOS process, the gate oxide is typically thinner, thereby requiring a lower voltage differential across the capacitor 106.

0023With R₁ and C₁ being the selected row and column, consider the impact on the FPGA cell 102 at the crosspoint of a selected row and an unselected column ("SR/UC"), e.g., R₁ and C₂. As shown on line 305, the voltage on the wordline R₁ is 0 volts, the voltage at the unselected read bitline (V_blr) is 0 and the voltage on the unselected bitline C₂ is 0. Because 0 volts is on the gate of the select transistor 104, the FPGA cell 102 does not program under these conditions.

0024With R₁ and C₁ being the selected row and column, consider the impact on the FPGA cell 102 at the crosspoint of a selected column and an unselected row ("UR/SC"), e.g. R₂ and C₁. As shown on line 303, the voltage on the unselected wordline R₂ is 3.3 volts, the voltage at the selected read bitline (V_blr) is 3.3 volts, and the voltage on the bit line C₁ is 8 volts. Because the read bitline voltage is 3.3 volts, this causes the select transistor 104 to be in an "on" condition, allowing the 3.3 volts on the word line to be placed on one terminal of the capacitor 106. The other terminal of the capacitor is tied to the voltage on the bitline, or in this case 8 volts. This causes a potential difference of 4.7 volts across the gate oxide of the capacitor 106. The memory cell 102 is designed to not program under these conditions.

0025With R₁ and C₁ being the selected row and column, consider the impact on the FPGA cell 102 at the crosspoint of an unselected column and an unselected row ("UR/UC"), e.g. R₂ and C₂. As shown on line 307, the voltage on the unselected wordline R₂ is 3.3 volts, the voltage at the unselected read bitline (V_blr) is 0 volts, and the voltage on the unselected bitline C₂ is 0 volts. Because 0 volts is on the gate of the select transistor 104, the select transistor is in the "off" position, leaving one terminal of the capacitor 106 floating. The other terminal of the capacitor 106 is at 0 volts by virtue of its connection to the bitline. FPGA cell 102 does not program under these conditions.

0026After a FPGA cell 102 has been programmed by breaking down the gate oxide of the capacitor 106, the physical characteristics of the cell 102 is changed. In particular, the capacitor 106 becomes a resistive element. Note that, during programming, although the voltage across the select transistor oxide (i.e. 3.3 volts on the gate of the select transistor 104) is larger than is typical (1.8 volts for 0.18 µm CMOS), this higher voltage will not break down the gate oxide of the select transistor 104 because of the short programming time (normally less than a few seconds).

0027The FPGA array 100 is read in the following manner. A read select voltage of 1.8-3.3 volts is placed on the selected column bitline ("SC"), a read select voltage of 1.8 volts is placed on the selected bitline read (V_blr), and a voltage of 0 volts is placed on the selected row wordline ("SR"). Note that these voltages are for a typical 0.18 µm CMOS process. Lower voltages would be typically used for smaller more advanced CMOS processes. For example, for a 0.13 micron CMOS process, the read select voltage on the selected column bitline and on the selected read bitline may be approximately 1.2 volts.

0028Assume that R₁ and C₁ are the selected row and column ("SC/SR") and that the FPGA cell 102 at that crosspoint is programmed. As shown on line 309, 1.8-3.3 volts (a read select voltage) is applied via bitline C₁ to one terminal of the capacitor 106. Note that a higher bitline voltage will enable a higher read current to be sensed from the bitline. Further, the gate of the select transistor 104 has applied 1.8 volts via the read bitline and 0 volts is applied to the drain of the select transistor 104 via the wordline R₁. This causes the select transistor 104 to be "on". Nevertheless, even though the select transistor is "on", there is still some amount of resistance across the select transistor. Furthermore, there is 1.8-3.3 volts across the capacitor 106, which if programmed, will cause a leakage current (typically over 10 microamps) to flow from the selected column bitline to the selected row wordline. In effect, the programmed capacitor 106 and the select transistor 104 form a voltage divider, whose middle node is connected to the gate of the switch 108. The result of this voltage divider is that a voltage is placed on the gate of the switch 108 sufficient to turn on the switch 108. If the cell 102 has not been previously programmed, the capacitor 106 has a much higher resistance than the select transistor 104. Further, because the select transistor 104 is turned on, this causes the wordline voltage of 0 volts to be placed on the switch 108, which causes the switch 108 to be off.

0029With R₁ and C₁ being the selected row and column for the read operation, consider the impact on the cell 102 at the crosspoint of a selected column and an unselected row ("'UR/SC"), e.g. R₂ and C₁. As shown on line 311, 1.8-3.3 volts (a read select voltage) is applied via bitline C₁ to one terminal of the capacitor 106. Further, the gate of the select transistor 104 has applied 1.8 volts via the read bitline and 1.8 volts is applied to the source of the select transistor 104 via the wordline R₁. This causes the select transistor 104 to be "off".

0030With R₁ and C₁ being the selected row and column for the read operation, consider the impact on the cell 102 at the crosspoint of an unselected column and a selected row ("SR/UC"), e.g. R₁ and C₂. As shown on line 313, 0 volts (a read select voltage) is applied via bitline C₂ to one terminal of the capacitor 106. Further, the gate of the select transistor 104 has applied 0 volts via the read bitline and 0 volts is applied to the drain/source of the select transistor 104 via the wordline R₁. Under these circumstances, there will no current flow from the wordline to the bitline.

0031With R₁ and C₁ being the selected row and column for the read operation, consider the impact on the cell 102 at the crosspoint of an unselected column and a selected row ("UR/UC"), e.g. R₂ and C₂. As shown on line 315, 0 volts (a read select voltage) is applied via bitline C₂ to one terminal of the capacitor 106. Further, the gate of the select transistor 104 has applied 0 volts via the read bitline and 1.8 volts is applied to the drain/source of the select transistor 104 via the word line R₂. Under these circumstances, since the selected device 104 is at off state, there will no current flow from the wordline to the bitline.

0032During operation, the following voltages are used, first, a wordline voltage of 0 volts is applied. Next, a bitline voltage of 1.8 volts is applied and a bitline read voltage of 0 to 0.8 volts is applied. The V_blr of 0 to 0.8 volts is applied to make the select transistor 104 operate in a weakly on state, so only minimal leakage current occurs (on the order of nA).

0033In the arrangement described in <figref idref="f0001 f0002 f0003">Figures 1-3</figref>, the gate voltage of the switch 108 will be generally equal or less that Y_cc (1.8 volts for a 0.18 µm CMOS process). Thus, the switch 108 can only pass Y_cc- Y₁. This will affect the speed performance of the FPGA circuit. In an alternative embodiment, the switch 108 and the select transistor 104 have their gate oxides formed thicker, such as that used in input/output devices. For example, the thickness of the gate oxides for the switch 108 and the select transistor 104 may be on the order of 6 nm or thicker. The gate oxide of the capacitor 106 should be maintained at the conventional thickness for the particular CMOS process, e.g. 3 nm for a 0.18 µm CMOS process. The programming and read voltages for this alternative embodiment are shown in <figref idref="f0003">Figure 4</figref>.

0034In this alternative arrangement, during the read and programming operations, the bitline voltage can be biased to 3.3 volts (compared to 1.8 volts in the embodiment of <figref idref="f0001 f0002 f0003">Figures 1-3</figref>). The switch 108 will have 3.3 volts on its gate, so it can fully pass a Y_cc with extra drive gate voltage. Thus, the programmed switch will have very low resistance so as to improve the speed performance

0035While the above description for an FPGA provides significant improvement over the prior art, further improvement can be made to include dynamic reprogrammability as shown in the embodiments of <figref idref="f0004 f0005 f0006 f0007">Figures 5-9</figref>. For example, turning to <figref idref="f0004">Figure 5</figref>, a FPGA array is shown that includes a write transistor (Tw), a programming capacitor (Cp) that is made of gate poly over the thin gate oxide, and a control switch device (Tsw) that is used to control the programming logic. The structure of <figref idref="f0004">Figure 5</figref> is similar to that of <figref idref="f0001">Figure 1</figref>; however, the operation is different when the memory array is used as dynamic memory. The operation for non-volatile memory operation is the same as described above. Note that new nomenclature is used in <figref idref="f0004">Figure 5</figref>, where Bw (equivalent to Blr) is the bitline for writing, Bp (equivalent to Bl) is the bit line for non-volatile programming, WL is the word line. Vg-sw is the switch gate voltage.

0036Importantly, before the "hard" non-volatile programming (applying voltage to the Cp to breakdown the Gox), the cell can be used as dynamic memory by constant writing or refreshing of the charge stored in the switch gate and Cp. In this manner, the array can also be used as a DRAM style memory device. Specifically, the cell operation is detailed in <figref idref="f0004">Figure 6</figref>.

0037The writing or refreshing (rewriting) is done column by column by selecting all WL (row word line) lines and one write column bit line (Bw). Write or refresh of "1's" is done by setting Vwl to high (Vcc), and "0's" by setting Vwl to low (0V). The unselected columns (Vbw) are all at 0V. Therefore, as seen in <figref idref="f0004">Figure 6</figref>, if a cell is to be written with a "1", then the Bw line is high, the Bp line is low, and the WL is high. If a cell is to be written with a "0", then the Bw line is high, the Bp line is low, and the WL is low.

0038In one example, Vcc=1.8V for a 0.18 µm process. In order for the write transistor Tw (which is typically implemented as a NMOS transistor) to fully pass the "1", it is advantageous in some embodiments to use a lower Vdd (for example from 1.8V to 1.2V) in the FPGA logic circuits for higher speed and lower power consumption.

0039For example, assume that the memory array of <figref idref="f0004">Figure 5</figref> is to be used as dynamic memory and written with dynamic data. In such an operation, the "bit stream" for a specific column (the bit stream may be, for example switch control data for the FPGA) is loaded from, for example, external or on chip EPROM or flash to WL shift registers (not shown). If there are 1024 WLs (rows), then 1024 bits data will be read into 1024 bit shift registers. Then, the selected column Bw is at V_cc and 1024 bits in the column are simultaneously written to by the shift registers, with "1's" at V_cc and "0's" at 0 volts.

0040After the first column has been written (synonymous with refreshing), the bit stream for the 2^nd column is loaded into the shift registers, and the 2^nd column (Bw) is selected and refreshed. This operation is repeated column by column until the last column has been written or refreshed. The process is repeated again from the 1^st column, 2^nd column ..., and so on. Thus, the columns are constantly refreshed in a serial manner.

0041The write or refresh time is on the order of a few nS to uS for one column. If there are 1024 columns, then the refresh cycle time is in the range of a few µS to mS range. In one embodiment, the charge stored on the switch gate will not be reduced more than 10% within this time range.

0042During the writing or refreshing process, the waveforms have a time sequence on the WL and Bw that are designed in such a way to avoid discharging the data stored on the gate of control switch Tsw. More specifically, a WL with "1" should be ready at Vcc before Bw (V_blr) becomes Vcc which turns on the select NMOS switch (Tw) for refresh. Further, the Bw (V_blr) should become 0 volts to turn off the selected NMOS before WL drops back to 0 volts. This is illustrated in <figref idref="f0005">Figure 7</figref>. Writing or refreshing a "0" is similar, but the Vwl is at 0V instead of Vcc. Further, during periods between sequential column refreshes, Vwl is set at a bias of Vwb (low) to reduce the leakage current resulting from the Tw source drain leakage (I_doff) common when a deep sub-micron process is used.

0043Note that since the write or refreshing process is done column by column, there is no need for a special decoding circuit for the columns (bit line) for the refresh process. Instead, a simple closed loop shift register chain can be used.

0044The dynamic memory of the present invention uses refresh or rewrite without reading. The reading of the dynamic memory of the present invention will typically destroy the stored data or disturb the controlled switch gate voltage. This is in contrast with conventional DRAM in which the data or charge stored has to be read out first then followed by a refresh operation to restore the data. This prevents DRAM cell from being directly used for FPGA switch control, which requires static states to provide "off" or "on" switches.

0045The dual nature of the memory array described above is useful in many applications. By having the capability to be dynamic memory and non-volatile memory, this will aid in prototyping applications. In those types of applications, the user needs to program the FPGA chip multiple times, and after finalizing the design, the user can permanently program the FPGA as described above.

0046As seen above, using constant writing to refresh the switch memory requires continuously reading the design bit map from a non-volatile memory (external or internal). For large FPGA chips which require very high density configuration non-volatile memory, the access speed, the data shift-in speed, refreshing cycling time and I/O errors, etc, will be limiting factors.

0047In order to resolve these issues, in accordance with the present invention, a small sense device (Ts), diode (Td), and sense bit line (Bs) are added in parallel as shown as in <figref idref="f0006">Figure 8</figref>. This embodiment eliminates the need to constantly refresh from non-volatile memory, but instead is self-refreshing.

0048The self-refreshing process has two operational steps: the sensing and refreshing. In the sensing operation, all selected WLs will be pre-charged to high (~Vcc), and the selected sense column (Bs) will be pulled down to low (between 0V and Vcc/2). Further, all unselected Bs's will be kept or pre-charged to high (between Vcc-Vt to Vcc) to prevent any leakage current from the WL lines to the unselected BS lines through cells with "1" (stored positive charge and the Ts is on). In this way, only one cell on one WL is selected. If the selected FPGA cell is at "1", the sense device (Ts) will be on, so it can conduct a sense current (Isn) and it will pull down the WL to low. If the FPGA cell is at "0" (off), the sense device will be off, there will be not current, and the WL will be kept at high. Thus, the sense and refresh circuitry can sense and remember (latch the data) the state of the cells on the selected column and refresh them. The bias conditions for sense, write and non-volatile programming are illustrated in <figref idref="f0006">Figure 8</figref>. A top layout view is seen in <figref idref="f0007">Figure 9</figref>.

0049It should be noted that the diode (Td) can be put on either side of the sense device (Ts), with the same polarity. It can also be placed on opposite polarity with proper bias arrangement to prevent the leakage current on the unselected columns.

0050It should also be noted that, the diode Td can either be made by a gated diode using a MOS device with its gate connected to either source or drain terminal, or by using a P-N junction. This FPGA cell array can also be made using PMOS devices sitting in either an N-well or on an N-type substrate.

0051As seen in <figref idref="f0008 f0009">Figures 10-11</figref>, the cell array without switch devices can be also used as DRAM memory based on standard CMOS process, but the sensing circuits are greatly simplified. Again, the diode can be put on either side of the sense device to prevent the leakage from unselected devices, and the diode can be made by gated MOS devices. It can also be made by PMOS devices.

0052The embodiments described above show various structures for dual mode memory, i.e., non-volatile and dynamic memory. In a non-claimed example, a solely dynamic array may be formed. For example, referring to <figref idref="f0010">Figure 12A</figref>, by removing the capacitor (Cp) and program bitline (Bp), a structure is provided that acts as a dynamic memory. Further, by removing the capacitor Cp, the size of the cell is significantly reduced. Still, the operation is the same as that described above. <figref idref="f0011">Figure 12B</figref> shows a top layout view of the circuit of <figref idref="f0011">Figure 12B</figref>.

0053As seen in <figref idref="f0010">Figure 12A</figref>, the floating point (FP) node stores charge dynamically when the Tw device is in the off state. Further, the FP can be connected to an NMOS or PMOS gate to make a dynamically programmable switch used in a FPGA chip. The FP node can also be used as an input control of an inverter, that in turn can be used in a look up table (LUT) as used in most SRAM based FPGA's. These extended applications are described in greater detail below.

0054The FP controlled polysilicon gate of the NMOS or PMOS transistor or inverter in combination with the polysilicon gate of Tsw (NMOS or PMOS or inverter), Ts, as well as the junction capacitor of FP serves as the "capacitor" of the dynamic memory. The combination of the FP and the polysilicon gate of the transistor (NMOS or PMOS or Ts) is typically large enough to hold the charge (signal) such that refreshing is only required at a relatively low frequency.

0055<figref idref="f0012">Figures 13A, 13B, and 13C</figref> show in schematic form the FP node connected to a NMOS transistor, a PMOS transistor, and inverter, respectively (i.e. switch T_sw). The notation "XDM" stands for super dynamic memory, and "X" also has the meaning of using thin gate oxide capacitor as the memory element, either for breaking down for non-volatile memory or for storing dynamic charge.

0056As an other example, the programming capacitor Cp can be left in the cell. This is seen in <figref idref="f0013">Figures 14A</figref>-C. This capacitor Cp is the same as used in Figure5A for non-volatile operation. The retention of the programmable gate capacitor Cp added to the FP makes the cell non-volatile.

0057The memory array of <figref idref="f0010 f0011">Figure 12</figref> can further be modified by removal of the FPGA switch Tsw. The result is seen in <figref idref="f0014">Figures 15A</figref> and <figref idref="f0015">15B</figref>, which is substantially a three-transistor dynamic RAM memory.

0058In an alternative implementation, as seen in Figure 16A, there is a dedicated read select wordline (WLr) added, such that the write and read functions have separate controls. This implementation is similar to early 3T DRAM cells, see e.g. <nplcit id="ncit0001" npl-type="b"><text>D.A. Hodges and H. Jackson, Analysis and Design of Digital integrated Circuits, Semiconductor Memories, 2nd ed. New York: McGraw Hill, 1988</text></nplcit>, except that a dedicated capacitor is connected to the nodes to store the charge. The gate of Ts or Tsw, along with the junction of FP serves as the dedicated capacitor.

0059This implementation is generally suited for deep submicron semiconductor process technology (&lt;0.25µm) where the gate capacitor of Ts can be made large enough due to reduced gate oxide thickness (Gox is 4.5-5nm for 0.25µm, 3~3.5nm for 0.18µm, 2~2.3nm for 0.13µm, 16~1.9nm for 90nm process, etc.). Further, note that the transistor Tr can also be connected between Ts and GND.

0060The cell array of Figure 16A writes and reads row by row, and data are input and output from the bit line. It is understood that the naming of the WL and BL can be interchangeable. In either case, the data lines are input to the drain of the Tw, and output from the drain ofTr.

0061A programmable gate capacitor Cp can also be added to the FP to make it non-volatile based on the similar idea of using dynamic memory to control a programmable switch or inverter can be have many varieties.

0062Yet another example can be seen in <figref idref="f0017">FIGS. 17A-17B</figref>. In this example, the cell structure of the memory is significantly made smaller in size. Further, the memory and FPGA switch can be turned on by consistently writing the configuration data into the floating point (FP) without sensing the data out. The configuration can come from a variety of sources, such as on-chip embedded memory, external SRAM, flash memory, or one-time programmable (OTP) memory. In the embodiment of <figref idref="f0017">FIG. 17A</figref>, only two transistors are required for each memory cell. Note that the voltage on the row wordline for writing a "0" (Vlb) or standby is a bias which is in the range of the select device's (Tw) threshold voltage (Vt) to present subthreshold leakage.

0063For advanced process technology, the core gate oxide is very thin (20A for 0.13um, and 17A for 90nm). Thus, the gate dielectric tunnelling current can make the charge retention time very short. This requires that a relatively high frequency refrash is needed. One implementation is to use an I/O style device (30A for 1.8V, 50A for 2.5V and 70A for 3.3V) device for the switch device (Tsw) to prevent the gate tunnelling leakage. Alternatively, a PMOS thin gate oxide device can also be used to reduce the gate leakage.

0064Although the charge stored in FP cannot be sensed out during the operation of FPGA chip, which requires that the FP have a relatively stable potential for "1" for switch "on" and "0" for switch "off", it can be read out destructively to verify the normal functional of the array.

0065The following points are observed with respect to the various embodiments of the present invention described above: <ol id="ol0001" compact="compact"><li>1. The floating point (FP) can be achieved by turning off Tw, and it can be used to control the polysilicon gate of an NMOS or PMOS or the input of a CMOS inverter.</li><li>2. A non-zero bias Vwb (~Vt) is applied to the drain of Tw (assuming the source of Ts is connected to the poly gate of Ts) to reduce the sub-threshold leakage of the write transistors (Tw) to increase the charge storing time or extend the refreshing cycle time.</li><li>3. The data is written and read from the row wordline.</li><li>4. The WL and BLs can be interchanged.</li><li>5. The floating node can be used to control a switch device gate to make a programmable switch for the application of FPGA.</li><li>6. The potential on the floating node (Vfp) should be higher than the switch source or drain voltage (usually logic Vcc) in order to fully pass the logical level and to reduce the effect of voltage coupling from the switch devices. In one embodiment, Vfp should be higher than (1+CR)*Vcc+Vt, where "CR" is the coupling ratio of the switch gate to the floating node.</li><li>7. The column write bitline potential is pumped to a higher voltage so as to pass full wordline write voltage to the floating node to achieve desired Vfg.</li><li>8. The NLDD implanted is blocked in the Td transistor so that there is no gate overlap between the polysilicon gate and n+ S/D diffusion. After the Cp has been programmed, there is a reverse diode formed between the floating node and the program bit line (Bp) to the dynamic memory and is still working even after the cell be programmed to non-volatile memory.</li><li></li></ol>

0066The description of the invention and its applications as set forth herein is illustrative and is not intended to limit the scope of the invention. Variations and modifications of the embodiments disclosed herein are possible, and practical alternatives to, or equivalents of the various elements, of the embodiments are known to those of ordinary skill in the art. For example, the various voltages set forth in the various examples are only illustrative, since one has some discretion as to the precise voltage to select within a range of voltages, and the voltages are in any event dependent on the device characteristics. The terms row wordline and column bitline have been used to describe types of lines commonly used in memories, but some memories may have alternatives thereto. Further, the various doping types may be reversed, such that an n-channel transistor described above may be replaced with a p-channel transistor. These and other variations and modifications of the embodiments disclosed herein may be made without departing from the scope of the invention.