Flash memory with improved erasability and its circuitry

0001The present invention relates to a so-called flash memory or a nonvolatile semiconductor memory from which existent information can be electrically erased concurrently or block by block. In particular, this invention is concerned with improvements in erasing a flash memory and improvements in the associated circuitry.

0002In recent years, a variety of semiconductor memories such as an electrically erasable programmable read only memory (EEPROM), which can be rewritten by electrically erasing data stored therein, have been developed in the field of nonvolatile memories.

0003The semiconductor memories are nonvolatile memories which are made rewritable by electrically erasing specified data previously written therein.

0004As for this kind of nonvolatile memory, in consideration of the use as a substitute for a magnetic storage or the like, efforts have been made to increase the capacity and decrease the cost. For use in a portable information processing terminal or the like, however, it is required to lower the voltage requirements.

0005Known semiconductor memories, which are nonvolatile memories that are made rewritable by erasing specified data previously written therein, include an erasable programmable read only memory (EPROM) and the EEPROM.

0006The EPROM has the advantage of a small cell size. However, data erasing is a nuisance because ultraviolet light must be used to erase data. As for the EEPROM, data erasing is easy to do because data can be erased electrically. However, the cell size in the EEPROM is larger than that in the EPROM, which makes it difficult to increase the capacity of an EEPROM.

0007As a semiconductor memory having the advantages of the foregoing memories, for example, what is referred to as a flash memory has been developed.

0008The term "flash memory" will be adopted in the succeeding description.

0009The nonvolatile semiconductor memory has an overall configuration, for example, as that shown in Fig. 1. In Fig. 1, N cell blocks 11₁ to 11_N (for example, eight blocks) each having multiple transistor cells are arranged in a row. The cell blocks 11₁ to 11_N are provided with bit line select circuits 12₁ to 12_N and sense amplifiers/sense buffers 13₁ to 13_N respectively.

0010A row address signal is supplied from a row address buffer to each of the cell blocks 11₁ to 11_N via a row decoder 15. A column address signal is supplied from a column address buffer 16 to each of the bit line select circuits 12₁ to 12_N via a column address decoder 17. A voltage V_S sent from a source power supply circuit 18 is applied to each of source electrodes of the transistor cells of the cell blocks 11₁ to 11_N.

0011Fig. 2 shows the cell block 11, one of the cells blocks in the aforesaid nonvolatile semiconductor memory, and its peripheral circuitry. In Fig. 2, components identical to those in Fig. 1 bear the same reference numerals, of which a description will be omitted. In Fig. 2, the bit line select circuit 12 comprises n-channel MOS field-effect transistors Q1 to Qn. Column address signals Y1 to Yn are fed from the column decoder 17 to the gates of the transistors Q1 to Qn.

0012The cell block 11 comprises a total of n by n field-effect transistors Q11 to Qnn having floating gates and control gates. The gates of n transistors Qil to Qin (where, i = 1, 2, etc., and n) arranged in tandem are provided with row address signals Xi sent from the row decoder 15 via word lines.

0013The drains of the n transistors Qli to Qni arranged in tandem are connected to the drains of the transistors Qi in the bit line select circuit 12 via bit lines. Voltage from the source power circuit 18 is applied to each of the sources of the transistors Qll to Qnn. A cell amplifier 13a and a write buffer 13b are connected to each of the sources of the transistors Q1 to Qn.

0014In the foregoing semiconductor memory, when row addresses Xi and column addresses Yi are selected for writing, data read from the write buffer 13 are written in the transistors Qij of the cell blocks 11₁ to 11_N. Writing is performed simultaneously on one bit per each of the cell blocks 11₁ to 11_N or a total of N bits designated with the row addresses and column addresses. Data erasing is performed concurrently on all transistors in the cell blocks 11₁ to 11_N.

0015In a flash memory, information is retained depending on the presence or absence of a charge in a memory cell. Fig. 3 shows an example of a structure of a memory cell. As shown in Fig. 3, a gate has a two-layered structure consisting of a control gate (CG) 25 and a floating gate (FG) 24. The control gate 25 is connected to a word line WLi and a drain (D) 23 is connected to a bit line BLi. Reference numeral 26 denotes a tunneling oxide film.

0016The flash memory is broadly divided into two types of what are referred to as NOR and NAND. These types differ from each other in a method of writing, reading, or erasing information into or from a memory cell. Taking the NOR type flash memory as an example, writing, reading, or erasing information into or from a memory cell will be described below.

0017When information is to be written in a memory cell having the aforesaid structure, as shown in Fig. 4, the word line WLi is set to Vpp (approx. 12 V), the bit line BLi is set to approx. 6 V, and the source S is set to 0 V. High voltage is then applied to the control gate CG and drain D. Current then flows into the memory cell. Part of the electrons flowing through the memory cell are accelerated due to the high electric field in the vicinity of the drain D, gain energy, and then goes beyond the energy barrier of an insulating film of the floating gate. The electrons are finally injected into the floating gate FG. The floating gate FG is not electrically coupled with other circuits, so it therefore can retain charges on a semi-permanent basis.

0018When information is to be read from a memory cell, as shown in Fig. 5, the word line WLi is set to Vcc (about 5 V), the bit line BLi is set to about 1 V, and the source S is set to 0 V. The memory cell is then selected by specifying the word line WLi and bit line BLi. The threshold value of the cell transistor varies depending on the charges retained in the floating gate FG. Current flowing through the selected memory cell varies depending on the information stored therein. The information therefore can be read out by detecting and amplifying the current.

0019The voltage levels of the control gate CG, drain D, source S, and substrate PS in the aforesaid operative states are set to the values listed in Table 1. <tables id="tabl0001" num="0001"><table frame="all"><title>Table 1</title><tgroup cols="5" colsep="1" rowsep="0"><colspec colnum="1" colname="col1" colwidth="31.50mm" /><colspec colnum="2" colname="col2" colwidth="31.50mm" /><colspec colnum="3" colname="col3" colwidth="31.50mm" /><colspec colnum="4" colname="col4" colwidth="31.50mm" /><colspec colnum="5" colname="col5" colwidth="31.50mm" /><thead valign="top"><row rowsep="1"><entry namest="col1" nameend="col5" align="center">Voltages in modes in a prior art</entry></row><row rowsep="1"><entry namest="col1" nameend="col1" /><entry namest="col2" nameend="col2" align="left">CG</entry><entry namest="col3" nameend="col3" align="left">D</entry><entry namest="col4" nameend="col4" align="left">S</entry><entry namest="col5" nameend="col5" align="left">PS</entry></row></thead><tbody valign="top"><row><entry namest="col1" nameend="col1" align="left">Reading</entry><entry namest="col2" nameend="col2" align="left">Vcc</entry><entry namest="col3" nameend="col3" align="left">to 1 V</entry><entry namest="col4" nameend="col4" align="left">0 V</entry><entry namest="col5" nameend="col5" align="left">0 V</entry></row><row><entry namest="col1" nameend="col1" align="left">Writing</entry><entry namest="col2" nameend="col2" align="left">Vpp</entry><entry namest="col3" nameend="col3" align="left">to 6 V</entry><entry namest="col4" nameend="col4" align="left">0 V</entry><entry namest="col5" nameend="col5" align="left">0 V</entry></row><row rowsep="1"><entry namest="col1" nameend="col1" align="left">Erasing</entry><entry namest="col2" nameend="col2" align="left">0 V</entry><entry namest="col3" nameend="col3" align="left">Float</entry><entry namest="col4" nameend="col4" align="left">Vpp</entry><entry namest="col5" nameend="col5" align="left">0 V</entry></row></tbody></tgroup></table></tables>

0020When information is to be erased from a memory cell, as shown in Fig. 6, the word line WLi is set to about 0 V and the bit line BLi is opened. In this state, the drain D is opened, about 0 volt is applied to the control gate CG, and a high voltage of about 12 volts is applied to the source S.

0021Since a high voltage is applied to the source S, deep diffusion is required in order to increase the resistivity of the diffused layer in the source. This hinders reduction in cell area.

0022For divided erasing, it is required that the Vss line in the source must partly have a different voltage. This leads to disconnection or an increased number of drive circuits. Eventually, chip size increases.

0023A solution to the above problem is to apply a negative voltage to the word line WLi. To be more specific, as shown in Fig. 7, a negative voltage (about - 10 V) is applied to the control gate CG and Vcc (about 5 V) is applied to the source S. The drain D is opened. Erasing is then executed.

0024In this case, since a low voltage is applied to the source S, the resistivity of the source need not be intensified. This contributes to reduction in cell size. Partial erasing is enabled by selectively applying negative voltage to the control gates CG.

0025The aforesaid erasing method is a source erasing method in which charges in the floating gate FG are routed to the source. A channel erasing method is also available, wherein charges in the floating gate are routed to a channel; that is, a substrate. Even in this method, negative voltage is applied to the control gate. The channel erasing method is sometimes employed for the aforesaid NAND-type flash memory.

0026Figs. 8 to 11 show the states of a memory cell with voltage applied according to various erasing methods. In Figs. 8 to 11, the memory cell is an n-channel transistor.

0027Fig. 8 shows a state in which positive voltage is applied according to a channel erasing method. The drain D and source S are opened, and the control gate CG is set to 0 V. The high voltage Vpp is applied to the P well equivalent to a channel. In channel erasing, a triple-well structure shown in Fig. 8 is adopted because positive bias is applied to the channel.

0028Fig. 9 shows a state in which a positive voltage is applied according to the source erasing method. The drain D is opened, and then the control gate CG is set to 0 V. The high voltage Vpp is applied to the source S. The substrate is opened or set to 0 V.

0029Fig. 10 shows a state in which a negative voltage is applied according to the channel erasing method. The drain D and source S are opened, and the control gate CG is set to a negative voltage V_BB. A positive voltage Vcc is applied to the p well equivalent to a channel. V_BB - V_CC is applied between the control gate CG and channel.

0030Fig. 11 shows a state in which negative voltage is applied according to the source erasing method. The drain D is opened, and then the control gate CG is set to the negative voltage V_BB. The source S is set to the positive voltage Vcc.

0031A flash memory includes an internal power switching circuit for switching supply voltages depending on a mode. A previously-proposed internal power switching circuit has a simple circuitry but is likely to cause a latch-up phenomenon. The switching speed is decreased in order to avoid the latch-up phenomenon. This contradicts efforts to speeding up processing.

0032Accordingly, it is desirable to realize an internal power switching circuit that can operate at a high speed with simple circuitry.

0033According to an embodiment of an aspect of the present invention there is provided an internal power switching circuit for use in a flash memory or the like, comprising a first MOS transistor of a first polarity one of whose drain and source electrodes is connected to a first power line and the other of whose drain and source electrodes is connected to a second power line; <ul id="ul0001" list-style="none" compact="compact"><li>a second MOS transistor of a second polarity one of whose drain and source electrodes and whose well are connected to a third power line having a higher voltage than the first power line, and the other one of whose drain and source electrodes is connected to a node; and</li><li>a third MOS transistor of the second polarity one of whose drain and source electrodes is connected to the node and the other of whose drain and source electrodes is connected to the second power line. The internal power switching circuit further includes a fourth MOS transistor of the first polarity one of whose drain and source electrodes is connected to the third power line and the other of whose drain and source electrodes is connected to the second power line.</li></ul>

0034In such an internal power switching circuit, when a low voltage is switched to a high voltage, the fourth MOS transistor assists in boosting the voltage of the second power line. This helps minimize the channel current in the second MOS transistor. A latch-up phenomenon can therefore be avoided.

0035Reference will now be made, by way of example, to the accompanying drawings, in which: <ul id="ul0002" list-style="none" compact="compact"><li>Fig. 1 shows an overall configuration of a flash memory in which the present invention is implemented;</li><li>Fig. 2 is a circuit diagram showing a major section of Fig. 1;</li><li>Fig. 3 shows a structure of a memory cell;</li><li>Figs. 4 to 7 are explanatory diagrams for methods of reading, writing, and erasing a flash memory; Fig. 4 shows the conditions for writing; Fig. 5 shows the conditions for reading; Fig. 6 shows the conditions for erasing; Fig. 7 shows the conditions for erasing based on negative-voltage application;</li><li>Fig. 8 is an explanatory diagram for channel erasing based on high-voltage application;</li><li>Fig. 9 is an explanatory diagram for source erasing based on high-voltage application;</li><li>Fig. 10 is an explanatory diagram for channel erasing based on negative-voltage application;</li><li>Fig. 11 is an explanatory diagram for source erasing based on negative-voltage application;</li><li>Fig. 12 shows a structure of a transistor cell;</li><li>Fig. 13 is an explanatory diagram for excessive erasing;</li><li>Fig. 14 shows an example of a previously-proposed internal power switching circuit;</li><li>Fig. 15 shows waveforms showing the operations of Fig. 14;</li><li>Fig. 16 is an explanatory diagram for occurrence of a latch-up;</li><li>Fig. 17 is an equivalent circuit of Fig. 16;</li><li>Fig. 18 is a configurational diagram showing the principle of an embodiment of the present invention;</li><li>Fig. 19 is an overall block diagram showing a flash memory embodying the invention;</li><li>Fig.20 shows components of a row decoder in the Fig. 19 embodiment;</li><li>Fig. 21 shows components of a column decoder in the Fig. 19 embodiment;</li><li>Fig. 22 shows components of a write amplifier in the Fig. 19 embodiment;</li><li>Fig.23 shows components of an erase amplifier in the Fig. 19 embodiment;</li><li>Fig. 24 shows components of blocks B1 and B2 of an internal power switching circuit in the Fig. 19 embodiment;</li><li>Fig. 25 shows components of blocks B3 and B4 of the high-voltage detector and internal power switching circuit in the Fig. 19 embodiment;</li><li>Figs. 26A to 26G show waves indicating the process of producing a signal G3 in the block B2 of the internal power switching circuit in the Fig. 19 embodiment; and</li><li>Fig. 27 shows waveforms representing the operations of the block B1 of the internal power switching circuit in the Fig. 19 embodiment. Prior to an embodiment , pre-erase writing will be described.</li></ul>

0036Fig. 12 shows a structure of a memory cell shown in Fig. 3 in more detail. A transistor serving as the memory cell has, as shown in Fig. 12, such a structure that n-type diffusive regions 32 and 33 are formed in a p-type substrate 31 with a certain space between them. A floating gate (FG) 34 and a control gate (CG) 35 are formed above the p-type substrate 31, and coated with an oxide film 36. The n-type diffusive regions 32 and 33 are coupled with a source electrode 37 and a drain electrode 39 respectively. The control gate 35 is coupled with a gate electrode 38.

0037An insulating film for isolating between the floating gate (FG) and control gate (CG) in the stacked gate type memory cell transistor may be an oxide film or oxide nitride film. However, an ONO film; that is, a three-layered structure of an oxide film over a nitride film over an oxide film is preferred in an effort to prevent a time-sequential leakage of charges from the floating gate (FG). Using the same process for producing the insulating film between gate electrodes, an insulating film for a gate in a single gate-type transistor which is used in any region of peripheral circuits may be produced.

0038In this case, the gate insulating film for the single gate-type transistor formed in a region of peripheral circuits may be the ONO film or oxide nitride film. The gate insulating films for transistors except memory cell transistors are not necessarily oxide films.

0039Writing in a memory cell is achieved by injecting electrons into the floating gate (FG) 34, whereby "0" will be written. Erasing is achieved by removing electrons from the floating gate (FG) 34, whereby "1" will be written.

0040When a source erasing method based on high-voltage application is performed to erase the transistor cell (to write "1"), as described previously, high voltage is applied to the source electrode 37, zero volt is applied to the gate electrode 38, and the drain electrode 39 is opened. The electrons retained in the floating gate 34 are thus removed.

0041When "0" has been written in the transistor cell; that is, when electrons have been retained in the floating gate 34, the threshold for the cell, I, in Fig. 13 -decreases to be lower than a threshold level for identifying data "1" for the period of an erasing time T.

0042In contrast, when "1" has been written in the transistor cell; that is, when no electrons have been retained in the floating gate 34, since the threshold level is low, as the threshold of the cell decreases as indicated with II in Fig. 13 due to the erasing, the threshold level becomes 0 for a shorter period of time than the erasing time T. During the above erasing, too many electrons are removed from transistor cells in which "ls" have been written. The transistor cells consequently enter a Normally On state.

0043In order to prevent the foregoing excessive erasing, it is necessary for a flash memory, in which cells are NORed, to write data "0" in each of the transistor cells in all cell blocks before concurrent erasing is performed. Writing data "0" is performed on all transistor cells, wherein the data values written in the transistor cells are not checked. Specifically, a source voltage Vs for the source electrode 37 in the transistor cell shown in Fig. 12 is set to 0 V, a gate voltage Vg for the gate electrode 38 is set to a high voltage, and a drain voltage V_D for the drain electrode 39 is set to a high voltage. High-energy electrons developed by applying high voltage between the source and drain then reach the floating gate 84 through the oxide film 36. Data "0" is thus written in the transistor cell.

0044As apparent from the previous description, in a flash memory, the electrode voltages of a memory cell transistor differ from the read mode to the write (or erase) mode. Therefore, the internal supply voltage must be changed for each mode.

0045Fig. 14 shows the components of a typical previously-proposed internal power switching circuit.

0046Reference numeral 650 denotes a power line (hereinafter, external power line) on which an external power supply Vcc of, for example, + 5 V is placed. 651 denotes a power line (hereinafter, external power line) on which an external power supply Vpp of, for example, +12 V is placed; and 652 denotes a power line (hereinafter, internal power line) for an internal power supply V_PI that is connected to a control electrode of a memory cell transistor via a row decoder which is not shown.

0047A MOS transistor 653, which is turned on when a signal G1 is high, is connected between the external power line 650 and internal power line 652. A MOS transistor 654, which is turned on when a signal G2 is low, and a MOS transistor 655, which is complementary to the transistor 653, that is, which turned on when the transistor 653 is off, are connected in series with each other via a node n60. The MOS transistor 653 is of the n-channel depletion type, and the MOS transistors 654 and 655 are of the p-channel enhancement type. The well (n type) of the transistor 654 is connected to the external power line 651, and the well (n type) of the transistor 655 is connected to the internal power line 652.

0048According to the foregoing circuitry, when the signals G1 and G2 are driven high, the transistor 653 is turned on and the transistors 654 and 655 are turned off. The voltage (Vcc of + 5 V) of the external power line 650 is reflected in the internal power line 652. On the contrary, when the signals G1 and G2 are driven low, the transistor 653 is turned off and the transistors 654 and 655 are turned on. The voltage (Vpp of + 12 V) of the external power line is reflected in the internal power line 652. When the levels of the signals G1 and G2 are controlled for each of read and write modes, an internal supply voltage V_PI representing a voltage (Vcc or Vpp) suitable for each mode can be generated.

0049Fig. 15 is an example of an actual timing chart for the Figure 14 circuit. In read mode, the signal G1 is high (Vcc of + 5 V) and the signal G2 is high (Vpp of + 12 V). As a result, V_PI equals to Vcc (+ 5 V).

0050For transition to the write mode, first, the signal G1 is driven low. This turns off the transistor 653 and turns on the transistor 655. Thereafter, the level of the signal G2 is decreased gradually to be low. The voltage at the node n60 becomes V_PI (+ 5 V) because the transistor 655 is turned on. After time elapses, when the level of the signal G2 decreases to such an extent that the transistor 654 is turned on, the external power line 651 and node n60 are connected to each other via the transistor 654. The node n60 is connected to the internal power line 652 via the transistor 655 which has already been turned on. Current therefore flows from the external power line 651 to the internal power line 652. Eventually, V_PI is switched to Vpp (+ 12 V).

0051In transition to the write mode, the level of the signal G2 is decreased gradually. The reason will be described below.

0052That is to say, when the level of the signal G2 (gate level of the transistor 654) is driven low (0 V) quickly, an enormous channel current i_ch2 proportional to a difference between the Vpp (+ 12 V) and gate level (0 V) flows through the transistor 654. At this time, only a limited channel current i_ch3 proportional to a relatively small difference between the voltage at the node n60 (+ 5 V) and gate level (0 V) flows through the transistor 655. A residual current (i_ch2 - i_ch3) flows from the p-channel diffused layer in the source region (node n60) of the transistor 655 into the n-type well. A large substrate current then develops.

0053As a result, a parasitic pnpn thyristor is turned on, and a MOS device is held at low impedance. Thus a fatal phenomenon; that is, a latch-up phenomenon is likely to occur.

0054A mechanism for causing a latch-up will be described briefly. Fig. 16 shows the structure of that portion of the internal power switching circuit in Fig. 14 corresponding to the transistors 654 and 655. Fig. 17 shows an equivalent circuit. The same components in Figs. 16 and 17 bear the same reference numerals. Referring to Fig. 17, a mechanism of occurrence of a latch-up will be described. In Fig. 17, - denotes an electron, and + denotes a hole.

0055When the signal G1 is 0 V, the signal G2 changes from Vpp to 0 V. Currents i_Q2 and i_Q3 flow. When the node n60 floats by about 0.7 V or more from the n-type well 662 whose voltage is V_PI, the p⁺n junction is biased in the forward direction. Holes are injected from the p⁺-type region into the n-type well 662 along a route 1. The holes are minority carriers within the n-type well. When diffused, the holes reach the well boundary and flow into the p-type substrate 660. Within the p-type substrate 660, the holes behave as majority carriers and flow into a p⁺-type contact 669.

0056At this time, a resistor 676 placed between the p⁺-type contact 669 and a source n⁺ 668 of an n-channel transistor 680 located on the other periphery causes the substrate voltage to partly rise. When the substrate voltage arising near the source n⁺ 668 rises by about 0.7 V, the n⁺p junction is biased in the forward direction. As a result, electrons are injected from the source n⁺ 668 to the p-type substrate 660 as indicated with 2. When diffused, the electrons reach the well boundary and flow into a contact n⁺ 667 of the n-type well 662. The voltage drop resulting from the current and resistance of the resistor 678 causes the voltage of the n-type well near a source p⁺ 665 of a p-channel transistor 655 to decrease. The p⁺n junction is then biased in the forward direction. Holes are injected again from the source p⁺ 665 into the n-type well 662. This increases substrate current. Positive feedback is thus repeated.

0057Due to the rise in substrate voltage, electrons injected from the n⁺-type source 668 to the p-type substrate 660 are diffused to move within the p-type substrate, and then flow into the n-type well 661 whose voltage is Vpp. Within the n-type well 661, electrons flow as majority carriers into an n⁺-type source 663. The voltage of the n-type well 661 partly decreases because of a voltage drop caused by a resistor 679 located between the contact of the n⁺-type source 663 and the p⁺-type source 664 of a p-channel transistor 654. When the voltage of the n-type well 661 in the vicinity of the p⁺-type source 664 drops by about 0.7 V, the p⁺n junction is biased in the forward direction. As a result, holes are injected from the p⁺-type source 664 to the n-type well 660 as indicated by 5. The holes reach the well boundary while diffusing, and then flow into the contact of the p-type substrate 660 with the p⁺-type source 669. The current and resistance of the resistor 676 boost the substrate voltage.

0058Consequently, the n⁺p junction is biased in the forward direction. Electrons are injected from the n⁺-type source to the p-type substrate again as indicated with 4. Positive feedback is thus repeated.

0059As a result, overcurrent steadily flows from the external power supply line Vpp to V_SS. This causes a latch-up.

0060In some previously-proposed circuits, the level of the signal G2 is decreased gradually in order to avoid an abrupt increase in channel current i_ch2. Substrate current is thus minimized to prevent occurrence of a latch-up phenomenon.

0061However, in this kind of previously-proposed internal power switching circuit for a semiconductor memory, a latency time must be present between the read and write modes in order to avoid a latch-up phenomenon. The latency time includes the time constant for decreasing the level of the signal G₁ and the time for absorbing a manufacturing error of each transistor. Mode change cannot be speeded up, which has obstructed realization of a high-speed external memory that is freely rewritable.

0062An internal power switching circuit embodying the present invention can resolve the aforesaid problems and can prove effective for a flash memory or any other semiconductor memory in which the voltage levels of an internal power supply are switched.

0063Fig. 18 is an explanatory diagram showing a principle of an internal power switching circuit embodying the present invention. As shown in Fig. 18,the internal power switching circuit comprises:a first n-channel MOS transistor 704, one of whose drain and source electrodes is connected to a first power line 701, and the other one of whose drain and source electrodes is connected to a second power line 702; a second p-channel MOS transistor 705,one of whose drain and source electrodes as well as whose well are connected to a third power line 703 of a voltage Vpp that is higher than the voltage Vcc of the first power line 701, and the other of whose drain and source electrodes is connected to a node n70; and a third p-channel MOS transistor 706, one of whose drain and source electrodes is connected to the node n 70 and the other of whose drain and source electrodes as well as whose well are connected to the second power line 702. The internal power switching circuit further includes a fourth n-channel MOS transistor 707 one of whose drain and source electrodes is connected to the third power line 703 and the other of whose drain and source electrodes is connected to the second power line 702.

0064In the foregoing circuit, if the n-channel transistor 707 is turned on during transition from the read mode to write mode, the voltage V_PI of the second power line 702 is lifted via the n-channel transistor 707 and exceeds the voltage Vcc of the first power line 702 (the lift-up voltage depends on the gate voltage of the n-channel transistor 707). The lifted V_PI is transmitted to the node n70 via the p-channel transistor 706 that is on.

0065During mode transition, the voltage at the node n70 is boosted to exceed at least Vcc. The drain-source voltage (equivalent to a potential difference between the n70 and Vpp) can therefore be decreased to minimize the channel current of the p-channel transistor 705. Furthermore, the gate-source voltage (equivalent to a potential difference between G1 and n1) of the p-channel transistor 706 can be increased to expand the channel current-carrying capacity of the p-channel transistor 706. As a result, the channel current in the p-channel transistor 706 is balanced with the channel current in the p-channel-transistor 705. Consequently, substrate current is minimized to prevent occurrence of a latch-up phenomenon reliably.

0066In such an internal power switching circuit even when the gate level of the p-channel transistor 705 is driven low abruptly, a latch-up phenomenon does not occur. The latency time at the time of mode transition can be minimized to be equal to the time required to lift up the level of the node n70. Such a circuit can thus provide an internal power switching art helpful in realizing a high-speed external memory that can be freely rewritable.

0067An embodiment of the present invention will be described in conjunction with the drawings.

0068Figs. 19 to 27 show this embodiment in which the present invention is implemented in a flash memory.

0069To begin with, the configuration will be described. In Fig. 19, reference numeral 711 denotes a flash memory (hereinafter, EEPROM). The EEPROM 711 comprises a row address buffer 712, a column address buffer 713, a row pre-decoder 714, a column pre-decoder 715, a row decoder 716, a column decoder 717, a memory cell array 718, a column gate 719, a sense amplifier 720, a write amplifier 721, a data-out buffer 722, a data-in uffer 723, an erase amplifier 723, a control circuit 725, a control buffer 726, an internal power switching circuit 727, and a high-voltage detector 728. Also in the figure, 729 denotes a row address signal terminal for handling multiple bits; 730 denotes a column address signal terminal for handling multiple bits; 731 denotes a data-in/out terminal; 732 denotes a control signal terminal for handling various control signals; and 733 denotes a high-voltage power (Vpp) terminal for writing.

0070The internal power switching circuit 727 selects either of external supply voltages Vcc and Vpp according to the output of the high-voltage detector 728, and outputs the selected voltage as an internal supply voltage V_PI. In this example, the internal supply voltage V_PI is supplied to each of the row decoder 716, column decoder 717, write amplifier 721, and erase amplifier 724.

0071Fig. 20 shows components of the row decoder 716. In the row decoder 716, multiple n-channel MOS (hereinafter, nMOS) transistors 760 to 762, which receive signals from the row pre-decoder 714 to the gates thereof, and an nMOS transistor 763 serving as a load element are connected between the internal supply voltage (V_PI) terminal and ground (0 V) terminal. When the transistors 760 to 762 are all on; that is, when the signals from the row pre-decoder 714 are all high, the output of a CMOS inverter gate 716a made up of a p-channel MOS (hereinafter, pMOS) transistor 764 and an nMOS transistor 765 represents the internal supply voltage V_PI. The output (V_PI or 0 V) of the CMOS inverter gate 716a is connected to a word line in a memory cell array 718, and supplied to control gates of memory transistors, which are arranged in a row in the memory cell array 718, via the word line.

0072Fig. 21 shows components of the column decoder 717. In the column decoder 717, multiple nMOS transistors 766 and 767, which receive signals from the column pre-decoder 715 to the gates thereof, as well as a nMOS transistor 768 serving as a load element are connected between the internal supply voltage (V_PI) terminal and ground (0 V) terminal. When the nMOS transistors 766 and 767 are all on; that is, the signals from the column pre-decoder 715 are all high, the output of a CMOS inverter gate 717a made up of a pMOS transistor 769 and an nMOS transistor 770 represents the internal supply voltage V_PI. The output of the CMOS inverter gate 717a (V_PI or 0 V) is supplied to the gate of a transistor for selecting a bit line in the column gate 719.

0073Fig. 22 shows components of the write amplifier 721. In the write amplifier 721, two nMOS transistors 771 and 772, which are turned on when a signal sent from the control circuit 725 to represent a write mode and write data fetched from the data-in buffer 723 are both high, and an nMOS transistor 773 serving as a load element are connected between the internal supply voltage (V_PI) terminal and ground (0 V) terminal. The write amplifier 721 further includes a CMOS inverter gate 721a made up of a pMOS transistor 774 and an nMOS transistor 775, and an nMOS transistor for output. When the two nMOS transistors 771 and 772 are turned on, the internal supply voltage V_PI is supplied to a bit line specified by the column gate 719 via the output transistor 776.

0074Fig. 23 shows components of the erase amplifier 724. In the erase amplifier 724, an nMOS transistor 777, which is turned on in response to a high-level signal that is sent from the control circuit 725 and represents an erase mode, and an nMOS transistor 778 serving as a load element are connected between an internal supply voltage (V_PI) terminal and a ground (0 V) terminal. A CMOS inverter gate 724a made up of a pMOS transistor 779 and an nMOS transistor 780, a CMOS inverter gate 724b made up of a pMOS transistor 781 and nMOS transistor 782, and a CMOS inverter gate 724c made up of a pMOS transistor 783 and an nMOS transistor 784 are connected between the internal supply voltage (V_PI) terminal and ground (0 V) terminal. When the nMOS transistor 777 is on, the internal supply voltage V_PI is supplied to all source electrodes of memory cell transistors in the memory cell array 718.

0075Figs. 24 and 25 show components of the internal power switching circuit 727 and high-voltage detector 728 respectively. B1 and B2 in Fig. 24, as well as B3 and B4 in Fig. 25 denote four blocks constituting the internal power switching circuit 727. The block B1 has the same circuitry as that in Fig. 18. The same reference numerals as those in Fig. 18 are used.

0076Reference numeral 701 denotes a power line (first power line) for an external power supply Vcc; 702 denotes a power line (second power line) for an internal power supply V_PI; 703 denotes a power line for an external power supply Vpp (where, Vpp > Vcc); 704 denotes a first n-channel MOS transistor one of whose drain and source electrodes is connected to the first power line 701, and the other of whose drain and source electrodes is connected to the second power line 702; 703 denotes a second p-channel MOS transistor one of whose drain and source electrodes and whose well are connected to a third power line 703, and the other of whose drain and source electrodes is connected to a node n70; 706 denotes a third p-channel MOS transistor one of whose drain and source electrodes is connected to the node n70 and the other of whose drain and source electrodes and whose well are connected to the second power line 702; and 707 denotes a fourth n-channel MOS transistor one of whose drain and source electrodes is connected to the third power line 703 and the other of whose drain and source electrodes is connected to the second power line 702. The fourth n-channel MOS transistor 707 is a key circuit element in this embodiment.

0077The blocks B2 to B4 generate signals G1, G2, and G3 for controlling the on or off operations of the MOS transistors 704 to 707. The block B3 generates the signal G1 whose logical amplitude ranges from 0 V to Vcc and the signal G4 whose logical amplitude ranges from 0 V to Vpp in response to an output of the high-voltage detector 728. The block B4 generates a delay signal G2 for the signal G4.

0078When the voltage Vpp exceeds the voltage Vcc at least by the sum of a threshold voltage Vth of a pMOS transistor 785 and a threshold voltage Vth of a pMOS transistor 786 (Vpp > Vcc), the high-voltage detector 728 provides a high-level signal equivalent to Vpp - 2Vth as an output of a CMOS inverter gate 728a made up of the pMOS transistor 786 and nMOS transistor 787, and outputs the high-level signal via two-stage inverter gates 728b and 728c with the phase unchanged.

0079When the output of the high-voltage detector 728 is high; that is, Vpp is larger than Vcc, the block B3 provides a low-level signal for two output node a and b of a state hold circuit 748 made up of a NOR gate 734, a NAND gate 735, inverter gates 736 and 737, and capacitors 821 and 822. The block B3 then fetches a signal from the node a via two-stage inverter gates 738 and 739 with the phase unchanged, and outputs the signal as a signal G₁. The block B3 also fetches another signal from the other node b via four-stage inverter gates 740 to 743 with the phase unchanged, and outputs the signal as a signal G₄. 790 to 793 denote nMOS transistors. In the figure, 794 to 797 denote pMOS transistors.

0080The supply voltage of the inverter gate 739 is Vcc or ground (0 V). The logical amplitude of the signal G1 ranges from 0 V to Vcc. The supply voltage of the inverter gate 741, 742, or 743 is Vpp or ground. The logical amplitude of the signal G4 ranges from 0 V to Vpp. The logical amplitude of a signal G2 fetched from the block B4 which is not shown also ranges from 0 V to Vpp. When Vpp is larger than Vcc, the signals G1, G2, and G4 are all 0 V. When Vpp is not larger than Vcc, the signal G1 is equivalent to Vcc and the signals G2 and G4 are equivalent to Vpp.

0081The block B2 constitutes a known bootstrap circuit, and generates a signal G3 whose logical amplitude is an extended one of the logical amplitude of the signal G4 ranging from 0 V to Vpp (180° out of phase). In Fig. 24, 800 to 810 denote nMOS transistors; 811 to 814 denote pMOS transistors; 830 and 831 denote capacitors; 744 denotes an inverter gate made up of transistors 805 and 811; 745 denotes an inverter gate made up of transistors 810 and 814; 746 denotes an inverter gate made up of transistors 808 and 813; and 747 denotes an inverter gate made up of transistors 807 and 812.

0082Figs. 26A to 26G show wave forms indicating a process of generating a signal G3 in the block B2. In these waveforms, when the signal G4 drops, the voltage at the output node n71 of the inverter gate 744 rises to Vpp. At the same time, the voltage at the output node n72 of the inverter gate 745 drops to 0 V. With the voltage fluctuation at the node n71, the voltage at the node n76 rises to Vpp - Vth (Vth denotes a threshold voltage of the transistor 804). A capacitor 831 is charged with the voltage. On the other hand, charges are released from a capacitor 830 with the voltage fluctuation (VPP to 0 V) at the node n72. The voltage at the node n73 starts decreasing along the time constant curve defined by the capacitance of the capacitor 830 and the channel resistance of the transistor 809. When the voltage at the node n73 becomes lower than the threshold value of the inverter gate 746, the voltage at the node n74 decreases to 0 V and the transistor 806 is turned off.

0083When the transistor 806 is turned off, the voltage at the node n75 rises to Vpp - 2Vth (2Vth denotes a threshold voltage of the transistors 804 and 800). This causes the voltage at the node n76 to rise to a voltage to which a voltage across the voltage capacitor 831 at the node n75 is added. That is to say, when the voltage at the node n76 rises, the voltage at the node n75 rises to a value provided by subtracting Vth (Vth denotes a threshold voltage of the transistor 800) from the voltage at the node n76. Eventually, the voltage at the node n75 becomes Vpp and the voltage at the node n76 rises to a voltage provided by adding a voltage across the capacitor 831 to the voltage at the node n75. As a result, the signal G3 whose logical amplitude is an extended one of the logical amplitude of the signal G4 ranging from 0 V to Vpp (herein, 0 V to Vpp + Vth) is generated.

0084Fig. 27 shows wave forms representing an operation of the block B1; that is, a process of switching the level of an internal supply voltage V_PI at the time of transition from the read mode to write mode.

0085In Fig. 27, during the read-mode period, the signal G1 is Vcc, the signal G2 is Vpp, and the signal G3 is 0 V. Only the transistor 704 is on. During the read-mode period, V_PI is therefore equal to Vcc.

0086For shifting to the write mode, during the P period, the signal G1 is set to 0 V, the transistor 704 is turned off, and the transistor 706 is turned on, and V_PI thus remains equal to Vcc. The voltage at the node n70 is raised to Vpp (= Vcc) via the transistor 706. During the Q period, the signal G4 is set to 0 V, the signal G3 is raised to Vpp + Vth, and the transistor 707 is turned on. This causes current to flow from the third power line 703 to the second power line 702. The voltage at the transistor 707 rises to V_PI + Vpp for the Q period.

0087The voltage at the node n70 during the Q period is V_PI; that is, Vpp. The potential difference between the source and drain of the transistor 705 is 0 V. Even when the signal G2 is set to 0 V, a large current will not flow through the transistor 705 immediately. The potential difference between the p-channel diffused layer and n-type well in the source (node n70) of the transistor 706 becomes 0 V. Substrate current that causes a latch-up phenomenon does not flow, either. With the signal G2 held at 0 V, if writing of a memory cell is actually started, the cell current flows uninterruptedly from the third power supply 703 through the transistor 705, node n70, transistor 706, second power supply, etc., to the memory cell array 718.

0088As mentioned above, in this embodiment, even if the gate level of the transistor 705 is decreased to 0 V abruptly, there is no possibility that a latch-up phenomenon will occur. The latency time required for transition from the read mode to write mode (equivalent to P + Q in Fig. 108) can be minimized to be as short as that required for stable operations of devices. This results in an internal power switching art permitting fast mode change and contributing to realization of a high-speed external memory that can be freely rewritable.

0089In the aforesaid embodiment , the high-voltage level of the signal G3 is set to Vpp + Vth so that the V_PI value will become a maximum voltage Vpp during the Q period. The high-voltage level of the signal G3 is not limited to Vpp + Vth. Alternatively, the high-voltage level of the signal G3 may be Vcc+ Vth, Vcc + Vth to Vpp + Vth, or Vpp + Vth or more, as long as the V_PI value will be higher than Vcc during the Q period.

0090Through use of such an internal power switching circuit embodying the present invention, the latency time occurring at the transition from the read mode to write mode can be minimized. This results in, for example, an internal power switching art that contributes to realization of a high-speed external memory which can be rewritten freely.

0091Flash memories are expected to substitute for magnetic storages. For use in portable equipment on behalf of magnetic storages, in particular, the properties of flash memories; that is, small size and light weight will prove very helpful. For actual use in portable equipment in place of magnetic storage, the flash memory must further be innovated to achieve downsizing, an increase in operational speed, reduction in power consumption, and decrease in voltage requirements. Embodiments of the present inventioncan provide innovative flash memory and solve some of the problems obstructing the widespread use of the flash memory.