Patent Application
20060192595
This application claims priority under 35 USC §119 to German Application No. DE 102005008516.4, filed on Feb. 24, 2005, and titled “Sense Amplifier,” the entire contents of which are hereby incorporated by reference.
The invention relates to a sense amplifier. The invention relates, in particular, to those sense amplifiers which have at least one pair of negative-feedback field effect transistors having identical conductivity type within the pair. The sense amplifier preferably serves for amplifying an electrical signal that is read out from a memory cell of a dynamic memory component.
Sense amplifiers serve for amplifying a weak electrical signal present at an input of the sense amplifier. They are also used in particular in the area of memory components, where they are used to amplify the charges that are stored and read out from dynamic random access memory cells (DRAM) as a signal. The strength of the signals of charges stored in trench capacitors, for example, is comparatively weak on account of the smallest possible dimensioning of the trenches.
In the area of nonvolatile memories, e.g., those having memory cells based on the magnetoresistive effect (MRAM) or on floating gate storage (charge trapping devices), weak cell signals usually have to be amplified.
There are various architectures for sense amplifiers. They typically have a respective pair of n-channel and p-channel field effect transistors which may be arranged, e.g., in cross-coupled fashion or in a manner coupled with negative feedback in pairs. In the case of memory components, these are usually arranged at the edges of the memory cell arrays on or in the substrate.
In this case, the threshold voltages of the n-channel and p-channel field effect transistors are of great importance for the amplification:
For example, a gate terminal of a first n-channel field effect transistor of the pair may be connected to the line carrying the signal (signal line hereinafter) and a gate terminal of a second one of the n-channel field effect transistors may be connected to a reference line. The reference line is biased to a comparison potential with which the signal is to be compared for the purpose of the direction of the amplification proceeding from the comparison potential.
In the case of the memories, the signal line corresponds to a bit line and the reference line corresponds to a reference bit line.
By contrast, the source terminals of the two n-channel field effect transistors are connected to the same means for feeding in a supply voltage. The means for feeding in the supply voltage may itself also in turn have a transistor with which the supply voltage is trimmed. In the case of the n-channel field effect transistors, e.g., a potential is run through from a maximum value to a minimum value. The gate-source voltages at the two n-channel field effect transistors (n-FETs hereinafter) then depend on the potentials present on the signal and reference lines in comparison with the trimmed supply potential.
What then matters is which of the continuously decreasing gate-source voltages at the two n-FETs falls below the threshold voltage of the transistors first. If, e.g., the potential on the signal line is higher, then that n-FET whose gate terminal is connected to the signal line turns on first.
The negative-feedback mode of switching connection of the two n-FETs consists in the fact that the drain terminals are in each case connected to that line which is precisely not connected to the gate terminal of the relevant n-FET. This means, however, that the gate terminal of the respectively precisely non-switching n-FET is connected to the decreasing potential, that is to say that until the minimum potential value is reached its conductivity decreases since its gate-source voltage vanishes.
Overall, one of the two lines is thereby run down to a minimum potential. The complementary task—of running the potential of the other line up to the maximum value—is performed analogously by the other pair of p-channel field effect transistors (p-FETs hereinafter).
It is important for the threshold voltages of the two n-FETs or p-FETs to be as far as possible equal among one another, if not in fact identical. In the case of the memories, a typical potential difference between bit line and reference bit line may be 60 millivolts (mV) given threshold voltages of 300-400 mV. If the threshold voltages differ to a sufficiently great extent, however, an amplifier operation in which the signal is amplified in the wrong direction may occur. In this case, the information actually stored in the cell could be interpreted incorrectly during read-out and amplification.
For an n- or p-FET, the threshold voltage is influenced by the purity and quality of the fabrication process. The quality of the formation of the n-type well (in the case of the p-FET) or the p-type well (in the case of the n-FET) is particularly important in this case. The two n-FETs or p-FETs of the sense amplifier are formed in principle on the substrate, in the region of the same n-type or p-type well in each case. The two n- or p-FETs depicted identically in terms of their geometry then generally exhibit a stochastic switching behavior that deviates from one another albeit only to a small extent (so-called “mismatch”). Electrical measurements of a multiplicity of such transistors in this case have a Gaussian distribution in the physical quantities respectively measured, e.g., the threshold voltage.
However, systematic, no longer acceptable differences between two respective transistors of the same conductivity type may occur as well. The term used in this context is a systematic “offset” or “mismatch”. It has thus been established, for example, that the distance between the transistor and the outer edge of the n-type or p-type well influences the measured distribution. The cause of this is an inhomogeneous or asymmetrical doping profile along the well.
The asymmetry arises as a result of a scattering or reflection of the dopant particles at the edges of the resist masks that define the well regions during the implantation. This is because this implantation, in order to avoid the so-called channeling effect along a crystal direction lying perpendicular to the surface of the silicon substrate, is performed obliquely at an angle of 7 degrees, for example.
One solution consists in implementing a particularly wide trench isolation, e.g., STI: shallow trench isolation in the region of the well edges, so that effective backscattering into the substrate cannot occur there. However, this gives rise to a further source for asymmetries in the well doping profiles, because the trench isolation, during the further processing of the semiconductor substrate, leads to mechanical stresses in the adjoining substrate areas. These stresses may likewise locally influence the electrical properties in a disadvantageous manner. If the trench isolation lies closer to selected active transistors, this gives rise to a further source.
The problem has been avoided hitherto by designing the cell signal to be sufficiently strong corresponding to a generously dimensioned storage capacitor. Further endeavors amount to reducing the line capacitances. It is apparent, however, that these measures will soon no longer suffice in the context of increasing packing density, advancing structure miniaturization, rising leakage currents on account of tunnel effects, transistor leakage currents, so-called “'sub-Vt-leakage”, etc.
The document U.S. Pat. No. 6,445,216 B1 describes a sense amplifier having two input transistors of identical conductivity type, which are used to assess data signals D, D′. The influence of varying channel lengths between the transistors on the respective threshold voltage is reduced by virtue of the fact that a switching unit that generates the so-called “forward body bias” applies the same potential to the bulk terminals of the two transistors, for which purpose it is connected to the two terminals and the latter are interconnected.
In accordance with the present invention, a sense amplifier comprises at least two field effect transistors of identical conductivity type, each including a gate terminal, source terminal, drain terminal and bulk terminal. With regard to a first one of the field effect transistors, the gate terminal is connected to a signal line, the source terminal is connected to a terminal for feeding in a first supply voltage and the drain terminal is connected to a reference line. With regard to a second one of the field effect transistors, the gate terminal is connected to the reference line, the source terminal is connected to the terminal for feeding in the first supply voltage and the drain terminal is connected to the signal line. The first bulk terminal of the first field effect transistor is formed in a first well of a substrate and the second bulk terminal of the second field effect transistor is formed in a second well of the substrate, the second well being electrically insulated from the first well.
The two field effect transistors of identical conductivity type may be two n-FETs or two p-FETs which are assigned to the sense amplifier. Over and above the gate terminal, source terminal and drain terminal, they also have a bulk terminal to the substrate, or more precisely: to the respective well in or above which the n- or p-FETs are formed.
The two n-FETs or p-FETs are connected up in a negative-feedback manner. The gate terminals are connected oppositely either to the signal line or to the reference line. The drain terminals connect the transistor to the respective other one of the two lines. Therefore, they are connected via the line to the gate terminal of the respective other transistor of the pair. The source terminals are both connected to the same terminal for feeding in a supply potential. The terminal may, in particular, also have a means with which the supply potential of a potential source can be varied, can also be trimmed in accordance with one design of the invention, for instance by means of a further transistor.
The bulk terminals of the two n-FETs or p-FETs have the particular property of being electrically insulated from one another. That is to say the depletion regions of the transistors or the p-channels of the p-FETs or the n-channels of the n-FETs are respectively embedded in different n-wells or p-wells which are not in direct electrical contact with one another.
In other words: each transistor of an n-FET pair or of a p-FET pair comprises its own well, which is electrically isolated from that of the respective other transistor of the negative feedback pair. The isolation is preferably effected by an isolation trench, but may also be brought about by nonconductive, undoped substrate. An electrical connection between the two wells, which can be produced via at least one further switching element, is not precluded, however.
By virtue of the isolation of the wells and thus also of the bulk terminals, it is possible to match the geometry of the individual field effect transistors of a negative-feedback pair with regard to the well edges, in particular the distance between the depletion region and the well edge, or with regard to adjacent isolation regions which exert mechanical stresses. This gives rise to the advantage that the threshold voltages are no longer impaired by gradients of electrical parameters along the common well. The systematic offset is therefore cancelled. The threshold voltages of the two n- or p-FETs are thus close together.
On account of this, the risk of an erroneous read-out and amplification of a charge signal from a memory cell does not occur. To put it another way, the permissible strength of a cell signal can be reduced further, which has a positive effect on the dimensioning of a memory and also on the required power consumption.
One exemplary configuration of the sense amplifier provides for connecting the first well to a first means for feeding in a first well potential for setting a first threshold voltage of the first field effect transistor, and for connecting the second well to a second means for feeding in a second well potential—which is different from the first well potential—for setting a second threshold voltage of the second field effect transistor.
As a result, the threshold voltages of the two transistors can be coordinated with one another. Stochastic effects resulting from the formation of different wells can thus be compensated for individually for each of the wells by means of an adapted well potential.
The so-called body effect, which is also called the substrate bias effect, is utilized in this case. Accordingly, the threshold voltage of a field effect transistor correlates with the voltage present between gate and bulk terminal. This correlation can be calculated in order then to obtain the same threshold voltage for both n- or p-FETs depending on this. The values for the well potentials that can be read from the calculated relation for the desired threshold voltage are provided to the well potentials.
Supplying well potentials themselves also in turn comprise transistor providing a potential which is present from a further supply voltage and supplies the customary substrate bias voltage can be varied by up to 100 mV.
A method for providing well potentials in a sense amplifier according to the invention comprises: determining the threshold voltage without application of a well potential in each case of the first and of the second field effect transistor by measurement or simulation, in each case for the first and the second field effect transistor calculating a relation for a change—which is to be brought about by the body effect—in the threshold voltage of the first or second field effect transistor depending on a well potential to be applied, comparing the two relations for determining a first well potential for the first field effect transistor and a second well potential for the second field effect transistor, so that the threshold voltages of the first and second field effect transistors match in accordance with the relation, and separately setting the first circuit arrangement including transistors and the second circuit arrangement including transistors so that the first circuit arrangement can feed in the first well potential and the second circuit arrangement can feed in the second well potential.
The above and still further objects, features and advantages of the present invention will become apparent upon consideration of the following detailed description of specific embodiments thereof, particularly when taken in conjunction with the accompanying drawings wherein like reference numerals in the various figures are utilized to designate like components.
An example of a sense amplifier which is connected up for amplifying a charge signal read out from a memory cell is illustrated in
The selection transistor 100 is shown on the left-hand side of the illustration, which selection transistor, in controllable fashion via the word line WL, can enable a charge stored in the memory node DT. Before this operation is begun, however, first the bit line BL and the reference bit line {overscore (BL)} are brought to a common bias voltage potential (“precharge”). For this purpose, a voltage VINT of 900 mV is passed to the source inputs of the transistors 102, 104 via a transistor 108. Via the gate terminals of the transistors 102, 104, the latter are turned on by a voltage signal VEQ generated by a further source. Via the drain terminals, the bit line BL and reference bit line {overscore (BL)} are brought to exactly this voltage potential of 900 mV. The transistor 106, which is likewise turned on by means of the signal VEQ, has the effect that, in the case of a turn-on operation at a different speed, a potential equalization takes place between bit line BL and reference bit line {overscore (BL)}.
Independently of this, a further source generates a voltage signal MUXL, which can connect up the sense amplifier I both for the bit line BL and for the reference bit line {overscore (BL)}. If the signal is present at the gate terminals, the transistors 110 and 112 are turned on. The background for these enable transistors 110 and 112 is that, on the right-hand side of
Afterward, the signal VEQ is ended and the transistor 108 can also be closed. The bit line BL and the reference bit line {overscore (BL)} now have the same potential, but are electrically isolated from one another. Then as described in the introduction through driving of the transistor 100, the charge is read out from the cell DT and is transferred onto the bit line BL. The resulting charge signal is limited according to strength and also with regard to the time duration. For example, a charge state of the memory cell DT “high” or “1” leads to a momentary increase in the voltage potential on the bit line by 160 mV to 1060 mV.
This potential is present at the input of the n-channel field effect transistor 114 (n-FET 114 for short). The n-FET 114 is one of the two transistors, connected in negative-feedback fashion, of a pair 10 having the same conductivity type which the sense amplifier 1 comprises. A voltage potential VBLL which can be trimmed by means of the signal nSET via the transistor 126 and initially has a value of 1800 mV is present at the source input of the n-FET 114.
The voltage potential is also present at the source input of the other n-FET 116 of the pair 10. The gate input thereof is connected to the bit line {overscore (BL)}, which still has a potential of 900 mV. The threshold voltage both of the n-FET 114 and of the n-FET 116 is 300 mV in each case in this example. The gate-source voltages of initially −740 mV (n-FET 114) and −900 mV (n-FET 116), respectively, result. Both transistors are closed.
The voltage potential VBLL is then continuously trimmed to 0 mV (“low”) by means of the transistor 126. At 760 mV for the potential VBLL, the n-FET 114 reaches the gate-source voltage of 300 mV, that is to say the threshold voltage, first. At the same point in time, the corresponding voltage at the n-FET 116 is still 140 mV. The n-FET 114 then opens, so that the reference bit line {overscore (BL)} is connected to the further decreasing potential VBLL. Its voltage potential is thus run down with it from 900 mV (bias voltage, precharge) to 0 mV (“low”).
The transistor 128 is controlled in a similar manner by means of a signal pSET in order to trim a potential VBLH from 0 mV to 1800 mV. In a manner opposite to the functioning of the n-FETs 114, 116, the p-FETs 118, 120 of the negative feedback pair 20 of transistors of the sense amplifier 1 are used in such a way that, in the present example, the voltage potential of the bit line BL is run up from 1060 mV (charge signal for “1”) to 1800 mV (“high”). The p-FET 120 turned on in this case. An initial difference between the voltage levels of 160 mV on account of the stored charge is thus raised to 1800 mV by the sense amplifier.
The previous example related to the ideal case of identical threshold voltages between the two n-FETS 114, 116 and/or the p-FETs 118, 120.
The transistors are formed above the same well 22. The resulting substrate or bulk terminal thus relates in principle to the same well potential.
As can be seen in
In this example, in accordance with the prior art, this results in a difference in the threshold voltages of 100 mV, that is to say the n-FET 114 has a threshold voltage of 350 mV and the n-FET 116 has a threshold voltage of only 250 mV. On account of a memory cell error lying in the tolerance range, during the read-out of a charge signal only a voltage of 1000 mV is supplied on the bit line (potential difference of 100 mV instead of 160 mV on average).
In this possible case, it can happen that the n-FET 116 opens first even though a “high” signal is present. As a result, a potential of 0 mV is then present on the bit line BL and a potential of 1800 mV is present on the reference bit line {overscore (BL)}.
As additionally shown in
In the example, the source regions 16a and 16b are also isolated from one another. Since they have to be at the same potential VBLL in accordance with the arrangement according to
For the contacts 14 of the gate electrodes and the drain regions 16, 18, in this exemplary embodiment there are no structural differences with respect to the prior art in accordance with
An example of the generation of separate well voltages for the n-FETs 114, 116 shown in
b shows a detailed example of the different potential sources. Via signals A, B, a basic potential Vgnd (e.g. 0 mV) is respectively switched to the well terminals 22a, 22b in a controllable manner. The potentials present can now be modified, however, specifically in that transistors which are controllable by sources of signals {overscore (A)}, {overscore (B)} can switch in a correction voltage potential VBIAS, “bias voltage”, for a respective one of the well terminals. A direct electrical connection between the potentials present at the different wells 22a, 22b no longer exists in this case.
In this exemplary embodiment, the signals {overscore (A)}, {overscore (B)} and also the associated two transistors constitute a source 72, by which the basic potential Vgnd present at the wells can be varied. However, the invention is not restricted to this specific embodiment for the source 72. Rather, other sources 72 are also conceivable which can be used to generate different bias voltages VBIAS for the respective wells.
c shows an optionally usable exemplary embodiment of a circuit arrangement by means of which the bias voltage VBIAS can be generated. Via voltage divider resistors 60, which make it possible to tap off potentials between +0.5 V and −0.5 V, a predetermined potential is applied to an operational amplifier OP. The operational amplifier OP has an operating voltage of between, e.g., −1.3 volts and +1.3 volts.
While the invention has been described in detail and with reference to specific embodiments thereof, it will be apparent to one skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope thereof. For example, the invention is not restricted to the feature combinations shown in the exemplary embodiments. Rather, it also encompasses those alternative features which the person skilled in the art, using his expert knowledge, would routinely interchange with the combinations presented here in order to achieve the same aim. In particular, the invention is not restricted to the area of application in memory components. Accordingly, it is intended that the present invention covers the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents.
| Number | Date | Country | Kind |
|---|---|---|---|
| 10 2005 008516.4 | Feb 2005 | DE | national |
