High Resolution Charge-Redistribution a/D Converter

1. Field of Invention

The present invention relates to analog-to-digital (A/D) converters and more particularly to a charge redistribution A/D converter in which the resolution of the converter may be extended without appreciably increasing the area occupied by the A/D converter on an integrated circuit chip.

2. State of the Art

Successive approximation A/D converters are well-known in the art. In general terms, such A/D converters use a collection of binary-weighted capacitors, resistors or some combination thereof successively switched into and/or out of the circuit so as to compare an unknown analog input voltage with binary-weighted fractions of a reference voltage. To form a collection of binary-weighted capacitors or resistors on an integrated circuit chip, typically an array of unit capacitors or unit resistors each having some specified minimum size is formed, after which the unit capacitors or unit resistors are suitably interconnected to form a collection of binary-weighted capacitors or resistors. The unit capacitor or unit resistor size is chosen in order to meet the accuracy requirements of the A/D converter in the face of process variations. For example, a unit capacitor is typically formed by a nominally square area of metal or polysilicon deposited on an integrated circuit wafer. Because of process imperfections, however, the edges of the metal or polysilicon area may, instead of being perfectly straight, exhibit some waviness, such that the actual capacitance of the unit capacitor will vary somewhat from the design value.

In a typical charge-redistribution A/D converter using a collection of binary-weighted capacitors, a least-significant bit of digital output is produced using a capacitor formed from a single unit capacitor, and the most-significant bit is produced using a capacitor formed from 2.sup.n-1 interconnected unit capacitors where n is the resolution, or number of output bits, of the A/D converter. Mathematically it may be shown that, to meet the accuracy requirements of the converter (viz., the digital output accurately reflecting the analog input to within .+-.1/2 LSB), the capacitor used to produce the least-significant bit may have a tolerance of 50%, the capacitor used to produce the next-significant bit must have a tolerance of 25%, and so forth, such that the capacitor used to produce the most-significant bit is required to have a tolerance of 100/2.sup.n %. For a 12-bit converter (n=12), the capacitor used to produce the most-significant bit must have a tolerance of 0.02%. Given a particular process, this strict tolerance imposes a severe limit on process variations as well as a minimum limit on the size of the unit capacitors. That is to say, if the unit capacitor is made very small, the variation in area (capacitance) produced by the waviness of the edge of the metal square will be greater than if the unit capacitor is made larger.

For every additional bit of resolution desired from the A/D converter, the number of unit capacitors must be doubled, substantially doubling the size of the A/D converter. The size penalty incurred to achieve higher resolution is therefor severe. What is needed is a way to increase the resolution of the A/D converter with a minimum increase in size while maintaining the required accuracy of the converter.

According to the present invention, an A/D converter integrated circuit includes a voltage comparator, an array of unit capacitors interconnected to form a first number of capacitors weighted such that the capacitance of a given capacitor is equal to the sum of the capacitances of all lesser capacitors, and a second number of binary-weighted capacitors, a greatest one of the second number of binary-weighted capacitors having a capacitance no greater than a capacitance of one of the unit capacitors of which the first number of capacitors is formed. One plate of each of the first number of capacitors and the second number of capacitors is connected in common to an input of the voltage capacitor. Switches connect another plate of each of the first number of capacitors and the second number of capacitors to one of circuit ground and a voltage proportional to a reference voltage. Control logic means is responsive to the voltage comparator for controlling the switches. A voltage divider apparatus produces different voltage signals proportional to the reference voltage, and the second number of binary-weighted capacitors have their bottom plates connected by the switches to either circuit ground or one of the voltage signals nearest circuit ground. The least voltage signal connected to the bottom plates of the first number of capacitors, on the other hand, is at least twice as great as the voltage signal nearest circuit ground. In an exemplary embodiment, two additional bits of resolution may therefore be obtained in the space of less than two additional unit capacitors by providing, as the second number of binary-weighted capacitors, a first capacitor having a capacitance equal to the unit capacitance and a second capacitor having half the capacitance of the unit capacitance. A capacitor having 1/4 the capacitance of the unit capacitance is also provided and is used to shift the analog input voltage by an amount represented by 1/2 the least significant bit in order to perform sign-magnitude to 2's complement conversion.

FIG. 1 is a schematic block diagram of a preferred embodiment of A/D converter of the present invention; FIG. 2 is a more detailed illustration of a portion of FIG. 1.

Referring now to the Figure, showing a preferred embodiment of the increased-resolution A/D converter of the present invention, the converter includes a first group of capacitors C1, C2, . . . , C7 weighted such that the capacitance of a given capacitor is equal to the sum of the capacitances of all lesser capacitors. The capacitance value of each capacitor in terms of a unit capacitance C is listed below each capacitor. It may be seen that C1, having a capacitance of 32C, is equal to C2+C3+. . . +C7, having a total combined capacitance also of 32C. Similarly, C2=C3+C4+. . .+C7, and so forth.

The converter also includes a second group of capacitors C8, C9 and C10 that are binary-weighted. A greatest one, C8, of the second group of capacitors has a capacitance equal to a least one C7 of the first group of capacitors, both having the capacitance of the unit capacitor. The capacitor C9 has a capacitance of 1/2C equal to half the capacitance of the capacitor C8, and the capacitor C10 has a capacitance of 1/4C equal to half the capacitance of the capacitor C9.

The top plates of all the capacitors C1-C10 are connected in common to the negative input of a voltage comparator 10. The top plates of the capacitors may also be connected to circuit ground through a switch S11 under the control of a signal P1. The positive input of the voltage comparator is connected to circuit ground. The bottom plate of all of the capacitors C1-C10 may be connected through switches S1-S10 (not explicitly shown) to circuit ground or to a voltage proportional to a reference voltage V.sub.R. In the case of the first group of capacitors C1-C7, the bottom plates of these capacitors may also be connected through the switches to an analog input voltage V.sub.in.

Although the bottom plates of all of the capacitors may be connected through the switches to a voltage proportional to a reference voltage V.sub.R, the specific proportion varies between the capacitors as follows. The capacitors C1-C6 may be connected to .+-.V.sub.R. Hence, for these capacitors, the proportionality factor is unity. The capacitor C7 may be connected to nV.sub.R /32 where n may assume any integer value from -32 to +31 (zero inclusive). Hence, for C7 the proportionality factor may have any one of 64 different values. The voltage signals corresponding to each of these values are produced using a voltage divider circuit 16.

The voltage divider circuit 16 uses a chain of 64 equal-valued resistors connected in series, 32 resistors designated R1P-R32P being connected between +V.sub.R and circuit ground, and 32 resistors designated RN-R32N being connected between circuit ground and -V.sub.R. Taps are taken on both sides of each of the resistors and the resulting voltage signals input to an analog decoder 14 which is controlled to select the appropriate voltage signal.

The capacitors C8 and C9 may be connected to +V.sub.R /64, designated V.sub.1P, or to -V.sub.R /64, designated V.sub.1N. The voltage signal V.sub.1P and V.sub.1N are produced at center taps of resistors R1P and R1N, respectively. (The portion of the circuit indicated by reference numeral 40 in FIG. 1 is shown in greater detail in FIG. 2.) Finally, the capacitor C10, if it is not connected to circuit ground, may be connected to the voltage signal V.sub.1N only.

The connections of each of the capacitors are controlled by a successive approximation register logic circuit (SAR) 12 responsive to the output V.sub.C of the voltage comparator 10. A strobe signal input to the SAR directs the A/D converter to begin a conversion, and a clock input signal causes the SAR to sequentially step through each step of the conversion operation. The SAR outputs timing signals for controlling the operations of the A/D convertor including a signal P1 that causes the analog input voltage V.sub.in to be sampled. The SAR produces as a final output a digital word representative of the analog input voltage signal. In an exemplary embodiment, the digital output word is 14 bits wide, the bits being designated B.sub.13 through B.sub.0 where B.sub.13 is a sign bit, B.sub.12 is the most significant magnitude bit and B.sub.0 is the least significant magnitude bit. By appropriately modifying the SAR logic, the connection of capacitors C8-C10 may be varied slightly such that each of these capacitors is connected to the same one of either V.sub.1P or V.sub.1N depending on the sign bit. The operational details of such a modification will be apparent to one of ordinary skill in the art.

With the exception of capacitors C8, C9 and C10 together with their associated circuitry, the operation of the A/D converter thus far described will be familiar to one of ordinary skill in the art. Briefly, to perform A/D conversion by successive approximation, capacitors, resistors or a combination thereof may be used. In an exemplary embodiment, capacitors C1-C6 are used to obtain the most significant six magnitude bits B12-B7, and capacitor C7 is used in conjunction with the voltage divider circuit 16 to obtain the five next most significant bits B6-B2.

When a strobe signal has been received directing the start of an A/D conversion operation, a signal P1 is issued from the SAR to cause the top plates of the capacitors to be grounded through the switch S1 and the bottom plates of the capacitors C1-C7 to be connected through the switches S1-S7 to the analog input voltage V.sub.in to sample the input voltage. The sample command P1 is then deasserted, causing the switch S11 to open. During a subsequent sign-bit (or input voltage reversal) phase, the bottom plates of capacitors C1-C7 are connected through the switches S1-S7 and the decoder 14 to ground, causing the negative of the analog input voltage, -V.sub.in, to appear at the negative input of the voltage comparator 10 as V.sub.X.

According to whether the analog input voltage is positive or negative, the output signal V.sub.C of the voltage comparator 10 will exhibit a positive or negative transition, in response to which the SAR sets the sign bit (B13) in the successive approximation register appropriately. The sign bit in turn determines whether a positive or negative reference voltage V.sub.R will be used during the remainder of the conversion operation. If the sign bit is positive, then +V.sub.R is made available to the bottom plates of capacitors C1-C6, and the decoder 14 selects a voltage signal from among the positive voltage signals for capacitors C7-C9. If the sign bit is negative, then -V.sub.R is made available to the bottom plates of the capacitors C1-C6, and the decoder 14 selects a voltage signal from among the negative voltage signals for capacitors C7-C9. Capacitor C10 is switched to -1/64 V.sub.R irrespective of sign bit B13 for sign-magnitude to two's complement conversion.

To determine the most significant magnitude bit, the bottom plate of capacitor C1 is switched from ground to the reference voltage V.sub.R causing the voltage V.sub.X to shift by 1/2 the reference voltage. If no transition is observed in the output signal V.sub.C of the voltage comparator 10, then the input voltage is greater than 1/2 the reference voltage, and the bottom plate of the capacitor C1 remains connected to the reference voltage V.sub.R while the bottom plate of the capacitor C2 is switched from ground to the reference voltage V.sub.R as well. If a transition is observed in the output signal V.sub.C, then the input voltage is less than 1/2 the reference voltage, the bottom plate of the capacitor C1 is returned to ground, and the bottom plate of the capacitor C2 is switched to the reference voltage. Operation proceeds in this manner through capacitor C6, whereby the six most significant magnitude bits are accumulated in the successive approximation register.

The capacitor C7 operates differently from the capacitors C1-C6. Instead of producing a single magnitude bit, the capacitor C7 operates in conjunction with the voltage divider circuit 16 to produce five magnitude bits. Since the tolerances on the voltage increments added to the voltage V.sub.X are greatly relaxed at this point in the conversion, resistors, which are typically lower precision than capacitors, may be used to augment the operation of the capacitor C7.

When integrated on a semiconductor chip, resistors are typically formed by elongated regions of doped semiconductor having a periphery relatively large in relation to the enclosed area, as opposed to the generally square regions used to form capacitors, which regions have a relatively small periphery in relation to the enclosed area. Accordingly, the waviness of edges defined on the semiconductor chip has a greater adverse affect on the precision of resistors than of capacitors. The precision of the resistors is sufficiently good, however, to allow them to be used in determining lower order bits B6-B2 in the illustrated embodiment.

The aforementioned bits B6-B2 are determined using the capacitor C7 in conjunction with the voltage divider circuit 16 in the following manner. When the sampling command P1 goes low following the initial sampling phase of the conversion cycle, the bottom plate of the capacitor C7 is connected to the output of the decoder 14 which is set to select circuit ground. When the time arrives for the capacitor C7 to participate in the conversion, the sign bit B13 is used to select a voltage signal midway between circuit ground and +V.sub.R if the sign bit is positive or midway between circuit ground and -V.sub.R if the sign bit is negative. Accordingly, the voltage V.sub.X is shifted by half the amount as in the case of the capacitor C6 even though capacitor C6 and capacitor C7 have the same capacitance, namely the unit capacitance. The bit B is thereby determined and is input to the decoder 14 to select a further voltage signal, either 3/4 of the way between circuit ground and the reference voltage or 1/4 of the way between circuit ground and reference voltage depending on the value of B6. Operation proceeds in this fashion until five bits B6-B2 have been produced.

Using the techniques of the prior art, for each additional bit of resolution desired, either the total capacitance would have to be doubled, doubling the required number of unit capacitors, or the number of resistors would have to be doubled. In either case, the size of the overall A/D converter greatly increases. According to the present invention, additional bits of resolution may be obtained with an essentially negligible increase in the size of the A/D converter.

In the illustrated embodiment, two additional bits of resolution are obtained to realize a 14-bit A/D converter in the same area as a comparable 12-bit A/D converter. Of the capacitors C8, C9 and C10, included in the second capacitor array, the capacitor C8 has a capacitance equal to the unit capacitance. In the case of the capacitors C9 and C10, since the required accuracy of the capacitors are 25% and 50% respectively to achieve 1/2 LSB accuracy in the converter characteristics, these capacitors are realized by sub-dividing the unit capacitor. The capacitor C9 therefore has a capacitance of half the unit capacitance, and the capacitor C10 has a capacitance of 1/4 the unit capacitance. The increment by which the voltage V.sub.X is shifted is further divided in half by connecting the bottom plates of the capacitors C8-C10 to a center tap point of the appropriate one of the resistors R1P and R1N. Thus, if the bit B2 is 0, indicating that the voltage V.sub.X should be incremented positively, the switches S8a and S8b are connected to the center tap of the resistor RIP producing a voltage V.sub.1P. Conversely, if the bit B2 is 1 indicating that the voltage V.sub.X should be decremented, then the switches S8a and S8b are connected to the center tap of the resistor R1N producing a voltage V.sub.1N. The switch S8 then connects the bottom plate of the capacitor C8 to the appropriate one of V.sub.1P and V.sub.1N. Depending on the output V.sub.C of the voltage comparator 10, the switch S8 either remains as it is or is switched to ground. The switch S9 associated with capacitor C9 is controlled in the same manner to produce the least significant magnitude bit, B0.

The capacitor C10 does not produce an additional magnitude bit. Instead, the bottom plate of the capacitor C10 is grounded during the sampling phase and is connected to V.sub.1N otherwise. The result is to produce a 1/2 LSB shift of the input voltage in the positive direction to achieve sign-magnitude to 2's complement conversion.

The foregoing has described a novel circuit technique to achieve 14-bit resolution in a charge-redistribution CMOS A/D converter. The sign bit plus the six most significant bits are obtained using a 6-bit capacitor array, the next five bits are determined using a resistor array, and finally the last two bits are obtained by use of a second 2-bit capacitor array. Using this technique, the area of the 14-bit A/D converter is not appreciably larger than a corresponding 12-bit A/D converter. The same technique may also be used for A/D converters of different resolutions.

It will be apparent to those of ordinary skill in the art that the present invention may be embodied in other specific forms without departing from the spirit or essential character thereof. The disclosed embodiments are therefore considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the appended claims rather than the foregoing description, and all changes which come within the meaning and range of equivalence thereof are intended to be embraced therein.