AU640354B2 - Successive-approximation register - Google Patents

Description

SUCCESSIVE-APPROXIMATION REGISTER

FIELD OF INVENTION

The present invention relates generally to the field of Digital-to-Analogue (DAC) and Analogue-to-Digital (ADC) Converters. More particularly, the invention relates to a Successive-Approximation Register (SAR) for use in such converters.

Most specifically, the invention relates to a method of implementing such SAR, utilising a single register only, to provide both the addressing and data-storage functions (as more fully hereinafter described), requiring significantly less electronic circuitry than known prior art designs.

BACKGROUND OF THE INVENTION

In electronic instrumentation systems, it is frequently necessary to accept an analogue value (i.e. a continuously varying electrical signal) and to transform the value of that signal into a digital representation for subsequent processing.

Such a digital representation is commonly made as a binary fraction of some known "reference" value, that is the "quantity" is represented as a N-bit binary number, whose value X denotes the Quantity thus

Quantity = Reference x(X/2^N),

where the value X may range from zero to 2^N-1.

This notation may represent real quantities from zero to just less than the chosen Reference value. For example, if the Reference value is chosen as 10.24V (i.e. 2¹⁰ x 0.01V), and the digital representation is in 10 bits, we may represent voltages from zero to 10.23V, to within an accuracy of 0.01V.

Many techniques for executing Analogue-to-Digital conversion have been described in such standard texts as "Bipolar and MOS Analog Integrated Circuit Design" (Grebene A., John Wiley, 1984). The present invention is of specific application to converters employing the so-called "Successive Approximation" method, which is known. A technique of logic design practised in Very Large-Scale Integration (VLSI) circuits, is the so-called "dynamic" technique (see, inter alia. Mead and Conway, "Introduction to VLSI Systems", Addision-Wesley, 1980). This scheme relies upon the short-term storage of information as electrical charges in stray wiring capacitance. Adroitly used, this method can result in considerable reductions in circuit complexity and power consumption. Its disadvantage is that the electrical charges will slowly "leak" (due to imperfect insulation), resulting in loss of the stored information. It is therefore a characteristic of Dynamic circuits to refresh, that is any information so stored must be periodically read out and re-written to maintain the charge.

According to a known Successive Approximation method of conversion, there is provided at least a Successive-Approximation Register (SAR), a Digital-to- Analogue Converter (DAC), a Comparator, and some form of sequencing or control logic, or as is shown in Figure 1.

The DAC is adapted to develop an analogue output, equal to the Reference voltage multiplied by the binary fraction stored in the SAR, as described above. Many methods of achieving this are known, for example the so-called "R/2R Ladder" method as illustrated in Figure 2. As may be seen, the circuit comprises an array of switches adapted to switch between Ground and the Reference voltage, and an array of resistances, being alternately of some value (R) and twice that value (2R). By suitably setting the several switches, voltages between zero (Ground) and Reference may be developed at the output. In a practical ADC, the several switches are adapted to be controlled by the several bits of the SAR.

This output voltage is applied to the Comparator, which develops a logical (i.e. True/False) output indicating whether the DAC output is higher or lower than the unknown analogue input. This logical output is used by the Control Logic to adjust the binary number held in the SAR, so as to bring the DAC output as close as possible to the unknown analogue input. The operating principle is typically as follows : The Control Logic begins by setting all the SAR bits to Zero. The most significant SAR bit is then preset to One, and the Comparator output is tested. If the DAC output is found to be higher than the unknown input, the SAR bit is cleared or reset back to Zero, otherwise it remains set to One. The Control Logic then presets the next lower significance SAR bit, and proceeds similarly, until all SAR bits have been processed.

Conveniently, a given SAR bit is preset to One, simultaneously with the selective resetting of its predecessor.

When this process is complete, the required digital representation of the unknown input is held in the SAR bits. The requirement to maintain the SAR output for extended periods of time usually precludes the use of Dynamic logic

(as aforedescribed) in practical SAR designs.

This technique is widely used in practical devices, It will be apparent that at least two Data Storage Means are required namely, the SAR bits themselves, and some further means (assumed in the above example to be contained within the control logic) to keep track of which SAR bit is currently being processed.

Practical systems have utilised either a counter or a shift register circuit to perform this latter (bit addressing) function. An example of the shift-register technique may be found in the 74LS502 SAR integrated circuit (Fairchild Semiconductor, Inc.) whose operation is described in the manufacturer's data sheets.

It will be appreciated that when two shift registers are used, the second (addressing) register consumes a similar amount of electronic circuitry (area) to the actual SAR register itself. Counter-based systems consume a similar amount of extra circuitry for the requisite Counter and Decoder. Heretofore, it has not proved possible to eliminate this extra circuitry. Clearly such elimination would be desirable, leading to greater economy in manufacture than has heretofore been possible. There are numerous prior art pertaining to SAR design and to Analogue-to-Digital conversion and particularly orientating toward improving the speed of operation of the SAR rather than seeking to reduce the circuit complexity. Some known prior art are US 4527148, US 4764750, EP258840, EP 258841.

Others, for example US 4654584, relate to various forms of Analogue-to-Digital Converters rather than to the internal functioning of SAR.

In US 4688018, the problem of SAR design is addressed. However, it discloses and requires separate Shift Register (SHR1, SHR2, SHR3, SHR4) and SAR storage cells (12, 14) rather than a single storage cell to perform both functions of bit addressing and accumulating the result. Further, the use of separate Shift Register and SAR latches necessitates additional serial signal paths (8 in US 4688018) between successive stages.

Furthermore, the operation of the SAR latch (36) requires that the string of set/reset transistors (52, 54, 56) have a greater electrical conductivity than the latch transistors (inverter 38). This arrangement of differentially conductive transistors may preclude fabrication of circuits according to a form of semiconductor chip design on so called "Gate Array" devices, characterised by a multitude of transistors all of a standard size, thus able to offer low design and set-up costs. To modify such circuit arrangement to eliminate the requirement for differentially conductive transistors, significantly more transistors may be required for the modified design.

US 4441198 is intended to increase the operating speed of SAR using both edges of the clock signal. Two latch elements per bit of the SAR are disclosed.

OBJECTIVES OF THE INVENTION

It is an object of the present invention to alleviate some or all of the disadvantages of the prior art and to simplify the design of SAR.

It is a further object of the present invention to provide a method of implementing a SAR and its addressing function, within a single shift register, thereby utilising approximately half the circuitry needed by prior art devices.

It is a further object of the present invention to provide a means whereby such a register functioned as described above may conveniently be fabricated, especially on Very Large-scale Integration (VLSI) chips.

It is a further object of the present invention to provide a Register as described above further incorporating self-refresh features, thereby permitting its implementation using so-called dynamic logic (as hereinbefore described), and permitting further reductions in chip-area and operating power requirements.

It is a further object of the present invention to provide a Register as described above being capable of fabrication using transistors all of a single standard size, so facilitating the realisation thereof on Gate Array devices.

SUMMARY OF THE INVENTION

In the following description, it will be assumed that the SAR shifts from left to right. This will imply that the leftmost bit is the most significant, and the rightmost the least significant. Naturally, the converse arrangement is equally feasible, as would be understood by the skilled addressee and Ihus also falls within the scope of the present invention.

Likewise, the logic structures herein described are capable of many forms of practical embodiment in electronic circuitry, as would be understood by those skilled in the art. Such alternative embodiments, such as various forms of MOS, bi-polar, discrete or others also fall within the scope of the present invention. Accordingly, the preferred embodiment described hereinafter, which assumes a Complementary Metal Oxide Semiconductor (CMOS) fabrication process, is to be regarded as exemplary only. The present invention provides a successive-approximation register (SAR) having a single shift register for processing, that is presetting and selectively resetting, a number of bits, wherein said single shift register is arranged to provide bit selection for processing the bits and also to provide desired result accumulation in the processed bits.

The present invention may provide an SAR wherein processing of a given bit is arranged to start simultaneously, on the same clock signal, with the completion of processing its immediate preceding bit and the presetting of the given bit is arranged to occur simultaneously, on the same clock signal, with the selective resetting of its immediate preceding bit.

Conveniently, a processed bit is recognised by having at least one set bit in the shift direction, the bit being processed is recognised by having no set bit in the shift direction and is itself a set bit, an unprocessed bit is recognised by having no set bit in the shift direction and is itself not a set bit, and the next bit to be processed is recognised by having no set bit in the shift direction and having an adjacent set bit in the opposite direction to the shift direction and is itself not a set bit.

In practice, the status recognition of bits is performed by a Manchester Carry chain, said chain being arranged to pass information in the opposite direction to the shift direction.

The present invention may also provide an SAR wherein the shift register comprises an array of stages and the stages include a first stage, a last stage and a number of active stages equal to the number of bits of digital output.

Preferably, the first stage is adapted to provide an output for presetting the active stages, and the last stage is adapted to provide an output for controlling the bit selection for processing the active stages. Further, each active stage may comprise a storage cell for storing successive bits of the desired digital output and switches for selective processing of data for said storage cell, and each switch may be implemented in CMOS.

The SAR in accordance with the present invention may be implemented in VLSI, MOS or Gate Array devices. BRIEF DESCRIPTION OF THE DRAWINGS

Preferred embodiments of the present invention will now be described with reference to the accompanying drawings, wherein :

Figure 1 shows a prior art ADC using an SAR.

Figure 2 shows a prior art R/2R ladder DAC.

Figure 3 shows an SAR according to the present invention.

Figures 4A and 4B show switch logic designs implementing the present invention.

Figures 4C, 4D and 4E show schematically the different stages of Figure 3 in CMOS.

Figure 5 shows prior art static and dynamic storage cells.

Figure 6 shows an active stage adapted for rapid initialisation.

Figure 7 shows an active stage adapted for serial output.

Figure 8 shows the results of a PSpice simulation of 5-bit ADC having the SAR of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

An SAR according to the invention comprises a shift register, said register comprising a plurality of stages, each stage comprising a storage cell and additional logic functions, as hereinafter described. Conveniently, there will be provided as many stages as are the number of required bits (N) in the resulting digital value (N=10, in the example cited above). These stages are connected in a series chain, so that information may be caused to move from left to right along the chain. There are additionally provided according to the invention, dummy first and last stages, which define the ends of the chain, said first and last stages themselves not contributing to the digital output value.

The general arrangement of the present invention is illustrated in Figure 3. The CLOCK signal causes each stage to assume a new data value. The COMP input is the output of the comparator (of Figure 1), and indicates whether the current SAR value is above or below that required. The RESET\ input is used to initially set all SAR bits to Zero (the "\" suffix denoting that the signal is Active-Low).

As described above, operation begins by setting all storage cells to Zero, by simultaneous application of RESET\ and CLOCK signals.

Once the RESET\ signal is deactivated, the next CLOCK will cause the first stage to emit a logical One signal (for this one CLOCK event only). This One will shift into the first active stage, so presetting the most significant bit to One, as described above. The comparator presents its result on the COMP line, and the CLOCK is re-applied. This causes the next stage to the right to be preset to One, simultaneously the current active stage will be reset to Zero if the comparator indicates that the SAR value is too high.

Subsequent CLOCKS will repeat this process on each SAR bit in turn, until the shifted One bit reaches the last stage. This last stage is so designed that once set to One, it remains set (until the next RESET\), and so provides a Conversion Complete output signal.

The invention is embodied in the design of the logic within each said stage, said logic being adapted to recognise when its stage is due to preset to One or to selectively reset to Zero by COMP, and when its stage is not permitted to change. This permits a single array of stages both to provide its own addressing, and to accumulate the required output data. The invention stems from the observation that, once any given stage has been evaluated (i.e. preset and selectively reset), there will always be at least one stage to its right holding a One-bit. Such a given stage may not change its value again until the next RESET\. Thus, a procedure may be followed to preset the next stage to One-bit simultaneously with the selective resetting the One-bit of the given stage. Likewise any other stage having no One-bits to its right, and its immediate left neighbour also holding Zero, is not yet to be evaluated, and hence also should not change.

At the rightmost One-bit stage in the chain, the following rule holds:

the next stage to the right shall become preset to One, while the present stage itself shall selectively reset to Zero, that is, if the comparator shows the SAR value as high.

These rules may be summarised in the fol lowing decision table (as is customary, "X" denotes don't care):

The "One-bits to Right" test may conveniently be implemented using the so-called "Manchester Carry Chain", which is commonly used to determine when all preceding bits in a counter are set to One (see, inter alia, Mead & Conway, "Introduction to VLSI Systems", Addison-Wesley, 1980). The device comprises a two-way switch, or multiplexer,

associated with each stage, being adapted to pass a signal from the preceding stage when the current stage holds a Zero, and to connect instead to a constant (Zero or One) when the current stage holds a One.

In the present case, we require to detect a One among the following bits, and the carry chain is therefore connected in the reverse direction to the usual, passing the carry against the direction of register shift. The start of the carry chain is provided by the last stage, which is provided with special logic for this purpose.

The corresponding decision table for the last stage

is as follows :

The Shift + Enable\ signal is optional, and provides a means to inhibit all change in tie SAR, should this be required.

Referring to Figures 4A and 4B, switch logic designs implementing the "One-bits to Right" test and the "Manchester Carry Chain" for the Active stage are shown. Each switch is to be Set to the "0" position if its controlling input has the value O (low) or N and to the "1" position if its input has the value 1 (high) or Y.

The "One-bits to Right" test is controlled by a "ONE-BITS TO RIGHT" switch and a "PRESENT STATE OF THIS STAGE" switch. It will be seen that when the ONE-BITS TO RIGHT has the value 1 (Y), NEXT STATE OF THIS STAGE takes the value of its PRESENT STATE. When the ONE-BITS TO RIGHT has the opposite value 0 (N), NEXT STATE OF THIS STAGE takes either the value of the NEIGHBOUR AT LEFT or the COMPARATOR output depending on the value of its PRESENT STATE.

The "Manchester Carry Chain" is controlled by a

"PRESENT STATE OF THIS STAGE" switch. It will be seen that when the PRESENT STATE OF THIS STAGE has the value O, the signal from NEXT STAGE will be carried TO PREVIOUS STAGE. Otherwise, the signal is not carried and the RESET\ signal (a constant) will be carried TO PREVIOUS STAGE.

One embodiment of the present invention will be described with reference to a CMOS VLSI fabrication process. Exemplary circuit forms are shown in Figures 4C, 4D and 4E.

Figures 4C is a schematic circuit of the first stage of the present invention shown in Figure 3. It comprises a logic inverter, NAND gate and flip-flop driven by CLOCK signals. The RESET\ signal is provided as input to the flip-flop whilst its inverted signal and the flip-flop output are coupled to the NAND gate to provide the output of the first stage (Data Out).

Operation of the first stage will be as follows:

RESET\ D Q DATA OUT Reset

Its effect is to set DATA OUT high for exactly one cycle of CLOCK after RESET\ itself goes high (after having been low) .

Figure 4D is a schematic circuit of the active stage of the present invention shown in Figure 3. It embodies the switch logic designs for the "One-bits to Right" test and the "Manchester Carry Chain" of Figures 4A and 4B. This is realised by the use of six transistor pass-gates acting as analogue switches in the form of parallel pairs of complementary transistors.

It will be seen that the following are equivalents between the switch logic designs and active stage circuit.

Manchester Carry Chain Active Stage

RESET\ ≡ RESET\

FROM NEXT STAGE ≡ Carryin

TO PREVIOUS STAGE ≡ Carryout

PRESENT STATE OF THIS STAGE ≡ Data Out or Q

One-bits to Right Test Active Stage

ONE-BITS TO RIGHT ≡ Carryin PRESENT STATE OF THIS STAGE ≡ Data Out or Q NEXT STATE OF THIS STAGE ≡ D

NEIGHBOUR AT LEFT ≡ Data In

COMPARATOR output ≡ SARlow

The Data and Carry shift in oposite directions.

The value of the NEXT STATE OF THIS STAGE (D) is primarily controlled by the ONE-BITS TO RIGHT (Carryin) signal and secondarily by the PRESENT STATE OF THIS STAGE (Data Out) signal. The steering logic feeds back the original stored value Data Out (Q) to the storage cell (flip-flop), whenever no change is required (Carryin = 1). This advantageously permits (given a high enough clock frequency) the use of Dynamic circuit techniques in the storage cell. Reference to Figure 5 (which shows typical storage cells using Static and Dynamic methods) will show that the Dynamic cell requires about half the electronic circuitry (transistors) of its Static counterpart. This saving permits the extra logic of the present invention to be realised with little extra circuitry than would be required for a single register of Static design. Since the present invention requires but one such register to implement an SAR, while prior art requires two, there is a net saving of some 50% of the required circuitry.

Referring back to Figure 4D, when change is permitted, either the stored value from the previous stage (Data In) or the SARlow signal from the COMPARATOR is input to the storage cell (D).

When RESET\ is off (high), the Carry chain will be connected to a constant high when Data Out is also high, and the chain will also carry the constant high signal from the next stage to the previous stage when Data Out is low.

Operation of the active stage will be as follows:

RESET\ Carryin Data In Data Out SARlow D Carryout

It is noted that when RESET\ is low, Carryin (from the last stage) and Data In (from the first stage) will also be low. Further, the value of SARlow may be dependent on the present state of storage cell (Data Out to the Comparator for determining the value of SARlow).

In realising an SAR of more than a few Active Stages, it will be advantageous to periodically regenerate the carry-chain signal by use of a pair of inverters, to improve its speed. This device is described in Mead & Conway, op. cit., and elsewhere, in connection with counters. Figure 4E is a schematic circuit of the last stage of the present invention shown in Figure 3. It implements the "Manchester Carry Chain" for the last stage in accordance with the decision table therefor above. It will be seen that Complete (Q) represents the Present State and Carryout represents the Manchester carry start.

Both the Carryout and the Next State (D) are outputs from respective NAND gates enabled by Reset. the other inputs to the NAND gates are outputs from two separate NOR gates. The NOR gate for the Carryout is enabled by Shift + Enable\ for acquiring the value of the Present State (Q), whilst the NOR gate for the Next State (D) is enabled by the Present State (Q) for acquiring the value of the second last stage (Datain) when Shift + Enable\ is on low.

Operation of the last stage will be the same as the decision table therefor above.

The above described circuit for an SAR may be realised throughout employing only transistors of a single standard size (typically the minimum permitted by the chosen fabrication process). This makes the embodiment inherently compatible with Gate Array processes and devices.

Appendix 1 gives a simulation run of a 5-bit SAR according to the invention, on the industry-standard circuit simulation program PSpice (from Microsim Corporation, Irvine, CA).

Other embodiments of the present invention incorporate amendments or modifications which will be found advantageous in certain applications.

Figure 6 shows the active stage circuit of Figure 4D, modified including further CMOS transistor switches to permit all SAR stages to be set to Zero by a single Clock signal (upon the application of RESET\). This may be advantageous where maximum operating speed is required.

Figure 7 shows the active stage circuit of Figure 4D, further amended to permit the result (at completion of a conversion operation) to be shifted serially out of the SAR, while simultaneously shifting in Zeros, in readiness for the next conversion. In this design, the action of RESET\ forces the SAR to act as a simple shift register. If desired, the SAR contents may be shifted out (for example, for transmission over some suitable medium), and simultaneously shifted back into the SAR. In this case, the effect is to restore the original SAR data, should this be required.

Whilst the primary field of use the present invention is in ADC, SAR may also find other applications. For example, in a phase-locked loop in which such an SAR is employed to provide a self calibrating property, so providing tolerance to manufacturing process variations, as disclosed in copending application PCT/AU88/00445. Other uses will occur to those skilled in the art where SAR of the type hereinbefore described may be found suitable for other functions and purposes, besides those already described.

APPENDIX 1. PSPICE SIMULATION

The following two files give respectively, a library of simulated logic functions, and a 5-bit SAR according to the Invention, which utilises those simulated functions. The simulation results are given in Figure 8.

* "Pseudo-NMOS" logic Function D R Brooks Nov. 1988 *

* These functions provide quick simulations of NMOS-style logic

* functions.

* Being based on Voltage-Controlled Switches, rather than MOS

* transistors, they require a fraction of the computing resources

* otherwise required, at the expense of a less accurate simulation.

* They are intended to provide support functions for circuits under

* development, where the actual logic inplementation is simulated

* separately. A typical use is in mixed analogue/digital simulations

* where most of the computing power is required for the Analogue areas.

* General logic specs. :

* Levels: 0V false, 57 true.

* Zout: 1k/5pF

* Zin: infinite (voltage-controlled resistor)

* Thresholds: 1V, 4V

* Gate Delay: 7nS typical

*

********************************************************************************

*

* The basic Voltage-Controlled Switches (ie "NMOS transistors")

. model trans vswitch (ron=1 roff=1e9 voff=1.0 von=4.0) ;Pulldown

. model pass vswitch (ron=1e4 roff=1e9 voff=1.0 von=4.0) ;Pass-gate

*

********************************************************************************

*

* Standard output- load stage (pullup & limit switching rate)

. subckt load Vdd in out

rpullup Vdd in 100 ;Pullup resistor

rslow in out 1K ;Output Impedance

cslow out 0 5pF ;Output Delay

. ends

*

********************************************************************************

Claims (22)

THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS:

1. A successive-approximation register (SAR) having a single shift register for processing, that is presetting and selectively resetting, a number of bits, wherein said single shift register is arranged to provide bit selection for processing the bits and also to provide desired result accumulation in the processed bits.

2. An SAR as claimed in Claim 1, wherein processing of a given bit is arranges to start simultaneously, on the same clock signal, with the completion of processing its immediate preceding bit.

3. An SAR as claimed in Claim 2, wherein the presetting of the given bit is arranged to occur simultaneously, on the same clock signal, with the selective resetting of its immediate preceding bit.

4. An SAR as claimed in Claim 2, wherein a processed bit is recognised by having at least one set bit in the shift direction.

5. An SAR as claimed in Claim 2, wherein the bit being processed is recognised by having no set bit in the shift direction and is itself a set bit.

6. An SAR as claimed in Claim 2, wherein an unprocessed bit is recognised by having no set bit in the shift direction and is itself not a set bit.

7. An SAR as claimed in Claim 2, wherein the next bit to be processed is recognised by having no set bit in the shift direction and having an adjacent set bit in the opposite direction to the shift direction and is itself not a set bit.

8. An SAR as claimed in Claim 2, wherein

(a) processed bit is recognised by having at least one set bit in the shift direction,

(b) the bit being processed is recognised by having no set bit in the shift direction and is itself a set bit,

(c) an unprocessed bit is recognised by having no set bit in the shift direction and is itself not a set bit, and

(d) the next bit to be processed is recognised by having no set bit in the shift direction and having an adjacent set bit in the opposite direction to the shift direction and is itself not a set bit.

9. An SAR as claimed in any one of Claims 4 to 8, wherein the status recognition of bits is performed by a Manchester Carry chain, said chain being arranged to pass information in the opposite direction to the shift direction.

10. An SAR as claimed in Claim 1, wherein initialisation of the register occurs upon application of a single clock signal.

11. An SAR as claimed in Claim 1, wherein the register is adapted to provide a bit-serial output.

12. An SAR as claimed in any one of preceding claims, wherein said single shift register comprises an array of stages, said stages including a first stage, a last stage and a number of active stages equal to the number of bits of digital output.

13. An SAR as claimed in Claim 12, wherein the first stage comprises a storage cell and is adapted to provide an output for presetting the active stages.

14. An SAR as claimed in claimed in Claim 12, wherein the last stage comprises a storage cell and is adapted to provide an output for controlling the bit selection for processing the active stages.

15. An SAR as claimed in Claim 12, wherein each active stage comprises a storage cell for storing successive bits of the desired digital output and switches for selective processing of data for said storage cell.

16. An SAR as claimed in Claim 15, wherein each storage cell is implemented in Dynamic Logic and adapted to hold its data as long as clock signals are applied to the register.

17. An SAR as claimed in Claim 15, wherein each switch is implemented in CMOS.

18. An SAR as claimed in Claim 12, wherein

(a) the first stage is adapted to provide an output for presetting the active stages,

(b) the last stage is adapted to provide an output for controlling the bit selection for processing the active stages, and

(c) each active stage comprises a storage cell for storing successive bits of' the desired digital output and switches for selective processing of data for said storage cell.

19. An SAR as claimed in any one of preceding claims comprising only transistors of a single standard size and implemented in gate array devices.

20. An SAR as claimed in any one of preceding claims implemented in VLSI.

21. An SAR as claimed in any one of preceding claims implemented in MOS.

22. An analogue to digital converter (ADC) including an SAR as claimed in any one of preceding claims.