BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to the field of differential switches and, in particular, to differential switches operated under control of a single binary signal, as in the case of digital to analog converters (DACs).
2. Description of the Related Art
Binary-control differential switches operate as single-pole double-throw switches and are employed in many applications such as DACs. Within such a switch, a single control terminal effects contact between a first conducting terminal and a second conducting terminal while breaking contact between the first conducting terminal and a third conducting terminal. Although the utility of such switches will be described in reference to their application within DACs, they may be used for many other applications.
There are a number of conventional DAC architectures which employ differential switches. Some are current output, some are voltage output. For examples of both see, Analog - Digital Conversion Handbook, Daniel H. Sheingold ed., Prentice Hall, Englewood Cliffs, N.J., 1986, pages 191-206. A current output DAC is illustrated in FIG. 1 , but the new switches are applicable to voltage output DACs as well. Generally, a DAC is a device which converts a quantity specified as a binary number (this includes BCD, two's complement and other binary codes) into a current or voltage which is proportional to the value of the digital input. The digital input is typically held in a digital section 11 which may receive the digital input either serially or in parallel through a digital interface 13 . The binary number thus stored typically controls an analog output section 15 which comprises a set of differential switches S 1 -Sn, with each bit of the binary number controlling which one of two terminals within an associated differential switch S 1 -Sn is connected to a third terminal within the switch. In some cases a single switch is employed to “build up” an analog signal over some period of time.
More specifically, in the design of FIG. 1 , an array of differential switches S 1 -Sn connect binarily-weighted current sources I 1 -In to either a return, or reference, terminal 10 or output terminal 12 under control, respectively, of binary control inputs from the least significant bit LSB 14 to the most significant bit MSB 18 . Current for the sources I 1 -In is provided through terminal 9 . The sum of the currents appearing at the output terminal 12 provide a “stair step” approximation to the continuous signal represented by the binary control inputs MSB-LSB. Alternatively, the reference terminal 10 and output terminal 12 could be connected to high and low voltage references and the terminal 9 would then provide a voltage output. In this case, the taps of a resistor ladder are connected in place of the current sources I 1 -In.
As the binary values at the control inputs LSB-MSB vary, the switches S 1 -Sn route currents corresponding to the varying binary values of the control inputs to the output terminal, where the total current may be converted into a voltage. The switches are generally switched “simultaneously” to provide valid current levels at the output terminal 12 . However, as will be explained in greater detail in relation to FIG. 2 , there is often a delay introduced between the “make” and “break” actions of the switches S 1 -Sn, causing spurious signals, or “glitches”, to appear at the output terminal 12 .
The mid-scale glitch, produced by the transition of the control codes appearing at terminals 18 - 14 between 1000 . . . 0 and 0111 . . . 1, is usually the worst glitch because all the switches S 1 -Sn are switching at this transition. Glitches will also occur at other transition points, but they will generally be of lesser magnitude. Glitches are particularly onerous in waveform reconstruction applications such as direct digital synthesis systems.
Code-dependent glitches, such as those just discussed, will produce both out-of-band and in-band harmonics of the desired signal. For example, in reconstructing a sine wave, the midscale glitch occurs twice during each sine wave period, at each mid-scale crossing. In this manner the midscale glitches produce a second harmonic of the sinewave. Although filtering may eliminate or reduce to a tolerable level the contribution from some of the glitches, higher order harmonics, which alias back into the Nyquist bandwidth, cannot be filtered. To avoid filtering and to eliminate spurious signals that cannot be filtered, it would therefore be desirable to avoid introducing the glitches whenever possible.
The block diagram of FIG. 2A provides a more detailed view of a typical conventional switch, which may be employed as one of switches S 1 -Sn of FIG. 1 . As an example, switch S 1 includes a differential switch pair 20 , comprising switches swa and swb, which connect either terminal a or b to terminal c. Control terminals T and I are connected to receive complementary control signals developed within a latch L 1 . The latch L 1 accepts the binary control signal MSB and converts it into a complementary pair of control signals for use with the differential pair swa and swb. The utility of the latch L 1 derives from the fact that, at a system level, whatever device is driving, or controlling, the DAC, in all probability has other duties to perform and may address those other duties only if it stores its required digital patterns within the switches S 1 -Sn, and then proceeds to other operations.
The control input MSB provides a digital signal path for control inputs to the switch S 1 . An MSB signal enters the latch L 1 and, under control of enable signals ck and ckb, is transferred through an analog switch ASW 1 . It is then inverted, or complemented, by an inverter INV 1 to produce a control signal INVERTED which is applied to the control terminal of the switch swb. Analog switches are known in the art. A description of them may be found in, Paul Horowitz, and Winfield Hill, The Art of Electronics, Cambridge University Press, N. Y., 1989, pages 142-143. The output of the inverter INV 1 is connected to the input of a second inverter INV 2 which inverts the signal INVERTED to produce a control signal TRUE which is applied to the control input 24 of the switch swb, the other switch of the differential pair. Note that inversion of the INVERTED signal by inverter INV 2 produces a delay between the control signals applied to the differential pair. That is, the INVERTED signal will arrive at the control terminal 22 of switch swb one inverter's delay before the TRUE signal arrives at the control terminal 24 of switch swa. Consequently, a glitch impulse will be created at the S 1 output terminal c.
Returning to the operation of the latch L 1 , the enable signals ck and ckb are assumed to be complementary and non-overlapping. That is, more circuitry than a simple inverter is required to produce ckb from ck. During a first phase of the enable signals ck and ckb, the input signal from MSB is “clocked” through the analog switch ASW 1 . At the same time, because the control inputs to analog switch ASW 2 are connected opposite to the connection of ASW 1 , ASW 2 will be “off”, thus isolating the output of ASW 1 from the output of the inverter INV 2 .
However, during the second phase of the enable signals, analog switch ASW 1 is off and ASW 2 is on. With ASW 1 off, the MSB terminal is isolated from the circuit beyond the analog switch ASW 1 . With analog switch ASW 2 on, inverters INV 1 and INV 2 are “cross-coupled”. That is, the output of INV 1 feeds the input of INV 2 and the output of INV 2 feeds the input of INV 1 . In this conventional configuration, there are two stable states which the inverters may assume, i.e., INV 1 =1, INV 2 =0 or INV 1 =0, INV 2 =1 and, by feedback, they will remain in whichever state to which they are forced. In this way the cross-coupled inverters, coupled through the switches ASW 1 and ASW 2 , form a latch which provides TRUE(delayed) and INVERTED control signals for a differential switch pair from a single binary control signal i.e., that from the MSB terminal.
The switches swa and swb which comprise the differential pair may be any type of switch, including p-channel or n-channel MOSFETs, NPN or PNP bipolar transistor or analog switches. Employing analog switches for switches swa and swb provides some flexibility in choosing between current output or voltage output DACs. An implementation which employs PNP transistors as switches swa and swb is illustrated in FIG. 2B , for example. The emitters of two PNP transistors are connected to a current source such as I 1 in FIG. 1 A single control signal, e.g. MSB, is converted to a differential pair of control signals and applied to the respective control terminals 22 and 24 , i.e. the bases, of the transistors. With complementary control signals, only one of the transistors will conduct at a given time(ideally), switching current, for example, from the current source I 1 into either a return 10 or output 12 path.
An analog switch implementation of the differential switch pair, illustrated in FIG. 2C , operates substantially the same as the PNP transistor pair of FIG. 2 B. Differential control signals derived from a control input such as MSB are applied, cross-coupled, to the inverting and non-inverting control inputs of two analog switches. One input, or switch contact, of each analog switch is connected to a current source, the other input is connected to the return terminal 10 , the other input of swb is connected to output terminal 12 . Ideally, the differential control signals place only one of the switches in the conduction mode at a time, thereby switching current from the current source either into the return path or into the output 12 path. However, as described above, the delay between control signals TRUE and INVERTED sometimes place both switches swa and swb into conduction at the same time.
The advantage of employing analog switches for switches swa and swb lies in the fact that they conduct bidirectionally; therefore a voltage output may be produced by substituting voltage references at the return 10 and output 12 terminals and taking the output from the terminal 14 which, in the current output configuration, provides the reference current I 1 .
Another latch may be added “in front of” L 1 to produce a conventional master/slave latch which provides added isolation between input and output. This additional level of isolation may be used, for example, to update a binary input value by shifting a desired binary value into position at the inputs to a set of master latches, keeping the slave latches isolated, then shifting the updated value into the slave latches simultaneously.
As just described, conventional switches require a somewhat elaborate scheme to produce non-overlapping complementary enable signals to drive the control inputs of analog switches which, along with a pair of inverters, form a binary to differential control latch. Not only is an elaborate enable signal required, glitches, which may create unfilterable spurious signals, are produced by the delay between the generation of TRUE and INVERTED control signals for the differential switch pair.
SUMMARY OF THE INVENTION
The invention is directed to a differential switch that minimizes the complexity of a switch controller's clock generation circuitry and reduces spurious switching, thereby reducing the occurrence and duration of undesirable switch outputs, or glitches. These goals are achieved by a latched differential switch which inverts the control input and then simultaneously transfers the control input and its complement through transfer switches into storage elements. Although inverting the control signal introduces a delay between the control signal and its complement (referred to as the “TRUE” and “INVERTED” signals hereinafter), simultaneously transferring them into storage elements eliminates this skew. The storage elements' outputs are connected to the control inputs of a differential switch pair, thus providing “de-skewed” control for a differential switch pair from a single binary control input.
In one implementation, the novel switch includes a n intermediate set of transfer switches, operated from the same “enable” signal as the first set of transfer switches, and an intermediate set of storage elements. The intermediate sets of switches and storage elements are interposed between the TRUE and INVERTED inputs and the first set of transfer switches. During the first cycle of the enable signal, the TRUE and INVERTED signals are simultaneously transferred into the intermediate set of storage elements. As described above, this simultaneous transfer eliminates the skew between the TRUE and INVERTED signals. The TRUE and INVERTED signals are then transferred into the first storage element, which is isolated from the intermediate storage element, during the second cycle of the enable signal.
The isolation between the first and second storage elements prevents transitions at the input to the second storage element from appearing at the first storage element. The TRUE and INVERTED signals are therefore available from the second storage element without the skew between them that had been introduced by inverting the control signal. The outputs of the first storage element are connected, as described above, to the control inputs of a differential switch pair, thus providing “de-skewed” control for the switch pair. The novel switch may be used, for example, within a DAC to reduce the DAC's glitch energy output.
These and other features, aspects and advantages of the invention will be apparent to those skilled in the art from the following detailed description, take together with the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram, described above, of the interconnections between control bits, differential switches and current sources within a conventional DAC.
FIG. 2A described above, is a schematic diagram of one of the differential switches of FIG. 1 .
FIG. 2B described above, is a schematic diagram of an PNP transistor implementation of the differential switch block of FIG. 2 A.
FIG. 2C described above, is a schematic diagram of an analog switch implementation of the differential switch block of FIG. 2 A.
FIG. 3 is a block diagram of a preferred embodiment of the new differential switch provided by the invention, illustrating the switch's latch and the latch's inter-connection with a differential switch pair.
FIG. 4 is a schematic diagram of one implementation of the novel differential switch which employs cross-coupled inverters as storage elements.
FIG. 5A is a schematic diagram illustrating “switched” cross-coupled inverters, which may be used as the novel switch's storage elements.
FIG. 5B is a schematic diagram illustrating a gated inverter employed as one of a cross-coupled inverter pair, which may be used as one of the novel switch's storage elements.
FIG. 5C is a detailed schematic diagram of a gated inverter such as illustrated in FIG. 5 B.
FIG. 5D is a schematic diagram illustrating the use of a capacitor as a storage element within the novel switch.
FIG. 6 is a schematic diagram of a preferred master/slave embodiment of the novel switch.
DETAILED DESCRIPTION OF THE INVENTION
The new switch 30 of FIG. 3 may be employed within a DAC, for example, as one of the differential switches S 1 -Sn illustrated in FIG. 1 . The new switch 30 comprises a latch 32 with a binary input MSB and enable input ck. The latch 32 accepts a single binary signal at an MSB input, and produces de-skewed TRUE and INVERTED control signals at like-named outputs. These outputs are connected to the control terminals 34 and 36 of a conventional differential switch pair 38 which may be compose, for example, as described in the background section, of NPN or PNP bipolar transistors, N-channel of P-channel MOSFETs or CMOS analog switches.
The latch 32 accepts a binary signal at the input MSB, inverts the signal with an inverter INV 3 and, under control of transfer switches TSW 1 and TSW 2 , transfers the TRUE and INVERTED signals thus produced into storage elements 40 and 42 . Outputs from the storage elements 40 and 42 are connected to the control inputs 34 and 36 of a differential switch pair comprising switches swa and swb. The enable signal ck controls the simultaneous transfer of TRUE and INVERTED control signals into storage elements 40 and 42 , respectively. Although the inverter INV 3 introduces a skew between the TRUE and INVERTED signals as it produces the INVERTED signal from the TRUE signal, the simultaneous transfer of these signals into the storage elements eliminates this skew. In some applications, TRUE and INVERTED signals may be available as inputs to the novel switch(with an inverter's delay between them). In those cases, the inverter INV 3 could be eliminated from the switch 30 .
As indicated in FIG. 3 , the control signals TRUE and INVERTED are available at the control terminals 34 and 36 of the differential switch pair coincident with their transfer into the storage elements 40 and 42 . Once these signals are stored and are providing control over the differential switch pair, the transfer switches TSW 1 and TSW 2 are opened to provide isolation from the input MSB which may be subject to modification of its logic state due to the transfer of data into a DAC of which the novel switch 30 is a part.
An implementation of the novel switch 30 is illustrated in FIG. 4 , employing P-channel FETs as transfer switches TSW 1 and TSW 2 within latch 32 . Weak cross-coupled inverters INV 4 and INV 5 serve as storage elements 40 and 42 . As noted in the background section above, cross-coupled inverters are known in the art and, briefly, have two stable states which they may assume. Once forced into one of those states, they will maintain it. Because the two states desired for the differential switch control terminals 34 and 36 coincide with the two states available from the cross-coupled inverters, the inverters provide the functions of storage elements 40 and 42 , although they are not independent storage elements in the sense of a pair of memory cells, and cannot store binary patterns corresponding to the (undesirable) states which would simultaneously turn both switches swa and swb “on” or “off”. As described in greater detail in relation to FIG. 6 , the inverters INV 4 and INV 5 are “weak” in that they have limited drive current capability and therefore may be easily forced, by whatever device drives the MSB input and the inverter INV 3 , into a desired state.
Alternatively, a pair of cross-coupled inverters INV 6 and INV 7 , one of which (INV 7 ) is a gated inverter, may be employed, as a storage element, as illustrated in FIG. 5A , with one pair for each storage element 40 and 42 . As will be described, this implementation eliminates the need for weak inverters. The inverter INV 3 , inputs MSB and ck and outputs TRUE and INVERTED are as described in relation to FIG. 3 . In this implementation, the storage elements 40 and 42 are identical; the detailed description given below for storage element 40 will also apply to storage element 42 .
A signal ck is asserted to transfer a single-bit signal MSB through a transfer switch TSW 1 , when MSB is valid, to the input of inverter INV 6 . A signal ckb, the non-over-lapping inverse of ck, opens switch SWI 1 when ck closes TSW 1 . The switch SWI 1 is connected between the output of the inverter INV 7 and the input of the inverter INV 6 . Consequently the output of the inverter INV 7 is isolated from the device driving the input of the inverter INV 6 and, because there is no contention between the output of INV 7 and the device driving the input of inverter INV 6 , inverters INV 6 and INV 7 needn't be “weak” inverters. When the ck input is “de-asserted”, transfer switch TSW 1 opens and switch SWI 1 closes. During the short time that neither switch is closed, stray capacitance maintains the state of inverter INV 6 until the switch SWI 1 closes. With switch SWI 1 closed, the inverter INV 7 provides positive feedback to the inverter INV 6 , thereby “latching” the inverters and providing the TRUE and INVERTED control signals for the differential switch control terminals 34 and 36 (not shown). In some applications, the differential switch pair requires more drive current than “weak” inverters may provide. Employing standard inverters and the additional switch SWI 1 as illustrated in FIG. 5A eliminates buffers which would otherwise be required to drive the differential switch pair 38 .
Similarly, FIG. 5B illustrates an implementation of the storage element 40 (also applicable to storage element 42 ) which employs a gated inverter 46 to achieve the same isolation between the input to the inverter INV 6 and the output from a feedback inverter, in this case, gated inverter 46 . This isolation is achieved, as described below, by causing the output of the inverter 46 to “float”. The input to inverter INV 6 is the switched data input, i.e. the input to the storage element 40 , and its output TRUE drives the differential switch control inputs 34 , as illustrated in FIGS. 3 and 4 above. The gated inverter 46 , illustrated in detail in FIG. 5C , is cross-coupled with the inverter INV 6 . The same advantage accrues to the use of a gated inverter in this implementation as the use of the inverter switch SWI 1 in the implementation of FIG. 5A , i.e., the cross-coupled inverters needn't be weak inverters and, therefore, buffers which otherwise may be required to drive the differential switch pair's control terminals 34 and 36 may be eliminated.
The gated inverter 46 shown in FIG. 5C is controlled by the control signals ck and ckb previously identified. The inverter 46 is composed of two n-channel FETs, n 1 and n 2 , and two p-channel FETs, p 1 and p 2 . FETs n 2 and p 1 are connected as a conventional CMOS inverter, FETs n 1 and p 2 are connected in series between the inverter formed by n 2 and p 1 and positive and negative supplies V + and V, respectively. When the ck signal is asserted (driven HIGH), transferring the MSB signal into storage element 40 , it also forces p 2 into a non-conducting state. At the same time ckb is driven LOW, forcing n 1 into a non-conducting state. The inverter is “floating” in this state and whatever device drives the input to inverter INV 6 can also drive the output of gated inverter 46 without contention from the gated inverter 46 .
When ck is de-asserted, after the binary value from MSB has been transferred to the input of the inverter INV 6 , FETs n 1 and p 2 are turned on, thereby providing supply voltages to the inverter 46 . At this point the input 50 to the gated inverter 46 will be driven to the updated level by the output of inverter INV 6 , and the output 48 of the gated inverter 46 will drive the input of the inverter INV 6 to the same level transferred to it by the transfer switch TSW 1 . In this way an updated value of MSB is transferred to the input of the inverter INV 6 without contention from the output of the gated inverter 46 . This value is then latched, through positive feedback from the gated inverter 46 , at the output of the storage element 40 .
In some applications, notably those employing a DAC in waveform reconstruction, the DAC's differential switches' data inputs, LSB 14 -MSB 18 , are rapidly updated. As illustrated in FIG. 5D , a capacitor C 1 of sufficient capacity to hold its value throughout an individual switch's longest update period, i.e. the longest period during which transfer switch TSW 1 is left open, could be used as a storage element 40 . After transfer switch TSW 1 charges the capacitor C 1 to the binary value present at MSB, it opens so that only leakage paths are available to charge or discharge the capacitor C 1 (assuming that C 1 drives a high impedance input). So long as the voltage across the capacitor does not fall to the following device's low input threshold before being recharged, the capacitor can serve as a storage element. It may be necessary to buffer the capacitor C 1 , if, for example, storage element 40 is to drive one of the differential switch pair 38 of FIG. 4 and the pair 38 comprises bipolar transistors. This may be accomplished using an inverter INV 8 which provides sufficient drive current for one of the differential switch control terminals 34 or 36 .
In a preferred embodiment, the novel switch 30 employs a master/slave architecture, as illustrated by the schematic of FIG. 6 . All the inverters employed within this preferred embodiment are CMOS inverters, the relative strengths of which will be discussed below. A master latch 50 latches control signals during a first phase of the enable input ck. Then, during the second phase of ck, these signals are transferred to a slave latch 52 . As described in relation to FIG. 2A above, the master/slave architecture provides greater flexibility in re-loading the latches, in that the data value of the input MSB need be valid for only a short time around the transition of the enable signal ck. A buffer section 54 which follows the slave latch 52 increases the drive capability over that of the slave latch 52 . The outputs of the buffer section drive the control inputs 34 and 36 of the differential switch pair swa and swb which, in the preferred embodiment, are p-channel FETs.
The switch 30 accepts an enable signal at an input ck and data at an input MSB. Within a DAC, the data appearing at the input MSB would represent one bit of a digital code which is to be converted into an analog output. Data appearing at the MSB input is inverted by a CMOS inverter INV 9 and the true and complement signals thus formed are passed through switches P 5 and P 4 , respectively, when the signal ck goes low. In this embodiment, switches P 5 and P 4 are p-channel FETs. Inverters INV 10 and INV 11 form a cross-coupled weak inverter “master” latch that accepts signals passed through the switches P 5 and P 4 . A pair of inverters INV 12 and INV 13 buffer the outputs from the latch.
When the enable signal ck goes high, n-channel MOSFET switches N 4 and N 5 pass the outputs from the buffers INV 12 and INV 13 to a “slave” latch formed by cross-coupled weak inverters INV 14 and INV 15 . At the same time switches P 4 and P 5 turn off, thereby isolating the master latch from the MSB signal and its complement. The INVERTED and TRUE outputs from the slave latch are buffered by inverters INV 16 and INV 18 connected in series and INV 17 and INV 19 connected in series, respectively. The buffer inverters INV 18 and INV 19 drive the control terminals 34 and 36 of the differential switch pair 38 . In the preferred embodiment, switches swa and swb are p-channel MOSFETs. Since the signal at the MSB input is complemented before being latched, the skew created by the delay of inverter INV 9 is eliminated during the latching process. Consequently, a single binary signal is converted into de-skewed differential switch pair control signals available to drive inputs 34 and 36 .
To accommodate the difference in mobility between holes and electrons, the size (i.e., channel width to length ratio) of the p-channel FET within each inverter INV 9 -INV 17 is preferably about 3.2 times that of the corresponding n-channel FET, and the p-channel FETs of inverters INV 18 and INV 19 are about twice the size of their corresponding n-channel FETs. Additionally, the channel widths of the weak inverters, (INV 10 , INV 11 , INV 14 and INV 15 ) are approximately 0.4 that of the other inverters. This allows the devices which drive the weak inverters to do so with little contention from the weak inverters.
The forgoing description of specific embodiments of the invention has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed, and many modifications and variations are possible in light of the above teachings. For example, the novel switches may be employed within a DAC which is itself a part of an analog to digital converter, such as a successive approximation converter. The embodiments were chosen and described in order to best explain the principles of the invention and its practical application, to thereby enable others skilled in the art to best utilize the invention. It is intended that the scope of the invention be limited only by the claims appended hereto.
