This application is a division of my copending application Ser. No. 809,700 filed Mar. 24, 1969, now U.S. Pat. No. 3,685,045.
This invention relates to digital-to-analog converters. More particularly, this invention relates to such converters which are capable of high-speed conversion with stability and freedom from transient effects.
A wide variety of digital-to-analog converters have been provided for many purposes heretofore. Initially such converters used vacuum tubes, but as with most electronic devices, vacuum tubes have been replaced with later developed solid-stage elements. Since the design criteria of solid-stage elements are significantly different from vacuum tubes, this replacement process has presented a number of special problems. In addition, with the increasing speeds attainable by computers and other digital devices, there has been a corresponding demand for increased speed from digital-to-analog converters.
Accordingly, it is a principal object of the present invention to provide solid-state digital-to-analog converters with improved operating characteristics, particularly high-speed conversion capabilities together with reliable, accurate performance.
Other objects, aspects and advantages of the invention will in part be pointed out in, and in part apparent from, the following description considered together with the accompanying drawing in which
FIGS. 1A and 1B together show a circuit diagram of one presently preferred embodiment of the invention, and
FIG. 2 is a schematic diagram, partly in block, showing in simplified form the principal elements of the preferred embodiment which serve to maintain constant the current through the switching transistors.
Referring now to the upper portions of FIG. 1B, there is shown aconventional storage register 10 having a series of separate binary stages 12 (12A, etc.). Input leads 14 (14A, etc.) supply to the stages 12 the individual binary elements of a digital number to be converted to a corresponding analog signal level. These input leads may be connected to any digital source (not shown), such as a high-speed data processor.
The binary signals stored in the stages 12 are gated out essentially simultaneously by astrobe circuit 16 energized by conventional gate-generating means (not shown) producing periodic pulses of suitably high frequency. When the stages 12 are thus gated open, the stored binary signals are directed through respective coupling circuits comprising individual diodes 18 (18A, etc.). That is, each stage containing a stored binary bit produces a control pulse which passes through the corresponding coupling diode. This control pulse is of negative polarity, and is applied to the emitter 20 (20A, etc.) of a corresponding PNP buffer transistor 22 (22A, etc.) arranged so as normally to conduct current.
The bases 24 (24A, etc.) of all of the buffer transistors 22 are connected together to apower supply lead 26 providing a regulated bias voltage somewhat more positive than -15 volts. The emitters 20 of all of the buffer transistors are connected through respective resistors 28 (28A, etc.) to a secondpower supply lead 30 having a regulated voltage of about +15 volts. The collectors 32 (32A, etc.) of the buffer transistors are connected to corresponding NPN switching transistors 34 (34A, etc.) so as to control the output thereof in a manner to be described hereinbelow in detail.
The collector 32 of each buffer transistor 22 is connected to one end of a corresponding load resistor 36 (36A, etc.) forming part of the output circuit of the associatedswitching transistor 34. The remote ends of these load resistors are connected in common to apower supply lead 38 maintained at about -60 volts. When any buffer transistor is on, its output current flows through the associated load resistor 36, and the resulting voltage drop across this resistor causes the emitter 40 (40A, etc.) of the corresponding switching transistor to be biased to cut-off. Thus, no current will flow through a switching transistor while the associated buffer transistor is on.
When any buffer transistor 22 is cut off by a negative control pulse coupled through itsinput diode 18, the cut-off bias at the emitter 40 of thecorresponding switching transistor 34 disappears, and that transistor therefore immediately conducts. The load circuit of each switching transistor is so arranged that when the transistor is turned on, the magnitude of its output current will be virtually equal to that of the current previously passing through the series resistor 36 from the associated buffer transistor 22. Thus, the operating conditions of the switching transistor will be changed but very little during the switching transition, e.g. the voltage of the emitter 40 may change by only a little more than 0.7 volts, the normal voltage drop across a conducting transistor. This small change in operating voltages tends to assure smooth and rapid switching.
The buffer transistors 22 serve the important function of substantially isolating theswitching transistors 34 from the transient effects of the gating strobe control pulse. That is, such a control pulse, if applied directly to the switching transistor, would introduce relatively large momentary signal variations in the output circuit of the transistor, as a result, for example, of leakage capacitance coupling of the leading edge of the control pulse. Such transient effects introduce errors in the conversion operation, particularly as the conversion speed is increased to the point where there is insufficient time for the transient effects to settle out. The transient effects of a switching pulse are somewhat erratic, and are difficult to eliminate by conventional circuit arrangements.
The individual buffer transistors 22 significantly minimize the transient effects of capacitive coupling feed-through to the switch output. The result is a considerable improvement in accuracy of the conversion, especially at high speeds. In addition, the use of the buffer transistors makes it readily possible to strobe the converter with negative-going gate pulses, preferred in such logic circuitry.
The output currents of all of the first eightswitching transistors 34A-34H are pre-adjusted to be of exactly the same magnitude (approximately 1 mil), by selection of the appropriate value for the associatedload resistors 36A-36H. A portion of the output current of each conducting transistor is coupled through alead 42 to the summing input terminal 44 (FIG. 1A) of anoperational amplifier 46. The magnitude of this portion of current is fixed in accordance with a 2:1 weighting relationship to correspond to the order of the binary bit represented by the respective switching transistor. Specifically, the current contribution of thesecond transistor 34B is arranged to be one-half that of thefirst transistor 34A, the current contribution of the third transistor 34C is one-half that of the second 34B, and so forth.
The output of thefirst switching transistor 34 is connected directly to thesumming terminal 44 of theoperational amplifier 46, and thus this transistor contributes its entire output. The next threeswitching transistors 34B, 34C, and 34D are connected to the summing terminal by individual weighting networks comprisingcurrent dividers 48, 50 and 52. The preferred form of divider consists of two series-connected resistors the common junction of which is connected to the collector electrode 54 of the associated switching transistor, and the remote terminals of which are connected respectively to ground and thesumming input terminal 44 of theoperational amplifier 46. Thus, the amount of current contributed by any one of these latter threeswitching transistors 34B, 34C and 34D is determined by the ratio of the two resistors in the corresponding current-divider network 48, 50 and 52, in such a manner as to provide the required 2:1 ratio from one to the next.
The next four output transistors 34E-34H form a second discrete set, all coupled to thesumming input terminal 44 of theoperational amplifier 46 by means of a two-to-oneladder network 56 consisting of a series of four cascadedidentical stages 58, 60, 62 and 64. The intersection points 66, 68, 70 between the separate stages are connected respectively to thetransistor collector electrodes 54E, 54F and 54G, and the right-hand end terminal 72 (serving as the input terminal for the ladder network) is connected tocollector electrode 54H. In this ladder network, the ohmic resistance of eachseries resistance 74 is one-half that of the associatedshunt resistance 76. Thus, visualizing the signal flow as proceeding from right to left, each stage of the ladder network provides a 2:1 attenuation of any current supplied thereto, either from the associatedswitching transistor 34, or from the preceding (righthand) stage of the ladder network.
Although thisladder network 50 does introduce some distributed capacitance effects, and permits some interaction between the functioning of the associatedswitching transitors 34, these effects produce relatively small consequences in the overall conversion accuracy because the bits of data involved are several orders down from the most significant bit of the complete digital number. Moreover, such adverse effects are compensated for, at least to some extent, by arranging the switching transistors to produce the same magnitude of current output. This equal-current arrangement tends to minimize instability and other error effects.
The load resistors 36I-36M of the last set of five switching transistors 34I-34M are so proportional relative to one another as to provide the desired two-to-one ratio in current flow through the respective transistors. The load resistors 28I-28M of the corresponding buffer transistors 22I-22M are similarly proportioned. That is, each resistor in the sequence has a total ohmic resistance of approximately twice that of the preceding resistor of the sequence. Thus, the magnitude of the current supplied by each switching transistor 34I-34M is one-half that of the preceding transistor, i.e. the transistor to the left, as seen in the drawing.
The collector electrodes (54I-54M) of all five of this third set of switching transistors 34I-34M are connected together to theinput terminal 72 of theladder network 56. Any one of the transistors which is gated on will thus supply a corresponding weighted current contribution through the ladder network to the summinginput terminal 44 of theamplifier 46. Although the use of currents of different magnitude in each of the switching transistors 34I-34M introduces some asymmetries in the conversion operation, these asymmetries do not have any important effect on the final result because the five transistors of this third set provide digital bits corresponding to the lowest orders of the digital number, i.e. the five least significant bits of the group. The direct connection of this third set provides desirable economies of construction without important performance limitations.
For some applications, it is necessary to provide a sign-change capability, i.e. so as to develop either positive or negative anlog outputs corresponding to positive or negative digital inputs. Referring to the lower portion of FIG. 1A, such a capability can be provided by coupling the output of theamplifier 46 to an invertingamplifier 80, and by employing aselector switch 82 having twosections 82A, 82B to select either the direct output or the inverted output. Theswitch 82 is operated by aconventional switch driver 84 controlled by aninput lead 86 to which is directed a sign bit, i.e. a binary bit indicating whether the number to be converted is positive or negative. The sign bit is gated by a strobe circuit (not shown) synchronized with the converter strobe. When so gated, thedrive 84 opens eitherswitch section 82A orswitch section 82B, but not both at the same time. The selected analog signal is coupled to anoutput amplifier 88 which provides the final analog output signal of the converter.
Since it is not readily possible to assure exact synchronism between the operation of theswitch 82 and the strobing of thestorage register 10, transient errors may be developed in the converter output during the transition between negative and positive outputs. The problem cannot be solved simply by arranging the circuit so that thesign switch 82 is always actuated slightly before or slightly after strobing of the storage register, because a momentary error effect such as an overshoot can result under either circumstance, depending upon the beginning and ending voltages of the analog output. In accordance with a further aspect of the present invention, this problem has been solved by a special arrangement for insuring that the analog output signal, whenever a sign change is to be made, will first be brought to zero potential. Since every sign change requires the analog voltage to pass through zero, automatically shifting the voltage to zero whenever a sign change is to take place insures that the output will not vary in the wrong direction at the outset of a change. Holding the output on zero until all of the switching has been completed prevents overshooting of the final voltage.
In more detail, the converter includes (referring now to the upper left-hand corner of FIG. 1A) asign change detector 90 which in the present embodiment includes a conventional flip-flop 92 arranged to receive the sign bit as the controlling input. The set and reset outputs of this flip-flop are coupled through respectivederivative circuits 96, 98 and isolatingdiodes 100, 102 to acommon load resistor 104. Thus, whenever there is a change of sign (where the sign bit changes from "zero" to "one" or vice-versa), a sharp positive spike will be developed at theload resistor 104. This spike momentarily turns on atransistor switch 106 which, in turn, momentarily disables atransistor 108 serving to establish the bias voltage for thepower supply lead 26.
Power supply lead 26 thereupon goes negative and holds the buffer transistors 22 "on", producing a flow of current through all of the resistors 36 for a brief peiod. This current flow causes all of the switchingtransistors 34 to be turned off momentarily, thereby making the output voltage ofamplifiers 46 and 80 to be held momentarily at zero. Thus, even though the sign-change switch 82 is not exactly synchronized with the gating of theregister 10, the converter output will momentarily be shifted to zero during a sign change.
After the spike has subsided at the input totransistor 106, the buffer transistors 22 are all returned to normal operating conditions, and the strobed control pulses fed to these transistors fromregister 10 will activate the switchingtransistors 34 in a pattern representing the stored digital number. Thus, the output of theamplifier 88 will be shifted to the proper level, and transient errors during the sign change transition avoided.
Another source of error is changes in ambient temperature, which alter the operating characteristics of the switchingtransistors 34, and tend to vary the magnitude of current produced thereby. In accordance with a still further aspect of the invention, means are provided to minimize such effects of ambient temperature. More specifically, all of the bases 110 (110A, etc.) of the switchingtransistors 34 are connected to abias lead 112 the voltage of which is regulated so as to maintain the current through the switching transistors substantially constant with changes in temperature.
The voltage of thebias lead 112 is primarily determined by atransistor 114 in series with aresistor 116.Bias lead 112 also is connected to thebase 120 of acontrol transistor 122 matched to thefirst switching transistor 34A, particularly in having a "beta" which tracks the corresponding parameter oftransistor 34A with changes in temperature. Theemitter 124 ofcontrol transistor 122 is connected through aload resistor 126 to thepower supply lead 38, and thecollector 128 of this transistor is connected through aresistive network 130 to a positivereference voltage terminal 132. The circuit elements are so selected so as to produce a predetermined flow of current through theresistive network 130 and thecontrol transistor 122, and resulting in a zero potential at acontrol point 134 between theresistive network 130 and the control resistor. The magnitude of the current throughtransistor 122 is set to equal the flow of current through the switchingtransistor 34A when the latter has been turned on.
If there is a change in ambient temperature, the result typically will be a change in operating characteristics of the switchingtransistor 34A, so as to alter the normal current flow therethrough. By positioning thecontrol transistor 122 physically adjacent the switchingtransistor 34A, the same temperature effect will be experienced by the control transistor. The change of current produced by a change in temperature is detected by anoperational amplifier 136 having one input terminal connected to controlpoint 134, and its other input terminal connected through aresistor 138 to ground. The output of thisamplifier 136 is connected through aresistor 140 and an isolatingdiode 142 topower supply lead 38.
When there is a change in the current supplied to theamplifier 136 fromcontrol point 134, there will be a corresponding change in the amount of current drawn by this amplifier from thepower supply lead 38. Since this power supply lead is connected through aresistor 150 to thepower supply terminal 152, the change in current drawn by amplifier .[.134.]. .Iadd.136 .Iaddend.will cause a corresponding change in the voltage ofpower supply lead 38. Thus, the amplifier .[.134.]. .Iadd.136 .Iaddend.provides an amplified negative feedback action which automatically alters the voltage oflead 38 in such a way as to maintain constant the current flow through thecontrol transistor 122. Sincetransistor 122 is matched to switchingtransistor 34A, the change in potential ofpower supply lead 38 will have a similar effect on the functioning of this switching transistor, i.e. it will compensate for the change in ambient temperature oftransistor 34A, and assure that the current through that transistor is maintained effectively invariant with changes in temperature. Moreover, this result can be achieved with apower supply 152 of relatively modest complexity and cost, because the power supply need not be closely regulated internally.
The same controlling influence tends to maintain constant the current through theother switching transistors 34B, etc. However, as a practical matter these latter transistors need not be so identically matched in characteristics to thefirst transistor 34A, because they represent binary information of progressively less significance to the ultimate analog output voltage.
Typical values and types of elements used in a preferred embodiment of the invention as described above are as follows:
______________________________________ Diodes 1N4149 Buffer transistors 222N4250 Switching transistors 34 SE4010 Operationalamplifiers MC1539G Resistors 28A-28H 12.7K Resistor 28I 25.5K Resistor 28J 51.1K Resistor28K 100K Resistor28L 200K Resistor28M 390K Resistors 36A-36H 50K Resistor 36I 100K Resistor36J 200K Resistor36K 400K Resistor 36L 800K Resistor 36M 1.6MCurrent divider 48 2.5K and 2.5KCurrent divider 50 5K and 1.666KCurrent divider 52 3.5K and1K Series resistor 74 500Shunt resistor 76 1K Resistor 77 500 ______________________________________
It will be apparent from the foregoing description that various changes can be made to the preferred embodiment without departing from the spirit of the invention. For example, the values of elements herein disclosed should not be construed as limiting. Other changes suited for particular applications will be apparent to those skilled in the art.