Low Leakage CMOS D/a Converter

1. Field of the Invention

This invention relates to digital-to-analog converters of the CMOS type. More particularly, this invention relates to minimizing leakage current in such converters.

2. Description of the Prior Art

CMOS D/A converters have been available commercially for quite some time, and provide important advantageous features. Referring now to FIG. 1, conventional converters of the CMOS type include a thin-film R-2R ladder network 10 the shunt resistors of which are connected to N- channel CMOS switch-pairs 12A/B; 14A/B; etc. The individual switches of each pair are complementarily driven so that one switch is off while the other is on, and vice versa. The switches typically have a voltage across them of 10 mV for a 10 V reference voltage and a 10 k.OMEGA.-20 k.OMEGA. ladder network.

There are three commonly-used configurations of conventional CMOS D/A converters. In order to explain leakage current, each one of these configurations will be examined separately.

Referring first to FIG. 2, the converter shown there comprises three separate lines for I.sub.OUT1, I.sub.OUT2 and AGND (analog ground). The ladder termination resistor 16 is tied to AGND. When an all zero's code is applied to the converter, all of the current in the R-2R ladder 10 (except for the one bit of current flowing through the ladder termination resistor to AGND) is steered to the I.sub.OUT2 terminal. Ideally no current should under these conditions flow to the I.sub.OUT1 terminal. However, in practice a finite current does flow in the I.sub.OUT1 terminal and is know as the leakage current.

When an all one's code is applied to the converter of FIG. 2, all of the current in the R-2R ladder (except, again, for the one bit of current through the ladder termination resistor to AGND) is steered to the I.sub.OUT1 terminal. Any current flowing in the I.sub.OUT2 terminal is defined as leakage current.

Referring now to FIG. 3, again there are three separate lines for I.sub.OUT1, I.sub.OUT2 and AGND. The ladder termination resistor 16 now is tied to I.sub.OUT2. When an all zero's code is applied to the D/A converter all of the current in the R-2R ladder is steered to the I.sub.OUT2 terminal, and any current flowing in the I.sub.OUT1 terminal is defined as leakage current.

When an all one's code is applied to the converter of FIG. 3, all of the current in the R-2R ladder (except for the one bit of current flowing through the ladder termination resistor to I.sub.OUT2) is steered to the I.sub.OUT1 terminal. Because of this one bit of current flowing along the I.sub.OUT2 line, it is not possible to measure leakage current. Hence, leakage current is only defined along the I.sub.OUT1 line for an input code of all zeros.

Referring now to FIG. 4, there is an I.sub.OUT line and an AGND line, where AGND incorporates AGND and I.sub.OUT2 of FIGS. 2 and 3. The ladder termination resistor 16 is tied to AGND. When an all zero's code is applied to the D/A converter, all the current in the R-2R ladder is steered to the AGND terminal. Any current flowing in the I.sub.OUT line is defined as leakage current.

Leakage current, as defined above for the circuit configurations of FIGS. 2 through 4, causes errors in the D/A transfer characteristic, particularly at temperatures greater than 100.degree. C. where the effect is most pronounced. Thus it is desired to reduce such errors.

The leakage current described above actually comprise two separate components which arise for different reasons. To explain this, reference is made to FIG. 5 which shows somewhat schematically the IC structural arrangement of the switches 12A/B for a conventional CMOS DAC. The N- substrate 20 normally is at the most positive supply voltage (5 V to 15 V). The P- well 22 is at zero potential (ground). Leakage current in such a structure comprises the following two components:

(a) Subthreshold leakage current from the R-2R ladder across the "off" switch to I.sub.OUT1 or I.sub.OUT2, depending on along which line the leakage current is being measured. This is shown as I.sub.1.

(b) Collector-emitter leakage current of the substrate NPN bipolar transistor with grounded base. The transistor is formed by a reverse-biased diode from the N- substrate (collector) to the P- well (base), and a diode from the P- well (base) to the N+ source diffusion (emitter). The N+ source diffusion (emitter) referred to is either to I.sub.OUT1 or I.sub.OUT2, depending on along which line the leakage current is being measured. This is shown as I.sub.2.

This invention is directed to means and techniques for eliminating or significantly minimizing the leakage currents described above.

It has been found that if, in the FIG. 5 arrangement, the P- well 22 is biased negative (e.g. to about 200 mV) with respect to the I.sub.OUT1 and I.sub.OUT2 terminals, the bipolar transistor leakage currents (I.sub.2) flowing into the I.sub.OUT lines are virtually eliminated. The subthreshold leakage current (I.sub.1) also is simultaneously reduced slightly because the negative bias on the P- well effectively increases the threshold voltage of the N- channel switches.

It further has been found that subthreshold leakage current (I.sub.1) is eliminated if the logic low drive to the switches 12A/B is biased negative, e.g. by reducing the "off" gate voltage from 0 V to -200 mV.

FIGS. 1-4 are circuit diagrams showing conventional CMOS DA circuit configurations;

FIG. 5 illustrates schematically the arrangement of CMOS switches in a diffused integrated circuit chip;

FIG. 6 is a circuit diagram illustrating one preferred embodiment of the present invention;

FIG. 7 is a circuit diagram showing one arrangement for developing a bias voltage for the embodiment of FIG. 6;

FIG. 8 is an equivalent circuit corresponding to FIG. 7;

FIG. 9 is a schematic illustration of an integrated circuit implementation of the FIG. 7 bias voltage circuit; and

FIG. 10 is a block diagram showing another implementation providing features of the invention.

Referring now to FIG. 6, showing one preferred embodiment of the invention, the N- channel switches 12A/B and the switch drivers 30, 32 are placed in a common P- well generally indicated at 34 (e.g. corresponding to P- well 22 in FIG. 5). This common P- well is maintained at a negative potential of -200 mV, as by means of a bias voltage source 36 connected to the well terminal 38, and forming part of the same IC chip.

An exemplary circuit for the bias voltage source 36 is shown in FIG. 7. A negative supply voltage V.sub.S is applied across a resistor R.sub.3 in series with a regulating diode D.sub.1, producing a regulated voltage V.sub.D of about -700 mV. This voltage is divided by resistors R.sub.1 and R.sub.2 to provide the -200 mV bias voltage (V.sub.TUB) for the P- well.

The actual bias voltage selected will represent a compromise between too high a voltage which will increase the N+ source diffusion to P- well leakage and on-switch resistance (producing a gain error), and too low a voltage which will be insufficient both to stop the leakage current crossing the N+-P- boundary and to turn off the subthreshold leakage current from source to drain. Presant indications are that the lower limit of the bias voltage is about -150 mV. The upper limit is not so definite, but preferably is in the range of about -300 mV to -500 mV.

The value of resistors R.sub.1 and R.sub.2 should be selected so as to avoid any problems caused by transient currents (I.sub.TRANS) through the switch drivers. To illustrate the relationships involved, the following somewhat simplified analysis is presented: ##EQU1## The equivalent circuit seen looking from the P- well into the FIG. 7 circuit is shown in FIG. 8. In order that the tub potential not rise above 0 V at the time of switching, the voltage drop across the equivalent internal resistor 0.714 R.sub.X must be less than, or equal to, 0.2 V. ##EQU2## Assuming that the transient current is 10 mA - ##EQU3## The value of R.sub.3 is chosen to assure that there is always a forward biasing current through the diode D.sub.1.

A more analytical approach to calculating component values would take into account variation of diode voltage with temperature, plus resistor variations with temperature (whether thin-filmed or diffused).

FIG. 9 shows a suggested integrated circuit implementation of the circuit of FIG. 7. The P- well 40 which normally contains the digital logic (not the Well for the switch drivers 30, 32 and switches 12A/B) is enlarged to provide a ground tub extension 42 to allow room for the P--N+ diode D.sub.1 plus a diffused N+ resistor 44. The series combination of R.sub.1 and R.sub.2, 98 .OMEGA. in all, is divided in two, with 50 .OMEGA. in the digital ground tub 42, and 48 .OMEGA. in a separate P- well 46. The reason for splitting the resistors between separate tubs is to avoid any more than 0.35 V forward bias across P--N+ junctions, thereby ensuring all parasitic diodes are turned off.

As the circuit is shown, R.sub.3 is external to the package. It could also be included on-chip in the form of a thin-film resistor, its value depending on the voltage to be put on the V.sub.CONTROL pin.

It will be noted that in the FIGS. 6 and 9 embodiment, both forms of leakage (I.sub.1 and I.sub.2) are eliminated simultaneously, but the effect of transient currents in the switch drivers must be taken into account. To avoid such transient current effect, the N- channel switches 12A/B, etc., may alternatively have their thresholds increased by ion-implantation, with only the back-gates of the switches being biased negative (as by -200 mV). The P- well containing the N- channel switches would in such a modified arrangement be separated from that containing the switch drivers 30, 32 which can be placed in the grounded P- well 42 used by the digital logic. The bipolar transistor leakage current is as before eliminated by the -200 mV bias of the back gates. The subthreshold leakage current is eliminated by the ion-implanting of the N- channel switches.

The circuit of FIG. 7 can be used for generating the -200 mV bias for the P- well containing the N- channel switches. In this case, the equivalent resistance seen looking from the P- well is not critical since there will be no current flowing from the well. The biasing circuit can easily be integrated onto the D/A converter. Resistors R.sub.1, R.sub.2 and R.sub.3 could be thin-film and, as before, an external control voltage is required to bias this circuit into operation.

Still another embodiment of the invention is shown in FIG. 10. This arrangement includes a conventional CMOS DAC 50 with its I.sub.OUT1 and I.sub.OUT2 both offset in a positive direction with respect to DGND and AGND (zero volts). Illustratively, such offset is produced by connecting a positive bias voltage source 52 (e.g. 200 mV) to the positive terminal of the output op-amp 54. The output of the op amp is connected to the feedback resistor R.sub.F/B terminal which, in conventional DAC configurations, leads to an on-board feedback resistor in the DAC structure (as in FIG. 6).

The FIG. 10 configuration eliminates bipolar transistor leakage current, and somewhat reduces subthreshold leakage current as a result of the effective increase in N- channel threshold voltage. The ladder termination resistor (not shown in FIG. 10) must be connected to I.sub.OUT2. Connecting the ladder termination resistor to AGND would cause D/A converter non-linearity.

Although several preferred embodiments of the invention have been disclosed herein in detail, it is to be understood that this is for the purpose of illustrating the invention and should not be construed as necessarily limiting the scope of the invention, since it is apparent that many changes can be made by those skilled in the art while still practicing the invention claimed herein.