Digital to Analog Converter Linearity With Mismatched Current Sources

1. Field of the Invention

The invention relates to the tolerance differences between current sources in an a digital to analog converter employing the use of R-2R ladder networks. These current differences affect the linearity of the output of this type of converter. This invention demonstrates how to reduce the differences between current generators without adjustments while dramatically improving the linearity performance of digital to analog converters of the type using R-2R ladder networks.

2. Description of Related Art

The linearity performance of digital to analog converters (DAC's) are greatly affected by variations in circuit items which are not common to the total signal flow. These circuit items include the current sources, ladder network resistors and the different ground reference points between the ladder network and the output amplifier. One way to obtain the desired linearity is to adjust the current sources to be identical, trim the similar resistors to have the same value and make the ground or reference leg of the ladder network be the same as the current summing operational amplifier. Whereas, this will produce the desired results it is expensive.

Referring to U.S. Pat. No. 5,036,322 (Barrow et al.), the variation between current sources has been reduced by developing the knowledge of the variability of the manufactured current sources over a statistically significant product sample. Then apply this knowledge to constructing a desired accurate current source from the summation of several smaller and less accurate current sources. In this way producing a final current source of larger magnitude and having a smaller variability than the individual current sources that make up the final current source. In today's world of integrated circuits the tracking of variations on a single device may need to be taken into consideration. Tight tracking between like elements of a semiconductor device is desirable, but it can provide a source of yield loss when the particular devices line up near the acceptable limits and all have closely the same value. The mean value in this case would be off from the expected nominal value predicted from variations across several devices.

When the ground reference between the ladder network and the current summing amplifier are different, the voltages at the nodes can affect the currents flowing in the ladder network. Thus causing an inaccuracy in the final voltage of the digital to analog converter. Making the two references electrically the same is difficult because of lead resistance and variations in ground reference current from DAC input signal to input signal. Referring to U.S. Pat. No. 4,034,366 (Memishian), one method used to correct this problem is to sense the R-2R ladder network ground reference, compare it to the reference of the current summing amplifier and adjust the ground reference of the ladder network to be the same for various current generator inputs.

In a digital to analog converter using a R-2R ladder network the mismatch of current sources is one of the causes of non-linearity of the analog signal output. Matching the number of current sources in such a converter can be more difficult than matching the resistors in the ladder network. This invention uses a novel approach to improving current matching from a set of unmatched current sources. The basic idea is to connect every current source to each node of the ladder network during the conversion to an analog signal. In this way all nodes will see the same current depending on the corresponding bit value (1or 0) of the digital input. In this way the currents can be averaged together and all nodes corresponding to the same bit value will receive the same current. The averaging is accomplished by closing each switch connecting the current sources to a different node of the R-2R network for 1/nth of the conversion cycle. Where "n" is the number of bits in the digital input to the DAC. For every input bit equaling a logical "1" the current into the corresponding node of the R-2R ladder network will be ##EQU1## otherwise the current is zero. A sequencer is used to cycle the connection of the current sources to the various nodes of the ladder network, and a low pass filter network built with an operational amplifier averages the currents to provide a smoothed output. In this way the invention provides a means to feed each node of a R-2R ladder network with the same current and producing improved output linearity. Because there is a finite time in which to produce a valid output of the DAC, the minimum time to complete a conversion can be longer than other techniques. However, the other techniques may be more difficult and costly to produce the same results.

This invention will be described with reference to the accompanying drawings, wherein:

FIG. 1 is a schematic of a digital to analog converter according to the invention, and

FIG. 2 shows the sequenced timing of the of the conversion cycle according to the invention, and

FIG. 3 is a diagrammatic example of the sequence of current sources connected to each of the ladder network nodes according to the invention.

Referring to FIG. 1, a DAC is shown in accordance with the invention. There is a set of current sources 1 having a number of independent sources typified by 2, varying in quantity depending on the resolution of the DAC, and equal to the number of binary bits in the digital data input 3. Digital data input 3 is connected to control block 4 which represents a collection of logic circuits that are to control and sequence a set of switches 6 through the block output 5. The set of switches 6 connects the set of current sources 1 to the nodes typified by 7 of an R-2R ladder network 8. The set of switches 6 is arranged in row and columns. There is one row for every independent current source typified by 2. Similarly there is one column of switches for every ladder node typified by 7. The switches connect the rows to the columns, and therefore, connect the set of current sources 1 to the nodes of the ladder network 8. Typically there are 8 bits in the binary input 3 but not restricted to this number. Thus there are typically eight independent current sources 2 and eight network nodes 7, which can be greater or less in number depending on the desired resolution of the DAC. The outputs 9 and 10 of the ladder network 8 are connected to the inputs nodes 11 and 12 of a current summing operational amplifier 13 and low pass filter 15. The output 14 of the amplifier is also the output of the DAC and is fed back through a resistor and capacitor of the low pass filter 15. As known to those schooled in the art of operational amplifier design, the feedback maintains the two input nodes 11 and 12 at equal potential and produces a voltage at output 14 which is proportional to the current flowing into the amplifier input node 11.

Referring to FIG. 2, the sequencing and control circuitry contained within block 4 divides the conversion cycle 16 into a number of time slots typified by 17. The number of time slots are equal to the number of current sources typified by 2 in FIG. 1. During each time slot a different current source typified by 2 is connected through the set of switches 6 to different ladder network nodes typified by 7. This is done such that there is an exclusive connection of only one current source typified by 2 connected to only one node typified by 7 at the same time. When the last time slot typified by 17 in the conversion cycle 16 is completed, each current source typified by 2 will have been connected once to each ladder network node typified by 7. In addition the current connected to each node typified by 7 during each time slot will be further controlled to be on or off depending on the value of the corresponding digital bit of the digital input 3. This results in a current flowing between the ladder network output 9 and the operational amplifier input 11 that is a function of the average value of all of the current sources contained within 1 and proportional to the binary value of the digital input 3.

To further aid in the understanding of this invention, shown in FIG. 3 are equivalent circuit diagrams for each of the three time slots needed for a three binary bit DAC input 3. The required three ladder network nodes 18, 19, and 20 are connected to each of the three current sources in a different time slot as shown in FIG. 3a, FIG. 3b and FIG. 3c. The current connected to node 18 for the total conversion cycle 16 is the sum of 21, 22 and 23 divided by three. The adjustment by dividing by three is because each time slot is valid for one third of the conversion cycle and each current 21, 22 and 23 are only applied to each node for one third of the time. Similarly the currents for nodes 19 and 20 are also the sum of 21, 22 and 23 divided by 3. Thus identical currents are applied to each node improving the linearity of the converter without having to adjust each individual current source 21, 22 and 23.

While the invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention.