Philips Semiconductors Linear Products

Product specification

AM6012

12-Bit multiplying D/A converter

August 31, 1994

779

CIRCUIT DESCRIPTION

The AM6012 is a 12-bit DAC which uses diffused resistors and

requires no trimming to guarantee monotonicity over the

temperature range. A segmented DAC design guarantees a more

uniform step size over the temperature range than is normally

available with trimmed 12-bit converters. The converter features

differential high compliance current outputs, wide supply range, and

a multiplying reference input.

In many converter applications, uniform step size is more important

than conformance to an ideal straight line. Many 12-bit converters

are used for high resolution rather than high linearity, since few

transducers are more linear than

±0.1%. All classic binarily weighted

converters require

±1/2LSB (±0.012%) linearity in order to guarantee

monotonicity, which requires very tight resistor matching and

tracking. The AM6012 uses conventional bipolar processing to

achieve high differential linearity and monotonicity without requiring

correspondingly high linearity, or conformance to an ideal straight

line.

One design approach which provides monotonicity without requiring

high linearity is the MOS switch-resistor string. This circuit is actually

a full complement to a current-switched R-2R DAC since it is slower,

has a voltage output, and, if implemented at the 12-bit level, would

use 4096 low tolerance resistors rather than a minimum number of

high tolerance resistors as in the R-2R network. Its lack of speed

and density for 12 bits are its drawbacks.

With the segmented DAC approach, the 4096 required output levels

are composed of 8 groups of 512 steps each. Each step group is

generated by a 9-bit DAC, and each of the segment slopes is

determined by one of 8 equal current sources. The resistors which

determine monotonicity are in the 9-bit DAC. The major carry of the

9-bit DAC is repeated in each of the 8 segments, and requires eight

times lower initial resistor accuracy and tracking to maintain a given

differential nonlinearity over temperature.

The operation of the segmented DAC may be visualized by

divided into 512 levels by the 9-bit multiplying DAC and fed to the

selected for each 512 counts. The previous segment is fed to output

monotonicity independent of segment resistor values. All higher

With the segmented DAC approach, the precision of the 8 main

resistors determines linearity only. The influence of each of these

resistors on linearity is four times lower than that of the MSB resistor

in an R-2R DAC. Hence, assuming the same resistor tolerances for

both, the linearity of the segmented approach would actually be

higher than that of an R-2R design.

The step generator or 9-bit DAC is composed of a master and a

slave ladder. The slave ladder generates the four least significant

bits from the remainder of the master ladder by active current

splitting utilizing scaled emitters. This saves ladder resistors and

greatly reduces the range of emitter scaling required in the 9-bit

DAC. All current switches in the step generator are high-speed

fully-differential switches which are capable of switching low currents

at high speed. This allows the use of a binary scaled network all the

way to the least significant bit which saves power and simplifies the

circuitry.

Diffused resistors have advantages over thin film resistors beyond

simple economy and bipolar process compatibility. The resistors are

fabricated in single crystal rather than amorphous material which

gives them better long term stability and tracking and much higher

moisture resistance. They are diffused at 1000

°C and so are

resistant to changes in value due to thermal and chemical causes.

Also, no burn-in is required for stability. The contact resistance

between aluminum and silicon is more predictable than between

aluminum and an amorphous thin film, and no sandwich metals are

required to enhance or protect the contact or limit alloying. The initial

match between two diffused resistors is similar to that of thin film

since both are defined by photomasks and chemical etching. Since

the resistors are not trimmed or altered after fabrication, their

tracking and long-term characteristics are not degraded.

DIFFERENTIAL VS INTEGRAL NONLINEARITY

Integral nonlinearity, for the purposes of the discussion, refers to the

“straightness” of the line drawn through the individual response

points of a data converter. Differential nonlinearity, on the other

hand, refers to the deviation of the spacing of the adjacent points

from a 1 LSB ideal spacing. Both may be expressed as either a

percentage of full-scale output or as fractional LSBs or both. The

graphs in Figure 1 define the manner in which these parameters are

specified. The left graph shows a portion of the transfer curve of a

DAC with 1/2LSB INL and the (implied) DNL spec of 1 LSB. Below

this is a graphic representation of the way this would appear on a

CRT screen where the AM6012 is used as a display driver. On the

right is a portion of the transfer curve of a DAC specified for 1/2LSB

INL with LSB DNL specified and the graphic display below it.

One of the characteristics of an R-2R DAC in standard form is that

any transition which causes a zero LSB change (i.e., the same

output for two different codes) will exhibit the same output each time

that transition occurs. The same holds true for transitions causing a

2 LSB change. These two problem transitions are allowable for the

standard definition of monotonicity and also allow the device to be

specified very tightly for INL. The major problem arising from this

error type is in A/D converter implementations. Inputs producing the

same output are now represented by ambiguous output codes for an

identical input. Also, two LSB gaps can cause large errors at those

input levels (assuming 1/2LSB quantizing levels). It can be seen

from the two figures that the DNL-specified D/A converter will yield

much finer grained data than the INL-specified part, thus improving

the ability of the A/D to resolve changes in the analog input.