DAC8760 accuracy questions

Hello Arnold,

Welcome to the E2E Community. Thanks for your interest in the DAC8760 family of devices.

Arnold Gordon said:

1) The spec sheet description area for both the DAC7760 and DAC8760 shows "±0.1% FSR Total Unadjusted Error (TUE) Max". But within the spec sheet it states "Total unadjusted error, TUE TA = –40°C to +85°C –0.06 +0.06 %FSRTA = +25°C –0.04 ±0.015 +0.04 %FSR. Which of these is correct?

The front-page of the datasheet does a little bit of marketing for the device. It is specifying that all output spans realize TUE within 0.1% FSR at room temperature. The electrical characteristics table gives specifications for the TUE of each output over different spans of temperature such that the effects of drift are also included.

One additional comment - these are measured specifications, not calculations based on the cumulative errors specifications for the device as discussed here.

Arnold Gordon said:

2) If that is the UNadjusted error, how can we adjust that error?

The goal of a TUE specification is to express how accurate the device is on it's own with no external influences. If  you want to adjust the error you can implement analog or digital calibration that corrects for offset error and gain error. Typically linearity error is impractical to calibrate because it is very cumbersome for the calibration solution and very small relative to the other errors.

Arnold Gordon said:

3) The spec sheet shows: "Relative accuracy, INL A = –40°C to +85°C ±0.022 %FSR" Is it possible to adjust the accuracy so we can achieve the relative accuracy spec?

Relative accuracy is synonymous with Integral Non-Linearity or INL, as is noted in the electrical characteristics table (all the DAC DC errors are briefly discussed here). As I mentioned in the previous point, calibrating linearity error is usually impractical so calibration solutions frequently just calibrate offset and gain error. If the system is calibrated (under ideal or perfect calibration conditions) the only remaining error is the linearity error or relative accuracy.

Arnold Gordon said:

4) The differential non-linearity spec shows as +/- 1 lsb for both the DAC7760 and DAC8760. That implies that the DAC8760 has a non-linearity spec which is 16 times better than the DAC7760. That should mean that the TUE might also be lower for the 8760. Can you respond to this? 

Most modern DNL specifications are, at best, 1 LSB to express that the design is fully monotonic and there are no missing codes.

DNL is not used in a TUE calculation since it only applies to the performance between two adjacent codes which isn't very helpful to estimate the performance of the entire transfer function. INL is the integration of all DNL errors and is more relevant for estimating TUE since it expresses the cumulative effects of DNL.

Given that DNL within +/-1LSB expresses monotonicity and no missing codes there isn't any real motivation to specify anything tighter.

Please let me know if anything is unclear.