DAC7513: DAC total unadjusted error

Katlynne

Ok thankyou. I have a further question; with any DAC the offset error, INL and DNL are usually fixed although they may vary with temperature. However the gain error is always specified as % of FSR e.g. 0.15 % FSR. Therefore, if you had an input to a DAC of 1/2 FSR then the error would be 0.075 % and this should be input to the TUE formula. Interestingly for ADC the errors are fixed i.e. offset, gain, INL and DNL are fixed,

Br

Edba

PS. by fixed I mean it is not dependent on the input level to the DAC or ADC

Hi Edba,

Edward Banner said:

However the gain error is always specified as % of FSR e.g. 0.15 % FSR. Therefore, if you had an input to a DAC of 1/2 FSR then the error would be 0.075 %

It's true that the mid-code error will be half as much of the full-code error, but dividing down the percentage itself causes some confusion in my opinion. Just apply the percentage to the voltage you want to know the error at. If we have a 5V FSR we'd see 7.5mV of error at full-code (0.15% of 5V), and 3.75mV at mid-code (0.15% of 2.5V).

ADCs have the same behavior. Figure 3 in section 2.2 of the article you shared shows this. The error in terms of V is higher at higher codes and lower at lower codes, not fixed. Gain error is an error in the slope of the transfer function. That section also shows how to convert between the units. You should use the voltage or code at which you want to know the error for the TUE calculation. 

%FSR to LSB

Error (LSBs) = Code * (%FSR/100)

%FSR to V

Error (Vs) = Voltage *(%FSR/100) 

Best,

Katlynne Jones

Katlynne

Q1. Offset error, INL and DNL if they are converted to % FSR do they scale with the voltage you want to know the error?

Q2. If you had an  ADC immediately followed by a DAC where they both had a TUE of 0.15%, then ignoring the conversion to digital and back to analogue, the ADC could be said to have a gain of 0 dB +/- 0.013 dB and likewise for the DAC, so that on a block diagram the ADC/DAC could be reduced to an amplifier with gain 0 dB +/- 0.026 dB?

Br

Edba 

Q3. Could the error due to the VREF be included in the TUE calculation?

Hi Edba,

Q1: Depends how you want to look at it. Each code in the transfer function has the potential to have up to x LSBs of INL error and x LSBs of DNL error. At a smaller voltage/code 1LSB is going to be a larger percent of the total error at that code. In that sense the percentage would scale. 

Q2: The TUE of the ADC and DAC would be uncorrelated errors so you should also use an RMS sum to add those as well. And then you could convert to dB. 

Q3: Yes, you can add initial reference error into the RMS sum for TUE. 

Best,

Katlynne Jones

Katlynne

What about negative values and TUE; say you have a gain error of +/-0.04 % and Vref error of +/-0.2 %, would it simply be for positive case √0.04^2 + 0.2^2 and for the negative case √-0.04^2 + 0.2^2,

Br

Edba

Sorry that should be:

positive case +√0.04^2 + 0.2^2 and for the negative case -√0.04^2 + 0.2^2,

Br

Edba

Hi Edba,

Yes, that's right. This is a pretty conservative estimate though. In a real DAC some errors may be positive while others are negative. So, you can consider your calculations as worst case if all errors happen to be positive, or all happen to be negative. 

There is a TUE plot in this datasheet (at room temp only, so a typical plot) that will more accurately reflect how these errors typically combine together:

Best,

Katlynne Jones