I'm wondering how to select a reference IC when I design ADC sometimes.For expample, looking at AD7918, 10bit-ADC external referene and the gain error itself is 0.5LSB MAX.Reference accuracy is said to be less than 1% according to the comment of datasheet.
Ive attached a file on this mail.Please refere to it.But I'm wondering that the accuracy of reference is 1% sounds larger.Because the resolution of 10bit is equivalent to 0.1% and the gain error of this device is only 0.5LSB(0.05%).Please give me advice on the coneection between the gain error and the accuracy of reference voltage.(In this case, reference noise and tempareture coefficient are not considered to make it simple)And one more thing, I want to know how to estimate source current of ADC refinput to choose right reference IC.
The AD7918 device can be operated with 2.5 V +/-1% reference only. Such restriction is driven by the internal comparator in the part which starts to operate outisde its normal operating region when the reference is not withing datasheet limits. This situation can result in incorrect or inconsistent conversion results.
It should be noted that the reference variation doesn't influence the Gain Error specification as the reference voltage is measured and calibrated out during this test. Therefore the Gain Error specification is error of the AD7918 itself and not the reference source. Please be however aware that Offset Error, Gain Error, Linearity Error and Reference Error (absolute error, tempco) defines absolute error of the converter.
In terms of chosing the reference source for ADC. The AD7918 datasheet indicates suitable reference sources for this device as well as the REFin input impedance therefore it is relatively easy to calculate required current. At 1 MSPS the REFin input impedance is equal to 36 kOhm therefore reference input current is in tens of uA region.
Hi Jakub,Thank you for your advice.
Let me ask a little bit more about it.First, you said the Gain error specification is not infuluenced by the reference variation and mesured by test specification.So how much accurcy of Vref is used during the test?And sencondly, in AD7918, if we use +/-1% refrence that datasheet mentions, can we acheve the DC statics peformance such as gain error:+/-0.5LSB ?I think acutual gain error could be +/-10.5LSB if we use +/-1% refrence ( vref initail error and ADC gain error from the fomula of Code=Vin*2^n/Vfef)
Is my perspective correct?I'm not sure where this +/-1% refrence come from and what exactly happen if we use less accuracy of Vref (more than 1% accuracy).
Could you give me advice?
During the Gain Error test, tester is connected to the reference and measure its value. Using this value of reference, the ideal first and last transitions are calculated and compared with the test result therefre reference tolerance/variance/tempco cannot influence Gain specifications simply because tester takes into account actual reference value. As I said test specification focus on the Gain Error figure of the converter and the reference voltage error needs to be calibrated out. Of course reference tollerance affects DC accuracy of the converter however the Gain Error is not. If you like to calculate total unajusted error of the converter (TUE) you should take into account, Offset, Gain, INL, Reference tolerance and Reference tempco. Combination of these parameters defines absolute error of the converter.
This part is specified to operate with a 2.5 V +/-1 % external reference and the AC/DC specifications on the datasheet are determined using this value. If you operate the AD7918 with reference other than 2.5 V you may observe incorrect conversion resutls. The most critical situation is with the reference voltage higher than 2.5 V. The reason behind this is that the comparator input will go outside of the rails for certain values of the input voltage and this will cause errors in the DAC due to lost charge when diodes turn on in the usual way. Therefore it is not recommended to oparete this device with the reference voltage outside the specification.
Hi Jakub,Thank you for your quick reply.I understand what you mean.I really appreciate it!