SBAA534 March 2022 ADC128S102-SEP , ADC128S102QML-SP , ADS1278-SP , ADS1282-SP , LF411QML-SP , LM101AQML-SP , LM111QML-SP , LM119QML-SP , LM124-SP , LM124AQML-SP , LM136A-2.5QML-SP , LM139-SP , LM139AQML-SP , LM148JAN-SP , LM158QML-SP , LM185-1.2QML-SP , LM185-2.5QML-SP , LM193QML-SP , LM4050QML-SP , LM6172QML-SP , LM7171QML-SP , LMH5401-SP , LMH5485-SEP , LMH5485-SP , LMH6628QML-SP , LMH6702QML-SP , LMH6715QML-SP , LMP2012QML-SP , LMP7704-SP , OPA4277-SP , OPA4H014-SEP , OPA4H199-SEP , THS4304-SP , THS4511-SP , THS4513-SP , TL1431-DIE , TL1431-SP , TLC2201-SP , TLV1704-SEP , TLV4H290-SEP , TLV4H390-SEP
The final analysis verifies if the settling time at the ADC input is short enough. For each sampling, the ADC must charge its internal sample and hold capacitor. During the setup time, the charge is transferred over from the external buffer capacitor. After the setup time, the ADC driver must be strong enough to recharge the buffer capacitor on time.
The worst-case condition for the settling time is when the signal input is set to 10 V. This will generate the max Vdiff output of –2.5 V or –2.499541 V, to be precise. Figure 2-13 illustrates using the DC Analysis option in the TINA-TI simulator to find the exact voltage that the 10-V input voltage generates at the ADC input.
The ADS1278 simulation model provides the output of the sample-and-hold capacitor. If the voltage settles correctly it is exactly at –2.499541 V. Figure 2-14 shows this voltage is applied to this pin in the simulation via an extra voltage source, plus a voltage meter to measure the error from that expected voltage.
Figure 2-15 illustrates how the simulation is activated by selecting the TC transfer option from the analysis window.
Figure 2-16 shows a zoomed in view of the graph of the error voltage after the settling of –4.153 μV. This error appears as a gain error, and reduces to approximately 0 V at 0-V input. In relation to the full excitation of –2.5 V, these –4.153 μV translate into a gain error of 0.0001663% or –1.663 ppm. This represents roughly a 20-bit resolution, hence the settling error is small enough to meet the original design goal of > 16 ENOB.