AnsweredAssumed Answered

Bimodal error distribution at the AD8310 output?

Question asked by Snorlax on Aug 1, 2016
Latest reply on Aug 16, 2016 by Snorlax

The AD8310 datasheet provides comprehensive info on the device static errors. Are there any publications regarding the device dynamic errors?

Below is the description of the dynamic errors observed with the device. Any insight as to what is causing this behavior and any remedies possible would be highly appreciated.

The Setup: test circuit is as per Figure 40 on p. 20 of the datasheet, except there is no R3, C3, C7; C1, C2 and C6 are 0.1uF, R6 is 10K and W1 is closed. The circuit is powered by a clean VPOS=+3.3V. The input signal comes from a constant amplitude 3MHz sine wave AC voltage source with a 1K resistor between the source and C2 input. There is an additional 75K resistor connected across C6. The voltage across C6 is digitized by an ADC with ~5M input impedance every 0.1 sec. The typical trace of the output voltage is shown on the attached chart in both ADC counts and Volts (on the right).

The Behavior: The AD8310 output voltage predominantly stays at a certain constant level, but occasionally drops by, in this case, 4 ADC counts, stays at this lower level for a while, then comes back to the original higher level (please see Fig. 1 below). The difference between the two levels is somewhat proportional to the AD8310 output voltage: at the output level of, say 17000 ADC counts, the drop is about 2 counts. This effect is observed with input frequencies between 3 and 5MHz. It occurs without any discernible periodicity, the duration varies from 0.1s up to 2.6s. When such output trace is converted into a histogram, there are two distinct peaks, especially striking in log scale (please see Fig. 2 below).

ADC itself does not have this issue when measuring external test sources of DC voltage. This behavior could certainly be attributed to the AC source, but it occurs with two quite different AC sources. Could it be some workings of the internal offset-nulling circuitry? Something else?


Fig. 1

Fig. 2