QWhen applied signal input crosses the half-full scale output code,
either in the positive direction or negative direction, the ADC
output can erroneously output one or more full-scale output codes
($3FF). Wondered if you have seen this before. Do you know what
causes it? Do you know how it can be cured?
AIn the AD9050 the analog inputs are internally biased to 3.3V, therefore
applying an AC signal with a pp value of 1.024 V drives the ADC from zero to
full-scale on the o/p, this range (2.788 to 3.812) is the nominal input range
of the AD9050. If the input level is outside of this range, internal
comparators shut off the T/H, and the digital values are locked at their max or
min value...refer to page 9 (Overdrive of the analog input).
The input is protected internally to one volt outside of the power supply
rails... i.e. for nominal power (+5V and gnd), the analog input will not be
damaged with signals from +6V to -1.0V).
The common mode level on the input should be the same as the internal common
mode /internal biasing of the AD9050, which in this case is 3.3V, refer to the
applications circuits, where the AD9050 can either de AC coupled, or
alternatively DC coupled using the internal bias to level shift the input to
the level required by the AD9050 (3.3V), see fig 16.
Most of our adc inputs that self bias have some small range around the self
bias point that the part can safely operate in. typically this is on the order
of 200mV or so.
For the 9050 this implies that you can come in with a 1Vpp single-ended signal
centered at 3.1 to 3.5 V (the internal bias is 3.3V).. (we do not extensively
characterize this however )
Note that if you are driving it differentially the common mode range variation
is increased further as the signal swings are smaller.