M68HC11
REFERENCE MANUAL
ANALOG-TO-DIGITAL CONVERTER SYSTEM
MOTOROLA
12-19
to grossly different analog levels.
A subtle aspect of
Figure 12-5
is that the DAC capacitance is shared by all conver-
sions. This aspect results in the initial voltage on the DAC capacitance, just before a
sample period, being approximately equal to the voltage on the last channel convert-
ed. For individual conversions, this result usually does not produce any observable ef-
fect because the charge stored in the internal DAC capacitance is so small. Consider
what happens when a multiple-channel, continuous-scan conversion operation is oc-
curring (MULT and SCAN = 1). For an E-clock rate of 2 MHz, a particular channel is
sampled once every 64
μ
s. Each time, the initial voltage on the DAC capacitance just
before the sample is equal to the voltage on the previously sampled channel. Over
15,000 times a second a small amount of charge is removed from or added to the ex-
ternal capacitance on the pin. The charge is restored by charging or discharging
through the external source impedance. For some values of external R and C, the
charge added during the sample time cannot be fully bled off through the external RC
before the next sample time occurs. This problem causes a stair-step building of
charge in the external capacitance that builds until an equilibrium is reached, in which
the amount of charge added during a sample time is exactly offset by the charge bled
off during the period between samples. This condition is a secondary effect, which sel-
dom results in more than an LSB of error (even in the most extreme case). After study-
ing the mechanism, it should be fairly easy to avoid problems from this effect by careful
choice of the external R and C values, by avoiding channel assignments resulting in
grossly different levels on adjacent critical channels, and/or by avoiding the multiple-
channel, continuous-scan conversion modes when a high-frequency E clock is being
used.
Three types of cases must be considered, which result from the interaction of an ex-
ternal RC filter and the internal model of an A/D input pin. The method of determining
the error expected from a particular choice of external component values depends on
which of the three cases applies. Errors arise from leakage current acting through the
external series resistance or from system noise. If a very large external series imped-
ance is used, a problem can arise where the internal DAC capacitance cannot be prop-
erly charged within the 12-cycle sample time; however, errors due to simple leakage
through the external resistance usually do not allow using a large enough external re-
sistance to cause this effect. The following paragraphs describe the cases of interac-
tion of an external filter to the input model.
The first case arises when the external time constant is small compared to the length
of the 12-cycle sample period. In this case, all residual charge on the internal DAC ca-
pacitance is dissipated, and the pin settles at the expected voltage before the end of
the sample time. The problem with this case is that a filter with such a short time con-
stant provides very little filtering. In this case no errors result; thus, no calculations are
needed.
The second case arises when the external time constant is long compared to the sam-
ple period but is relatively short compared to the period between samples. In this case,
the residual charge on the internal DAC capacitance is redistributed to the external ca-
pacitance during the sample, but this charge is not dissipated through the external re-