When considering an analog-to-digital converter (ADC),
most electronic engineers think of applying a voltage to the input, sampling, and converting the sampled
voltage to a digital value. However, there are a wide variety of ADCs and a wide variety of input methods.
There are ADCs that accept a voltage input, others that accept a current input, and still others that accept
a variety of physical inputs (essentially, a transducer or sensor coupled to an ADC).
For those ADCs that accept a voltage input, there are three general methods of providing the voltage to
the ADC: single-ended, pseudo-differential, and differential. Within each of these methods, there is a
great deal of variability in regards to the limits imposed on the voltage range that the ADC will accept
and the performance of the source providing the voltage.
Single-Ended Input
The most common type of input is a single-ended input. Essentially, during the acquisition time (also called
the sampling period), the ADC captures a sample of the analog voltage at its input on an internal sampling
capacitor (see Figure 1). This voltage is referenced to the converter's analog ground. Some ADCs offer a
specific ground pin marked "analog ground" or AGND, or they may have one or more ground pins that are simply
marked GND. In the latter case, the ADC's analog and digital ground are probably tied together on the die
itself or are connected to the same package pad with two wire bonds.

Figure 1 - Basic Single-Ended Input
on a Typical ADC
The on resistance of the sampling switch (RON) and the impedance of the source driving the ADC (not
shown) work with the ADC's sampling capacitor (CSAMPLE) to form a low-pass filter. The time constant
of this filter and the resolution of the ADC set the required acquisition time as well as the input bandwidth.
While some situations are more complex, this simple analysis works as a good approximation most of the time.
This situation also provides an opportunity to discuss a simple point regarding source impedance: when it is
high, simply give the ADC more time to acquire the voltage. This works as long as the leakage current within the
ADC is low enough, relative to the source impedance, to be safely ignored.
A single-ended connection to an ADC assumes that the ultimate source of the analog signal is referenced to the
same ground as the ADC. If the difference in ground potential between the source and the ADC is significant
(larger than 1 least significant bit [LSB] weight of the ADC's digital output), then a single-ended connection
is not a good idea. The single-ended approach works best when both the source and the ADC are on the same
printed circuit board.
Keep in mind that the difference in potential between two widely separated ground reference points may not be
a static difference. More likely, the difference will be dynamic and the rate of change will extend into the
very high frequencies.
Pseudo-Differential Input
A pseudo-differential input is something of a hybrid between a single-ended input and a fully differential input.
This somewhat dubious compromise is further complicated by the ways in which different manufacturers have decided
to provide pseudo-differential inputs. Figure 2 shows two possible ways that a pseudo-differential input can be
implemented. Of the two, Method B is preferable. In reality, I do not believe that any converter actually uses
either methodeach is a simplification for discussion purposes only.

Figure 2 - Two Different Types of Pseudo-Differential ADC Inputs
The idea behind the pseudo-differential input is to provide not only a connection for the input, but also a
connection for the ground reference. The hope is that this offers better performance when the source of the signal
is not referenced to the same ground as the ADC, which might be the case for a small system comprised of multiple
boards or assemblies.
Method A is the simplest scheme using just a single switch. The sampling capacitor obtains a charge that is
proportional to the difference in voltage between +IN and -IN. Once the charge is captured, the +IN input is
disconnected. The -IN input remains as a ground reference.
For some converters, this approach is implemented in a scheme that is more complicated than shown in Figure 2, but
still uses just one switch. The +IN voltage is captured relative to the local ground potential and then -IN is
connected in order to subtract the difference between the local ground potential and the remote ground potential.
The major weakness with this implementation is that the remote ground potential must be stable with regard to the
local ground potential throughout the conversion. Any significant change between the sampling of +IN and the end of
the conversion will result in an error in the digitized result. While this scheme is "differential," I would argue
that it just barely qualifies as such.
I believe the design labeled as Method B is now more common than Method A. This circuit requires an additional
switch with a very low on resistance, and this increases the die area and the converter cost (note that in Figure 2
(Method B), RON represents the combined on resistance of both switches). However, the benefits are
substantial. For this configuration, the voltage difference between +IN and -IN is captured when the converter goes
into the hold mode. During the conversion, both +IN and -IN are "free" to vary. In addition, a good design will
reject even a somewhat high-frequency common-mode voltage or noise that might have been coupled onto the signal on
its way from the source to the ADC.
Next
>>