Chapter 2: Autocalibration and compensation techniques 19
they are not applicable directly because the system must be disconnected from
the signal path during offset and noise sampling. This problem can be over-
come by using the ping-pong technique.
The principle is to duplicate the signal processing circuit, in this case an
amplifier. One of them is calibrated, while the other is amplifying the input
signal [11]. When calibration is done, the role of both amplifiers is reversed:
The just calibrated amplifier is used for signal amplification, and the second is
calibrated. The same procedure goes on and on, hence the name “ping-pong”.
Figure 15 presents an improved ping-pong technique [12] that reduces the
glitches during the transition from one amplifier to another. On the left,
amplifier A amplifies the signal and A’ is under calibration. During the transi-
tion phase in the middle, amplifier A’ copies the amplified output of A. This
prevents the system from generating an important transient spike as if A‘
directly replaces A. On the right, the roles of both amplifiers are exchanged:
A’ processes the signal and A is calibrated.
Sl. a p>
p- [oe bl.
Figure 15, Ping-pong amplifier system
Left: A* calibration; Middle: Transition; Right: A calibration
Even if the transition between both amplifiers is done cautiously, there
still remain two undesirable effects as presented in figure 16 [13]. First, the
amplitude of the spike at the instant of swapping towap is reduced by using an
intermediate phase, but not removed completely. Second, since the offset
voltages Voie, 4 aNd Vorger 4’ of both amplifiers A and A’ are not exactly
equal when the amplifiers are swapped, the output signal is slightly shifted
after the transition by:
A Voorvet. a‘( swap) Vortset, ACswap) @N)
The amplitude of the step A can be reduced by increasing the precision of
the offset correction. But if the time interval between two successive swaps is
long, the offset of the active amplifier can drift, or the low-frequency noise