Techniques of EEG Recording and Preprocessing - Quantitative EEG Analysis Methods and Clinical Applications - page 33

This latter characteristic, the amplifier's ability to suppress signal components that
are commonly present at all input terminals, is called the common mode rejection
ratio (CMRR) and is given in decibels (at a given frequency), with higher decibel
values representing better noise suppression.
Another method by which modern amplifiers help to achieve optimal SNR is
called active shielding , which is as simple as effective. The active shielding circuit
consists of a special electrode cable with a shield mesh wrapped around the inner
core (the EEG lead). The mains noise (and other ambient noises) is capacitively cou-
pled onto the electrode lead due to the lead's relatively high impedance. However,
the output of the amplifier carries the same signal in an amplified and low-imped-
ance variant. This output can then simply be fed onto the shield mesh and effectively
insulates the lead core with the EEG signal from the ambient noise. If this method is
extended to feeding back the average of all amplifier outputs, then the shield is
being driven actively with the common mode signal, which is an efficient mecha-
nism for achieving better SNR from the signals measured.

2.3 EEG Recording and Artifact Removal Techniques

2.3.1 EEG Recording Techniques
The amplifier parameters chosen to record the EEG have a large impact on the qual-
ity of the data derived. Central acquisition parameters are the sample rate, the gain
(vertical resolution), the highpass and lowpass filter characteristics, and the notch
filter that can be used to eliminate residual mains noise. All of these parameters have
to be set with respect to the signals to be derived from the recordings and with
respect to the demands of the experimental paradigm at hand.

Sample Rate
According to the sampling theorem, the sample rate should be at least twice as high
as the highest frequency of interest contained in the signal. However, the question is
how this rule relates to the event-related EEG signal of interest. A good rule of
thumb for ERPs is to consider the temporal extent of the shortest ERP component of
interest and to adjust the sample rate (SR) so that this component is acquired with a
minimum of 20 points. For example, if the N1 component of the ERP is the shortest
target component with an extent of around 100 ms base to base, then the calcula-
tion SR 20 would result in a minimally required sample rate of

approximately 200 Hz. The sample rate choice should also consider any prior
knowledge regarding the temporal extent of typical statistical effects for the compo-
nents under investigation. However, higher sample rates only make sense within the
spectral bounds of the neuronal circuitry under investigation and a trade-off should
be sought between information gain and file size.

= (1,000/100) ×

Gain
The gain or vertical resolution of the signal should be chosen with two aspects in
mind. First, the gain is directly coupled to the maximum positive and negative volt-
age the amplifier can resolve without saturation. This is particularly important with
dc recordings, where even profound drift of the signal is tolerated. Second, the gain