PRACTICAL INSTRUMENTATION
Daniel Dumitru, M.D, PhD.
University of Texas Health Science Center at San Antonio
San Antonio, Texas
Clearly, a primary objective of the electrodiagnostic
medicine evaluation is to obtain accurate responses from
both nerve and muscle. If the derived data is in any
way inaccurate, then false positive and false negative
diagnoses are likely inevitable. It is, therefore, incum-
bent upon the practitioner to ensure that any technical
shortcomings due to the electrophysiologic instrument
are eliminated. This objective is achieved by having
at least, a rudimentary working knowledge of how the
electrophysiologic instrument and its associated elec-
trodes function. In my estimation, the best approach
to understand how the instrument processes data is a
combination of, theory conceptualized in very practical
ways, and manipulating the electrodes/instrument so as
to observe commensurate direct alterations in the docu-
mented response. This is exactly the approach utilized
in this discussion.
Of course, it is an impossible task to discuss the topic of
instrumentation in totality within the confines of this
assessment. As a result, we shall consider a few key con-
cepts addressing particular aspects of the electrophysio-
logic instrument: electrodes (inter-electrode separation),
amplifiers (differential amplification), filters (high and
low frequency), the stimulator (cathodal/anodal stimula-
tion, and stimulus artifact), and signal averaging. In my
opinion, it is these particular instrument subcomponents
that have produced the most confusion amongst novice
as well as seasoned practitioners.
Electrophysiologic Instrument: Overview
It can be helpful to simplify the electrophysiologic
instrument into a series of individual parts in order to
better grapple with the manner in which each instru-
ment subcomponent in turn handles the physiologic
signal of interest (Figure 1). We will follow a derived
biologic signal generated in the patient to its eventual
display on a screen with respect to how each aspect of
the instrument processes the biologic waveform. Along
the way, clinically relevant pitfalls will also be addressed
as it relates to the manner in which the waveform and
instrument interact. First, we must recognize that the
physiologic signal of interest is obviously generated and
subsequently contained within our subject. This signal
Figure 1. Electrophysiologic Instrument. Electrodes (A) located on
or in the patient, will detect the minute biologic signals generated and
subsequently convey them to the differential amplifier (B). The sum-
mated electrical signal is then filtered (C), undergoes analog-to-digital
conversion (D), and is displayed to us (E) as well as heard acoustically
(F). Time-locked signals can also be generated in the body with the
stimulator (G). Reproduced with permission: Dumitru D, Walsh NE:
Electrophysiologic instrumentation. In: Clinical Electrophysiology;
Physical Medicine and Rehabilitation State of the Art Reviews. Phila-
delphia, Hanley & Belfus Vol 3 No. 4, 1989, pp 683-699.
is either voluntarily or spontaneously generated by the
patient’s tissue, or induced by external means through
the instrument itself, i.e. through the use of an electrical
stimulator. In either case, it is our task to detect, record,
and display this signal for analysis so as to formulate an
accurate clinical diagnosis.
ELECTRODES
In order for us to detect and record electrophysiologic
signals within the human body, we must first have some
type of interface between the patient and recording
instrument. The interface required for our purposes
consist of three electrodes: active (E-1), reference (E-2),
and a ground. From a terminology perspective, the
designations E-1 and E-2 are preferred because both
electrodes are active, but to different degrees. The terms
“active” and “reference” were initially chosen because
the so called “active electrode” was purposefully located
close to the signal generator of interest, while the so
called “reference electrode” was purposefully positioned
in a particular location believed to have literally a zero
chance of recording electrical activity, i.e. in a zone of
“zero” potential. This “zero” potential zone was chosen
1
to ensure there would be minimal signal interference
from the reference electrode with that originating from
the active electrode. We have since come to appreci-
ate that there is no such thing as “zero” potential in
the body, rather, areas of relatively more or less signal.
Further, the terms G1 and G2, which are unfortunately
still designated in the literature today, are completely
antiquated and should be forgotten since they refer to
“Grid 1” and “Grid 2” of vacuum tubes originally used
in the old type of amplifiers no longer used today. In
light of the above, I will use the terms active and refer-
ence as long as we appreciate the fact that by the term
“active” I imply that this electrode is in close proximity
to the electrical generator of interest (E-1), while “refer-
ence” (E-2) implies that this particular electrode is in a
region of relatively less, but not necessarily zero electrical
activity. In this way, you will not have to continually
think: “Now what was E-1 and E-2 again?”
The third electrode, or ground electrode, is no less
important than the active and reference electrodes.
Specifically, we can conceptualize the ground electrode
as specifying a zero potential for the internal workings
of the instrument. I am sure you can appreciate the
importance of this issue. For example, if the instrument
cannot determine what zero potential is, then it cannot
relate the voltages detected in the body to any refer-
ence point. An analogy could be as follows: if you are
in the middle of some empty zone in outer space with
no earth, but just distant stars, then what does the term
up or down mean? If there is no ground (in this case
literally the earth), then terms such as up and down have
no meaning because there is nothing to reference up
and down to. Similarly, if the instrument doesn’t define
a zero, then whatever it records in terms of voltage is
relative. Therefore, we always connect a ground to the
patient to create a stable baseline from which our de-
tected voltages can either move up (defined as negative
voltage) or down (defined as positive voltage) tracing out
a particular waveform as a voltage change over a specific
time-frame. The ground electrode also serves an electri-
cal safety function. We do not want any stray currents
potentially injuring the patient.
In general, the electrode should be made of a metal
that is capable of conducting an electrical current easily.
Recall, we are purposefully locating our active electrode
over the electrical generator site while our reference
electrode is some distance away in a region of relatively
less electrical activity, at least with respect to our active
2
electrode. Our intent is to have the active electrode de-
tect the voltage generated by our signal of interest while
the reference detects as little as possible of this same
signal. In other words, we are purposefully setting up a
situation where the active and reference electrodes detect
a voltage difference between their respective location for
the signal of interest. As you know from basic physics,
if there is a voltage difference between two regions, then
current will flow. If the metals in our electrodes can
detect the voltage generated by our signal, these elec-
trodes will transmit a corresponding voltage difference
to our instrument. Theoretically at least, we should then
be able to detect our intended signal. Our electrodes,
therefore, must be clean and free of debris (e.g. rust,
caked on dried electrode paste, etc.) as well as intact (no
broken wires) in order for us to accurately detect the
desired signal. Of course, we should always thoroughly
clean our non-disposable electrodes if for no other
reason than a sanitary one. As a patient, I certainly
wouldn’t want previously used electrodes to be placed on
me without them being thoroughly washed, would you?
We shall return to these issues below as the electrodes
and amplifiers form an integral unit.
AMPLIFIERS
As we are acutely aware, the signals we wish to detect
are incredibly small and on the order of microvolts (µV)
or millivolts (mV). If we are to detect them, it is im-
perative to increase the magnitude of these signals with
respect to the background noise (intrinsic instrument
noise as well as environmental noise: TV signals, cell
phone signals, radio waves, etc.), i.e. we need to amplify
the biologic signals of interest (Figure 1B). For the
purposes of this discussion, we can define an amplifier
quite simply as a device that is capable of increasing the
magnitude of the signal detected by way of the particular
electrode connected to it. Further, we can conceptual-
ize the amplifier used by our electrodiagnostic medicine
instrument to be comprised of two paired and relatively
identical amplifiers designated as an inverting amplifier
(negative [-] amplifier), and a non-inverting amplifier
(positive [+] amplifier). Quite simply, the non-inverting
or positive amplifier accepts whatever signal is fed to it,
and magnifies it the desired amount as set by the practi-
tioner (Figure 1B). However, the inverting or negative
amplifier also magnifies whatever signal is fed to it by
the same amount, but inverts it, in other words, turns it
upside down electrically.

Figure 2. Differential Amplification. A.) Two amplifiers, a non-in-
verting (Triangle with “+” sign) and inverting amplifier (triangle with
“–“ sign) detect signals from E-1 and E-2 respectively, magnify them
by a factor of 2, after which they are then electrically summated (“S”).
B.) E-1 detects a signal of magnitude “1” while E-2 detects a signal of
magnitude “0”. The non-inverting amplifier magnifies the signal by
a factor of 2 and then passes it to the summation unit (S), while the
inverting amplifier obviously magnified the signal by a factor of two as
well. The net result is an output of “2” (2 – 0 = 2). C.) E-1 detects a
signal with a magnitude of one, which is then magnified and sent onto
the summation unit. E-2 also detects the same signal and magnifies it
for a result of “2” but then inverts it to a “-2” which is added electrical-
ly to the (+2) resulting in a net zero signal (2 – 2 = 0). D.) E-1 again
detects a signal with a magnitude of 1 and amplifies it by two, while
E-2 detects a signal with a magnitude of one-half (1/2) and magnifies
by two. The ½ is inverted to a -1 which is then summated to the +2 of
the non-inverting amplifier for an output of 1 (2 – 1 =1).
We can appreciate the above amplifier effects by consid-
ering a few examples. As described above, we have two
amplifiers that are essentially identical from an electrical
perspective (Figure 2A). The active (E-1) electrode is
connected to the positive, or non-inverting amplifier,
while our reference (E-2) electrode is connected to the
negative or inverting amplifier. Each amplifier will
magnify the signal and send it to the summation circuit
(the electrical activity from both amplifiers are simul-
taneously added together from an electrical perspec-
tive: designated “S” in Figure 2A). The net waveform
result of this electrical summation is then sent on to be
filtered and subsequently displayed and/or listened to
acoustically (e.g. electromyography [EMG] waveforms).
If we can position our active electrode over the area of
interest (e.g. mid-belly of the abductor pollicis brevis
(APB) muscle) and the reference electrode over an area
of relative zero potential, then what is generated at the
active electrode should be amplified without waveform
modification, i.e. little in the way of influence from the
reference electrode (Figure 2B). On the other hand, if
both electrodes record the identical information, then
the resultant output of the instrument is zero, since the
same data is subtracted from itself to yield no observable
waveform (Figure 2C). However, if the active electrode
records a signal while the reference electrode records the
same waveform but with half the amplitude as that of
the active electrode, then the output is reduced by half
of what it could have been (Figure 2D). This situation
may occur when our reference electrode is not com-
pletely off, but perhaps partially located on the distal
portion of the APB muscle. A reduced APB amplitude
compound muscle action potential (CMAP) can then be
anticipated from this electrode arrangement compared
to the situation where the reference electrode is posi-
tioned off the muscle tissue.
Differential Amplification. The above examples define
the concept of differential amplification. Specifically,
differential amplification represents the amplified electri-
cal output difference between the signals recorded by
the active and reference electrodes. As the examples
in Figure 2 demonstrates, differential amplification is
an extremely powerful way in which to simultaneously
display very small signals while eliminating unwanted
large signals, i.e. getting rid of the surrounding sea of
electrical noise.
First, let us attempt to unpack the above definition a
bit, and then provide some specific examples to help
us problem solve a few clinical situations once we are
armed with the above conceptualization of differential
amplification. The primary idea of differential amplifi-
cation is to detect small electrical signals even if there are
very large unwanted electrical signals (noise) simultane-
ously coinciding with our signal of interest. As we all
know, there are enormous electrical signals surround-
ing us at all times originating from television stations,
cell phone towers, and even equipment in the hospital
(ventilators, beds, monitors, etc.) or, nearby within our
office. Therefore, when we try to detect a microvolt
signal in the body, our electrodes are also simultaneously
detecting all of the large environmental signals which
can easily overwhelm the tiny biologic signal.
As an example, let us consider what is actually happen-
ing whenever we stimulate a nerve and want to record
the ensuing response. Both of our electrodes simultane-
ously and easily detect the environmental noise sur-
rounding us (Figure 3; large sine wave). Also, the active
electrode is detecting our signal of interest coinciding
with the large sine wave noise (Figure 3; superimposed
3

Figure 3. Differential Amplification. This is a typical example of
what is occurring within our instrument’s amplifiers every time we
attempt to record a biologic waveform. The active electrode (E-1) is
simultaneously recording the large environmental signal (big sine wave
representing environmental noise) with an incorporated induced signal
(small spike “s” with an arbitrary magnitude of 1; e.g. SNAP) while
the reference electrode (E-2) is also simultaneously recording the same
environmental “noise” that the E-1 electrode is detecting. This large
environmental signal is essentially eliminated through differential
amplification since it is subtracted out, while the biologic signal which is
only detected by the non-inverting amplifier is magnified and subse-
quently displayed. The amplifiers are magnifying everything by a factor
of 2 in this example.
spike: “S”). In this instance, the non-inverting amplifier
detects this signal (noise with enveloped biologic signal)
and magnifies it by a factor of 2. Also, the inverting
amplifier detects the same large environmental sine wave
(noise) without the signal of interest because we moved
it far enough away from the nerve. Therefore, the invert-
ing amplifier magnifies the sine wave by a factor of 2 just
like the non-inverting amplifier, but inverts it compared
to what the active electrode/non-inverting amplifier
detects. The electrical summation unit then summates this
information resulting in a relatively complete cancellation
of the environmental signal (basic physics for wave amplifi-
cation/subtraction) and a magnification of the desired signal
by the designated amplification factor (Figure 3). In this
way, the large environmental signal that is common to both
electrodes/amplifiers is eliminated as a so called “common
mode signal” while the biologic potential of interest is in
effect separated out and magnified as the designated “dif-
ference mode signal”. Because the difference between
the two amplifiers is detected, we thus derive the term
“differential amplification” and “common signal/com-
mon mode” rejection. Simply, the difference between
the two electrodes is amplified, and that which is com-
mon to both is subtracted out (Figure 3). How well the
4
same signal is cancelled out is a way of measuring how
“good” or alike the two amplifiers are, and referred to as
the “common mode rejection ratio”. Today, amplifiers
are built not identically but for all practical purposes, al-
most identical so as to allow us to no longer worry about
the common mode rejection ratio. We couldn’t change
it anyway, as this is a fixed parameter of the amplifier
itself.
We can now return to Figure 2, and explore it in a
bit more detail so as to prepare ourselves to engage in
some practical clinical situations. Yes, I fully appreci-
ate it is taking awhile to get practical. However, if we
don’t appreciate the above concepts, we can only apply
a “cookbook” instead of “critical thinking” approach to
problems, thereby rendering ourselves less capable of
solving unique situational issues that we have either not
read about, or not previously encountered.
Inter-electrode Separation Nerve/Muscle. Although
Figure 2 is rather schematic in appearance, it is ripe
with both valuable and practical information. As we
can clearly see in Figure 1A, both electrodes detect some
signal and then convey it to their respective amplifiers.
As previously described, the information from the refer-
ence electrode is then subtracted from the active elec-
trode through the process of differential amplification,
and the resulting information is passed on to be further
processed and displayed. The information in Figure 2B
represents the theoretical ideal we are striving to achieve.
Namely, an active electrode placed exclusively over our
signal generator of interest, while our reference electrode
is located in a place of zero electrical activity so that the
end result represents a pure amplification of our desired
biologic signal. Figure 3 has shown us that any environ-
mental or biologic noise common to both electrodes will
be eliminated. Unfortunately, real recordings in clinical
practice seldom if ever achieve this ideal. Rather, the
situation in Figure 2D more commonly represents those
situations typically encountered. As noted above, E-2
has the possibility of recording some portion of what
E-1 detects and hence, there is really never a zone of
absolute zero potential.
As an example, suppose we want to record an anti-
dromic median nerve SNAP from the third digit while
stimulating the median nerve 14 cm proximal to the
active electrode. This situation seems simple enough
until you critically ask the question: “Where should the
reference electrode be placed?” Of course, most prac-
titioners would now respond with: “4 cm distal to the
active electrode.” But wait, lets go back in time when
that question was first asked. Now what? In order to
properly answer that question, we can attempt a simple
empirical approach.
Figure 4 depicts what can be anticipated when we alter
the location of the reference electrode with respect to the
active recording electrode. One must recognize that the
reference electrode is a crucial and integral component
of the electrophysiologic system and its placement, rela-
tive to the active electrode, can have significant signal
altering effects. It is not just any electrode stuck some-
where with little in the way of consequences, particularly
with respect to its location as it relates to the active elec-
trode. We can clearly observe distinct and characteristic
alterations in the waveform as the reference electrode is
moved further along the nerve course and away from
the active recording electrode. The waveform’s ampli-
tude undergoes a dramatic increase until approximately
a 4 cm inter-electrode separation is achieved (Figure
4). Also, the onset latency doesn’t change, but the peak
latency gets longer. Why are these effects observed?
The information in Figure 2 addresses not only ampli-
fier effects, but also provides us answers to questions
regarding the relationship of recording electrodes to each
other. Figure 2C demonstrates that if we were able to
locate both the active and reference recording electrodes
at the same location at the same time over the electri-
cal generator of interest (our biologic signal), then we
would record nothing. Also, Figure 2D reveals that if
the reference electrode records some of, but not exactly
the same data as the active electrode, we can anticipate a
reduction in, but not complete obliteration of, the wave-
form. Finally, if we can locate the reference electrode far
enough away from the active recording electrode so as to
record as little as possible of the desired signal, we would
anticipate a minimal if any drop in amplitude, i.e. a
more accurate representation of the potential in totality
as generated in the body (Figure 2B).
Therefore, relating the information in Figures 2 and 4,
we can now better appreciate that when the recording
electrodes are within 1 cm or less of each other, they are
recording very similar data resulting in a reduction in
the overall potential’s amplitude. As the reference elec-
trode is moved further away from the active electrode,
less similar data originating from the biologic signal
is recorded by both electrodes, and there is less com
Figure 4. Inter-electrode Separation: Nerve. We are stimulat-
ing the median nerve antidromically 14 cm proximal to a recording
electrode placed in proximity to the base of the third digit. A reference
electrode is sequentially located at 0.5cm, 1 cm, 2 cm, 3 cm, 4 cm, 5
cm and 6 cm more distal to the active electrode along the third digit.
The active electrode position is never changed. The resulting waveform
parameters are displayed in the accompanying data table. From: Du-
mitru D, Amato AA, Zwarts MJ: Electrodiagnostic Medicine, 2nd ed.
Philadelphia, Hanley & Belfus, 2002.
mon mode signal cancellation with an increase in the
waveform’s magnitude. Once we reach approximately
4 centimeters of inter-electrode separation, there is no
longer an appreciable increase in the waveform’s ampli-
tude. This is because the total rise time of the potential
is now only observed by the active electrode and has
not yet reached the reference electrode. The waveform’s
full amplitude can now be appreciated. Similarly, since
the reference electrode was subtracting out data similar
to the active electrode at 3 cm of separation or less, the
waveform was truncated prematurely in its amplitude
resulting in a shortening of its true peak latency. Obvi-
ously, the onset latency never changed since the wave-
form’s combined action potentials arrived at the active
electrode first, prior to them arriving later in time at
the reference electrode. Recall, we never altered the
active electrode’s location and hence its distance from
the stimulating electrode never changed. We can see,
therefore, that the reference electrode location is very
important with respect to defining our waveform of
interest. We also know that when the electrodes are 4
cm apart and the amplitude no longer increases, that we
have allowed the full potential magnitude to develop.
The time from potential onset to peak magnitude has
a rise time of about 0.8 ms. We can use the equation
nerve conduction velocity (NCV)=D/T to predict the
best inter-electrode separation as well. Specifically, a
wavefront traveling at roughly 50 M/s that requires 0.8
ms to fully manifest its peak amplitude has a spatial
5
expanse along the nerve of roughly 4 cm (50 M/s = 0.8
ms/D; D = 4.0 cm). In other words, the actual space
taken up along the nerve (spatial extent) required from
the action potential’s onset to its peak, is 4 cm. This
means the longitudinal extent of the action potential
along the nerve at any point in time needs 4 cm of space
to fully manifest from its onset to peak. That is, if you
can picture the traveling SNAP onset at any point in
time (initiation of depolarization), the SNAP’s peak
depolarization for all its fibers is 4 cm behind this onset.
Therefore, if we locate a reference electrode anywhere less
than 4 cm from the active electrode, both electrodes will
see some portion of the rising action potential and elimi-
nate it as a common signal resulting in a reduction in
the waveform’s amplitude (Figures 2D & 4). In other
words, some portion of the desired waveform is now
detected by both electrodes and eliminated as a common
signal through the process of differential amplification.
Therefore, for antidromic sensory nerves, we should be
using an ideal distance for the inter-electrode separation
of 4 cm. Of course, this is not always practical as you
may have a person with short digits, or certainly the fifth
digit in many persons may preclude this ideal separation.
The answer, of course, is the establishment of reference
values with an inter-electrode separation specified by the
investigator. As we know, there are a number of sensory
(e.g. radial, sural, and superficial peroneal) and mixed
nerve (median, ulnar, medial/lateral plantar) studies that
use bar electrodes with only 2 or 3 cm of inter-electrode
separation, and this is fine since we are using the refer-
ence data developed by the investigators for electrode
separations less than 4 cm.
Unfortunately, the above noted information regarding
sensory nerves and SNAPs is uncritically applied to
muscle. This is clearly a misapplication of information
destined to result in erroneous results. In particular,
although the concept of an inter-electrode separation of
4 cm is applicable for the median SNAP to the 2nd-4th
digits in most persons, it is not usually applicable for the
first and fifth digits (this is why SNAPs from toes are not
practical) and certainly not for (CMAPs) for any muscle.
The issue of maximizing the rise time for a nerve action
potential conducting at 50 M/s is clearly not relevant for
muscle tissue which conducts action potentials at 3-4
M/s. The critical issue regarding the proper inter-elec-
trode separation for muscle is not a specific distance, but
rather a specific location. In particular, the active
electrode should be positioned over the muscle’s end-
plate (motor point) region so as to record an initial
6
Figure 5. Inter-electrode Separation: Muscle. A) A CMAP is
derived from the abductor pollicis brevis muscle following median nerve
wrist stimulation with the active electrode over the muscle’s end-plate re-
gion and a reference electrode positioned at the musculotendinous junc-
tion on the first digit. B) Relocating the reference electrode onto the
muscle belly itself results in a reduction in the CMAP amplitude since
some of the same signal is detected from the muscle and thus eliminated
as a common mode signal. From: Dumitru D, Walsh NE: Practical
instrumentation and common sources of error. Am J Phys Med Rehabil
67:55-65, 1988.
negative deflection. The reference electrode should be
located off the active muscle tissue where a location in
close proximity to the musculotendinous junction of
the muscle under investigation typically suffices. Once
again, the primary principle is to have the active record-
ing electrode over the signal generator site (muscle end-
plate region), and the reference electrode in a location
that is minimally active from an electrical perspective.
If one locates the reference electrode over any portion
of the activated muscle, a reduction in the true CMAP
amplitude can be anticipated (Figures. 2 & 5). The
explanation is similar to that for nerve tissue. When the
reference electrode is on the musculotendinous junction,
it records little electrical activity in common with the
active recording electrode thereby generating little com-
mon mode subtraction. On the other hand, when the
reference electrode is located on the muscle tissue, the
degree of a common signal increases as does the ensuing
common mode subtraction with a resulting reduction
in the CMAP amplitude. Therefore, the concept of
4 cm is totally irrelevant when recording from muscle
tissue. The primary issue is to ensure that the reference
electrode is not located on the same muscle that is being
activated so as to ensure minimal common signal sub-
traction thus avoiding CMAP amplitude reduction.
Optimal Electrode Placement. Of course, the above
discussion addressing inter-electrode separation clearly
begs the question: “Ok, so just where should I locate
the electrodes to optimally record the signal of interest
with minimal contamination from both environmental
noise and coincident activity from the tissue of interest?”
An alternative question might be: “Why not just locate
a single reference electrode on the patient’s big toe and
use this location for any and all recordings?” In order
to address these questions fully, we first need to return
to the fundamental process of differential amplification,
but this time consider what the electrodes are recording
simultaneously from the environment as well as the tis-
sues of interest.
First and foremost, we must understand that the elec-
trodes do not “know” what we want to record, but will
in fact record whatever electrical activity is within their
recording territory. In effect, our electrodes are antennas
and will act to detect any signal of sufficient strength.
Let us consider the example of stimulating the median
nerve at the wrist and recording a CMAP from the
abductor pollicis brevis (APB) muscle. When we locate
our active and reference electrodes on the muscle’s end-
plate region and musculotendinous junction respectively
with a ground electrode nearby, what then is happen-
ing when we stimulate the nerve? Initially we activate
our stimulator with a minimal or submaximal current
which is insufficient to depolarize the nerve. What do
we see on the instrument’s screen? If our electrodes are
securely placed on the patient we should see a flat line,
i.e. no signal is detected. Let’s stop here and address
the completely silent screen. We know there are huge
electrical signals surrounding us, but yet the instrument
does not display them. Why not? As noted above, this is
because these distant signals are detected equally by each
of our recording electrodes and eliminated completely as
a so called common mode signal rendering a nice quiet
trace so that we can then detect even a very small bio-
logic signal. When the current is increased to a strength
sufficient to fully depolarize the nerve, we can now
document a fully formed CMAP from the APB (Figures
3 & 5). We can rest assured that if the active electrode
is on the muscle belly and the reference electrode off the
muscle, a maximal amplitude CMAP will ensue with
minimal CMAP degradation, or contamination from
large environmental noise generators.
With respect to the two questions posed above, we can
reformulate a single question regarding optimal active/
reference inter-electrode placement: “How close is too
close, and how far is too far?” Recall, in order to opti-
mally record a biologic signal of interest, we need suffi-
cient electrode separation to ensure that the two elec-
trodes are not recording similar data with respect to the
signal’s parameters of interest, i.e. amplitude, rise time,
etc. If the electrodes are too close, then some portion of
our signal of interest will be common to both electrodes
and subsequently eliminated as a common mode signal
(Figure 4). On the other hand, what would happen if
the two electrodes were very far apart, but still on the
patient, i.e. the “big toe question”?
As noted above, we can detect small biologic waveforms
because the two electrodes are capable of simultaneously
recording the large environmental noise and eliminate
it as a common signal. In order for this process to be
successful, it is imperative that the two electrodes detect
these large signals identically. You can imagine what
would happen if there were even the slightest difference
in detection for such large signals. Specifically, this small
difference would be detected by the two amplifiers, and
then this difference signal would be magnified. Recall,
any difference between the two electrodes is considered
a “difference signal” and subsequently amplified. As the
separation between our two electrodes increases, so does
the opportunity for each of them to not detect the vari-
ous environmental noise generators identically. Even a
separation of a few meters for somewhat distant signals
(big toe) or a few centimeters for close by signal noise
can result in a significant enough amplification of these
environmental waveforms to preclude biologic signal
detection. Specifically, if there is some type of machine
generating electrical noise nearby (next room or a floor
up or down), a few meters or even centimeters of elec-
trode separation may be all that is required to detect and
subsequently amplify this signal thereby interfering with
our biologic signal of interest.
Quite simply then, our electrodes should be far enough
apart to minimize common mode signals from the
biologic tissue of interest, but close enough to maximize
the common mode signals from those sources we wish
to eliminate. Therefore, placing the reference electrode
on the big toe may indeed be far enough apart to mini-
mize recording from the nerve or muscle of interest
compared to the active electrode, but also insufficiently
close to the active electrode so as to eliminate some
environmental or even biologic sources of noise par-
ticularly those nearby. The end result, in this situation,
7
would be sufficient contamination of the biologic signal
of interest so as to preclude its proper detection. Don’t’
forget, the body may also be generating undesired signals
or noise. If the electrode is on the big toe for example,
EMG potentials from contracting muscles near the ac-
tive recording electrode may not be detected by the big
toe reference electrode, and subsequently amplified as a
difference signal as opposed to eliminated as a common
signal resulting in too noisy a trace for optimal detection
of the desired waveform.
Electrode Materials/Cleanliness. Although not fre-
quently discussed, the materials of which electrodes are
made, as well as how clean they are, can directly influ-
ence the success of our recording the signals of inter-
est. Prior to addressing these two important issues and
appreciating their practical implications, we need to first
consider how the electrodes and amplifiers interact with
each other so as to influence what we see on the instru-
ment’s screen.
As previously noted, biologic signals are relatively small
and require placement of recording electrodes on, or
in the patient so as to approach, detect, and record the
desired signal. The signal is then passed onto the ampli-
fier and processed through differential amplification.
The two electrodes are actual physical entities and hence
are made of various metals that will, by their very nature,
degrade some of the signal. Specifically, no electrode
passes the signal onto the amplifiers without degrading
that signal somewhat. Whatever portion of the signal
is not adversely affected by the electrode, is then passed
onto the amplifier to be processed and subsequently
displayed. In electrical terms, the electrode and ampli-
fier form a “voltage divider”. Simply, this means that the
total biologic signal is “divided” or shared between the
electrode and amplifier. That is, some of the signal drops
across the electrode and that which remains passes on to
the amplifier for us to observe. Ideally, we want as little
of the signal as possible to drop at the electrode so that
most of it passes onto the amplifier to be analyzed by the
practitioner. In this way, the magnitude of the response
most accurately represents that which is occurring in the
body, i.e. total number of axons or muscle fibers. At this
point, it may help to try to express this issue with some
simple mathematics and then apply a conceptualization
to the mathematics so to have a practical understanding
that can be implemented clinically by the practitioner.
8
We can designate the biologic signal’s voltage in the
body as Etotal. Some of the signal’s voltage is going to
be lost at the electrode (Eelec) and what is left over will
be detected at the amplifier (Eamp). In other words, the
total amplitude of the biologic signal in the body can be
represented by the summation of the signal at the elec-
trode and that at the amplifier: Etotal = Eelec + Eamp.
According to Ohm’s law, the current (I) across the
electrode and amplifier is the same, but the voltage drop
across each component is different depending upon the
impedance (Z)/resistance (R) of each component. For
our purposes, we will use the term impedance (Z) and
consider it equivalent to resistance. From basic physics,
Ohm’s Law, therefore, is defined as: E = I X Z. This is
true for any circuit in series (series circuit) of which our
electrode and amplifier constitute such a series circuit.
Specifically, the voltage observed at any location is
defined by the current over that location multiplied by
the impedance for that location. For our voltage divider
concept of electrode and amplifier, the current is as-
sumed be the same across the recording electrode and
amplifier, while the impedance differs for the recording
electrodes and amplifier. As a result, the amount of volt-
age “dropped” at each location depends upon the respec-
tive impedance of the electrode and amplifier. Whatever
portion of the signal is not lost at the electrode, subse-
quently remains for us to observe.
Therefore, applying Ohm’s law for the electrode and
amplifier, the voltage drop across each component is:
Electrode: Eelec = I X Zelec, and
Amplifier: Eamp = I X Zamp
Substituting the above terms into : Etotal = Eelec + Eamp
results in:
Etotal = I X Zelec + I X Zamp or;
Etotal = I X (Zelec + Zamp)
Since we are particularly interested in the signal’s volt-
age at the amplifier as this is the signal we will eventually
analyze, we can solve the above equation for that portion
of the voltage detected at the amplifier.
The equation from above for the current flow across the
amplifier is:
I = Eamp/Zamp; and this term can then be substituted for
current (I) in the previous equation to get:
Etotal = Eamp X (Zelec + Zamp)
Zamp
Solving the above equation for what we see, the signal’s
voltage detected at the amplifier becomes:
Eamp = Etotal X Zamp
Zelec + Zamp
The above equation has a number of particularly relevant
clinical implications operational every time we assess a
patient with our electrodiagnostic equipment. As we
have previously stated, we would like the voltage de-
tected at the amplifier to be as close as possible to that
originating in the body so as to most accurately reflect
what is actually occurring biologically. As can be seen
in the above equation, the desired goal of accurate signal
detection can be best achieved when there is little drop
of voltage across the electrode, i.e. the impedance of the
electrode is very tiny compared to that of the amplifier.
If Zelec is minimal (say 1% of Zamp: i.e. Zelec = 0.01Zamp)
compared to the Zamp in the above equation, then Eamp
equates to Etotal:
Eamp = (Etotal X Zamp) / (0.01Zamp + Zamp)
Eamp = (Etotal)(Zamp) / 1.01Zamp
Eamp = (Etotal)(0.99)
Eamp = 0.99Etotal
We can now fully appreciate that if the electrode has
a minimal impedance compared to the amplifier, then
most of the biologic signal’s amplitude should be de-
tected by the amplifier since little of it was degraded
by the electrode. In the above hypothetical situation,
we were able to detect 99% of the signal generated in
the body since only 1% “dropped” across the electrode.
Practically, we do not need to worry about modern
day amplifiers as their “input impedance” is quite high
compared to all new electrodes. However, what if we
were using electrodes that perhaps where dirty, rusted,
caked with paste, or otherwise less than optimal. Obvi-
ously the electrode’s impedance could rise dramatically.
If the impedance of the electrode does indeed increases
substantially, and begins to become a significant percent-
age of the amplifier’s impedance, the above equation
clearly shows us that the detected signal at the amplifier
would be reduced and hence no longer accurately reflect
the total number of axons or muscle fibers depolarized.
Similarly, the large environmental noise may be detected
differently. This difference would be amplified and pos-
sibly overwhelm our signal of interest.
Clinical Considerations: Inter-electrode Separation.
OK, let us now attempt to use the above theoretical/con-
ceptual information to address a number of real world
clinical considerations. What would be anticipated if
our sensory recording electrodes were too close together?
As we have previously discussed, a reduction in ampli-
tude with a shortening of the peak latency would be
anticipated. If we were performing an antidromic me-
dian nerve sensory study utilizing wrist and mid-palm
stimulation sites in a patient with possible carpal tunnel
syndrome, and our electrodes were too closely spaced,
we could document a number of possible erroneous
findings. First, as an example, let’s say the patient either
had a mild peripheral neuropathy or a cold hand with
a prolonged mid-palm latency. If our electrodes were
properly spaced, we would easily detect that there was
preferential slowing of the mid-palm-to-digit segment
with no slowing across the remaining carpal tunnel seg-
ment. But, because our electrodes are too close, we have
a slight shortening of the peak latency for both responses
eliminating any hint of distal slowing. In this case, we
could erroneously conclude there is no abnormality
present if the reduction in amplitude was not that great,
particularly in a cold hand (focal temperature cooling
tends to increase SNAP amplitudes).
As another example, if our SNAP electrodes were too
close, we could record normal latency, small amplitude
SNAPs. Suppose our electrodes were properly placed
for the CMAP studies yielding normal amplitude
CMAPs. This finding could suggest a non-demyelinat-
ing axonal loss type of pure sensory neuropathy (small
amplitude normal latency responses). On the other
hand, if our SNAP electrodes had a proper inter-elec-
trode separation, but the CMAP electrodes were both on
the muscle tissue, normal SNAPs and abnormal CMAP
amplitudes with normal onset latencies and NCVs
would ensue. This could be misconstrued as a possible
myopathy, motor neuron disease, neuromuscular junc-
tion disorder, or pure motor neuropathy and possibly
even acute motor axonal neuropathy (AMAN). These
erroneous findings could certainly lead one to embark
on an expensive and fruitless clinical investigation. Similar-
ly, a combination of motor and sensory electrode mis-place-
ments can produce a combination of the above possible
clinical disorders leading to considerable confusion.
9
I am confident you could mix and match the above pos-
sibilities to clearly see that a full compliment of potential
false positive and false negative results can ensue from
a poor understanding of proper electrode placement.
In short, your diagnostic capabilities are only as good
as your ability to ensure an accurate collection of data.
This is certainly consistent with the old axiom: garbage
in, garbage out (GIGO).
Clinical Considerations: Electrodes. In addition to
the above noted inter-electrode spacing issues, prob-
lems with the electrodes themselves can produce similar
effects even if the electrodes were properly positioned.
Specifically, if the electrode’s impedance increases
enough to degrade the biologic signal’s detection at the
amplifier, we would then observe a signal with a reduced
amplitude. If this dirty electrode is the active electrode
we could see a number of possibilities. For sure, the
amplitude of the evoked SNAP and CMAP would be
reduced leading us to conclude that there was an axonal
sensory and motor neuropathy operational with little
in the way of demyelination if both sensory and motor
electrodes were dirty. However, if the sensory electrodes
were fine, but the motor electrodes were dirty, we could
interpret this data as a combination of normal SNAPs
with reduced CMAP amplitudes of normal latency sug-
gesting a pre-ganglionic lesion (motor neuron disorder
or radiculopathy) or a particular type of post-ganglionic
lesion (neuromuscular junction disorder, myopathy, or
even AMAN). Also, what if the sensory electrodes are
dirty, but the motor electrodes are fine. In this case, we
could conclude that a primary sensory axonal neuropa-
thy with no motor component is operational. As can be
clearly seen, something as simple as dirty or otherwise
adversely affected electrodes (rust, frayed wires, etc.) with
elevated impedances can result in false positive studies
with serious adverse diagnostic consequences.
An additional serious issue with respect to dirty, or even
broken wires within plastic sheathing, arises with re-
spect to the environmental noise. If both electrodes are
sufficiently different enough in their impedance charac-
teristics (one is dirty while the other is clean or one of
the wires is unknowingly broken) then the instrument’s
ability to accomplish common mode signal rejection and
hence differential amplification is compromised. The
net result is the amplification of environmental noise
leading to the possibility of large 60 Hz interference
or other unwanted signals either completely or par-
tially obliterating the desired biologic signal. Therefore,
10
whenever large 60 Hz interference is detected, this in
effect signifies an inability of the instrument to success-
fully perform common mode rejection. This situation is
problem solved by ensuring the following:
-- all electrodes are connected to the patient
-- all electrodes are plugged into the proper ports and
proper channel
-- all electrodes are thoroughly cleaned
-- a ground electrode is connected to the patient and
instrument
-- the skin is not dirty or covered with debris which
may lead to different signal detection between the
two electrodes
-- no obvious broken wires are present
-- the same type of electrodes with respect to metallic
composition are used
-- the preamplifier is located as close to the patient as
possible (minimize electrode antenna effect)
-- short as opposed to long electrode leads are used
(minimize antenna effect)
-- the active and reference wires are braided together
to assist in common mode rejection (each electrodes
detects the common noise identically all along their
length)
-- any nearby equipment that can be unplugged should
be unplugged
-- lights should be turned off
-- touching the patient with your bare hand/finger
-- consider using the 60 Hz notch filter
Problem Solving: Clinical Situation. Let us use a
hypothetical case to specifically address some possible
instrumentation/amplifier/electrode issues related to
differential amplification. Suppose we are asked to enter
one of the most difficult electrophysiologic environ-
ments in which we will ever be asked to work so as to
assess a patient with a possible neuropathy/myopathy,
i.e. the Intensive Care Unit (ICU). These are very chal-
lenging locations in which to perform an electrophysi-
ologic study because they are frequently high up in the
building (great place to detect transmitted TV/Radio
signals), filled with electronic instruments generating
wonderful interference, much of the equipment cannot
be turned off, the patient may not be able to cooperate
fully with the assessment, physical space is a limitation,
we may not be able to position either the instrument or
the patient optimally, and the list goes on.
Suppose, for this discussion, we begin our assessment
with a monopolar needle examination of some of the
limb muscles. We note that there is significant 60 Hz
interference. We first make sure all our electrodes are
plugged into the proper amplifier ports. Although
employing the 60 Hz notch filter can distort some of
our waveforms, we use it anyway but there is only a
small improvement. We check all our electrodes and
they seem to be intact and connected to, or inserted
into the patient. We then touch the patient with a
non-gloved hand and the interference is reduced some-
what but not completely. It is unclear why touching
the patient helps, but it certainly can, and does make a
difference in this case. Looking around the room reveals
that the only things we can unplug are the bed and calf
sequential compression device. Not much of a change
is noted. We clean the skin under the surface reference
and ground electrodes and dry it off, and them employ
a small amount of paste with a good but not complete
reduction in the interference. We next move the pre-
amplifier as close as we can to the patient to shorten the
electrode leads minimizing their “antenna effect”( long
leads act like good antennas) and again a further small
reduction in interference is achieved. At this point, it is
now possible to detect spontaneous and voluntary activ-
ity, but the interference is still annoying. We move the
surface reference electrode close to the needle insertion
site and this helps a bit. The intent is to have our refer-
ence electrode as close as we can to the needle electrode
to eliminate common environmental signals. Since the
needle is deep in the tissue and the reference is on the
skin surface, it is unlikely they will result in a reduction
of the EMG signal because they are too far apart with
respect to the biologic signal. We change our electrodes
in case there is a faulty wire but no change is observed.
As a last ditch effort we disconnect the surface reference
electrode and replace it with another monopolar needle
inserted just into the subcutaneous tissue with our ac
Figure 6. Differential Amplification-Clinical Setting. A) A
monopolar recording demonstrating significant environmental interfer-
ence despite employing a number of mitigating strategies (see text). B)
Insertion of a second monopolar needle (reference electrode) just into the
subcutaneous tissue close to the active electrode, monopolar needle in the
muscle, results in a dramatic reduction in the interference.
tive needle electrode deeper in the muscle (Figure 6).
Amazingly all of the interference disappears and we now
have a completely quiet baseline with easily detectable
biologic waveforms.
In the above example, any one of the maneuvers we tried
could have eliminated the problem. In this case, the
most helpful alteration was to use a second monopolar
needle as the reference electrode. The instrument was
telling us that differential amplification of the common
mode signal was insufficient because we could clearly de-
tect the large 60 Hz interference. By replacing the sur-
face electrode with a second monopolar needle, we were
now using two electrodes with identical metallic compo-
sitions permitting both electrodes to record the nearby/
distant environmental signals with equal magnitude,
thereby eliminating them as a common signal. Recall
from the above equation, if the impedances of the two
electrodes are different for whatever reason, we may not
be able to effectively record either the signal of interest
or eliminate unwanted common environmental signals.
An alternative approach could have been to use a con-
centric needle electrode. The idea here is that although
the two recording surfaces are of different metallic com-
position, their close proximity may have also helped the
situation. However, having tried this approach in the
past, I have found that the two monopolar needles work
better than a concentric needle electrode suggesting
the difference in metals for the concentric needle’s two
recording surfaces adversely outweighs their close prox-
imity. This isn’t to say, that a concentric needle electrode
may not suffice to a sufficient enough degree so as to
record the desired activity. The above described example
11
of performing a needle EMG in the ICU is a real situa-
tion I encountered and problem solved as noted above.
The utilization of two monopolar needles was dramatic
and should be considered in any situation when more
common solutions prove to be ineffective.

Concentric/Monopolar Needles. As an aside, in my

practice I use both concentric and monopolar needles
with regular frequency. The concentric electrode is used
exclusively for performing a quantitative needle EMG
assessment particularly for diagnosing myopathic condi-
tions. The closely spaced active electrode (central core)
and surrounding reference (stainless steel) cannula act
quite well to eliminate distant motor unit potentials and
environmental noise as a common signal. This is im-
portant as the crucial, but subtle parameter of waveform
duration needs to be measured accurately. Too much
biologic or environmental noise can act to deteriorate
one’s ability to accurately measure motor unit action Figure 7. Amplifier Sensitivity Effect on CMAP onset latency. A reduc-
potential (MUAP) waveform onset and termination. tion in the amplifier’s sensitivity (A-D) produces a sequential prolonga-
We use monopolar needles for all other routine record- tion in the onset of the CMAP. From: Dumitru D, Walsh NE: Practi-
cal instrumentation and common sources of error. Am J Phys Med
ings. It is true they are more “noisy”, but the somewhat
Rehabil 67:55-65, 1988.
lessened differential amplification can work in your favor
by not eliminating nearby signals of importance. I have
20,000 microvolts. The amplifier gain in this example
found it somewhat more difficult to detect subtle/sparse
is 1000. In other words, the input signal was amplified
positive sharp waves and fibrillation potentials with
1000 times. Alternatively, we can express the magnifi-
concentric needle electrodes when compared to mono-
cation of a waveform as an increase or decrease in the
polar electrodes exactly because of the better differential
waveform’s ratio of the input voltage, to the size of the
amplification of the concentric needle electrode. We
ensuing waveform deflection on the instrument’s screen,
recently examined a patient with mild polymyositis and
and measured in microvolts/millivolts per screen divi-
had difficulty documenting positive sharp waves and so
sion, i.e. the sensitivity of the amplifier. Therefore, an
called pseudomyotonic discharges with the concentric
amplifier with a sensitivity of 10 microvolts per divi-
needle, but they were relatively easy to document with
sion implies that an input signal of 10 microvolts will
the monopolar needle. I also find the concentric needle
result in a 1 division deflection on the screen. Altering
to produce more bleeding secondary to its cutting edge
the amplifier’s gain/sensitivity results in the anticipated
and is comparatively more painful. I am fully aware
waveform change. Specifically, increasing the gain/sen-
of what the literature has found regarding pain, but I
sitivity will result in a large waveform displayed on the
respectfully disagree based on my experience and that
instrument’s screen while reducing the gain/sensitivity
reported to me by my residents. As a result, regarding
displays a comparatively smaller waveform (Figure 7).
monopolar vs concentric needle electrodes, your experi-
The primary issue with respect to altering the gain is a
ences may differ from that of mine. They are both excel-
commensurate alteration in the waveform’s measured
lent tools for specific jobs and should be utilized as such.
onset latency. Specifically, as one increases the gain/
One electrode is not better than the other, they are just
sensitivity of the amplifier, there is an “apparent” short-
different, and should be used when their particular char-
ening of the waveform’s onset latency. In reality, the
acteristics can be most useful clinically.
onset of the waveform is physiologically defined by the
conduction velocity of the fastest conducting fibers, and
Gain. We may define amplifier gain as a unit-less
the distance between the stimulus and recording site.
quantity of the output signal amplitude divided by the
However, as we increase the amplifier’s gain/sensitiv-
input signal amplitude. For example, an input signal of
ity the potential’s onset shifts to shorter times because
20 microvolts is amplified such that the output signal is

12
we can perceive earlier and earlier (smaller and smaller)
waveform departures from the baseline (Figure 7). The
important clinical aspect to this finding is a requirement
to utilize the same gain/sensitivity settings originally
defined by the reference data, and all stimulation sites
between different locations. That is, if our reference data
utilizes a sensitivity of 2 mV/division and we are using
500 µV/division, all of our detected latencies will have
earlier onsets than that defined by the reference data.
This could result in a borderline prolonged distal motor
latency detected earlier than would be documented at 2
mV/division producing an erroneously normal result. Of
note, the issue of gain and waveform onset is much more
relevant to CMAPs than SNAPs since such a relatively
high gain is already employed for measuring SNAPs.
FILTERS
In my estimation a functional understanding of filters
and how they affect electrophysiologic signals is arguably
the least understood, and/or most misunderstood con-
cept in electrodiagnostic medicine. This is not due to
an inability of clinicians to understand signal filtration,
but rather a result of such poor attempts on the part of
electrophysiologic textbooks at explaining filters and
their effects on waveforms. Specifically, either the topic
is not addressed at all, or electrical engineers are typically
asked to write chapters on filters utilizing incomprehen-
sible mathematics with absolutely no, or precious little
clinical relevance. In this section of the handout, I will
employ basic concepts with no mathematics so as to pro-
vide information that is preferentially directed toward a
functional and clinical application of filters.
We may define a filter as a device with the specific
purpose of limiting the frequency content of waveforms
recorded by the electrodiagnostic medicine instrument.
In other words, we want to observe a bandwidth of fre-
quencies consisting of only those frequencies contained
within our electrophysiologic waveform of interest (e.g.
SNAP, CMAP, EMG, single fiber EMG [SFEMG], etc.)
and exclude all other frequencies. We may consider all
those frequencies not contained within our waveform
to be noise. Therefore, through the process of filtration,
we are attempting to create recorded waveforms that
contain only those frequencies for the waveform
we want and remove all others producing a relatively
noise-less recording.
Quite simply, each waveform can be thought of as be-
ing comprised of low frequencies and high frequencies.
Although the terms high and low are rather arbitrary, for
our purposes, a low frequency is in the neighborhood of
2-5 Hz and a high frequency approximates 10,000 Hz,
with those frequencies below and above these respective
values considered for the most part as noise. We are
hereby defining “noise” as any signal or frequency that
does not contribute to the clinically relevant portion of
our waveform. The frequencies between 2-5 Hz and
10,000 Hz for the most part are the frequencies primar-
ily contained in the majority of the different waveforms
we routinely detect. Those frequency values between
these somewhat arbitrary values constitute our desired
bandwidth. The bandwidth is derived and maintained
through the application of a low frequency filter and a
high frequency filter.
A low frequency filter permits frequencies above its
limit to pass through the instrument while block-
ing those frequencies below its limit. This is why low
frequency filters are also sometimes referred to as high
pass filters. That is, they permit high frequencies to pass.
For example, a low frequency filter of 5 Hz will act on a
waveform so as to preclude all those frequencies approxi-
mating 5 Hz or less from contributing to the observed
waveform. In short, those frequencies of 5 Hz or less are
“extracted” out of the waveform. The frequencies greater
than 5 Hz are not affected.
A high frequency filter permits lower frequencies below
its limit to contribute to the observed waveform, but
extracts out any frequencies at or above its limit. As a
result, a high frequency filter is also referred to as a low
pass filter. That is, it permits low frequencies to pass.
For example, a high frequency filter of 10,000 Hz will
allow all frequencies below 10,000 Hz to contribute to
the waveform of interest but “extract” out those frequen-
cies of approximately 10,000 Hz and above.
It is the manipulation of the above noted high and low
frequency filters which acts to create the bandwidth
comprised of only frequencies we want contained in our
waveform and remove all others considered to be noise.
The high and low frequency filters act in concert to for-
mulate a “window” (bandwidth) of observed frequencies.
An inability to conceptualize what effects filters have on
clinically derived waveforms will have adverse conse-
quences regarding the creation of false positive and false
negative diagnoses. We will consider these consequences
at the end of this filter section.
13

Figure 8. Subcomponent Waveform Frequencies. A) A series of
differing waveforms with different amplitudes, frequencies and phases
are all summated to produce a waveform with a specific configuration
(square wave). B) Altering the above noted individual frequencies
results in a completely different looking waveform resembling a more
biological waveform. From: Dumitru D, Amato AA, Zwarts MJ:
Electrodiagnostic Medicine, 2nd ed. Philadelphia, Hanley & Belfus, 2002.
Subcomponent Waveforms. Prior to delving into the
intricacies and clinical effects of filters, we need to dis-
cuss the concept of a signal and its composite subcom-
ponent waveforms. For our purposes, we will consider
all waveforms detected by the electrodiagnostic instru-
ment to be comprised of the summation of multiple
individual subcomponent waveforms, each with their
own frequency, amplitude, and phase (Figure 8). In this
discussion, the most important issue to consider, is that
each recorded biologic waveform is comprised of mul-
tiple subcomponent waveforms of both high and low
frequencies all summating to generate a composite wave-
form. Clearly, altering any of the subcomponent wave-
forms will surely result in an alteration in the observed
waveform as well. That is, if we were able to remove
some of the subcomponent waveforms of our choosing,
then the composite waveform would no longer appear
as it did prior to the subcomponent waveform removal.
Further, depending upon which waveforms were re-
moved, those of high or low frequency, characteristic
changes in the composite waveform would ensue. Spe-
cifically, if we were able to remove preferentially the low
frequencies, then the remaining high frequencies would
predominate and result in a waveform with characteris-
tics more consistent with a higher frequency waveform,
i.e. occurring sooner in time, shorter duration, more
phases per unit time. Similarly, if we were to remove
some of the high frequency subcomponents, then a
composite waveform would result that is more reminis-
cent of a low frequency waveform, i.e. occurring later in
time, longer duration, and possibly fewer phases per unit
time. In other words, the composite waveform we ob-
14
served on the instrument’s screen is in essence a balance
or summation of high and low frequency subcomponent
waveforms all contributing to the observed waveform.
Extracting any of the subcomponent waveforms from
the overall signal results in a changed composite wave-
form because of a now different subcomponent summa-
tion of frequencies with those extracted subcomponents
no longer interacting with those that remain.
Therefore, whenever we alter a high or low frequency
filter, we in turn produce a different distribution of high
and low frequency subcomponent waveforms. The
remaining subcomponent waveforms subsequently sum-
mate differently resulting in a waveform with different
characteristics. The observed waveform characteristics
(amplitude, duration, phases, rise time, etc.) are directly
dependent upon those frequencies that remain and
hence predominate. In other words, a new balance of
summations and subtractions of subcomponent wave-
forms is established with a subsequent alteration in the
detected waveform’s parameters compared to before. In
essence, we create a totally new waveform by altering the
frequency subcomponent interactions based on a dif-
ferent frequency composition. Which ever remaining
frequencies are left, now predominate and the observed
waveform takes on those remaining frequency character-
istics noted above regarding waveform parameters. Let
us now attempt to put this all together and demonstrate
specific waveform alterations based upon low and high
frequency filter changes. Finally, we shall address the
clinical implications of waveform filtration and possible
false positive and false negative diagnoses.
For this discussion, I will specifically refer to low fre-
quencies as “slow stuff” and high frequencies as “fast
stuff”. My intent is to create a teaching tool that uses
colloquial language without mathematics to improve ac-
cessibility. You can equivalently substitute low frequen-
cies and high frequencies respectively when the terms
slow stuff and fast stuff are encountered, if you prefer.
My primary point is to demonstrate that the proper
utilization of simple language to convey a complex mes-
sage is preferable as long as it aids comprehension and
permits a day-to-day clinical application of important
concepts.
Low Frequency Filter. In this discussion we can
consider the low frequency filter to be operational from
1 Hz to 500 Hz. Therefore, if we act to elevate the low

Figure 9. Low Frequency Filter Elevation On SNAP. A sequential
elevation in the low frequency filter without altering the high frequency
filter produces: 1) amplitude reduction, 2) total potential duration
shortening, 3) negative spike duration shortening, 4) peak onset latency
shortening, 5) no change in onset latency, and 6) addition of a phase.
The specific waveform parameter alterations are documented in the
figure’s accompanying table. From: Dumitru D, Walsh NE: Practical
instrumentation and common sources of error. Am J Phys Med Rehabil
67:55-65, 1988.
frequency filter on our instrument from 1 Hz to 300
Hz, what type of ensuing waveform alterations can we
expect (Figure 9). In order to answer this question us-
ing our subcomponent waveform model in conjunction
with the previously noted “slow” and “fast” stuff termi-
nology, we can ask ourselves a series of questions so as to
arrive at the most appropriate answers.
First, we note that we are “elevating” the low frequency
filter from 1 Hz to 300 Hz. We then ask ourselves,
“Are we taking something out of the waveform or put-
ting something into the waveform?” By elevating the
low frequency filter to 300 Hz we are “taking away” or
extracting stuff from the waveform. If we “take away”
something, do you think the waveform will get big-
ger or smaller? Obviously, if there is less of something,
it MUST get smaller. Hence, we would anticipate
a reduction in waveform’s amplitude. Next, we ask
ourselves, “What did we take away, “fast stuff” or “slow
stuff”?” Since we elevated the low frequency filter we
took away slow stuff from the waveform. Now, “What
is left?” Clearly, if we removed some “slow stuff”, then
“fast stuff” must remain, or at least be comparatively
Figure 10. Low Frequency Filter Elevation On CMAP. Elevat-
ing the low frequency filter while not altering the high frequency filter
on a CMAP produces similar changes to those observed for a SNAP but
with comparatively more dramatic alterations documented. In addition
to the significant amplitude drop as the high frequency filter is elevated,
note how clearly the waveform changes from a biphasic to triphasic
configuration. From: Dumitru D, Walsh NE: Practical instrumentation
and common sources of error. Am J Phys Med Rehabil 67:55-65, 1988.
more predominant as less of the “fast stuff” is balanced
by the previously extracted “slow stuff”. Therefore,
“What would a waveform with a lot of “fast stuff” look
like?” Well, we can anticipate that its peak latency
would be shorter (fast stuff occurs sooner in time, that is
why it is called fast stuff) than before, its negative spike
duration would be shorter because “fast stuff” starts and
stops sooner, and the total waveform duration should be
shortened for the same reason. “Would the onset laten-
cy be expected to shorten?” No, this is because the onset
of a waveform is typically where the waveform changes
from the baseline (zero potential) to the waveform’s peak
within a relatively short time-frame. In other words, this
is a rapid change dominated by “fast stuff”. “Did we al-
ter the “fast stuff” by elevating the low frequency filter?”
Of course not. Therefore, there is no observed change in
the waveform’s onset latency. In summary, elevating the
low frequency filter for any recorded waveform (EMG
motor unit action potential (MUAP), somatosensory
evoked potentials [SEPs], SNAP, CMAP, etc.) will result
in the following: 1) amplitude reduction, 2) peak latency
shortening if it is a time-locked signal, 3) negative spike
duration shortening, 4) total potential shortening, 5)
no change in the onset latency if the waveform is de-
rived from neural stimulation, and finally 6) a possible
increase in the number of phases. This last waveform
15
alteration arises because the waveform is now predomi-
nated by high frequencies (fast stuff) and high frequen-
cies have more phases per unit time compared to low
frequencies. In other words, elevating the low frequency
filter produces waveforms that are smaller, shorter, more
polyphasic, with faster rise times, and if a stimulation
is required to elicit it, no change in onset latency. Of
course, the degree of the waveform alteration depends
entirely upon the amount of low frequencies contained
in the waveform. The more low frequencies contained
in the waveform, the greater the reduction in magnitude
as well as the other changes noted above. For example, a
CMAP can be anticipated to have much more dramatic
changes to similar elevations in the low frequency filter
compared to a SNAP because the CMAP is a waveform
with a comparatively greater amount of slower changing
parameters (more slow stuff), i.e. longer total potential
duration and slower rise time (Figures 9 & 10). Spe-
cifically, a CMAP is a waveform comprised of signifi-
cantly more low subcomponent frequencies (slow stuff)
than subcomponent high frequencies (fast stuff). As a
result, a similar but much more dramatic manifestation
of waveform alterations can be anticipated for CMAPs
compared to SNAPs when the low frequency filter is
elevated, since comparatively more stuff (slow stuff) is
extracted from the waveform (Figure 10).
“What would you expect if we “lowered” the low fre-
quency filter from 300 Hz to 1 Hz?” Well, since we are
now not taking out, but putting back in, low frequen-
cies, the exact opposite changes to those previously
described would occur. The high frequencies (fast stuff)
would now be balanced by the added slow stuff (low
frequencies). The waveform would increase in ampli-
tude while the peak latency would increase, rise time
would increase, total potential duration would increase,
a reduction in the number of phases may occur, and
finally, as previously noted, there would continue to be
no change in onset latency.
High Frequency Filter. We can use a similar approach
to understand the operation of high frequency filters
as that used above for low frequency filters. For our
discussion, we will consider the high frequency filter to
function between 10,000 Hz and 500 Hz. We may then
ask, “What effects will lowering the high frequency filter
from 10,000 Hz to 500 Hz have on a SNAP?”
As before, we ask ourselves, “Are we taking something
out of, or putting something into, the waveform?”
16
Clearly, when we lower the high frequency filter from a
Figure 11. High Frequency Filter Reduction On SNAP. As we
lower the high frequency filter without altering the low frequency filter
for a recorded SNAP, we note there is a reduction in potential magni-
tude (see Table), but now the onset and peak latencies increase as does
the rise time and total potential duration. From: Dumitru D, Walsh
NE: Practical instrumentation and common sources of error. Am J
Phys Med Rehabil 67:55-65, 1988.
high to relatively lower level, we are removing those high
frequencies, or “fast stuff”, from the waveform (Figure
11). “If we take out the fast stuff, or remove anything
for that matter, do you think the waveform will get
bigger, or smaller?” Once again, if there is less of some-
thing compared to before, how can there possibly more
of it, i.e. how can the amplitude get bigger? This may
seem like common sense, and it is, but unfortunately,
the problem with common sense is that it is not very
common. Hence, the waveform must get smaller, i.e.
decline in amplitude. We then ask, “What did we take
out of the waveform, fast stuff, or slow stuff?” Since we
are lowering the high frequency filter and taking out fast
stuff, the remaining slow stuff must now influence the
waveform to a greater degree than previously as it is no
longer balanced by the extracted fast stuff. “What then,
would be anticipated for a waveform now predominated
by slow stuff?” Well, slow stuff by its very nature takes
longer to initiate, stays around longer, i.e. takes longer
to disappear. Therefore, the waveform’s onset would be
anticipated to take longer to manifest, the negative spike
duration would take longer to form and last longer, the
total potential duration should also increase as there is
less fast stuff to terminate the potential comparatively
sooner. Also, low frequencies have fewer phases per unit
time compared to high frequencies, and the waveform
under investigation may lose a phase if it contains a suf-
ficiently large number of subcomponent high frequen-
cies. In the final analysis then, a reduction in the high
frequency filter will produce a comparative waveform
with the following changes in its parameters: 1) reduced
amplitude, 2) prolonged onset latency, 3) prolonged
peak latency, 4) increase in negative spike duration, 5)
increase in the total potential duration, and 6) a pos-
sible reduction in the number of phases (Figure 11).
Of course, the degree of the previously noted altera-
tions is directly dependent upon the frequency content
of the waveform and the relative amounts of low and
high frequencies. Those waveforms with primarily high
frequencies (SFEMG potential, SNAPs, and EMG mo-
tor unit potentials) will demonstrate the greatest effects
while those waveforms with little in the way of high
frequency subcomponents (CMAPs) will demonstrate
very little changes.
Combined Effects. It is to be understood that simul-
taneously elevating the low frequency filter while lower-
ing the high frequency filter will act to take out both
the fast and slow stuff. As the two filters approach each
other, there are fewer and fewer remaining subcompo-
nent waveforms comprising the observed potential. This
combined effect may render the waveform very small in
amplitude and rather distorted. Consideration of the
above filter effects suggest a number of very plausible
clinical effects.
Clinical Consequences Of Filter Errors (False Posi-
tives/False Negatives). One could quite correctly argue:
“Look, if we set the filters correctly in the first place,
don’t mess with them, then everything should be OK.”
I certainly agree with this sentiment, but unfortunately
there are individual practices where circumstances may
conspire to diminish this laudable argument. Specifi-
cally, one may be required to travel from one hospital to
another, or from one office to another. This may neces-
sitate the practitioner using an instrument previously
used by other practitioners who may purposefully or
inadvertently alter the filter settings. Failure to recognize
“inappropriate” filter settings can result in errors. This is
of course, assuming the practitioner is aware of so called
“correct settings” in the first place.
As we have seen above, an instrument with too high a
low frequency filter setting can result in a SNAP with a
normal onset latency, a relatively normal peak latency
(remember we look for prolonged and not “shortened”
peak latencies), but a reduced amplitude. If one were to
encounter this finding clinically and not realize that the
low frequency filter setting was too high, an erroneous
conclusion of a primary sensory axonal neuropathy with
little in the way of demyelination could be entertained.
If this same filter setting were used for both upper and
lower limb sensory nerves, a generalized axonal sensory
neuropathy could be considered. If consideration were
not given to the appropriate filter settings, a potentially
expensive and fruitless workup could be initiated with
no success in achieving an accurate diagnosis.
Similarly, if the low frequency filter is set too high when
obtaining CMAPs, we would document CMAPs with
normal distal motor latencies, normal conduction ve-
locities, but reduced amplitudes. These findings could
suggest that a demyelinating peripheral neuropathy is
not present. If the filter setting for the SNAPs were op-
timal in this case, a pure axonal motor neuropathy may
be considered. Also, one would have to also consider
a neuromuscular junction disorder (reduced CMAP
amplitudes and normal SNAPs), a myopathy, or even
AMAN. These possibilities clearly demonstrate that a
lack of knowledge regarding filters can result in a mul-
titude of erroneous possibilities. Once again, an expen-
sive and fruitless assessment could be undertaken not to
mention the unnecessary inconvenience and potential
worry to the patient.
Also, as noted above, these filter effects also apply to
the needle EMG. If the low frequency filter is set too
high, the recorded MUAPs could display durations that
are too short (recall that the total potential duration of
waveforms is reduced). Specifically, the MUAP’s initial
positive onset and positive termination are predomi-
nated by relatively low frequencies. Removing the low
frequencies to a sufficient degree can reduce the duration
of the MUAP’s onset and termination thereby reduc-
ing the overall duration of the MUAP thus simulating
a myopathic waveform. This is specifically why one
should never attempt to define if a myopathy is present
by performing a qualitative needle EMG with either a
monopolar or concentric electrode as the low frequency
filter for most routine EMG programs is purposefully
set at between 10 Hz and 20 Hz specifically to reduce
the significant baseline deviations produced by needle
insertions. This low frequency filter setting (20 Hz)
is great for producing a stable baseline during needle
insertions, but terrible for defining the true MUAP
17
duration. As a test, try lowering the low frequency filter
during routine needle insertions to 3 Hz (necessary for
accurate quantitative MUAP duration measurements). I
think you’ll be amazed at just how hard it is to now as-
sess insertional activity. Additionally, reducing the high
frequency filter can produce a reduction in the overall
amplitude of the MUAP as the main spike component
is preferentially comprised of primarily high frequencies
that accounts for most of the MUAP amplitude. Also,
reducing the high frequencies from a MUAP may also
produce a slight prolongation of the MUAP duration.
It is conceivable that normal amplitude, normal dura-
tion MUAPs could be transformed into longer duration
smaller amplitude waveforms resulting in some diagnos-
tic confusion. Clearly filter parameters for qualitative
vs quantitative needle EMG are set for specific purposes
and using them interchangeably or without regard for
clinical consequences will, without doubt, lead to er-
roneous conclusions.
If one performs SEPs extreme caution must be exercised
with respect to the low frequency filter in particular.
The SEP is comprised primarily of low frequencies.
Elevating the low frequency filter much above 10 Hz (a
relatively low setting) can produce a dramatic reduction
in the SEP’s amplitude. Obviously, a marked reduction
in the SEP amplitude can suggest a number of patholog-
ic entities at numerous points along the neuraxis. The
high frequency filter for SEPs can be relatively low (500
Hz) and produce little in the way of adverse effects on
the waveform.
The above effects are purposely kept somewhat simplistic
primarily for discussion purposes. All of the above ef-
fects are described for patients without pathology affect-
ing the neuromuscular system. If pathology is present,
then multiple effects can be anticipated. For example, if
a patient has a mild carpal tunnel syndrome with some
degree of demyelination but not much axonal loss, and
the low frequency filter is set too high, the normative
data for peak latencies is no longer valid as the peak
latencies are erroneously short secondary to the filter ef-
fects. In this instance, a mild disease state can be missed
and we arrive at a false negative study when indeed the
patient has pathology. Similarly, a patient with neuro-
genic motor units (long duration) may display normal
duration motor units if the low frequency filter is set
too high, again producing a false negative study. On
the other hand, if the high frequency filter is set too low,
borderline short duration MUAPs may be erroneously
18
Table 1
elongated into the lower limit of normal range, again
producing a false negative study. As noted above, the
degree to which these effects are noticed depends en-
tirely upon the specific frequency content of individual
waveforms which may vary considerably including
MUAPs.
Notch Filter. Most if not all current electrodiagnostic
medicine instruments permit the practitioner to employ
a 60 Hz (50 Hz) notch filter. This is a specific filter
designed to preferentially eliminate the above noted line
frequency noise. Personally, I find this filter can be quite
helpful and do not hesitate to use it for any study. As
long as the practitioner is aware that some waveform dis-
tortion may occur, the filter can at times permit a study
to be performed when otherwise the interference would
preclude any assessment. This may be particularly true
when examining patients in the intensive care unit.
Recommended Filter Settings. The literature is replete
with filter recommendations for various studies. In
this document a table of such filter settings is provided
(Table 1). There is nothing particularly special about
this table and is effectively derived empirically by record-
ing a waveform with a rather low frequency setting and
a particularly high, high frequency setting. The low
frequency filter is slowly elevated until the waveform
shows demonstrable changes and then it is lowered
to the previous setting. Similarly, the high frequency
filter is sequentially lowered until the waveform shows
some degree of alteration and the high frequency filter
is returned to the next highest value. This same process
is repeated for each type of study (motor NCV, Sensory
NCV, needle EMG, SEP, and SFEMG). As a result, a
particular bandwidth is established for each study.
The most important aspect of filter settings is to use
those bandwidths for particular studies from which the
normative data is taken. If one uses different filter set-
tings from those utilized to derive a reference data base,
erroneous results will ensue. As has been demonstrated
above, different filter settings can produce significant
alterations in those waveform parameters used to diag-
nose pathology.
How To Answer Filter Test Questions: Two typical test
questions one may encounter on various electrophysi-
ologic specialty boards are listed below. Consideration
of the above “fast stuff” and “slow stuff” explanation
should permit the practitioner to now answer these types
of questions with ease.
1. A SNAP is recorded antidromically from the third
digit. The low frequency filter is elevated from 10
Hz to 100 Hz. Which of the following changes in
the ensuing waveform can be anticipated?
A. Amplitude increase, unchanged onset latency,
reduced negative spike duration
B. Amplitude increase, reduced peak latency, re-
duced negative spike duration
C. Amplitude decrease, unchanged onset latency,
possible increase in phases
D. Amplitude decrease, unchanged onset latency,
increased peak latency
Answer: C
Recall, that if the low frequency filter is increased, we
are taking something away from the waveform. If we re-
duce the waveform’s content of anything, can it possibly
get bigger? Of course not. Therefore, any answer that
suggests an increase in waveform magnitude is simply
wrong. As a result, possible answers A and B are wrong
irrespective of whatever else is contained in the supposed
answer. This question then boils down to an effective
true/false response, it is either C or D. Since we elevated
the low frequency filter, we extracted slow stuff from the
waveform leaving mostly fast stuff. Fast stuff happens
sooner in time. Because the waveform’s onset latency
is fast stuff, and we didn’t fool with fast stuff, the onset
latency should remain unchanged. We know that fast
stuff has more phases than slow stuff per unit time, so a
possible increase in phases is correct. Also, we know that
fast stuff happens sooner in time, so one would antici-
pate a reduction and not increase in peak latency. The
only correct response based on an analysis of fast stuff
and slow stuff is C.
2. A SNAP is recorded antidromically from the third
digit. The high frequency filter is lowered from
2,000 Hz to 300 Hz. Which of the following
changes in the ensuing waveform can be anticipated?
A. Amplitude increase, prolongation of the onset la-
tency, prolongation of the total potential duration
B. Amplitude increase, possible reduction in the
number of phases, prolongation of the peak
latency
C. Amplitude reduction, potential onset prolonga-
tion, peak latency prolongation
D. Amplitude reduction, shortening of the total
potential duration
Answer: C
Again, we first ask ourselves if we took stuff out or put
stuff in. By lowering the high frequency filter we are
removing high frequencies and, therefore, taking out fast
stuff. So once again, if we take something away from
the waveform, can it possibly get bigger? Of course not.
So, possible answers A and B are immediately eliminated
no matter what else is in the response. This leaves C or
D and the question is once again effectively a true/false
question. If we take out fast stuff, then slow stuff is left
and predominates so as to influence the resulting wave-
form. Since slow stuff takes longer to happen, we would
anticipate a prolongation of the onset latency and peak
latency as well as an increase in the total potential dura-
tion. Only response “C” fulfills all of the criteria noted
previously for lowering the high frequency filter.
As a cautionary note regarding test questions, care must
be taken to read the question carefully. If the ques-
tion stem states that a high frequency filter is elevated
and not lowered, then a reverse process of thinking
must be employed. In this instance, we are not taking
something out, but putting something in. The ensuing
waveform’s amplitude must then increase. If we put fast
stuff in, then it balances the slow stuff already present
and the waveform takes on more of the high frequency
characteristics: shortened onset latency, shortened peak
latency, etc. Similarly, lowering the low frequency filters
should be thought of as adding back in slow stuff with
appropriate consequences.
As can be directly observed from the previous discus-
sion, filters can be relatively easy to conceptualize, and
apply both clinically and for examination purposes. No
19
complex mathematics, guess work, or magic is required.
Just a simple understanding of waveform subcomponent
frequencies consisting of low frequencies or so called
“slow stuff” and high frequencies or so called “fast stuff”.
If the concepts described in the above section on filters
are worked through slowly and completely, they will
never let you down. Also, you can now finally explain
filters to your friends and relatives.
STIMULATOR
The stimulator typically utilized in electrodiagnostic
medicine evaluations can either be a constant current, or
constant voltage stimulator. Either can be used equally
well for most if not all evaluations. A constant current
stimulator implies that irrespective of the impedance
between the stimulator and patient changing over the
course of an evaluation, the current is maintained at the
same level. That is, if the impedance changes, the instru-
ment automatically alters the voltage delivered so as to
maintain the same current delivered for each stimulus.
Similarly, if a constant voltage stimulator is used and the
skin/stimulator interface impedance changes under the
cathode/anode, then the current is also altered com-
mensurately so as to maintain the same voltage for each
stimulus delivered.
We can appreciate the above concepts by simply apply-
ing Ohm’s law (E = IZ). Let us suppose we are stimulat-
ing a nerve using a constant voltage stimulator and set
the voltage at 100 V and our skin resistance is 5,000
ohms. The amount of current delivered is 20 milliamps
(100 V = 20 milliamps X 5,000 ohms). If the skin im-
pedance suddenly increases to 10,000 ohms, the instru-
ment will deliver 10 milliamps to maintain the stimula-
tor’s output at 100 V (100 V = 10 milliamps X 10, 000
ohms). On the other hand, if we are utilizing a constant
current stimulator and applying 20 milliamps through
a skin impedance of 5,000 ohms, the stimulator will
deliver 100 volts (see right). If the skin impedance now
increases to 10,000 ohms, the stimulator will output 200
volts to maintain a current flow of 20 milliamps.
The typical electrophysiologic instrument stimulator
consists of a cathode and anode. The cathode is the
stimulator’s negative pole (attracts cations or positive
ions) while the anode is the stimulator’s positive pole
(attracts anions or negative ions). When we activate the
stimulator, charge will build up on the respective poles
of our stimulator and induce a current flow between the
two poles. An intervening nerve will be activated and
20
subsequently induced to propagate an action potential
if the nerve’s threshold voltage value is achieved. The
instrument’s sweep is simultaneously triggered and we
thereby observe a time-locked response from the nerve
or muscle induced by the stimulator. If the stimulator
location and recording electrodes do not change between
stimuli, then the response will occur at the same latency
following each stimulation thereby defining the above
concept of an induced response being “time-locked”
with the stimulus.
There are two issues explored in this document with
respect to the electrophysiologic instrument’s stimulator.
The first is “anodal block” while the second is “stimulus
artifact”. Both of these concepts are important in elec-
trodiagnostic medicine and in my opinion, often misun-
derstood by novice as well as seasoned practitioners.
Anodal Block
Anodal block DOES NOT HAPPEN IN ROUTINE
ELECTROPHYSIOLOGIC STUDIES! PERIOD.
END OF STORY. In fact, anodal stimulation rou-
tinely happens and can result in erroneous diagnoses
if the practitioner is not aware of this fact.1,2 An even
cursory search of the cardiology literature quickly reveals
discussions of cathodal vs anodal stimulating currents.
Further, direct muscle stimulation by both the cathode
and anode has been known since the turn of the previ-
ous century. In order to better appreciate the above
categorical statement, we need to further delve into how
a stimulator produces neural activation by the cathode
and then appreciate what is going on about the anode.
If we locate a cathode over a nerve, initially with no
current flowing, the nerve is in the resting state with the
intracellular region approximately -90 mV with respect
to the extracellular space (Figure 12A). If we begin to
apply a current to our stimulator, the cathode begins to
acquire a negative charge (Figure 12B). The sodium
voltage-gated channels imbedded in the axon’s axolem-
ma begins to “sense” this voltage shift. If an increasing
amount of current is applied, the extracellular region
about the cathode continues to increase in its negativ-
ity. At some point the extracellular space will acquire a
sufficiently large negative charge and the transmembrane
sodium voltage sensor will detect an extracellular voltage
that is now relatively more negative than the intracellular
voltage. In other words, the intracellular voltage (-90
mV) is in effect less negative, or relatively more positive,
Figure 12. Cathodal Stimulation. A) A cathode (negative pole) is
located over a nerve. In the resting state the nerve’s intracellular region
is negative with respect to the extra-cellular environment. B) Applica-
tion of current to the stimulator begins to generate a negative electric
field about the cathode. C) An increase in the negative field at some
point cause the voltage-gated sodium channels to open producing a local
depolarization sufficient to induce sodium activation in the tissue sur-
rounding the cathode thereby generating a propagating action potential.
than the extracellular region (Figure 12C ). At some
threshold voltage value, the transmembrane sodium
voltage-gated protein will respond through sodium
activation because there has, in effect been a transmem-
brane voltage difference relatively more positive inside
compared to outside in effect acting to initiate sodium
activation. The net result is the creation of a negative
sink beneath the cathode with a traveling wave of depo-
larization now initiated because the neural tissue im-
mediately adjacent to the cathode is induced to undergo
sodium channel induced depolarization through local
circuit current flows. The sodium channels beneath
the cathode are “tricked” into opening because of the
cathode generating a sufficiently large negative voltage
such that compared to the inside of the cell, the intra-
cellular voltage (-90 mV) is actually relatively positive
compared to the outside of the cell. The net result is
cathodal depolarization with an ensuing propagating
action potential.
Now, let us consider what is happening under the anode
(Figure 13A). As previously noted, the anode will take
on a positive charge once the stimulator is activated (Fig-
ure 13B). As the current to the stimulator is progres-
Figure 13. Anodal Stimulation. A) An anode (positive pole) is
positioned over a nerve. B) Increasing the current to the stimulator
produces a positive electric field about the anode. C) Further increas-
ing the current generates an increasingly intense positive field with a
relative negative zone adjacent to the anode in effect creating a sur-
rounding “virtual” cathode. D) When the positively generated anodal
field is sufficiently large and creates an associated accompanying virtual
cathode large enough, it will act to depolarize the neural tissue similar
to the previously describe “real” cathode (see Figure 12). In effect, a
peri-anodal zone of depolarization is generated.
sively increased, the anode will commensurately increase
in its positive charge (Figure 13B). As the positive
charge increases, the relative negative transmembrane
voltage across the membrane will also increase tending
to result in a relative hyperpolarization of the nerve im-
mediately about the anode. Clearly, as far as the trans-
membrane voltage-gated sodium channels are concerned
beneath the anode, the detected transmembrane voltage
has moved further from the depolarization threshold, i.e.
hyperpolarization. However, if we continue to increase
the current intensity to the anode, its associated posi-
tive electric field is now much more positive than the
extracellular tissue surrounding the anode. That is, the
extracellular tissue’s previously positive voltage adjacent
to the anode is now becoming relatively more negative,
i.e. less and less positive (Figure 13C). In effect, the
anode is acting to generate a surrounding peri-anodal
21
“virtual cathode”. A further increase in the anode’s
field strength will subsequently increase the intensity
of the virtual cathode zone. This in turn will produce
a relative peri-anodal negative region sufficient enough
to exceed that of the intracellular negativity detected
by the transmembrane voltage-gated sodium channels
at this peri-electrode location, which can again “trick”
those channels into sensing a relative shift in the trans-
membrane voltage. That is, the extracellular peri-anodal
region is now sufficiently negative enough, so that the
intracellular region is relatively less negative and reaches
the transmembrane threshold value required for sodium
activation induction and subsequent action potential
propagation (Figure 13D). Once this state is estab-
lished, a propagating action potential is then initiated in
the peri-anodal region, i.e. not anodal block but rather,
anodal stimulation.
The above explanation can be quite easily demonstrated
clinically. Let us locate an active electrode on the APB
with the reference electrode located on the APB’s mus-
culotendinous junction. We then position a disc elec-
trode over the median nerve at the wrist with a second
disc electrode 15 cm more proximal and on the dorsal
aspect of the forearm. The disc located over the median
nerve is plugged into the cathode port for the stimulator
and the second disc is plugged into the stimulator’s
anode port. The current is sequentially increased until a
just supramaximal CMAP is obtained (Figure 14A).
Then, the two stimulating electrode leads are switched so
that the disc over the median nerve is plugged into the
anode port while the forearm disc is plugged into the
cathode port. The current is again increased until a just
supramaximal CMAP is obtained (Figure 14B). The
electrodes’ locations are specifically chosen to ensure that
only the electrode over the nerve can activate it while its
“partner” is too far away and no where near the nerve to
participate in the nerve’s activation (proximal dorsal
forearm). In this example, we first notice that both the
isolated cathode and anode are quite capable of generat-
ing a supramaximal CMAP from the APB. We do note
however, that the amount of current required by the
anode to achieve a similar CMAP amplitude as that of
the cathode is approximately doubled. Further, the
CMAP arising from the anode is shorter in latency than
that derived with cathodal stimulation. Therefore, we
document the following: 1) no anodal block but anodal
stimulation, 2) the anodal CMAP onset is shorter in
latency confirming the site of depolarization is further
from the activating electrode (anode) than the same
22
Figure 14. Cathode VS Anode Responses CMAP. A) A median
nerve CMAP for the APB is generated with a cathode over the median
nerve at the wrist. The APB onset latency is 3.7 ms with an amplitude
of 7.2 mV requiring a current delivery of 33 mA. B) Transposing the
cathode and anode once again produces a median nerve CMAP with
an amplitude of 7.2 mV but now with an onset latency of 3.4 ms and
requiring 66 mA to produce this comparable CMAP.
stimulation location for the cathode (displaced peri-
electrode anodal depolarization compared to the relative
coincident site of the cathode/nerve site for neural ac-
tivation), and 3) more current is required for the anode
(a need to create a sufficiently large peri-anodal rela-
tive negative field (virtual cathode) for transmembrane
depolarization). Certainly, we know that increasing the
current for the cathode at some point will also result
in a shortening of the CMAP’s onset latency (current
spread). This confirms that the detected CMAP for the
anode is generated by a current spread away from the
hyperpolarization zone (closer to the recording elec-
trode) thus inducing neural depolarization, and that
there is no hyperpolarization induced anodal block.
Repeating the above demonstration of anodal stimula-
tion for a mixed nerve response (median/ulnar 8 cm
mid-palm study) reveals a number of interesting find-
ings. Let us first perform a routine 8 cm mid-palm
study locating active recording electrodes over the
median and ulnar nerves at the wrist while activating
each corresponding nerve in the mid-palm (Figure 15
A&B). We detect the anticipated mixed nerve responses
with comparable peak latencies (median peak latency:
1.9 ms; ulnar peak latency: 1.8 ms). We now utilize the
identical setup for the median nerve with the exception
of transposing the anode and cathode (cathode is now
several centimeters more distal to the unchanged wrist
recording electrodes). We easily obtain a response
identical to that previously recorded for median nerve
stimulation but with a peak latency of 2.2 ms (Figure 15C).

Figure 15. Cathode VS Anode Responses Mixed Nerve. A) A mixed
nerve response from stimulating the median nerve in the mid-palm
while recording 8 cm more proximally over the median nerve at the
wrist (peak: 1.9 ms). B) A mixed nerve response from activating the ul-
nar nerve in the mid-palm while recording 8 cm more proximally from
the ulnar nerve (peak 1.8 ms). C) A mixed median nerve response from
the wrist when the anode and cathode are transposed at the mid-palm
stimulation site (peak: 2.2 ms). D) Same response obtained in C but
with more current applied. E) A further increase in the current now
generates a biphasic waveform with an initial peak latency of 1.7 ms.
F) Further increasing the current generates a similar biphasic response
with an even shorter initial peak latency (1.5 ms) and reduction in the
second peak’s amplitude.
The response is expectedly delayed secondary to the
increased cathodal distance away from the recording
electrode. We then increase the delivered current from
7 mA to 28 mA with no change in the waveform or
latency (Figure 15D). A further increase in the current
to 37 mA now generates a biphasic response with the
second peak identical to that obtained previously, but
with a much shorter and somewhat smaller initial peak
with a latency of 1.7 ms (Figure 15E). Increasing the
delivered current to 65 mA subsequently generates a
similar biphasic response with the initial peak now hav-
ing a latency of 1.5 ms and an amplitude that is greater
than the second peak (Figure 15F). We can see that a
continued increase in current following cathode/anode
transposition results in the eventual detection of a bipha-
sic response. It is clear that as the current is increased
a peri-anodal “virtual” cathode zone is generated that
depolarizes the nerve closer to the recording electrode as
demonstrated by the initial peak’s shorter latency. Fur-
ther, since neural propagation is induced to propagate in
both directions, a portion of the nerve begins to collide
with action potentials generated by the cathode resulting
in a progressive reduction of the second phase’s recorded
magnitude. Within this demonstration and the given
intensity of the current delivered (within subject toler-
ances), the anode never completely depolarizes the nerve
since complete cathodal blockade was not achieved, i.e.
the second peak never disappeared.
Clinical Relevance To Anodal Stimulation. There
are several general points to be made from the above
issues discussed regarding anodal stimulation. First,
it is certainly rather easy to activate the nerve with the
anode. Specifically, anodal stimulation is quite simple to
achieve, but anodal block does not occur. Further, from
the mixed nerve study (Figure 15), it appears the anode
is not capable of fully depolarizing all of the nerve fibers
even with rather intense stimulation intensities at least
within the tolerance of the subject examined.
Motor Studies. The demonstration above combined
with other studies clearly suggests there is no such thing
as anodal block from a clinical perspective. Therefore,
when performing F-wave studies it is not necessary to
relocate the position of the anode by locating it distal to
the cathode while keeping the cathode in the original
location. A previous investigation did not reveal any
statistically significant difference in F-wave latencies
irrespective of the anode’s location provided the cathode
was in the proper location at the wrist.1 If the current
is sufficient enough, it appears theoretically possible for
some F-waves to be generated by the anode with possi-
bly shorter latencies than that for the cathode, however,
as noted, no difference was detected. This may be be-
cause F-waves are generated solely by the cathode prior
to the current intensity required for reaching a sufficient
strength at the anode to “kick in” and initiate anodal
stimulation. Or, the natural statistical variation in F-
wave latencies may be greater than that generated from
a slightly (few centimeters) more proximally located
“virtual” cathode arising from the anode.
Of course, inadvertently reversing the anode/cathode lo-
cation will result in a prolonged latency for any response
because the cathode is located further from the recording
electrode. Erroneous distal motor latencies can certainly
lead to false positive results simply from this alteration in
distance. It is unlikely that any anodal stimulation will
occur in this situation when just supramaximal
currents are utilized. That is, the anode current intensity
23
appears to be insufficient to cause depolarization when
the cathode has already achieved its supramaximal
state (Figure 14).
Sensory/Mixed Nerve Studies. Stimulating the mixed
nerve in this presentation and examinations of pure
sensory nerves in previous studies have clearly shown
the biphasic response.1 The most important aspect
of anodal stimulation in this context is inadvertently
activating the sensory or mixed nerve with the anode.
It is clearly shown that if the anode and cathode are
physically reversed in location, a comparatively longer
cathodal induced response will be observed first (Figure
15A & C). The reason for the prolongation is obvious
as the cathode is further from the recording electrode.
Clearly, this inadvertent prolongation can produce a false
positive result as the nerve is normal, but an unrealized
longer distance than that measured has been introduced,
thereby prolonging the latency yielding a possible false
positive result.
On the other hand, if one inadvertently reverses the an-
ode and cathode, but starts off with rather large current
intensities, or the anode is on the nerve while the cath-
ode is physically off the nerve, an erroneously short and
possibly biphasic response can result (Figure 15 E &
F). In this instance, a comparatively shortened response
can result in a false negative study. That is, a slightly
prolonged response will be shortened possibly into the
normal range because of the shortened distance between
the recording electrode and displaced anodal stimula-
tion. Although it doesn’t always happen, the detection
of a biphasic sensory or mixed nerve response should
alert the practitioner to a possible artifact arising from
anodal stimulation.
Stimulation Artifact
Arguably, one of the most commonly encountered and
annoying issues relating to neural stimulation, is that
of stimulation/stimulus artifact. The stimulus artifact
is essentially always present as a relatively large baseline
deflection originating commensurate the stimulus onset.
It takes a variable amount of time for it to settle back to
baseline. The primary issue relates to how long it takes
to achieve baseline return with respect to the temporal
occurrence of the waveform of interest. If there is a
relatively long time before the waveform appears with
respect to the stimulus artifact, then the stimulus artifact
is essentially a non-issue. However, when the waveform
of interest coincides in time with the stimulus artifact,
24
we can encounter significant diagnostic dilemmas.
Specifically, if we do not have complete confidence in
the waveform’s onset or latency, erroneous diagnostic
conclusions are inevitable.
When the stimulator is activated, a necessarily large
voltage difference is produced between the cathode and
anode so as to in turn create a sufficiently large voltage
difference across the neural membrane conducive to
action potential generation (see above). The stimula-
tor’s voltage difference extends virtually instantaneously
throughout the body because our tissue fluids are such
good conductors. Because the stimulator’s voltage dif-
ference traverses the body’s fluid medium and does not
depend on neural conduction for its presentation, it is
recorded immediately upon stimulus induction by both
the active and reference electrodes. Given the above pre-
sentation of differential amplification, a reasonable ques-
tion would be: “OK, if differential amplification is such
a good deal, why doesn’t it get rid of the stimulus arti-
fact?” In fact, differential amplification does an excellent
job with stimulus artifact, it is just that the generated
artifact is so large that as previously noted, even a small
difference at the electrodes is amplified. Remember,
if a noise generator is in close proximity to the record-
ing electrodes, there may be a small difference of that
noise as it presents to both electrodes. This situation
can then lead to an amplification of the small difference
and be sufficient enough to obscure our intended signal,
if that signal is not particularly large. In other words,
because we see the stimulus artifact, we now know from
the above discussion that all things being equal, there
continues to be a small voltage difference detected at
each electrode as generated by the stimulator. As noted
above, the closer the noise generator is to the record-
ing electrodes, the more important small differences in
location become. Even though the recording electrodes
are only a few centimeters apart, the voltage difference
arises since the stimulator is typically a few centimeters
away from the recording electrodes. So, we will need ad-
ditional strategies to preclude the stimulus artifact from
adversely affecting our intended waveform of interest.
Problem Solving Stimulus Artifact. Since there is no
hope in separating a sufficient stimulation intensity from
its accompanying stimulus artifact, we need to think
about what else can be implemented to minimize its
effect on our recorded SNAP or CMAP (Table 2).
Obviously, we do not want to help the stimulus artifact
reach our recording electrodes. Therefore, we need to
Table 1
consider what is going on with our electrode paste. The
electrode paste should only be sparingly applied to our
cathode and anode so that it is specifically located only
as an interface between the stimulating electrodes and
the underlying skin. If we slop on the electrode paste
and it forms a pathway over the skin between the cath-
ode and anode, then a significant portion of the current
will flow across the skin between the cathode/anode
instead of into the tissues. This will cause us to input
more and more current in order to depolarize the nerve,
which in turn increases the magnitude of the stimulus
artifact. So, the concept of a just supramaximal current
input (about 10% greater than when a maximal response
if first observed) is valid to not only preclude current
spread away from the cathode site, but also control the
magnitude of the stimulus artifact. Further, we do not
want an aberrant current pathway across the skin be-
tween the stimulator and recording electrodes. If there
is excess paste, perspiration, body lotion/makeup be-
tween the cathode/anode and recording electrodes, the
current will find the path of least resistance and travel
across the skin straight to the recording electrodes
producing a horrible stimulus induced artifact. It is also
a good idea to clean the area to be stimulated if there is a
suggestion of dirt, debris, increased callous, etc. All of
these components can lead to an increase in skin resis-
tance to current flow requiring an increase in the
amount of current necessary for neural activation,
accompanied by a commensurate increase in the stimu-
lus artifact. At times, a needle cathode may be employed
to place the cathode beneath the stratum corneum to
assist in optimal current delivery. Touching the patient
in-between the stimulus site and recording electrodes
with a bit of paste on your hand can help to significantly
Figure 16. Anode Rotation: Clinical. Figure: An antidromic me-
dian nerve SNAP is generated as depicted. A) A waveform is generated
but the stimulus artifact obscures the waveform’s exact onset and true
magnitude. B-D) Sequentially rotating the anode in a clockwise direc-
tion eventually permits a significant reduction in the stimulus artifact
magnitude permitting a more accurate depiction of the waveforms onset
and amplitude. From: Dumitru D, Amato AA, Zwarts MJ: Electrodi-
agnostic Medicine, 2nd ed. Philadelphia, Hanley & Belfus, 2002.
reduce the stimulus artifact. I view this as sharing your
body’s volume conductor with that of the patient es-
sentially doubling the volume into which the stimulus
artifact can diffuse and subsequently dilute its intensity.
This is a very effective strategy and can act in two man-
ners: 1) reduce stimulus artifact, and 2) as noted above,
assist with grounding the patient and differential am-
plification. Finally, rotating the anode about a constant
cathode location can also work wonders for reducing
or even eliminating stimulus artifact (Figure 16). This
final concept requires a bit more explanation.
We can return to our understanding of differential
amplification to get a better handle on how rotating the
anode can help reduce stimulus artifact. Since we are re-
cording a stimulus artifact, we know that our recording
electrodes are detecting some difference between them.
If it were possible to alter the stimulus artifact in some
way so as to minimize the difference in voltage recorded
between our recording electrodes, then there might be
a chance of reducing it. If we rotate our anode about a
stationary cathode, we are in effect altering the voltage
distribution in the body away from the cathode/anode
and at the recording electrode. As long as our cathode
remains over the nerve, there should be little difference
where the anode is located as long as it is not positioned
between the cathode and recording electrode (see above
discussion regarding anodal stimulation). Therefore,
when we do rotate the anode about the cathode, the
25
Figure 17. Anode Rotation: Theoretical. A) A hypothetical voltage
distribution is depicted as originating from the cathode and anode
projected toward the recording electrodes (E-1 and E-2). The voltage
difference between the recording electrodes is amplified. B) Rotating the
anode about a stationary cathode causes a redistribution of the stimulus
artifact’s voltage at the two recording electrodes so as to minimize the
difference detected between them. The net result is an improvement in
differential amplification and a commensurate stimulus artifact reduc-
tion. From: Dumitru D, Amato AA, Zwarts MJ: Electrodiagnostic
Medicine, 2nd ed. Philadelphia, Hanley & Belfus, 2002.
voltage distribution about the recording electrodes ap-
pears to change (Figure 17). Rotating the anode in
small increments empirically in either direction results
in producing a more uniform stimulus artifact voltage
distribution in the proximity of both the active and
reference electrodes. The net effect is to maximize differ-
ential amplification so that the stimulus artifact becomes
a more common mode signal with a resultant reduction
in the recorded artifact, and less interference with our
desired signal of interest. When the anode is rotated in
one direction or the other, it is possible in some cases to
actually completely eliminate the stimulus artifact and
additional minute rotations in either direction results in
a stimulus artifact with opposite polarity. Of note, one
cannot predict in which direction the rotation must oc-
cur for signal optimization and as a result, it is primarily
a trial and error approach.
A relatively new approach to handling the stimulus
artifact can be rather remarkable (Figure 18). This is a
proprietary methodology offered by one of the equip-
ment manufacturers referred to as a “signal enhancer”.
The exact method employed is not entirely clear to me,
but uses in part a low frequency filtration on a so called
“moving average” of the initially acquired response. A
26
Figure 18. Subtraction Of Stimulus Artifact. A) A radial nerve
SNAP is recorded with a large positive deflection rising toward the
SNAP. B) Application of the “signal enhancer” method improves the
practitioner’s ability to record the waveform’s onset and magnitude.
waveshape approximating the stimulus artifact is created
which in turn is subtracted from the original response
containing the stimulus artifact and desired waveform.
Let’s suppose you obtain a response but the stimulus
artifact is rather large. Because the stimulus artifact and
response are coincident with each other, the intended
response’s true latency and amplitude are difficult to dis-
cern. Specifically, if the stimulus artifact is very positive
and rising (moving in the negative direction) while the
response is occurring, the biologic waveform’s latency is
artificially delayed and amplitude somewhat magnified.
On the other hand, if the stimulus artifact is descend-
ing from a large negative voltage while the biologic
waveform is manifesting, the amplitude is somewhat
diminished and its corresponding latency shortened.
The computer using the “signal enhancer” technique
effectively extracts out the waveform from the envelop-
ing stimulus artifact. The ensuing potential now more
accurately reflects its true magnitude and latency. In my
opinion, I find this new method very promising, but I
tend to use it only after all of the above noted methods
have been implemented. This is because no studies have
been published verifying exactly what is going on, and
what effects it can potentially have on the accuracy of
waveform recordings. Because some type of low fre-
quency filtration effect is apparently being utilized, we
know from the above discussions that the waveform will
be affected to some degree. The question, of course,
is to what degree is the response altered? The stimulus
artifact certainly looks like it is comprised of mostly very
slowly changing frequencies which are likely way below
those contained in the waveform. If this is indeed the
case, then the SNAP is most probably not going to be
affected significantly. Although the CMAP is predomi-
nated by low frequencies, it typically occurs after the
stimulus artifact has returned to baseline and, therefore,
this particular technique most likely will not be imple-
mented anyway. Until actual studies are independently
carried out to assess the possible statistical consequences
of this technique, caution should be exercised in its use
particularly in responses arising from pathologically
affected nerves.
ANALOG-TO-DIGITAL CONVERSION (ADC)
As can be appreciated above, the signal of interest has
been detected in the body by our electrodes following
electrical activation through a stimulator (e.g. SNAP,
CMAP, etc.), or secondary to arising in the body ei-
ther spontaneously or through voluntary efforts (e.g.
EMG signal). This signal must then be amplified and
extracted from the surrounding environmental noise
(differential amplification), and subsequently processed
through filtration (high frequency and low frequency
filters) to help isolate just the waveform of interest. The
signal is then forwarded to a speaker so that it can be
listened to if relevant (EMG signals), and obviously
needs to be displayed for our visual assessment. Up to
this point, the signal of interest is a continuous varying
voltage over time and is referred to as an “analog signal”.
In order for us to observe and analyze the signal, it must
be held stationary on the screen or “frozen” in time if
you will. Once the fleeting and time-varying signal is
held in position on the screen, we can then leisurely
measure it various parameters of interest (onset/peak
latency, amplitude, phases, etc.), and/or manipulate the
waveform in whatever manner deemed clinically relevant
(alter the gain, change the time base, etc.). However,
this “freezing” of the waveform on our screen can only
be accomplished if the instrument can convert the ana-
log signal of interest into a faithful digital reproduction.
This conversion process is referred to as analog-to-digital
conversion (ADC). In other words, a continuous vary-
ing voltage over time (analog signal) is converted into a
discrete digital number proportional to the magnitude
of the voltage in the original analog signal for its corre-
sponding time period (digital signal).
Simply, there are primarily only two issues we need
concern ourselves regarding ADC: 1) resolution and
2) sampling frequency. The resolution has to do with
how accurately the instrument can reproduce the ana-
log waveform’s voltage for any given interval, while the
sampling frequency reflects how accurately the instru-
ment can reproduce waveform changes between one
interval and the next. With respect to resolution, we
are referring to how many vertical voltage intervals there
are for our instrument, i.e. how many voltage steps can
be produced for the vertical height of our potential.
The ADC’s resolution is given in “bits” and is expressed
digitally as a power of 2. Specifically, an ADC with a
resolution of 8 bits equates to 256 vertical levels of po-
tential voltage (vertical) measurement (2 to the 8th power
= 256 discrete levels). If we have a waveform that has a
peak amplitude that falls between two vertical measur-
ing points, we will obviously not get an accurate mea-
surement as the peak falls between two of our points of
measurement. The net result is a waveform with an am-
plitude that is too small. As a result, the more vertical
measuring points we have, the more accurate our digital
signal will be with respect to the original analog signal.
The good news is that all modern day instruments have
more than enough ADC resolving power and we need
not concern ourselves with this issue.
The second issue of sampling frequency is also important
and can be thought of as the instrument’s number of
discrete horizontal or time points of resolution. Quite
simply, we want to sample our analog or continu-
ously changing waveform over time with a sufficiently
small enough time interval to ensure that no waveform
changes occur between our sampling intervals, i.e. the
sampling frequency. Clearly, if a waveform alteration
(e.g. phase change) occurs between our sampling points
then it will be missed. This could lead to a waveform
with fewer phases or otherwise instrument induced
configuration changes, thereby no longer accurately
representing the analog waveform’s shape. For example,
if we picture a sine wave with a peak and trough every
1 ms, then the waveform is undergoing an important
change that can be represented by a frequency of 1000
Hz. It is generally accepted that a minimally faithful
27
reproduction of the waveform can be accomplished only
if we employ a sampling rate for our ADC of twice the
highest frequency contained in our analog signal. In this
example, the ADC needs to provide a minimal sampling
frequency (number of samples per time) of 2,000 Hz (a
time sample every 0.5 ms). This sampling time inter-
val is referred to as the Nyquist frequency. As with the
vertical resolution, we now have instruments with ADC
converters that have no problems sampling the biologic
waves of interest with sufficient time intervals. However,
we can alter the instrument’s time base (changing the
sweep speed) to such an extent that the observed signal
begins to reveal a discrete “stair-step” appearance because
we are reducing the amount of time given to any inter-
val. Clearly, when we begin to see this effect, the time
base should be altered so as to eliminate this observation
because unintended waveform distortions can occur.

Averaging: Signal-to-Noise Ratio. Although ADC

is clearly an important aspect of the instrument, the
primary reason for briefly discussing it in this document
is to provide a background for the concept of “averag-
ing”. As noted above, once a signal has been digitized, it
is now a relatively trivial matter to investigate its various
parameters. One benefit of digitizing a waveform, is to Figure 19. Averaging A SNAP. A) A routine radial SNAP is at-
assist in its extraction from background noise once all tempted but the background EMG noise obscures the response. B) Av-
other means noted above have been implemented. It is eraging 70 trials helps to better delineate the desired response permitting
easiest to discuss the benefits of averaging when a typi- waveform detection and analysis. C) Working with the patient to quiet
the background EMG activity reveals the true response which compares
cally derived time-locked signal is considered.
favorably with that obtained from averaging.

For the purposes of this discussion, let’s assume that all

(stimulation). Therefore, as more and more stimuli are
aspects of the recording have been optimized (electrode
averaged, the random peaks and troughs of our MUAP
separation, differential amplification, filtration, etc.), yet
begin to cancel each other while the regularly occurring
there is still some biologic noise that precludes us from
SNAP begins to emerge from the background noise.
clearly observing the waveform of interest. A common
We are acting to improve the desired signal with respect
problem occurs when one is trying to record a SNAP
to the background noise, i.e. improving the signal-to-
for example, but the patient simply cannot relax enough
noise ratio. The signal improvement is proportional to
with the resultant MUAPs obscuring the SNAP. We
the square root of the number of averages implemented
can take advantage of the fact that the SNAP will occur
multiplied by the signal’s amplitude, and this quantity is
at the same time following each stimulation while the
in turn divided by the noise amplitude: /Noise Ampli-
MUAPs will manifest somewhat randomly with respect
tude.
to the stimulus. When we average a response, the digi-
For example, suppose we have a signal with an ampli-
tized waveform is summated digitally point-for-point
tude of 2 µV and a noise level of 4 µV. The S/N in this
with a previously obtained waveform and then math-
instance is ½, i.e. the noise is twice as big as the signal
ematically divided in half. We take advantage of a time-
and we are likely not able to detect it. If we average 4
lock compared to relatively random waveform occur-
sweeps, our S/N is now 1/1 (S/N ~ 2 X √4/4 = 1). Now
rences since relatively random waveforms (e.g. MUAPs)
our signal is the same size as the noise but still likely not
have peaks and troughs that do not always align in time,
easily detectable. However, if we average 64 trials the
while our SNAP occurs with the same waveshape at
S/N now becomes 4/1, in other words our signal is now
exactly the same time following each sweep initiation

28
4 times greater than our background noise and likely
to be easily observed (S/N ~ 2 X √64/4 = 4). I find
averaging can be a very useful technique and never
hesitate to use it when deemed necessary. It can be quite
helpful at times to extract a signal out of the surround-
ing noise (Figure 19 ). Of course, a number of stimuli
are required to obtain an averaged response, and the
patient should be appropriately warned with respect to
this issue. Also, it is a good practice to utilize a stimulus
frequency that is not divisible into 60, i.e. 60 Hz (or 50
if the line current utilizes a 50 Hz frequency). Unfor-
tunately 60 is divisible by quite a few possible stimulus
frequencies (2, 3, 4, 5, 6, etc.). Therefore, a common
strategy utilized is to implement a stimulus frequency of
2.7 Hz or something similar. In other words, a stimulus
frequency that is not exactly divisible into 60 Hz. The
point is to not amplify the 60 Hz line interference by
treating it as a regularly occurring signal, i.e. not random
and therefore not subject to the anticipated cancellation
of random waveforms.
CONCLUSION
As we have discussed at length above, the electrophysio-
logic instrument can have profound effects on waveform
parameters. There can be no doubt that the presented
material is important, clinically relevant, and complicat-
ed, but above all, accessible with a little bit of diligence.
I would strongly encourage all practitioners to expend a
little bit of time and alter the electrode separation/loca-
tions, filter settings, and cathode/anode relationship to
better appreciate the material presented above. Actually
observing how waveforms change in relationship to the
above instrument/electrode alterations can significantly
help understand the material presented. It is incumbent
upon the practitioner to possess a working knowledge
of how the instrument and electrodes function so as to
problem solve various issues that may arise during the
electrophysiologic assessment. We must assure that any
data collected is accurate and worthy of being utilized to
help arrive at an appropriate clinical diagnosis. Failure
to do so can quickly lead to possible false positive and/or
false negative results.
BIBLIOGRAPHY
Dumitru D, Walsh NE: Practical instrumentation and
common sources of error. Am J Phys Med Rehabil
67:55-65, 1988.
Dumitru D, Walsh NE: Electrophysiologic instru-
mentation. In: Clinical Electrophysiology; Physical
Medicine and Rehabilitation State of the Art Reviews.
Philadelphia, Hanley & Belfus Vol 3 No. 4, 1989, pp
683-699.
Dumitru D, Amato AA, Zwarts MJ: Electrodiagnostic
Medicine. Philadelphia, Hanley & Belfus, 2002.
REFERENCES
1. Dryer SJ, Dumitru D, King JC: Anodal block
versus anodal stimulation. Am J Phys Med Rehabil
1993;72: 10-18.
2. Winkler T, Stalberg E: Surface anodal stimula-
tion of human peripheral nerves. Exp Brain Res
1988;73:481-488.
29