Functional Testing

The Next Best Thing to Calibration

By W. Reed Edgel, PhD, PE

Introduction

This document discusses functional testing vs calibration of differentiating electromagnetic

sensors, including "D-dot" and "B-dot" sensors that respond to the first time derivative of the
incident electric and magnetic fields.

Theory of Transverse Electromagnetic Radiation

The basic theory of transverse electromagnetic radiation is discussed in PAN 895, Pages 2-4.

Testing vs Calibration

The term "calibration" is not appropriate, because there is no other standard except another
magnetic field sensor that can measure the incident field accurately. It is only possible to
compare the output of the subject sensor to the output of another magnetic sensor, preferably an
identical one. These sensors themselves are the standards. While they cannot be calibrated, they
can be functionally tested.

A major challenge over the last four decades has been to find a source of transverse
electromagnetic (TEM) waves that was consistent with its theoretical fields to the extent that it
could be used to calibrate EMP sensors. Large EMP simulators such as ALECS, ARES and
TRESTLE were built at Kirtland Air Force Base in Albuquerque. They were mapped with electric
and magnetic sensors like the subject sensor. The sellers of TEM cells of less than room size
dimensions provide maps of field strength versus position in the working volume. These maps are
generally made with some form of electric or magnetic field sensor.

Testing done in the 1970's when this class of sensor was developed was done in large simulators
with the distance between conducting elements h of the order of 5 to 20 meters, many times the
largest dimension of the item under test. In the early days of EMP testing, pulsers were based on
mercury wetted reed relays and physical charge lines instead of solid state switching devices and
electronic timing devices. Sensor outputs from the early models of the subject sensor were
integrated digitally or with mechanical planimeters. Sensor outputs were found to be accurate to
within the experimental error of the test setup, typically 4 to 7 percent. The equivalent area is
controlled within 1 percent, so these sensors were used to map or "calibrate" the simulators.

The distinction between a sensor and a transducer and the lack of need for calibration is
discussed in more detail on page 7 of PAN 895.

Smaller EMP simulators were designed for functional testing of sensors. These included the
conical simulator (appendix A [Stuart Olsen’s report]) and the smaller conical-elliptical simulator
(appendix B [Gary Sower’s report]). One advantage of these simulators over the plethora of small
parallel plate simulators and TEM cells currently available is that they have radial symmetry about
the axis of propagation of the EM wave and eliminate one dimension of distortion of the wave.
They also are large enough that the ratio of simulator working height, h to the height of the
sensor, d is larger than that for most small parallel plate simulators and TEM cells.
The presence of the sensor perturbs the electric field generated by the simulator. The
perturbation has the effect of reducing the electric field and thence the magnetic field (See PAN
895, page 3) the sensor senses. The degree of perturbation depends on the ratio of simulator
working height, h to the height of the sensor, d—the higher the simulator working height, the less
a sensor of a given size perturbs the fields.

The interaction between the sensor and the simulator also exhibits frequency dependence.
Location of the sensor in a simulator of limiting size is critical. Capacitative coupling between
sensor and simulator becomes significant—small changes in the location of the sensor with
respect to the simulator can have much larger effects on the effective electric and magnetic fields.
The typical EM sensor would not see these effects so profoundly in a large simulator such as
ALECS, ARES, or TRESLE, where h ranges from 5 to 25 meters and h/d ranges from 10 to 50.

The foregoing means that small parallel plate simulators or TEM cells cannot be used to
"calibrate" EM sensors because the incident fields are affected more by the presence of the
sensor than the response of the sensor is affected by inaccuracies in its manufacture.
Comparison of one kind of sensor to another only results in a comparison of the perturbations
caused by different kinds. The best functional tests performed in the 1970s showed that the
sensors were more accurate than the test setups.

For the benefit of the technologist who needs a number that describes the sensitivity of this class
of sensors to electric or magnetic fields, the following discussion should be helpful.

One of the most frequently asked questions we get at PRODYN is "What is the calibration factor
that gives B, D or I from the output of a B-dot, D-dot or I-dot sensor?" The answer is that there is
no such factor, because the output of this class of sensors (differentiating sensors) is proportional
to the first time derivative of the input to the sensor, not to the input itself. This is an advantage
when the phenomenon of concern responds to the first time derivative of the input, which is often
the case. There is an exception, when the input is sinusoidal, that arises from the fact that the
derivative of the sine function is the cosine function. This case will be discussed after the general
case wherein the input is not sinusoidal.

While the foregoing applies to electric and magnetic field sensors and to current sensors, we will
use magnetic field (B-dot) sensors to illustrate how the value of the field strength is obtained from
the output of a differentiating sensor when the field steps from zero to some level.

The process is idealized in figure x. The magnetic field, or more correctly, the magnetic induction,
B, is measured in Teslas, or volt-seconds per square meter. As shown in figure x(a), it rises from
zero to Bp in time t. The rise is shown as a steep ramp function, with a rounded transition from
zero rate of increase to the maximum rate of increase as the signal "takes hold" and a rounded
transition from the maximum rate of increase to the flat value of Bp. The field is rising fastest in
the middle of the "step".

The output of the B-dot sensor is proportional to the first time derivative of the magnetic induction
B(t), that is:

B-dot = dB/dt B/ t

and, rewriting equation (15),

Vo = Aeq dB/dt, or dB/dt = Vo/Aeq

The output of the B-dot sensor is measured in volts that are proportional to first time derivative of
the magnetic induction, which has units of volts per square meter. The constant of proportionality
between the sensor output Vo and the first time derivative of the magnetic induction dB/dt is the
equivalent area Aeq of the sensor. This is the calibration factor, but it is applied to the first time
derivative of the magnetic induction, not to the magnetic induction itself.

The sensor output resulting from the input of figure x(a) is shown in figure x(b). It is a single pulse
that rises as the rate of increase in B increases to a maximum value, and then decreases as the
rate of increase in B decreases to zero when B reaches its peak value.

A reasonable estimate of the maximum value of dB/dt is B/ t, which is obtained by measuring the
change in B over the time interval during which B is changing most rapidly, the middle of the ramp
function in figure x(a). This is the maximum output we expect from the sensor.

The integrated sensor output, shown in figure x(c), is obtained by integrating the sensor output
with respect to time. If the sensor output has been digitized, the integral can be obtained easily by
multiplying the numerical values of B/ t by t over the time interval 0 to t. As the number of discrete
points increases, B/ t becomes a better estimate of dB/dt and (B/ t) t becomes a better estimate of
dB = B = 1/Aeq* Vodt. Ideally, the function in figure x(c) is identical to that in figure x(a).

Sinusoidal field functions make it possible to integrate the output by simply dividing by the angular
frequency of the signal. We will use an electric field for this case. The process is idealized in
figure y. The electric field E, which in free space is the electric displacement D divided by the
permittivity constant o is measured in volts per meter. As shown in figure y(a), it is a continuous
sinusoidal function with amplitude Ep and frequency f.

The output of the D-dot sensor is proportional to the first time derivative of the electric
displacement D(t), that is:

D-dot = dD/dt D/ t

and, rewriting equation (14),

Vo = R Aeq dD/dt, or dD/dt = Vo/R Aeq

then, recalling that D = o E,

dE/dt = Vo/R Aeq o

The output of the D-dot sensor is measured in volts that are proportional to units the first time
derivative of electric displacement dD/dt, coulombs per square meter. The constant of
proportionality between the sensor output and the first time derivative of the electric field dE/dt (=
dD/dt o) is the permittivity constant o. The constant of proportionality between the sensor output Vo
and the first time derivative of the electric field dE/dt is the product of the output impedance R and
the equivalent area Aeq of the sensor, and the permittivity constant o. This is the calibration factor,
but it is applied to the first time derivative of the electric field, dE/dt, not to the electric field itself.

The sensor output resulting from the input of figure x(a) is shown in figure x(b). It is a cosine
function with amplitude Vop and frequency f, where:

Ep = 1/2 f (dE/dt)p = Vop/2 f R Aeq o

In the sinusoidal or CW case, the peak value of the input field can be obtained from the peak
value of the sensor output, the constant of being proportionality being the product of the general
constant R Aeq o and 2 f, which is constant if the frequency is constant when the measurement is
made, as in the CW case.

A reasonable estimate of the maximum value of D/ t or E/ t is obtained by measuring the change

in D or E over the time interval during which D or E is changing most rapidly. This is usually taken
as the T/6 duration when D or E is going from +0.5 (Dp or Ep) to -0.5 (Dp or Ep), as shown in figure
y(a). This is the maximum output we expect from the sensor.

The integrated sensor output, shown in figure y(c), is obtained by integrating the sensor output
with respect to time. If the sensor output has been digitized, the integral can be obtained easily by
multiplying the numerical values of D/ t by t over the time interval 0 to t. As the number of discrete
points increases, D/ t becomes a better estimate of dD/dt and (D/ t) t becomes a better estimate
of dD = D = 1/Aeq* Vodt. Ideally, the function in figure y(c) is identical to that in figure y(a).

In the case of continuous sinusoidal (CW) fields, the graph on page 11+x of PAN 895 can be
drawn for a specific sensor and then translated to a graph of sensor sensitivity as a function of
frequency by dividing the sensor output by the constant value of the incident field. Plotting this
multiple of the low frequency transfer function on log-log axes gives a nearly straight line that
inspires confidence in the user, but the restriction that this graph applies only to CW fields must
be kept clearly in mind.

W. Reed Edgel, PhD, PE

Principal