Professional Documents
Culture Documents
Objective
The PN junction semiconductor diode is a real device whose steady-state terminal volt-ampere
approximates that for the idealized diode definition. Although they have significantly different
terminal characteristics it is nevertheless common practice for both devices to be represented by
the same icon. Hence some care is needed to distinguish which diode the icon implies;
generally however this is unambiguously clear from context. The diode is an important circuit
element in its own right, as well as being important in understanding the operation of other
devices, in particular but not limited to the bipolar junction transistor.
As is true generally for semiconductor devices the diode terminal volt-ampere relation is
nonlinear. Approximating the diode by a simpler PWL model is very useful for reasons noted
elsewhere. In this note on diodes a qualitative description of junction phenomena leading to
the basic diode terminal relation is presented. This is intended to provide a conceptual
background against which to better appreciate junction characteristics. Then the characteristic
of a representative1N4004 diode is examined (via a computer analysis using a nonlinear diode
model) to illustrate theoretical expectations. Some often useful generalizations are made about
diode characteristics. Then the idealized diode is used to describe modeling of real device
characteristics. Finally a useful computer model approximating an idealized diode is described,
primarily for a limited pedagogical use.
Addendum:
a) For the present we assume that the diode internal physical processes are fast enough to
be able to adapt essentially instantaneously to changes in diode terminal voltage. The practical
meaning of instantaneous is that physical changes in the diode state are more or less
complete before terminal voltages and currents change significantly. However the internal
response time although it usually is very short still is finite, and in a number of increasingly
common applications a diode may not respond quite fast enough for the reaction to be
'instantaneous'. This aspect of the dynamic behavior of the diode we consider separately later.
Introduction
A detailed technical examination of the fabrication technology and the physical phenomena associated with
a semiconductor junction is no simple undertaking, and it is not to be found here. On the other hand a
broad qualitative appreciation of semiconductor junction physics can be of enormous help in the application
of semiconductor devices in electronic circuits. Hence a review of a qualitative discussion of
semiconductor physics such as is found in the ECE311 text is recommended.
While there are several semiconductors of technological interest it is sufficient for our needs to consider
only silicon semiconductors. Silicon exhibits a remarkable confluence of metallurgical, chemical, electrical,
and mechanical properties that give it general commercial preeminence as a semiconductor material.
The silicon semiconductor diode is a two-terminal element exhibiting which unlike (say) a resistor exhibits
non-bilateral behavior, i.e., current flow through the device is easier in one direction than the other.
Forward Bias
The computed forward-bias volt-ampere characteristic of a 1N4004 diode (PSpice model used to compute
the characteristic) is shown below for several temperatures; this is representative of small-geometry diode
characteristics. Note that the forward-bias current does not reach a current about the milliampere level until
a threshold (knee) voltage of roughly 0.5 volt is crossed. It is useful to note, as is generally true for
diodes of the 1N4004 type, that roughly 0.1-volt change in forward bias increases the current by about an
order of magnitude. Note further that the forward-bias voltage needed for a given current decreases as the
temperature increases, i.e., as the average thermal energy of the carriers increases. For silicon the bias
voltage needed for a given current is reduced roughly 2 millivolt per C temperature increase.
Because of the large current range for a forward-bias junction a logarithmic re-plot of the characteristics as
shown below is informative. Note the close adherence to an exponential dependence of current on forward-
bias voltage over the several orders of magnitude plotted. For diode operation in the milliampere current
range around room temperature the bias will be roughly 0.6 to 0.7 volt. Note again that around room
temperature a bias voltage increase of 50 to 100 millivolts or so increases the current by roughly an order
of magnitude.
Reverse Bias
If the diode voltage polarity is negative the diode is said to be reverse-biased. In the theoretical
expression for the diode current this corresponds to the exponential becoming much less than 1 with a few
tenths volt reverse bias. The following figure plots the reverse-bias current for several temperatures. Note
the considerable difference between the reverse-bias current level and the forward-bias current range.
Idealized Diode
Basic circuit analysis courses introduce the concept of idealized lumped-circuit elements. These are
fictional constructs that embody some essential aspect of real device operation. For example the idealized
linear resistor embodies the property of power consumption implicit in Ohms Law. There is a not entirely
unnatural tendency to identify the canonical resistor icon with, for example, a familiar commercial
cylindrical composite device with colored bands identifying resistance value and tolerance. In fact however
the icon represents only the dissipation characteristic of that or in fact most other physical devices. Real
devices, in which several physical phenomena generally operate concurrently, are described to some degree
of approximation by a combination of idealized circuit elements. Thus a real resistor may need to be
described for some uses by a combination of idealized resistors, idealized capacitors, and idealized
inductors. The idealized resistor icon need not even represent a specific component; it might for example
represent collective power consumption by an unspecified device, e.g., by a motor under load, or by a TV
set.
At this point we review a special idealized circuit element the idealized diode. The idealized diode is an
abstract approximation to a property of a real diode, in the same general sense that an idealized resistor is
an abstract approximation to a property of a real resistor. That is it abstracts a particular aspect of the real
diode operation. In later discussion we compare the idealized diode volt-ampere relation to that of real
diodes, and use idealized diodes as part of the representation of the volt-ampere relation of various devices
incorporating diodes, e.g., the bipolar transistor. Temporarily however the idealized device can be viewed
simply as a fictional construct for purposes of an abstract discussion, a construct that later will turn out
fortuitously to be useful in the real world.
The idealized diode is piecewise linear (PWL), i.e., it is linear except for the discontinuity in slope (i.e.
first derivative) at V=I=0. What this means insofar as the analysis of an idealized diode circuit is
concerned is that instead of analyzing (with attendant difficulties) one non-linear circuit we may use two
relatively simpler linear analyses, one for each segment of the characteristic. Latter we consider this
partitioning in relation to the mathematical technique of approximating a non-linear curve by a set of
secants; in the limit as the number of secants approaches (and secant length -> 0) a highly accurate
approximation to direct use of the non-linear curve is obtained. This is basically the procedure generally
ECE 414-Diodes 4 M H Miller
used, for example, to estimate the area under a curve. There are of course some bounds to the charm of this
simplification; having to perform too large a number of linear calculations, for example, tends to make a
direct numerical analysis more attractive. However experience has demonstrated (and will do so again later)
that even a crude linearized circuit approximation often can provide truly remarkable effective assistance in
a circuit design.
Since the circuit topology remains unchanged unless and until a diode changes state circuit variables are
linearly related, i.e., the relationship between any pair of variables graphs as a segment of a line. In general
as the strength of an independent-source variable is increased monotonically from - to + branch
voltages and currents change, and eventually a diode (generally) will change state. At that point the
relationship between the variables plotted changes (the circuit topology and so the application of Kirchoffs
laws has changed), and a different line segment is involved. Because of the monatomic character of the
assumed source change it follows that in a linear circuit a given diode will change state no more than once if
at all. Hence for a circuit involving N diodes there can be at most N changes of one or another diode state,
and so at most N+1 line segments are involved in a complete description of the circuit performance. Note
however that there may be fewer segments in special circumstances although there cannot be more.
While in general there are no more than N+1 segments needed in describing the performance of a
particular N-diode circuit the number of conceivable combinations of diode states is 2N. If there is more
than one diode in the circuit then (since 2N > (N+1) for N > 1) there are more conceivable diode state
combinations than there are segments. For example, for N = 2 the maximum number of segments is 3, but
there are four possible combinations of diode states to consider. It follows that not all conceivable
combinations of diode states actually are realized in any specific circuit involving more than one diode. (As
a simple illustration consider a pair of back-to-back diodes; the diodes cannot both be short circuit
concurrently, nor can they both be open-circuit concurrently. (Parenthetical comment: consider the
relationship between the current through and the voltage across a back-to-back diode pair. Is this
relationship piecewise linear?)
For a voltage or current discontinuity to occur at the point where one segment ends and another begins
violates the energy conservation requirement (for voltage) or the charge conservation requirement (for
current), and cannot occur. The discontinuity always must be in the slope of the two adjacent segments.
Note that with zero voltage across and zero current through the diode at the point where it changes state a
physical distinction at that point between forward-bias (ON) and reverse-bias (OFF) operation is moot.
As noted before a mathematical existence theorem proves that a solution of a valid linear circuit exists and
moreover that solution is unique. It follows from this that within the domain of each segment individually
there is a unique solution. By induction then there is one and only one result for the analysis of a PWL
circuit. This unique solution can be identified by just three necessary and sufficient conditions:
Within the domain of each segment
a) KVL must be satisfied;
b) KCL must be satisfied;
c) The volt-ampere (constitutive) relations of the individual circuit elements must be satisfied.
It is with respect to item c) that uncertainties arise in a PWL circuit analysis, since the portion of the diode
volt-ampere relation applicable depends on its initially unknown circuit state. Thus it appears that the
ECE 414-Diodes 5 M H Miller
results of a circuit analysis really must be known a priori to determine the appropriate diode state to use for
the analysis to obtain the results. Fortunately there is a straightforward if indirect means to resolve this
difficulty. Simply assume a diode state for each diode (it can be simply a random guess, but an educated
guess can simplify the analysis), and analyze the corresponding linear circuit under the assumed
circumstances. The solution process itself, absent a procedural error, assures the satisfaction of KVL and
KCL. Similarly no question arises about satisfaction of the circuit element volt-ampere relations other than
those for the diodes. Absent an error the only uncertainty in the analysis is whether or not the correct
assumption was made about diode states initially.
Determine separately for each diode whether the assumed diode state is consistent with the solution based
on that assumption. This means simply that a diode assumed to be in an open-circuit state must be reverse-
biased by the calculated circuit voltages. Similarly a diode assumed to be short-circuit must carry
calculated currents in its forward direction. If for a given diode the appropriate condition is not satisfied
then the necessary and sufficient conditions for the unique solution are not satisfied (diode volt-ampere
relation is not satisfied), and the assumption made about the diode state must be wrong. But then since a
diode has just two possible states the correct diode state assumption to make immediately becomes clear.
Consider, for example the circuit drawn to the left. There are two (idealized)
diodes, and hence in principle four combinations of diode states to consider. But
in this particular circuit it is not difficult to determine directly that D2 must be
closed; if it were not there would be no current through the 5K, no voltage drop
across this resistor, and an inconsistent 9V forward-bias for D2. Hence D2 must
be closed (i.e., short-circuit), and so carry a current of 1.8 ma in the forward
direction for D2.
But D1 also must be closed, since if it were not there would be an inconsistent
forward-biasing voltage of 9V across it. Hence D1 carries a forward current of 0.9
ma. And despite what intuition may suggest KCL then requires I = -0.9ma (note
the polarity of I carefully).
A somewhat more intricate circuit involving two diodes is drawn to the right. For the
present we use a straightforward brute force approach to analyzing this circuit; each
possible combination of diode states is considered individually for validity. That is,
we make use of the fact that a unique solution exists, as well as the known
conditions to establish the correctness of a proposed solution. There are four
combinations of diode states to evaluate, and we do each in turn. The methodology
is straightforward albeit tedious. A specific combination of diode states is assumed;
the diodes are replaced by an equivalent open- or short-circuit as appropriate. The
circuit then is analyzed, and the results are used to validate (or not) the assumed diode states.
Since more accurate nonlinear diode models would be used for computer calculations ordinarily there is no
particular computational advantage to temporizing with a less accurate computer model. It may appear
peculiar therefore to degrade a nonlinear diode model to approximate an idealized diode for a computer
computation.. But this apparent incongruity should be evaluated in terms of objectives. At the very least
for instructional purposes, it is often convenient to use a computer analysis to compare the results of an
idealized diode PWL circuit analysis with those for an analysis using a nonlinear device model. For this
and similar purposes a computer model for the idealized diode is useful. This requires rather more care to
develop that might appear to be the case at first. The discontinuity in slope at the origin creates some
computational difficulties. It turns out, almost perversely, that a continuously differentiable nonlinear model
for the idealized diode is to be preferred. (This may be considered as one indication of important
distinctions between a hand calculation and a computer analysis.) There are several ways of devising a
satisfactory model, but probably the simplest way is to modify an existing nonlinear diode model
appropriately. The following remarks, although generally applicable, are expressed specifically for PSpice
use.
Suppose the diode emission parameter N is made very small, say 10-6. In effect this reduces the voltage
required for a given forward bias current by a factor of one million! If about a 50 mV change ordinarily
causes an order of magnitude current change in a real diode then the modified diode requires only 50
nanovolts for the same change. Hence the forward-bias diode characteristic rises quite sharply, not
abruptly but continuously, within a few nanovolts of the origin. (Keep in mind that this is a computational
trick; the modified nonlinear model does not correspond to a real diode.) The reverse current is not the
idealized zero value but it is quite small (and can be made smaller if the need arises). There is an additional
catch however that has to be considered to avoid computational problems. Because of the extraordinarily
large increase in forward current with small forward-bias voltage it is possible for the computed current
magnitude to exceed the storage size limitations of the analysis program or to become so large as to cause
mathematical convergence problems. To avoid this another parameter in the computer diode model can be
used; the nonlinear model includes a parasitic resistance (RS) in series with the real diode, representing
for example lead resistance. This can be used as a current limiting resistance; a milliohm will do nicely to
limit current magnitudes to less than about 10 kiloamperes. (Again note this is just a computational trick to
avoid computational problems.)
To realize this idealized diode computer model in PSpice place and select a 1N4002 diode. Then select
Model from the EDIT menu, and in the dialog that opens select the middle Edit item. In the list of model
parameter that is presented change Rs to 1m and N to 1u. PSpice will use the built-in default diode model
with the parameters N and RS changed as prescribed. Examples of this usage appear below.
The 1N4002, on the other hand, must wait for the knee to be passed before enough current flows to cause
a significant voltage drop across the series resistor. The 1N4002 bias voltage then increases slightly as the
diode current increases, corresponding to small junction voltage changes needed to obtain large changes in
current.
A different perspective is afforded by plotting the diode currents rather than voltages. VS and the series
resistor determine the current through the idealized diode. Since the forward-bias voltage is zero; the slope
corresponds as it should to a 1 k resistance. Except for the 0.7v offset the same behavior is shown by the
modified diode branch.
Incidentally note that the idealized diode should be a good approximation for the real diode in circuits in
which the 0.7volt (approx.) forward-bias voltage is small compared to other significant circuit voltages.
Further improvement in the approximation is obtained by adding the offset voltage.
When Vout reaches 2v (and until Vout = 3v) the first two diodes both are ON. Write (say) a node equation
at the output node: (Vin Vout)/R = (Vout 1)/R/2 + (Vout 2)/R/2, and simplify to Vout = (Vin 5)/6.
This secant is valid from the point (4,2) to the point (9,3).
It is left as an exercise to verify that the remaining secants as predicted correspond to the remaining diodes
being turned ON in succession. As each diode is turned ON it adds an additional shunt branch.
Of course idealized diodes are hard to come by in practice, and so for an evaluation of circuit performance
we use 1N4002 diodes (PSpice model). First however to bring the idealized diode design closer to a
1N4002 the circuit is modified as shown below. The offset voltage in each branch is divided into two parts,
with a nominal 0.7v part associated with the diode. It is the combination of the fixed offset voltage and the
idealized diode that is replaced for the PSpice computation by the 1N4002 diode model (indicated by the
dashed rectangles).
. Sketch a rough plot of the anticipated diode currents, and then compute the currents for comparison.
(2) The circuit diagram (below) shows an improved experimental realization of the theoretical Square
Root transfer function described above; V(2)2 VS for 0 VS < 25 volts. Note that the theoretical value
will agree with an experimental value for VS = 0 volts. A second point of agreement can be realized by a
calibration step; adjust V1 (experimentally) so that experimental and theoretical agreement is obtained for
VS = 25 volts. The value so obtained for this calibration step is 10.6v. Making the theoretical and
experimental (virtual) characteristics agree at the two ends tends to force better agreement at intermediate
points as well.
Compute the transfer relation v(out) * v(out) vs. v(in); also plot v(in) on the same axes. Suggestion: Plot
v(in) on the same axes to judge the efficacy of the design.
Assume VS is a triangular waveform with a peak value of 4v and a nominal period of 1 millisecond, and VC
is a square wave of amplitude 10v. Use R1=1K, R2=R3=47K. Use 1N4148 switching diodes for the
computer analysis rather than idealized diodes: Compute and plot Vo. Keeping in mind the back-to-back
diode connection evaluate the influence of the threshold voltage for the diodes?
(4) Diodes may be used as a kind of voltage reference in integrated circuits. Although there are
other generally preferable means to serve the purpose illustrated in the example nevertheless it is
instructive to examine the principles involved.
Compute the load voltage for the circuit as the load current varies from 0 to 20ma. Use VS = 10v, RS =
200, RB = 150. Note that for the maximum 20 ma current the terminal voltage will have dropped by
7 v from its no-load value. Use 1N4002 diodes. Compare the source current to the diode current.