Wiki › 6.002.

1x › Linear Ordinary Differen… ›  Add article


View
Linear Ordinary Differential Equations  Edit


Solving linear ordinary Changes

differential equations Last

modified:
April 19,
This paper is copied from the wiki from 6002x Spring 2022,
2012 (author unknown) 12:17 p.m.
This is a short paper on linear ordinary differential
equations (ODE or ODEs or ODE's), of the type that you'll See all
encounter in MIT 6.002x. You should visit MIT 18.03 or
18.03SC in MIT OCW for more in-depth treatment, or the children
Khan Academy videos here.
First of all, some terminology: An "ordinary" differential
equation (ODE) has only one independent variable, let's
say . For example, if you have a function with
derivative , second derivative , etc..., and
you combine some of these derivatives, itself, and other
time functions, say , in an expression, you are building
an ordinary differential equation. Here are some examples
of linear ODEs (see below the explanation of linear):

The concrete name of the variable described by the ODE is

irrelevant: , , etc... In the first example, the "extra" time
function, , in the ODE would be .
The order of the ODE is the highest derivative order
present in it. The two examples above are second order
ODEs.
The variable in the ODE, , can be an N-vector, that is
. This means you are
governing several scalar dependent variables, ,
within the same ODE. In electrical circuits these variables
are typically capacitor voltages, , and inductor
currents, . When is a vector, the constant
multiplying coefficients in the ODE ( and in the
examples above) become matrices of suitable dimensions.

If you have y=f(x,t) and you have partial derivatives

and combined in your expression, then you have to
deal with partial differential equations.
In MIT 6.002x, you only need to worry about linear
ordinary differential equations. Linear means that if
is your ODE variable, you only would see things
like , , , etc..., but you would NOT see things like
, in your differential equations.
An example of a nonlinear second order ordinary
differential equation is that which governs the oscillation of
a pendulum:

since it makes use of a nonlinear transcendental function,

. Here is the angular displacement of the
pendulum from the vertical, is its length and is the
gravity acceleration (constant).
Let be the relevant dependent function, and the
time and 1-D position be the independent variables. A
well-known partial differential equation is the wave
equation:
where is the velocity of propagation of the wave. For
instant, the deformation propagating along a rope can be
described by the wave equation.
Fortunately, in this course you will not deal with nonlinear
ODEs; neither will you ever have to solve partial differential
equations here.
For the rest of this section, we are concerned with time
as the independent variable, so we are interested in, for
example, some circuit voltage . For short hand, I will
write .
A homogeneous differential equation is one where there is
a combination of functions of and of its derivatives, , ,
etc..., in the left-hand side, and there is a 0 in the right-
hand side. To solve it, you need to account for some
boundary or initial conditions.

Homogeneous first order linear

ODEs
Here is a homogeneous linear ordinary differential
equation:
,
with the boundary condition that at , . is
a constant coefficient; is some constant initial value. To
solve explicitly, we separate the two sides like this:

Then, isolate the differentials:

Integrate both sides of the equation:

The result is:

where is the constant of integration. Finally, to find ,
we exponentiate both sides of the equation, and get:

We find the constant of integration by applying the

boundary condition (initial value) that , so,
setting , we can find the constant . Similarly, if your
boundary condition was , you would use your
solution for to find the constant .
Note that , so in fact,
.
Don't forget your units: If has units of volts ,
then has units of volts/second , and must have
units of inverse-seconds . Also, note that one
Hertz equals one cycle/second equals
(radians/cycle)*(cycles/second), i.e. radians/second
.

Inhomogeneous first order ODEs

Suppose we had an inhomogeneous ODE, for example:

In general (and this is painful to do!), multiply both sides of

the equation by , then, integrate to solve. Note that:

So, we have:
Now, separate the differentials and integrate:

Now, isolating ,

And there you are! This is painful. For MIT 6.002x, the
material is limited to a few specific cases, so rather than
asking you to integrate (heck, I can't integrate this
myself!), we will present some specific solutions to
commonly encountered inhomogeneous equations:

Specific inhomogeneous first order

equations:
One first order equation that you will see is:
, but with the initial value that . We
are not interested in what happens before time , but
we want to know what happens for . For example,
see this circuit, where you want to solve for .
After applying KCL and moving all the terms in to the
left hand side, you would get:

Which you can rearrange to:

, or you can also write it as:

could be an initial condition, that is, it might be a value

until time =0, then, it is equal to zero. Recall that the
general solution is . We simply match the
boundary condition that , and we have:
(for ).
This satisfies both the original differential equation, and
the initial value. Note that for this circuit model, is always
a positive number ( ). Here is a plot of what
looks like (I took values of and for
this plot):
If is an impulse, that is, for , but at
, , such that the integral , you
would see the same results as the response to an initial
condition as above.
Now, let's look at another very simple inhomogeneous
equation:
(with the initial value for . Again,
we are not interested in what happens before time .
Notice that I've sort-of-rearranged the original first order
equation (dividing through both sides by ) before
applying the "1". Anyway, this sort of inhomogeneous
equation is what you would see if you modeled a circuit
attached to a switch to a battery (or Thevenin source). For
time , the switch is open. At time , you close
the switch, and the circuit sees a sudden change in
voltage.
If you did all the integrals in the previous section, you
would see that the solution is

I leave it to you to verify that if you take the derivative to

get , then, apply , you would get the right hand
side "1".
Here's the response:
Last, let's consider: {it's left from here}

Now, we're not interested at time near t=0, but we want to

see what's the behavior at
In general, your solution in the case would
be (you can actually get this by doing that multiplication by
e raised to the integral of a dt!):

So, you can visually see there's a short term response with
that will decay away as , and a response to the
sinusoid, multiplied by the value (which I'll call a
magnitude), and offset with a phase .
Without proof: The magnitude (specifically, the relative
magnitude) is:

And the phase is:

Note that in the limiting case where (so

), we're back to inputting a "1" function (also called a step
function). The result is , and using the equation
 , we get the result
, which we'd already plotted above.
Below are four plots that illustrate sinusoid response.
in the circuit sketched above; and I apply
four frequency sinusoids: 0.5 rad/s, 1.0 rad/s, 2.0 rad/s,
4.0 rad/s. Observe that at the low frequency, the output
response follows the input pretty well, but it
lags--that is, it peaks a bit behind (later than) the input. As
the frequency goes up, the magnitude of relative to
falls, and the phase lags some more. In the limit, the
magnitude of will go towards zero, while the phase will
go towards .

I probably need to write a short math review for working

with complex numbers, and Euler's equation
where .
2nd order homogeneous linear
ODE's:
On to 2nd order linear ODE's:

Subject to initial conditions and

. Mathematicians like to see patterns and
apply them. Since the exponential function worked out
so well with first order linear ODE's, we guess that an
exponential function would also be the solution for 2nd
order linear ODE's.
Guess , and substitute your guess into your
ODE. Note that then, and . Then,

Factoring out ,

Since is always , then . If we

solve the quadratic for , we would
have our solution.

A couple of notes: Without rigor, 2nd order linear ODE's

encountered in MIT 6.002x will always have , and
and .
So, there are several possibilities:
(1) are real, but different (that is, and
)
 (2) are complex conjugates (that is,
)
(3) are real, but the same values (that is,
)
We will investigate each of these cases. Again, without
rigor, these 3 cases are the only ones you will encounter in
MIT 6.002x, and in MIT 6.002x, the real part of is always
.
So, for cases (1) and (2) your homogeneous (or general)
solution will always be:

where and are constants determined by your initial

conditions and .
For case (3), your homogeneous solution will always be:

You will recall from the lecture that the complete solution is
going to be the superposition of the homogeneous
solution and a particular solution. We will graphically
explore the solution of second order solutions to a step:
That is, the inhomogeneous part will be 0 for and 1
for . The initial conditions are that at t=0,
and .
Let's look first at case 3, that is:

Clearly, the homogeneous solution is of the form

, with the roots of
.
Here is the response to the step input (note that when the
roots are equal, this is called "critically damped"):

Now, let's look at case 1: (Two distinct real roots). We'll

consider this equation:

Again, initial conditions for and are zero. Your roots are
and . The step response is below. This
is basically two first order responses; one is feeding into
the other.

Lastly, let's look at the step response of a system with

complex conjugate :
The roots of are
where . Remember Euler's equation?
, so:

The homogeneous solution is a decaying exponential

envelope, multiplying an oscillation!
Here's what it looks like for the step:

And that should take care of you until week 9 of the

class!!!!
Still needed, but this wiki is getting awfully long, and
could use some splitting up?
Breaking down a higher order linear ODE into a
system of 1st order ODE's?
Numerical solutions and matrix methods?
Frequency response of 2nd order systems
Linear ODE's; the substitution of and
transformation of ODE's to algebra (the Laplace
transform method)?