Answer: Verify f(x(n)) for n = -5 to 5

[RaduAndrei]

Consider that:

f(x(n)) = n*x(n)

and that x(n) is 1 for n = 0,1,2,3 and 0 everywhere else.

Thus, f(x(n)) will be for n = -5 to 5 like this:

0 0 0 0 0 0 1 2 3 0 0

Am I right?

 

[Mark44]

Consider that:

f(x(n)) = n*x(n)

and that x(n) is 1 for n = 0,1,2,3 and 0 everywhere else.

Thus, f(x(n)) will be for n = -5 to 5 like this:

0 0 0 0 0 0 1 2 3 0 0

Am I right?

Looks fine to me. It might be simpler to write this:
f(0) = 0
f(1) = 1
f(2) = 2
f(3) = 3
f(n) = 0 elsewhere

 

[RaduAndrei]

Thanks for the answer. So, ok.

Now, can I write that:

f(x(n-m)) = (n-m)x(n-m) ?

For example,

f(x(n-15)) = (n-15)x(n-15)

 

[jbriggs444]

f(x(n)) = n*x(n)

That is improper notation.

Probably what you have in mind is:

Let x(n) = 1 for x in {0, 1, 2, 3} and 0 otherwise.
Let f(n) = n*x(n).

 

[RaduAndrei]

But I want to view this function as something that takes as input a function and returns another function.
So, is like a discrete-time system. You apply a discrete-time waveform at its input and at its output you see another discrete-time waveform which is the result of the system acting on the input.

Can't I do this?

 

[jbriggs444]

But I want to view this function as something that takes as input a function and returns another function.
So, is like a discrete-time system. You apply a discrete-time waveform at its input and at its output you see another discrete-time waveform which is the result of the system acting on the input.

Ahh, that makes more sense then. I do not want to weigh in on the proper notation for defining functions whose domain and co-domain are both sets of functions.

 

[RaduAndrei]

So, the real reason why I made this post and my confusion.

I can write x(n) = u(n) - u(n-4), where u(n) is the unit step function.
Thus:

f(x(n)) = f(u(n)-u(n-4)) = f(u(n)) - f(u(n-4)) = nu(n) - (n-4)u(n-4).

And this function, plotted with wolfram alpha (x*unitstep(x)-(x-4)*unitstep(x-4)) gives the following result:

0 0 0 0 0 0 1 2 3 4 4 4 4 ...

So, this is my confusion. What am I doing wrong? That is a linear function and superposition principle applies.

If, on the other hand, I am writing:

f(x(n)) = n*x(n) = n*(u(n)-u(n-4))

it gives the same result as in the first post.

 

[davidmoore63@y]

The problem contradicts itself, as follows:

We have f(x(n))=n x(n) and x(n)=1 for n=0,1,2,3 and zero otherwise.

Plugging in n=0, we have f(x(0))=0, so f(1)=0
Plugging in n=1, we have f(x(1))=1.x(1), so f(1)=1

Contradiction.

 

[RaduAndrei]

Ok, I see the contradiction.

So, I cannot write f(x(n)) = n*x(n).

But I can write y(n) = n*x(n). So, the system takes as input some discrete-time waveform x(n) and produces another discrete-time waveform y(n).

But, can you explain to me why I cannot write f(x(n)) = y(n) = n*x(n) ?

If someone could shed light on this, I would be very happy. I'm thinking about this for days now.

 

[jbriggs444]

The problem contradicts itself, as follows:

We have f(x(n))=n x(n) and x(n)=1 for n=0,1,2,3 and zero otherwise.

Plugging in n=0, we have f(x(0))=0, so f(1)=0
Plugging in n=1, we have f(x(1))=1.x(1), so f(1)=1

Contradiction.

That was where I was coming from in post #4. However, since the formal parameter "x(n)" on the left hand side is being treated as a function and since the formula on the right hand side is being treated as a function definition for the resulting function, the notation is not unreasonable and not contradictory.

The intent is that it be read as:

[ Let f be a function from the set of one-parameter functions to the set of one-parameter functions such that ] f(x(n)) = [the function y defined by y(n) =] n * x(n)

We could then meaningfully ask about f(x)(n): the function y obtained by applying f to x, applied to argument n. This would produce an integer. e.g. f(x)(3) = 3.
Or we could ask about f(x(n)), treating n as a dummy variable. This result would be a function and could be presented as a list of values as was done in the original post.

 

[RaduAndrei]

What about my confusion on post #7?

 

[jbriggs444]

What about my confusion on post #7?

The relevant definition being...

f(x(n)) = f(u(n)-u(n-4)) = f(u(n)) - f(u(n-4)) = nu(n) - (n-4)u(n-4)

I am in deep water here. Fighting to interpret the notation and understand your difficulty both. So bear with me while I try to parse this thing out.

f(x(n)) = f( u(n) - u(n-4) )​

So we have this operator f. It takes in a function x and spits out a function y. This first part of your assertion looks solid. It is mere definition.

f( u(n) - u(n-4) ) = f(u(n)) - f(u(n-4))​

This part looks dodgy. It seems to assume that f is linear. But let's skip past that.

f(u(n)) - f(u(n-4)) = nu(n) - (n-4)u(n-4)​

This part looks wrong. Try evaluating f(u(n-4)). You should not get (n-4)u(n-4)

 

[davidmoore63@y]

Here 's another illustration of the fact that you can't use the notation f(x(n)) = n x(n) to map the function x(n) to y(n):

Evaluate f(1). The problem is that you can't tell whether n is 0,1,2 or 3. Each of these gives a different result.

 

[jbriggs444]

Here 's another illustration of the fact that you can't use the notation f(x(n)) = n x(n) to map the function x(n) to y(n):

Evaluate f(1). The problem is that you can't tell whether n is 0,1,2 or 3. Each of these gives a different result.

If we interpret "1" as the constant function which takes all inputs to 1 and take f() as defined in the original post then f(1) is function g(x) = x * 1(x) = x. That is f(1) is the identity mapping.

f(1)(1) = 1. f(1)(2) = 2, f(1)(n) = n

 

[RaduAndrei]

f(u(n)) - f(u(n-4)) = nu(n) - (n-4)u(n-4)​

This part looks wrong. Try evaluating f(u(n-4)). You should not get (n-4)u(n-4)

But surely:
if y(n) = n*x(n) then y(n-4) = (n-4)*x(n-4)

I mean, view the function as a black box. If I input x(n), then I will get n*x(n). If I input instead at a later time, so I input x(n-4), then I should get (n-4)x(n-4).
So, the moment of time where I input the function should not matter.

You can view the waveforms on wolfram alpha.
If you write unitstep(x) then you will see a unit step function. And to this it corrresponds x*unitstep(x).
If I write unitstep(x-4), then this function starts at x=4. but is the same as unitstep(x). And to this it corresponds (x-4)*unitstep(x-4) which has the same form as x*unitstep(x), only delayed in time.

But x*unitstep(x-4) has a very different form.

 

[jbriggs444]

f(u(n)) - f(u(n-4)) = nu(n) - (n-4)u(n-4)
This part looks wrong. Try evaluating f(u(n-4)). You should not get (n-4)u(n-4)

But surely:
if y(n) = n*x(n) then y(n-4) = (n-4)*x(n-4)

f, as you have claimed to define it is a function that transforms one function into a different function.
y as you have defined it above is a function that transforms values into different values.

Let's evaluate f(u(n-4)).

u() is the unit step function. u(x) = 1 for x >= 0 and u(x) = 0 for x < 0.
u(n-4) is the unit step function shifted right by 4. so u(n-4)(x) = 1 for x >= 4 and u(n-4)(x) = 1 for x < 4.

Note that because f is a function that takes functions as input we must interpret its argument as a function and not as an integer value. That is why I have chosen to use the above notation -- to stress that "u(n-4)" is a function.

f(u(n-4)) transforms this right-shifted step function by multiplying each of its computed results by n.

f(u(n-4))(1) = 1*u(n-4)(1) = 1*u(1-4) = 0
f(u(n-4))(2) = 2*u(n-4)(2) = 2*u(2-4) = 0
f(u(n-4))(3) = 3*u(n-4)(3) = 3*u(3-4) = 0
f(u(n-4))(4) = 4*u(n-4)(4) = 4*u(4-4) = 4
f(u(n-4))(n) = n*u(n-4)(n) = n*u(n-4) = [n for n >= 4 and 0 for n < 4]

 

[RaduAndrei]

f(u(n-4))(1) = 1*u(n-4)(1) = 1*u(1-4) = 0
f(u(n-4))(2) = 2*u(n-4)(2) = 2*u(2-4) = 0
f(u(n-4))(3) = 3*u(n-4)(3) = 3*u(3-4) = 0
f(u(n-4))(4) = 4*u(n-4)(4) = 4*u(4-4) = 4
f(u(n-4))(n) = n*u(n-4)(n) = n*u(n-4) = [n for n >= 4 and 0 for n < 4]

Ok, I agree. I asked other people and they all say is n*u(n-4).

Now, let's take it a step further and talk about linear, time-invariant systems.

Consider that the response of the system to u(n) is s(n) = (1-a^(n+1))*u(n).

Basically, is the same as n*u(n). I just replaced n with the exponential (1-a^(n+1)), where a is given.

So to u(n) it corresponds s(n).

Now if I want to find the response to x(n) = u(n) - u(n-4), I apply the superposition principle and do the following thing:

- I calculate the output of the system separately: so to u(n) it corresponds s(n), and to u(n-4) it corresponds s(n-4)
- I add the outputs to obtain y(n) = s(n)-s(n-4) = (1-a^(n+1))*u(n) - (1-a^(n-4+1))*u(n-4).

So, observe that to u(n-4) it corresponds (1-a^(n-4+1))*u(n-4) and not (1-a^(n+1))*u(n-4).

This is what happens in an LTI system. I can't say that to u(n-4) it corresponds (1-a^(n+1))*u(n-4) because if I take not 4, but 100, then at n = 100, when the step down happens, the exponential would have died and I don't want this. I don't want this because when the step goes down, I want it to go down exponential and not suddenly.

So, what happens? Aren't linear, time-invariant systems described by functions but by other mathematical objects?

 

[RaduAndrei]

I figured it out. y(n) = nx(n) is linear but not time-invariant. That is why it does not work what I was trying to do.

For example, the RC circuit is LTI but the relationship between the input and output is not time-invariant.

Thanks for your time.