unit_step (was Re: New piecewise package indevelopment)

Barton,

I want to kidnap your unit_blip code for the case when pwexpand= 'signum if you don't mind.  Then everything will be signum 
including kron_delta, abs(x) and anything else I can think of.  BTW the derivitive of abs(x) is signum(x) except maybe at zero.  At 
least that one is easy.

Rich



----- Original Message ----- 
From: "Richard Hennessy" <rvh2007 at comcast.net>
To: "Barton Willis" <willisb at unk.edu>; "Maxima List" <maxima at math.utexas.edu>
Sent: Thursday, November 20, 2008 9:38 PM
Subject: Re: [Maxima] unit_step (was Re: New piecewise package indevelopment)


Hi,

You can use the kron_delta(x,a) function for unit_blip(x-a).

As for closed and open intervals, unit_step is right closed and left open.  There is another way to fix that.

For both open

expr*(unit_step(x-4) - unit_step(x-8) - kron_delta(x,8))

For both closed

expr*(unit_step(x-4) - unit_step(x-8) + kron_delta(x,4))

For right open and left closed.

expr*(unit_step(x-4) - unit_step(x-8) +kron_delta(x,4) - kron_delta(x,8))

This is how I have decided to do it in pw.mac.  I also have a switch to use,

pwexpand:unit_step

causes Maxima to use unit_step and

pwexpand:signum

causes Maxima to use signum.  I am still working on the simplification rules for signum and kron_delta.  I have most of
them right but there are still some problems with some of them.

Also (abs(x)+x)/(2*x) is another way but it explodes at zero, so I am not using that.

Rich



----- Original Message ----- 
From: "Barton Willis" <willisb at unk.edu>
To: "Maxima List" <maxima at math.utexas.edu>
Sent: Sunday, November 09, 2008 12:39 PM
Subject: unit_step (was Re: New piecewise package in development)


>Also, I believe that the orthogonal polynomial package has a "unit_step"
>function, though I don't know whether its left-continuous nature is
>compatible with what you need.

Maybe unit_step in orthopoly should be changed to exclusively use
signum. Otherwise, we get too many related functions with non-local
simplifications. Something like (lightly tested):


unit_step_right_continuous(x) := block([s : signum(x)], s *(1-s) / 2 + 1);

unit_step_left_continuous(x) := block([s : signum(x)], s * (1 + s) / 2);

unit_ramp(x) := (x + abs(x))/2;

unit_blip(x) := block([s : signum(x)], (1 + s) * (1 - s));

unit_pulse(x, left ,right) :=
  if is(left = 'open) then (
     if is(right = 'open) then unit_step_left_continuous(x) -
     unit_step_right_continuous(x-1)
     else if right = 'closed then unit_step_left_continuous(x) -
     unit_step_left_continuous(x-1)
     else error("The third argument to unit_pulse must be closed or open"))
  else if is(left = 'closed) then (
    if is(right = 'open) then unit_step_right_continuous(x) -
    unit_step_right_continuous(x-1)
    else if right = 'closed then unit_step_right_continuous(x) -
    unit_step_left_continuous(x-1)
    else error("The third argument to unit_pulse must be closed or open"))
  else error("The second argument to unit_pulse must be closed or open");


Short test:

 (%i78) map(lambda([s], unit_pulse(s, 'closed, 'open)),[0,1]);
 (%o78) [1,0]

 (%i79) map(lambda([s], unit_pulse(s, 'closed, 'closed)),[0,1]);
 (%o79) [1,1]

 (%i80) map(lambda([s], unit_pulse(s, 'open, 'open)),[0,1]);
 (%o80) [0,0]

 (%i81) map(lambda([s], unit_pulse(s, 'open, 'closed)),[0,1]);
 (%o81) [0,1]

 (%i82) unit_step_left_continuous(0);
 (%o82) 0

 (%i83) unit_step_right_continuous(0);
 (%o83) 1

 (%i84) map('unit_blip, [-1,0,42]);
 (%o84) [0,1,0]

unit_pulse(x, 'closed, 'closed) is a mess, but so it goes:

 (%i85) unit_pulse(x, 'closed, 'closed);
 (%o85) ((1-signum(x))*signum(x))/2-(signum(x-1)*(signum(x-1)+1))/2+1

 (%i86) unit_pulse(x, 'closed, 'open);
 (%o86) ((1-signum(x))*signum(x))/2-((1-signum(x-1))*signum(x-1))/2


A nonlocal simplification (that would be hard to do without uniformly
using signum for all these functions:

(%i93) is(equal(unit_step_right_continuous(x) - unit_step_left_continuous
(x), unit_blip(x)));
(%o93) true

Barton



_______________________________________________
Maxima mailing list
Maxima at math.utexas.edu
http://www.math.utexas.edu/mailman/listinfo/maxima

_______________________________________________
Maxima mailing list
Maxima at math.utexas.edu
http://www.math.utexas.edu/mailman/listinfo/maxima