X''(t) required for reciprocal of X'(t) to be constant

In summary, to have the ETA be constant, the download speed would have to be zero and the acceleration would have to be a function where 1/integral(f(x))=c.
  • #1
NotASmurf
150
2
Hey all, was watching my downloads today and saw that my download speed was dropping at such a rate that the eta of the download was almost constant, so I was wondering what negative acceleration would be required for the eta to be constant, if time left is size/speed, speed is first derivative of time and
acceleration is second derivative of time what acceleration is needed for time left to be constant? So a function where 1/integral( f(x) ) = c Any help appreciated.
 
Last edited:
Physics news on Phys.org
  • #2
The download rate has to be zero.
\begin{align*}
\frac{1}{\int_0^t f(u)\,du}&=C\\
\frac{1}{C}&=\int_0^t f(u)\,du\\
\frac{d}{dt}\frac{1}{C}&=\frac{d}{dt}\int_0^t f(u)\,du\\
0&=f(t)
\end{align*}
 
  • Like
Likes NotASmurf
  • #3
Thanks, 0 is a solution but is there no f'(t) for f(t) where f(t) will approach zero at a rate that 1/integral (f(t)) is constant until it approaches 0? Or perhaps for a certain interval?
 
  • #4
The differential equation is wrong. If we write the correct DE then there are nonzero solutions. There are two possibilities that occur to me:

Possibility 1: remaining time R is estimated as size of remaining download, divided by current download rate. Assume total download size is S, f(t) is the instantaneous download rate at time t, and R is the estimated remaining time, which is constant. Then the download will never complete, only asymptotically approaching completeness, and we require that ##S=\int_0^\infty f(t)\,dt##. So our DE is

$$R = \frac{S-\int_0^\infty f(t)\,dt}{f(t)}$$

This is easily manipulated to a simple, standard DE, whose solution is ##f(t)=\frac{S}{R}e^{-\frac{t}{R}}##.

Possibility 2: remaining time is estimated as remaining download, divided by average download rate so far. Then the DE is:

$$R=\frac{S-\int_0^\infty f(t)\,dt}{\left(\int_0^\infty f(t)\,dt\right)\ /\ t}$$

We can re-arrange this and express it as

$$(R+t)F(t)=St$$

where ##F(t)\equiv\int_0^t f(u)\,du##.

That has solution ##F(t)=C-\frac{SR}{t+R}##, whence ##f(t)=SR(t+R)^{-2}##.
 
  • Like
Likes NotASmurf
  • #5
My many obligations, thank you so much.
 

Similar threads

Back
Top