Not all points descend at the same rate, but here's some data for the NW corner...
http://femr2.ucoz.com/_ph/1/2/143855524.jpg
(click to zoom)
http://femr2.ucoz.com/_ph/1/2/172155712.jpg
(click to zoom)
I'm sure you can extrapolate to determine a good approximation of full descent time. Makes for some interesting numbers, especially how close to *g* it is
I already posted the correct numbers
HERE
The last column of that table is "percent of G".
Over the first 2.25 seconds, it rises, continuously & gradually, from 0% to about 85% of G.
___
If you want to do this with your own data, here is the RIGHT way to do it. (Most specifically NOT the way that Chandler did.)
Note that NIST, with access to REAL experts, found that the best that they could do with video footage (when trying to determine the speed of Flight 175) was ±6 feet. You might be able to get slightly better than this, due to smaller distances between the cameras & WTC7, but not much. And you don't have the knowledge or background of the professionals.
Get the best distance vs. time that you can. Generate an empirical equation. If your equation starts to diverge at the boundaries of your time of interest, then introduce an artificial control point well outside of the real data. Use this control point to draw your curve up or down, as needed to fit the raw data as best you can, and to keep it well-behaved.
If you can't get a single equation to fit all your data, do it with a couple of piece-wise empirical equations that have overlapping domains. Try to minimize any discontinuities in slope or magnitude as you transition between one curve & the next by using those artificial control points.
Then do your analytic single & double time differentiation for velocity & acceleration.
This will be the best representation of the "real" velocity & acceleration of the wall.
If you want to post your real data, I can do this for you in about 10 minutes.
An accurate error analysis - in integral part of any study - would take significantly longer. And requires knowledge of the performance details of the video camera systems, their positions, etc. From our discussions in the past, I am fairly convinced that you would not be able to execute this.
But you could get real close by using NIST's ±6' and then "scaling" the distances between camera & object. But only as long as you have video that is as high quality as that to which NIST had access. If you use compressed video, then the errors will be much higher.
____
Here is the germane point:
Look at NIST's 12-76 of NCSTAR1-9. (pg. 602 of Vol 2).
The raw data - the CORRECT raw data - is the downward displacement vs. time, taken from the videos. This is high quality data, as shown by the excellent quality of the fit all the way out to about 5.25 seconds.
The slope of this curve is also an excellent representation of the velocity at all points within the domain of the raw data (i.e., out to about 5.25 seconds). The last data point @ 5.4 seconds is off a bit.
Now look at Fig 12-77 on the next page.
The black curve is a plot of the calculated time derivative of the empirical curve in Fig 12-76. The QUALITY of this curve's reflection of the real value is given, NOT by any info in curve 12-77, but by the quality of the fit between the empirical curve & the raw data in Fig 12-76.
The single BEST representation of the "real" downward velocity is the black curve. All the way out to about 5.0 seconds. (Again, the empirical equation often starts to diverge at the edges of the data domain.)
The black dots in Fig 12-77 instantaneous velocity are NOT raw data. They are NOT "right". They are mediocre approximations calculated from the raw data.
The curve is a much better representation of the true downward velocity than black dots are.
This is why Chandler's refusal to provide his raw data is so egregious.
Now consider Chandler's "prize": his linear approximation.
He uses mediocre approximations to produce a POOR linearization. And produces no error analysis.
Excuse me for not getting all a-twitter about either his technique or his conclusions.
Tom