The 9/11 Forum

I'll be posting all new cam#3 data here initially, as and when I generate it.

Any comments along the way, no problem.

Cam#3 - Top Field - Static Region - Vertical - Raw - 29.97 samples/s

(Click to enlarge)

Is the larger version big enough ? Can use any size up to around 4096*4096.

1) The spike around the 95s mark appears to be caused by someone walking past the camera. His head is seen momentarily.
2) There doesn't seem to be any obvious shake caused by descent of the building.

femr2 wrote:Any comments along the way, no problem.

Beautiful.

Cam#3 - Bottom Field - Static Region - Vertical - Raw - 29.97 samples/s

(Click to enlarge)

Cam#3 - Both Fields - Static Region - Vertical - LowPass - 29.97 samples/s

(Click to enlarge)

1) Excellent correlation between upper and lower field motion after filtering.

Q) Anyone got any high-quality high/low pass filters for excel ? I use the following rejigged for excel...

http://en.wikipedia.org/wiki/Low-pass_filter

Code: Select all

// Return RC low-pass filter output samples, given input samples,
// time interval dt, and time constant RC
function lowpass(real[0..n] x, real dt, real RC)
   var real[0..n] y
   var real α := dt / (RC + dt)
   y[0] := x[0]
   for i from 1 to n
       y[i] := α * x[i] + (1-α) * y[i-1]
   return y


http://en.wikipedia.org/wiki/High-pass_filter

Code: Select all

// Return RC high-pass filter output samples, given input samples,
// time interval dt, and time constant RC
function highpass(real[0..n] x, real dt, real RC)
   var real[0..n] y
   var real α := RC / (RC + dt)
   y[0] := x[0]
   for i from 1 to n
     y[i] := α * y[i-1] + α * (x[i] - x[i-1])
   return y


Graph above...

Low pass : dt = 1/(30*1000/1001) : RC = 1

Cam#3 - Top Field - Static Region - Horizontal - Raw - 29.97 samples/s


Cam#3 - Bottom Field - Static Region - Horizontal - Raw - 29.97 samples/s


They are different

Before posting building feature position data, and normalised position data (with static region data subtraction), a quick note...


The best static region accuracy is determined using a large region on the left of the frame, which contains a building fairly close to the camera.

As the amplification of the movement increases with distance from the camera, numerous additional static region traces have been performed to determine a scalar.

The static region trace of the building immediately next to WTC 7 has been compared to the static region traces posted.

The quality (variance) of the further building is a fair bit lower than the close building, and is expected to be so, as the far building video data is much *greyer* and contains much more video noise.

Graphs will be posted to clarify, but a scalar of around *1.2 will be applied to subsequent near-building static region traces before subtraction from building feature traces.

Clear ?

A quick normalised view of NW corner horizontal motion...


Suggest gathering a few comments before proceeding.

Do I trust the slight motion between 60 and 160s ? (Tops out about 2 inches)

If so, then slight motion far in advance of initial twisting.

If not, then solid during that time.

femr2 wrote:Do I trust the slight motion between 60 and 160s ? (Tops out about 2 inches)

You tell me!

If so, then slight motion far in advance of initial twisting.

If so, one of the more interesting things I've seen in a while.

OneWhiteEye wrote:

femr2 wrote:Do I trust the slight motion between 60 and 160s ? (Tops out about 2 inches)

You tell me!

No direct reason not to, but I'll look at additional points to cross-check.

Cam3 - NW Corner - Top Field - Normalised


Top field shows same behaviour.

If another location shows the same thing, then, yes, reasonable to assume true motion. (It's not a *drift*, as it curves from *0* to max and then back to *0*...so it's not a simple side-effect of static point subtraction)

The displacement, if any, seems to happen over about 1.5 minutes? NW corner being pulled in?

Major_Tom wrote:The displacement, if any, seems to happen over about 1.5 minutes?

Yes. Not a displacement, as such, but a to-n-fro.

NW corner being pulled in?

I suggest a long timespan twist, as for the other motion near to descent.

If real, this is pretty big stuff.

You deserve a big congrats for doing this work, regardless of outcome.

It would be necessary to calculate wind flexure, though, I suppose. Doubt it very much. It would have to be a sustained gust and should be obvious in the smoke plumes.

Cam3 - NW Corner - Bottom Field - Normalised - Zoom


Bit of a closer view.

Not sure how to go about wind flexure, but I'd agree unlikely.