The 9/11 Forum

OneWhiteEye:

Thanks for the post! On the question of the correct labeling of the floors take a look at Figure 5-13 on page 105 of NCSTAR 1-9. It's a very nice drawing of the north face of WTC 7 with the floors all labeled. It shows the issue with the 47th floor and the roof parapet and allows one to make more accurate estimates of "true" floor heights. If you print out a copy of the Figure you will find it scales as 1 mm on the drawing represents 1.036 meters on the building. It suggests that you need to add an extra 6 meters at the top of WTC 7 to allow for the parapet wall.

I have also done a nice little exercise in trig to see what portion of WTC 7 should be visible above 101 Barclay St. First of all check out the link below for a good photo of WTC 7 with the 101 Barclay Street building, (Bank of New York), in front of it:

http://wirednewyork.com/wtc/images/wtc_wsh.jpg

Clearly the CBS video was taken from a location that was further north from WTC 7 than the linked photo, but the image is still very useful to see the overall geometry of the buildings of interest.

According to the map reproduced by NIST in Figure 5-183, (on page 263 of NCSTAR 1-9), the CBS video was taken at a lcation 0.63 km north of WTC 7. I have also determined that the lowest step on the roofline of 101 Barclay visible in the CBS video is 84 meters high. From large scale maps I estimate that the front of 101 Barclay was 490 meters from the CBS camera location, so the angle subtended by the 84 meter-high section of 101 Barclay is calculated from tan(theta) = 84/490 = 0.171, or theta = 9.728 degrees. Now from similar triangles, the height of WTC 7 obscured by 101 Barclay is 630 tan(9.728) = 630 x 0.171 = 108 meters. Hence the lowest level of WTC 7 that should be visible in the CBS video is (186 - 108) meters = 78 meters. Now if we allow 6 meters for the parapet wall we have 72 meters-worth of floors at 4 meters per floor = 18 stories, as observed!

Here are four frames of roofline, at times 7, 8, 9, and 9.33s:



At frame 295, the roofline's deviation from a straight line corner-to-corner is shown here (the vertical distance from the line to the kink is 9 pixels):



The last frame number was chosen because its right after the penthouse roofline disappears, allowing unambiguous sighting of the main roofline. It also represents more position change than the other one-second intervals preceding it.

Manual measurements of position change, in pixels, of the two corners:

Code: Select all

Pixel deflection; right and down are positive

NE Corner
Frame   dVert   dHorz
-----------------------
225   -----   -----
255      1     -1
285      2     -9
295      6    -11


NW Corner
Frame   dVert   dHorz
-----------------------
225   -----   -----
255      0      1
285      1      0
295      3      0


Observations, all of them I think obvious:

- Vertical pixel deflection is reasonably synonymous with true vertical, horizontal is not (necessarily) so
- The NW corner (very roughly) goes right a small amount and returns, does not move left of original location
- The NE corner goes left immediately and quite far
- The NE corner drops sooner and farther than the NW corner
- The kink leads the progression, eventually leading the corner average by 9px at the end of the interval

There is a slight impression, when stepping through frames in this interval, that the roofline curves a bit before the kink becomes pronounced. Certainly, in the last still, the roofline segments from kink to corners both seem pretty straight. There is a distinct impression of the east end of the north face rotating counterclockwise as either the kink moves towards the building interior or the corner moves towards the camera. It's a wrapping motion. The fact that discernable vertical motion is not observed at the NW corner until the interval 8-9 seconds means my first approximation choice of t0 at ~8.77s is good. I need check to see how that compares with 0.6 seconds into einsteen's data. With NIST difference images, a good choice of t0 is around 7.5 seconds because those images emphasize the center sag.

Distortion is of sufficient magnitude that tracking a feature in 2D is the only easy way of even getting the actual vertical pixel displacement of a point on the face or corner; a one dimensional slice one pixel wide will rapidly veer off in most cases. Unless the target is a line running horizontally, which areas near the NW corner roofline approximate, raw data must have correction for the change in intersection point as the feature moves laterally.

What follows is that vertical smearograms done on the roofline will have deviation/distortion due to the non-uniform vertical, lateral and axial motion of the target. Before the drop, there will be fluctuations reflecting the building's oscillations and distortion. At drop initiation, this 'drift' could be either additive or subtractive with non-zero slope. During the fall, as the upper section deforms and shifts laterally, the vertical slice intersects the roofline at different locations, factoring in the segment angles and any change in perspective which results.

The descent can be treated as a group of individual vertical slices of building, each following its own 1D approximation, t0, and curve. The descent of the roofline could be examined by aggregating the slices. Or a corners/kink average could be done in both dimensions. The two roofline segments could be mapped by linear regression. The entire visible face could be tracked.

I was working on the last post when you put in yours, Dr. G, I need some time to digest. All of the above was written before reading your post, which I will now do.

Dr. G wrote:According to the map reproduced by NIST in Figure 5-183, (on page 263 of NCSTAR 1-9), the CBS video was taken at a lcation 0.63 km north of WTC 7. I have also determined that the lowest step on the roofline of 101 Barclay visible in the CBS video is 84 meters high. From large scale maps I estimate that the front of 101 Barclay was 490 meters from the CBS camera location, so the angle subtended by the 84 meter-high section of 101 Barclay is calculated from tan(theta) = 84/490 = 0.171, or theta = 9.728 degrees. Now from similar triangles, the height of WTC 7 obscured by 101 Barclay is 630 tan(9.728) = 630 x 0.171 = 108 meters. Hence the lowest level of WTC 7 that should be visible in the CBS video is (186 - 108) meters = 78 meters. Now if we allow 6 meters for the parapet wall we have 72 meters-worth of floors at 4 meters per floor = 18 stories, as observed!

Excellent! You've got it nailed down.

This is more like what NIST did in its series of images. Here traces of the building top are overlaid, with frame 225 blue and 295 green.



The corners have barely moved but the inner portion has sagged considerably. This illustrates well the need to consider each vertical slice independently, and not expect them to move at the same time or in the same amount. The NW corner stays up the longest and falls almost collinearly with vertical in the image. It's the best choice for smear, but how much of the upper building does it represent?

This compares frame 225 (7 seconds) with 255, 285, and 295. The black trace is frame 225 and the translucent green is later frames, which allows the black to show through from underneath as darker green.

OneWhiteEye:

It looks for sure like the top of WTC 7 was "a-rockin' and a-rollin" in the minutes before the building collapsed.

Appendix C of NCSTAR 1-9 provides data on the pre-collapse horizontal east-west oscillations of the NW corner of WTC 7. These oscillations were over a range of 14 inches and periods up to 4 seconds. I am quite sure there were north-south oscillations too, but these would be much harder to detect because of the orientations of the cameras in all the well-known videos of WTC 7, (i.e. videos recorded from the north to the northeast of WTC 7).

I also think significant deviations from verticality were observable during the collapse of WTC 7, as seen in several videos.

One of the most interesting deviations from verticality is visible in Figure 5-212 of NCSTAR 1-9 which shows a deviation of about 5 degrees to the east of vertical in the alignment of the NW edge of WTC 7 about 3.5 seconds into the building's collapse. This occurs when the roofline has dropped about 50 meters. Deviations in the verticality of the NW and NE edges of WTC 7 are also shown in Figure 5-205 on page 279 of NCSTAR 1-9. Unfortunately these deviations are not entirely consistent with Figure 5-212.

Dr. G:

Yes, it will be difficult to apply meaningful interpretations to the potentially large amount of data to be taken. Great! This is going to be much more fun than I thought.

While the data I originally posted seems to be bad, and is too much pain to either validate or refute, there may be some real basis to some part of the drift, given there is some flex and sway through that time. There's no way to use the sub-pixel data, but this part which I did post later:



is nearest integer pixel and is probably OK.

This is the vertical coordinate of the bottom-most pixel classified as part of the dark rectangle. It uses a fixed intensity threshold mask so is very similar to what einsteen recently did for the NW corner. The primary differences: the target is the left bottom of the rectangle from a different video and it's a 2D method, which theoretically means it follows the bottom point despite lateral wandering. As such, it doesn't suffer some of these problems we've noted would affect a smear. I've no visual confirmation of its accuracy.

Good data from the same locations should allow the videos to be synced but NIST's t0 must also be found.

One of the NIST definitions of t0 is from the caption of Figure 5-205, to which Dr. G referred:

NIST: Fig. 5-205 wrote:Times are referenced to the time the east penthouse began to descend into the building.

In the CBS video, there is some slow jitter prior to the penthouse roof descent but, ignoring that, no apparent drop in the first 40 frames. The roof begins to sag noticeably from 43 to 44. Frame 42-43 is therefore where I'd put the NIST t0. Have a look at the penthouse in frames 0, 40, 45, 50, 55, and 60 - spanning the first two seconds of the video but jumping through the first 1.3 seconds:



Through at least frame 60, all of the motion is bowing of the roofline and the corners have not dropped perceptibly (at least the right hasn't), so it is possible that their t0 is actually later than this; it depends on what 'descend' means to them. I know I've run across the definition in other places and maybe it's more specific but I'll look later. The problem with waiting until the penthouse as a whole descends is that there's not really a specific objective time where that occurs. I'd say between frames 60-74, but that spans a half second, and someone else might judge it to be even later. First appearance of gross bowing can be judged to the nearest frame without ambiguity, that's why I'd expect 42 = 1.4 sec to be their t0.

If that's the case, there should be good correspondence between frames 261, 276, and 291 of the video and the first three images of Fig 5-205 labeled 7.3, 7.8, and 8.3s respectively. Looking at it, I can only say that it appears to be a reasonable match between them. There is more to be done to confirm or refine the time, but I'm ready to work with NIST t0 = CBS 1.4 sec for now.

Dr. G, any opinions on t0 placement as NIST defines it?

Dr. G wrote:One of the most interesting deviations from verticality is visible in Figure 5-212 of NCSTAR 1-9 which shows a deviation of about 5 degrees to the east of vertical in the alignment of the NW edge of WTC 7 about 3.5 seconds into the building's collapse. This occurs when the roofline has dropped about 50 meters.

Have you seen the video this frame is from? I have a dog-eared copy, I was going to trot it out later. Here are some extracted images, the first second of what I have (cropped):









Unfortunately, the stills do not convey the downward motion well since the upper portion of the view is washed out. You have to follow the smoke trails on the left. The corner on the left folds over the face in the wrapping motion I described earlier.

I would not necessarily suggest use of the data/curve above, neither would I discourage it. Just in case, here are the numbers. At the horizontal location of the measured point, 240/11 = 21.8 px/fl. With 3.96 m/fl, that's 0.182 m/px. Based on that, and assuming 30fps (it could be 29.97 - I have some question), the data above translates to:

Code: Select all

time_s,dy_m
0.000,0.000
0.033,0.000
0.067,0.000
0.100,0.000
0.133,0.000
0.167,0.000
0.200,0.000
0.233,0.000
0.267,0.000
0.300,0.000
0.333,0.000
0.367,0.000
0.400,0.000
0.433,0.000
0.467,0.182
0.500,0.000
0.533,0.000
0.567,0.000
0.600,0.000
0.633,0.000
0.667,0.000
0.700,0.182
0.733,0.000
0.767,0.000
0.800,0.182
0.833,0.000
0.867,0.000
0.900,0.000
0.933,0.182
0.967,0.000
1.000,0.000
1.033,0.182
1.067,0.000
1.100,0.000
1.133,0.000
1.167,0.182
1.200,0.000
1.233,0.000
1.267,0.182
1.300,0.182
1.333,0.182
1.367,0.182
1.400,0.182
1.433,0.182
1.467,0.182
1.500,0.182
1.533,0.182
1.567,0.182
1.600,0.182
1.633,0.182
1.667,0.182
1.700,0.182
1.733,0.182
1.767,0.000
1.800,0.000
1.833,0.182
1.867,0.182
1.900,0.000
1.933,0.000
1.967,0.000
2.000,0.000
2.033,0.000
2.067,0.000
2.100,0.182
2.133,0.182
2.167,0.182
2.200,0.000
2.233,0.000
2.267,0.000
2.300,0.000
2.333,0.000
2.367,0.000
2.400,0.000
2.433,0.000
2.467,0.000
2.500,0.000
2.533,0.000
2.567,0.000
2.600,0.000
2.633,0.182
2.667,0.182
2.700,0.182
2.733,0.182
2.767,0.182
2.800,0.182
2.833,0.182
2.867,0.182
2.900,0.182
2.933,0.182
2.967,0.182
3.000,0.182
3.033,0.182
3.067,0.182
3.100,0.364
3.133,0.182
3.167,0.364
3.200,0.364
3.233,0.364
3.267,0.364
3.300,0.364
3.333,0.182
3.367,0.364
3.400,0.182
3.433,0.364
3.467,0.182
3.500,0.182
3.533,0.182
3.567,0.364
3.600,0.364
3.633,0.364
3.667,0.364
3.700,0.364
3.733,0.364
3.767,0.182
3.800,0.364
3.833,0.364
3.867,0.364
3.900,0.364
3.933,0.364
3.967,0.364
4.000,0.182
4.033,0.182
4.067,0.182
4.100,0.182
4.133,0.182
4.167,0.182
4.200,0.364
4.233,0.182
4.267,0.182
4.300,0.182
4.333,0.182
4.367,0.182
4.400,0.182
4.433,0.364
4.467,0.364
4.500,0.364
4.533,0.182
4.567,0.182
4.600,0.182
4.633,0.182
4.667,0.182
4.700,0.364
4.733,0.182
4.767,0.182
4.800,0.182
4.833,0.182
4.867,0.182
4.900,0.182
4.933,0.182
4.967,0.182
5.000,0.000
5.033,0.000
5.067,0.000
5.100,0.000
5.133,0.000
5.167,0.000
5.200,0.182
5.233,0.182
5.267,0.000
5.300,0.000
5.333,0.000
5.367,0.000
5.400,0.182
5.433,0.182
5.467,0.182
5.500,0.182
5.533,0.000
5.567,0.000
5.600,0.182
5.633,0.182
5.667,0.000
5.700,0.182
5.733,0.182
5.767,0.182
5.800,0.000
5.833,0.000
5.867,0.182
5.900,0.182
5.933,0.182
5.967,0.182
6.000,0.000
6.033,0.000
6.067,0.182
6.100,0.000
6.133,0.364
6.167,0.000
6.200,0.182
6.233,0.182
6.267,0.364
6.300,0.364
6.333,0.364
6.367,0.182
6.400,0.182
6.433,0.364
6.467,0.364
6.500,0.364
6.533,0.364
6.567,0.182
6.600,0.182
6.633,0.182
6.667,0.182
6.700,0.000
6.733,0.000
6.767,0.364
6.800,0.364
6.833,0.182
6.867,0.364
6.900,0.182
6.933,0.182
6.967,0.182
7.000,0.182
7.033,0.182
7.067,0.182
7.100,0.182
7.133,0.182
7.167,0.000
7.200,0.182
7.233,0.000
7.267,0.182
7.300,0.000
7.333,0.000
7.367,0.000
7.400,0.000
7.433,0.000
7.467,0.182
7.500,0.000
7.533,0.000
7.567,0.000
7.600,0.000
7.633,0.000
7.667,0.000
7.700,0.000
7.733,0.000
7.767,0.000
7.800,0.000
7.833,0.000
7.867,0.000
7.900,0.000
7.933,0.000
7.967,0.000
8.000,0.000
8.033,0.000
8.067,0.000
8.100,0.000
8.133,0.000
8.167,0.000
8.200,0.000
8.233,0.000
8.267,0.000
8.300,0.000
8.333,0.000
8.367,0.000
8.400,0.000
8.433,0.000
8.467,0.000
8.500,0.000
8.533,0.000
8.567,0.000
8.600,0.000
8.633,0.000
8.667,0.000
8.700,0.182
8.733,0.000
8.767,0.182
8.800,0.000
8.833,0.000
8.867,0.000
8.900,0.182
8.933,0.182
8.967,0.182
9.000,0.364
9.033,0.364
9.067,0.364
9.100,0.364
9.133,0.364
9.167,0.364
9.200,0.546
9.233,0.728
9.267,0.728
9.300,0.728
9.333,0.728
9.367,0.910
9.400,0.728
9.433,0.728
9.467,1.456
9.500,1.092
9.533,1.092
9.567,1.456
9.600,1.456
9.633,1.456
9.667,2.730
9.700,1.820
9.733,1.820
9.767,2.002
9.800,2.366
9.833,2.002
9.867,2.184
9.900,2.366
9.933,3.458
9.967,2.912
10.000,2.912
10.033,3.094
10.067,3.094
10.100,3.458
10.133,4.186
10.167,4.004
10.200,4.186
10.233,4.004
10.267,5.096
10.300,5.096
10.333,5.096
10.367,5.278
10.400,6.188
10.433,6.006
10.467,6.916
10.500,6.734
10.533,7.280
10.567,7.462
10.600,8.008
10.633,8.008
10.667,8.372
10.700,9.282
10.733,9.282
10.767,10.374
10.800,10.556
10.833,11.102
10.867,12.012
10.900,12.558
10.933,13.468
10.967,13.286
11.000,14.378
11.033,14.196
11.067,15.470
11.100,16.016
11.133,16.744


Counting down from the top, the 8th light horizontal floor band has two dark dots appear over several frames, offering a perspective-free means of crudely synchronizing the times between the two of the videos einsteen (and dictator cheney) posted.

CBS (29.97 or 30fps):
F292 - No dots
F293 - Left dot light, right dot dark
F294 - Both dark, some deformation on the floor below

Naudet (50fps):
F047 - No dots
F048 - Right dot light
F049 - Right dot dark
F050 - Both dark
F051 - Both dark, some deformation on the floor below

As a first approximation, take these frames to roughly correspond:

F047 <=> F292
F051 <=> F294

and use the midpoint between these.

-----

CBS: (use 30fps for now): 293/30 = 9.77 sec
Naudet: 49/50 = 0.98 sec

The Naudet video clip start time is then ~8.79 seconds into the CBS clip. Relating both to NIST (global not simulation) t0:

CBS clip start time: -1.4 sec
Naudet clip start time: +7.39 sec

Now, with the frame times of the two videos approximately related, the data einsteen collected can be compared to what might be the one decent automated set of mine. Notes:

- einsteen's time is offset by -0.02 to start at zero
- einsteen's time is further offset by 7.38 to start at NIST time
- my time is offset by -1.4 to start at NIST time
- my displacement is multiplied by (174/186) to match his scaling
- my data doesn't go as far because of smoke contamination
- we are looking at two radically different locations

Mine is red, his are green and blue.




They are different.

OWE:

I really can't comment on how NIST arrived at its t(0)......

I would say this however: Although NIST claim that roofline motion is detectable 6.9 seconds after the East Penthouse collapsed, you wonder why the set of images in Figure 5-205 start at t = 7.3 seconds. Interestingly, the next image, (at 7.8 seconds), still shows no real downward motion of the roof at the NE and NW corners, but we do see the start of a roofline bowing motion near the center of the north face.

The plot you posted showing your, (near-center of roofline motion), data vs. Einsteen's (NW corner of roofline motion) data is very revealing in this regard. I would say the center motion shows a t(0) ~ 7.3 seconds, while the NW corner motion has a t(0) ~ 7.8 seconds - or 1/2 second later.

I would also say that the bowing which started at 7.3 seconds eventually pulled the center of the north face roofline down by as much as 2 meters. This behavior is nicely reflected in your curve vs. Einsteen's curve, which shows that after t = 8.3 seconds, the drop of the center of the roof was always about 2 meters ahead of the NW corner drop.

I thought I posted this afternon, but guess I pressed the wrong button. I had no idea that you guys are hiding here, a lot of info here. Cool graphs, it is not strange that the blue and green one are almost the same, they are made using 2 videos from the same Naudet DVD, DC had some problems with ripping them, the only difference is the area at which the smearograms are taken, the offset in both directions should only be a translation in the plot. Indeed the tower moves also a little bit to the right during the collapse and therefore its track is longer and therefore it should be a small correction if it didn't tilt, it also tilts in the other dimension. the NIST uses some professional techniques in their appendix, but I couldn't really find (using the search function...) how they estimated the collapse itself ? OneWhiteEye, your red line is interesting, is that the exponential one that dr. G fitted, I've to re-read this thread I think

Here as promised the primitive method to create smearograms, OneWhiteEye gave me some ruby scripts but after I restored an O/S image I never used it again, sorry OWE...you're the man of that!

http://www.megaupload.com/?d=JNMLLCLW