The 9/11 Forum

For once, I'm at loss for words! It is truly amazing. This requires serious thought. No question it could be real, though there is yet the possibility of a less interesting origin and I do want to rule that out quickly... not entirely sure how. Of course your explanation is plausible and does make sense immediately. I still need time to digest this.

The intrigue level is increasing as the square of time. Thanks for taking this up a notch. Man, you are to data what Popeye is to spinach.


Nice analogy, by the way. Twang.

Thanks OWE!

You know it's hard for me - not being able to do my own smearograms and all that fancy stuff - since I lack the software, the windows XP (or better!) operating system and (of course) the expertise! You see, I really am a computer dinosaur. So the best I can do is copy some of your numbers into an excel spreadsheet and, after some manipulation, plot the results. I would also like to post some of my more interesting plots but the few times I've tried using the image function the plot simply gets lost in cyberspace!

Anyway, I'm glad you find the oscillating roofline at least plausible, but I am well aware of the dangers of misinterpreting computer artifacts. As the saying goes: "Needs more work!"

You know it's hard for me - not being able to do my own smearograms and all that fancy stuff - since I lack the software...

Just about everything I use is free. I don't know what's up with the software field, they're the only bunch that routinely gives away all their work. Maybe after a foreclosure they'll wished they charged something. Anyway, a lot of great stuff is out there for the taking - although I'm still looking for a plotting program that fits like a glove.

...the windows XP (or better!)...

This is more of a problem. I'm not sure what will run on 98, though I'll bet we can come up with a few things. The issue may be slowness, to the point of exasperation.

...and (of course) the expertise.

Compared to solving a differential equation, any equation, all of this is simple. If the system you're on weren't an issue, we could get you off and running in a few days. As it is there may yet be some things available to bolster your toolset.

I would also like to post some of my more interesting plots but the few times I've tried using the image function the plot simply gets lost in cyberspace!

Are you trying to upload the image from your computer to the forum? I've never tried that. What I do is use an image host and then post the link. This may work for you, too. To test it out, go to this page:

http://tinypic.com/

click the 'Browse' button (which will open a file selection dialog), select a picture on your machine , then click the 'Upload' button on the page. Hopefully, after a short wait, your picture will uploaded and you'll be taken to a page that has a bunch of links to your newly hosted picture. Just click on one of those text fields and it will automatically copy the link so you can then do a regular paste into the text here. Give it a try. If it works, you're in business.

You see, I really am a computer dinosaur.

Dinosaurs were scary. They kept the mammals down for a long time, and it wasn't the mammals that took them out. If you figure out how to do all this stuff, whatever shall I do?

Anyway, I'm glad you find the oscillating roofline at least plausible...

Just so you know, my 'plausible' is most people's 'Eureka' so don't let my caution dampen your enthusiasm. This is going to be an interesting exercise no matter the outcome.

I am well aware of the dangers of misinterpreting computer artifacts.

Nest post I'll list some possible sources and remedies. I think we can dispatch with anomalous interference pretty rapidly.

OWE:

Thanks for the tips on posting images..... unfortunately, when I tried using the tinypic.com site to upload a graph from one of my excel files as you suggested, I get a "failure to upload" message. ???!!!

Dr. G:

I'm going to guess here, so I may be way off. I really hope I'm right because, if I am, I'm going to see some of your plots today. Bear with me, this is not the best medium for step-by-step explanations, but it's worth a try.

The file will have to be in one of the formats jpg, png, or gif. It could be that you're attempting to upload the file in bitmap (bmp) form, or even the excel file directly, neither of which will work. Windows programs like Paint tend to default to saving files as bitmaps, but these cannot be displayed on a web page directly.

Unless the above is inapplicable, here's what I suggest you try:

- Open up your Excel file in Excel.
- Select the plot you want by clicking on it once (clicking again may select only a component of the graph; we want the whole thing).
- Copy by pressing Ctrl-C or selecting 'Copy' from the 'Edit' menu.
- Open whatever graphics program you have available. As a minimum, your Windows machine will have Paint which, while execrable, may do for these purposes. Paint is located in: All Programs | Accessories | Paint, on the Start button.
- In the graphics program, paste the plot image by pressing Ctrl-V or selecting 'Paste' from the 'Edit' menu.
- If you are using Paint, it may tell you the image is too large and ask if you want to increase size to accommodate it - say OK/yes
- You should see the plot in the graphics program
- Save by selecting 'Save' from the 'File' menu (just like Excel)

This is where it gets sticky. I don't remember if the version of Paint distributed with windows 98 can save in anything other than bitmap format. Seem to recall it could do either JPG or GIF. If it only does bitmaps, then you're going to have to take a detour and get another program before proceeding. If this is case, let me know, and I'll guide you through that detour. For now, I'll proceed as if you can save the image as JPG.

- When you select 'Save' a file save dialog comes up; at the bottom, there should be a dropdown list labeled 'Save as type' which should let you choose a JPG or GIF file type, hopefully. Save it as one of those, of course paying attention to where it's located, because...

-That's where you're going to look when you go to Tiny Pic and try to upload it!

It's possible there's a different problem, but this is a likely culprit. Let me know. If any additional explanation is required, do not hesitate.

And thanks for the plot, great work guys, incredible how a oscillation can be obtained, how averaging can lead to a much higher resolution than one would expect from a video which is in fact a grid. It will be difficult to determine the real error margin, from the gray past I remember something that has to do with the square of the amount of samples... I hope you guys will once create a publication. Dr. G, why don't you invite Bazant, DBB etc to also concentrate on wtc7 ?

einsteen wrote:And thanks for the plot, great work guys, incredible how a oscillation can be obtained, how averaging can lead to a much higher resolution than one would expect from a video which is in fact a grid.

The key is that additional position information can be derived from the color data.

It will be difficult to determine the real error margin, from the gray past I remember something that has to do with the square of the amount of samples...

I have to admit that error margin is troublesome, haven't really looked into yet but it cannot be avoided.

I hope you guys will once create a publication. Dr. G, why don't you invite Bazant, DBB etc to also concentrate on wtc7 ?

And you? The elegance of the smears is what lured me into this.

Dr. G:

I have only started to examine this and something already has alarm bells going off. There are two things I've done wrong that may be responsible for some of this.

In all my posting, I don't think I mentioned that there's a penthouse included in most of the smears. There are two penthouse structures visible in this video, what I've heard called the 'east penthouse' and a lower one that stretches almost to the NW corner, which I'll call the west. Column 260 lies between the two penthouses while column 460 intersects the west penthouse towards the western end. In 260, there's a brief glimpse of the base of the east penthouse rotating up and tearing out as the penthouse tips and descends but, otherwise, the smear depicts true roofline throughout. Column 460 has an upper edge that first follows the west penthouse as it descends. Once it disappears at around frame 293, the edge of the smear is formed by the roofline which begins its drop shortly afterward. This makes two humps at the elbow of column 460 smear.

It probably wouldn't have mattered if I hadn't made mistake #2, which is forgetting the penthouse was there. If I'd remembered, I would have extracted the data in a different way, namely scanning up from the bottom to try to catch the roofline edge before encountering the penthouse edge. This is not as easy as scanning down through the sky and smoke and might not even work without significant tweaking. Not that naive scanning for threshold crossing is only thing possible, it's just the only thing rolling right now.

I do need to look closer before saying for sure, but I wanted to throw that out right away and have you re-examine with this info in mind. The later oscillations are something else, don't know about that yet.

If it turns out this is at least part of the phenomena, I accept full responsibility for sending us down a blind alley.

OWE:

Thanks for the detailed instructions!

Here's the plot (I hope!):



Thanks for the caveat emptor. But don't worry, I too have some problems with my plot. The main one comes from looking at Figures 5-205 and 5-206 of NCSTAR 1-9 which shows some nice overlays of Camera 3 images. What is troubling me is that it sure looks like the NW corner is always above the center of the roofline! My plot suggests otherwise, so I'm not sure how the simple subtraction of the two smearograms could oscillate about zero. Nevertheless, the oscillations could still be real - I just have a (non-physical) offset.

OOPS!

Looks like I messed up!

I have now re-checked my data and what I plotted was [Center Drop - NW Corner Drop], ....... Hence a positive value means that at that point in time the center drop was greater than the NW corner drop.

Also, OWE, based on the further information you have just posted, I am wondering if the early negative portion of the curve is something to do with the penthouses dropping. If the negative excursion is removed from the plot you are left with a more or less steady increase in the difference between the center drop and the NW corner drop which could simply be due to the development of the roofline kink!

Dr. G:

It's so good to see one of your plots here! This should open a new dimension in communication.

(I am reminded of situation a long time ago, when an engineer came by and asked where the data was. I handed him a ream of paper with tabular numeric data on it and he just stood there, wide-eyed. Plots, he said. Why? I asked. You've got everything you need there: initial transient, there's the peak, here's a local minimum, etc. Now, I'm a deep devotee of exploratory data analysis and visualization.)

Also, OWE, based on the further information you have just posted, I am wondering if the early negative portion of the curve is something to do with the penthouses dropping. If the negative excursion is removed from the plot you are left with a more or less steady increase in the difference between the center drop and the NW corner drop which could simply be due to the development of the roofline kink!

Can't say definitively yet, but I think it may be the case that the penthouse has thrown a wrench in the works. At this point, I think the quickest recovery will be to move to a slice closer to the corner, beyond the end of the west penthouse. Scanning upwards gives too many false triggers before the roofline is encountered. The only easy edge is the very top of the building in the image, whether that be rooftop or penthouse. Color and intensity-wise, there's a real deep trough running along the top of the building, very easy to distinguish (refer to the edge detection output I posted after the smears); that's what simple threshold scanning is going to pick up easily. Everything else is much harder.

What I'd like to do now is get some data in the area closer to the corner. The data einsteen posted from the other video is from somewhere in this region, so it should be fairly close after scaling. This will eliminate interference from the penthouse and allow direct comparison. Right from the start, the penthouse adds extra elevation to the baseline. These curves could be compared legitimately after the point the penthouse disappears, but only if the additional height is first subtracted off the C460 data. I don't even know what that is.

PS not to say the penthouse data isn't valuable, just that this is a mixed dataset when raw, and we need to disentangle the two.

I have to give some setup and context before serving up more data. Most of this thread is me thinking out loud, peppered with occasional injections of really interesting stuff like einsteen's video, Dr. G's discussion of Charles Beck's work and faster than g, chek's computer survival advice, and so on. Consider the next post a "READ ME.txt" file that accompanies some important set of files you've received.

The text will explain details specific to the upcoming data as well as some necessary info about general error in video and smears in particular. Also, some demythologizing of the data reduction process. I promise to be succint, but it's not going to be short.

Thanks OWE, ..... could always do with more data!

By the way, if we (you, me, Einsteen, etc) ever manage to write a paper on all this stuff we should call it:

THE MYSTERY OF THE SMEARS

I think that would be a rather cool title ......

Perspective Corrections to Smearograms

I thought it might be useful to post some of my musings on perspective corrections to smearograms. I have worked with the case of the Camera 3 video of the collapse of WTC 7. I am assuming the camera was at street level and 600 meters from the base of Building 7. Here is a schematic of the situation with some important distances marked and labeled:



H = height of selected point on face of building; this is 186 meters for WTC 7
d = drop distance or some reference length such as the height of one story
f = foreshortened length or the apparent value of d as seen from the camera location
R = distance from camera to the base of the building; this is 600 meters for Camera 3

Then, from the geometry, we have:

tan (theta1) = H/R
tan (theta2) = (H-d)/R
tan (theta1 – theta2) = f/(h1 – b1)
h1 = Sqrt[ R^2 + H^2 ]
b1 = H sin (theta1)

I have considered the case of a drop of 4 meters (or the height of one story) at the start and the end of the Camera 3 video. For the start of the motion I took H to be 186 meters, the full height of the building, and for the end of the (visible) motion I assumed the roofline had dropped 60 meters so the effective height of the building was then 126 meters. With these input values I have calculated the foreshortened length, f, of the selected distance (4 meters) at the start and the end of the visible motion.

If my math is correct, the initial foreshortened drop distance, (or apparent floor height) is calculated to be 3.49 meters and the final foreshortened distance comes out to 3.76 meters. This corresponds to a change in the foreshortening over 60 meters of 7.45 %. – not an insignificant change!

OWE, I guess (in theory) you could set up a variable pixel-to-vertical height correction factor that could automatically be applied to each smearogram as it was being generated. Alternatively, the necessary correction factors could be applied to the data as required when drop vs. time plots were being created.

Dr. G wrote:Thanks OWE, ..... could always do with more data!

That's why I need to generate some instead of saying I'm only going to talk about while doing neither!

By the way, if we (you, me, Einsteen, etc) ever manage to write a paper on all this stuff we should call it:

THE MYSTERY OF THE SMEARS

I think that would be a rather cool title ......

A dash of intrigue with a hint of salaciousness, not quite reaching the level of tawdry - I think it will sell well!