The 9/11 Forum

The saga begins here with unhyphenated's post and continues at physorg much later, with Trippy's post. Didn't want the members here left behind the curve, so to speak.

Dr. G wrote:OWE:

Thanks for posting this. I do believe you too found the lost jolt some time ago and did so independently of Trippy. If this discovery is confirmed you deserve some credit for your work, (and I'll buy you a beer next time I visit the pub!)

You're welcome, and I'm sorry we didn't get a chance to sink our teeth into it before it got out. I'll take you up on that beer!

---

Dr. G has known of this for some time and agreed that it deserved closer attention, but it came at a time when neither of us could give it the attention it deserves. At some point, I hope we can put together a solid picture of what's going on here. It comes as a surprise (it could have been known over a year ago, but I didn't look!) and it remains to be seen what can be learned from it. It may turn out to be simply a curiousity, but it seems pretty amazing to be able to see the mechanics in such detail and, personally, I believe it will be useful.

It's possible that this is all a mistake. Doubt it. But even the slightest doubt is reason to hold back; let that be a lesson to any having... premature... tendencies. Once Trippy started talking about it, there was no reason to hold back. However, the real work to verify and then extract meaningful information is yet to come.

Ultimately, Tony Szamboti deserves some credit for finding The Jolts. But only so much, since he didn't recognize them and may come here to argue they're not there. I say 'may', Tony, because you could surprise me. How much credit do you deserve while 'Missing' is still in the title?

femr2 wrote:

OneWhiteEye wrote:The saga begins here with unhyphenated's post and continues at physorg much later, with Trippy's post. Didn't want the members here left behind the curve, so to speak.

OWE, do you want me to upload the resolution enhanced copy of the Sauret footage ? (Using the unfolded view of each interlace frame, and aligning the two sets of data to a known shared point (perhaps your antenna band) would provide accurate placement with a true double-framerate, and allow more confidence in any jolt identification.
(That resolution at double framerate)

Yes, please! But don't knock yourself out, take your time. I've got an even greater backlog of higher priorities now but, more important, neither extreme accuracy nor resolution is required for this one. The existing data, including Tony's (at 6 samples/sec), are just fine for this.

Incidentally, I prematurely freaked about the quality of my original Sauret rip; pretty sure I was stepping through one of the other rips when I saw the nasty. Sure enough, it could be better and that's desired but is second order at this point. I'm only beginning to appreciate just how good the existing stuff is. Even if the rip used suffered from the problems I thought it did, it would suffice for all but the most demanding task, but it doesn't seem to have gross interlacing issues anyway.

I fear I must take this cum grano salis.

First of all, using OneWhiteEye's dish measurements from the antenna mast, there is no evidence of any jolts outside the standard deviation of 0.110--0.122 meters obtained for various assumptions in fitting the data to the B&V crush-down equation. Indeed, one easily calculates from the known building dimensions that at the zone C tilt obtained, that 3--4 floors were in the process of being crushed simultaneously, each from south to north; no tilt to be expected then.

The antenna mast sitting on the hat truss can be considered as a lightly damped extended mass-spring system. As such, each jolt should be considered to be an impulse producing vertical oscillations. If such exist, they are buried in the noise.

Second, taking first differences from noisy data is fraught with hazard. The data needs to be smoothed with an appropriate low-pass filter, such as lowess; this tends to reduce measurement noise. Then heed the advice given on page 186 of Numerical Recipes in C, 2nd edition: first differences are almost guaranteed to produce inaccurate results. Probably in this case the choice is Savitsky-Golay filters to obtain the derviative of a fitted (low order) polynomial. I have never tried that, but have had nothing but terrible failures using the routine on page 188, which attempts to do lots better than mere first differences. Be warned.

David B. Benson wrote:I fear I must take this cum grano salis.

I expect no less; I would, too, if I weren't on this side of the fence. I've made similar arguments to yours in the past, when the salt in my eye kept me from seeing things that were there.

First of all, using OneWhiteEye's dish measurements from the antenna mast, there is no evidence of any jolts outside the standard deviation of 0.110--0.122 meters obtained for various assumptions in fitting the data to the B&V crush-down equation.

How much processing of the data was done before obtaining that result?

Indeed, one easily calculates from the known building dimensions that at the zone C tilt obtained, that 3--4 floors were in the process of being crushed simultaneously, each from south to north; no tilt to be expected then.

That's what one would suppose. However, if the data shows it, and no other explanation can be found, then that supposition needs to be abandoned.

The antenna mast sitting on the hat truss can be considered as a lightly damped extended mass-spring system. As such, each jolt should be considered to be an impulse producing vertical oscillations. If such exist, they are buried in the noise.

First and foremost, it is irrelevant in this case what particular motion might be displayed by the antenna since Tony's measurements are at the roofline and his data shows the dips in the same places. Cannot be antenna-specific. Before objecting on the basis that he measured the washer apparatus, and that's unreliable, consider these points:

- what is the chance that both structural components would display similar pronounced variations in the same place a half dozen times without there being a common underlying cause?

- The two datasets are a very close match, as close as is reasonable given the difference in locations; if you proclaim Tony's data bad, you might as well throw mine away, too

Second, taking first differences from noisy data is fraught with hazard.

This I know. The important thing to remember is the potential hazard is not synonymous with 'guaranteed to be bogus' and I'm sure you know that. The question is, can first differences be meaningful in this case? I say yes. Post-difference averaging will also take care of noise about the mean.

The data needs to be smoothed with an appropriate low-pass filter, such as lowess; this tends to reduce measurement noise.

And, if the frequency and roll-off are inappropriately chosen, real information can be thrown away. Nonetheless, this is NOT high frequency data we're talking about and, unless the filtering is way off, there's no reason to believe it got processed away. But, if you filtered it out, no wonder you aren't seeing it. Whether or not you believe it's observable, if you filter out the possibility of seeing it, don't be surprised if you don't see it.

To me, the most important observation, aside from both datasets showing the same effects, is that the amount of variance is HUGE. At the tail end of the data set, the curve goes essentially flat. At that speed, do you comprehend how much camera shake or noise or whatever-you-want-to-come-up-with would need to be present to account for that? The picture would be unrecognizable! As it is, it's crystal clear and rock steady.


These things have already been covered at physorg so I assume you didn't buy the explanations then. What would it take to convince you?

OneWhiteEye wrote:How much processing of the data was done before obtaining that result?

Only the conversion from (frame, pixels) to (seconds, meters). The fit is betwen that unsmoothed data and the crush-down equation.

However, if the data shows it, and no other explanation can be found, then that supposition needs to be abandoned.

Not at all. From south to north the resistance goes from small to large (over the core) to small again. A speed bump might show up.

First and foremost, it is irrelevant in this case what particular motion might be displayed by the antenna since Tony's measurements are at the roofline and his data shows the dips in the same places.

Nope. Roofline connected to perimeter walls and the latter connected to hat truss. No big dampers in between.

Cannot be antenna-specific.

I make no such claim.

... both structural components ...

? Both roofline and window washer?

The two datasets are a very close match, as close as is reasonable given the difference in locations; if you proclaim Tony's data bad, you might as well throw mine away, too

Yours is obviously of superior quality in comparison to Tony's.

What would it take to convince you?

Send me the raw pixel data, if you have it. If not, then Tony's (seconds,meters). Not the first differences! I'll put it through the B&V crush-down engine and see if the speed bumps significantly show up, outside the noise.

A speed bump might show up.

Well crikey, David, that's what we're talking about! Just because Tony's erroneous calculations indicated it should practically arrest doesn't mean that's the nature of the true jolt. Or even that the jolts match correct energy calcs for an axial impact, it wasn't axial. There are a bunch of such speed bumps. If I were a passenger in the upper block, I dare say I'd call them jolts.

Nope. Roofline connected to perimeter walls and the latter connected to hat truss. No big dampers in btween.

You're talking about real underlying motion, and that's all I was ever talking about. So, what you're saying is, the entire hat truss and everything attached to it can alternate between periods of near free-fall and constant velocity, and that's NOT jolting? To me, anything that's delivered to the roofline and is measurable is significant. If you think the hat truss can do this much shaking without the bulk of mass undergoing a similar jolt, then it seems to me you're arguing the thing was shaking apart. At least hardly what could be called rigid. A premise I don't necessarily disagree with.

I make no such claim.

Sure seems that way when you referred to oscillations of the mast. I didn't know you were also including much of the observable exterior, roofline and such, when you said 'mast'.

? Both roofline and window washer?

I manually verified a few of Tony's points against the roofline, but I've yet to go back to my old manual roofline data.

Yours is obviously of superior quality in comparison to Tony's.

Thank you, and I agree. But his shows the same thing. High sample frequency is not required to see these dips, that's why I'm not sure why you haven't seen them already.

Send me the raw pixel data, if you have it. If not, then Tony's (seconds,meters). Not the first differences! I'll put it through the B&V crush-down engine and see if the speed bumps significantly show up, outside the noise.

Will do. Meantime, go over to physorg and look at the graph I most recently posted to see how a part of the two sets compare, with 80% of my points removed. Shocking, I tell ya.

OneWhiteEye wrote:If I were a passenger in the upper block, I dare say I'd call them jolts.

No evidence of this in your data, just noise.

I didn't know you were also including much of the observable exterior, roofline and such, when you said 'mast'.

I wasn't. The mast might well have had oscdillations of its own, distinct and separate from the rest of zone C. I don't see any evidence of that either. Probably happened, but too small to register above the noise.

? Both roofline and window washer?

I manually verified a few of Tony's points against the roofline, but I've yet to go back to my old manual roofline data.

Now I'm thoroughly confused. Did Tony measre at two locations or only one? Was his measuring object the window washer?

High sample frequency is not required to see these dips, that's why I'm not sure why you haven't seen them already.

Because there is no 2sigma indication of anything except noise.

Meantime, go over to physorg and look at the graph I most recently posted to see how a part of the two sets compare, with 80% of my points removed. Shocking, I tell ya.

Which two data sets? I did look and I was completely confused; not enough context offered.

No evidence of this in your data, just noise.

Then either it's noise or the definition of noise in this context includes what could be real motion. I favor the latter because I'm looking right at it! What I distrust, at this point, is any processing that says that this much of this kind of variance is simple noise. That's too loose a standard for this data. Even Tony's data. They ripped from DVD, too, and while the process was crude and has low time resolution, the spatial is probably good to +/-2 pixels at the very worst.

=>> That's random, magnitude-limited oscillation about a mean.

When the target is moving several pixels per second, there is NO WAY to be that far off in any measurement! Unless you're half blind, or a ******* monkey, as I pointed out at physorg. The dish is only a few pixels in size, even a pixel off-center for three or four frames would stand out like a sore thumb. That's why I produce a validation animation. Here it is:

http://i33.tinypic.com/2zgw2sj.jpg

Nowhere near the kind of excursion necessary, the dot tracks the target exceedingly well. That's where the video says the dish is. What you're saying is you accept noise levels that would have that green dot fly off the picture!

So, whatever math is telling you it's down in the noise should not be your guide. The transcriptions are accurate. If it's a matter of the video signal allowing the apparent position of the dish drift around like that, then you need to square that with the fact that the picture does not wander around all over the place as this would require. It's solid.

Look at this first difference plot:

http://i37.tinypic.com/5wh4q9.png

Tony's data and mine together. Obviously mine is the high-freq. Note the areas circled in green. Tell me what you see.

This is Tony's data in frames and pixels. Note that it's the roofline level and, IMPORTANT, there's about a 15 frame difference in the frame numbers versus mine, you'll see.

Code: Select all

929,0
934,1
939,2
944,4
949,6
954,9
959,13
964,17
969,23
974,29
979,37
984,44
989,52
994,61
999,71
1004,81
1009,92
1014,104
1019,117
1024,130

Whoa.

A few things I'd like to know as data is released:

1) Acceleration until the first jolt.

2) Evenness of jolt: WTC1 from the north seemed to fall very symmetrically. I'm interested in detection of the jolt along the roofline at multiple points along an east-west line. Did the first floor slab contact occur simultaneously along the entire roofline? Did the jolt detected at the NW corner roofline happen at the same time as the NE corner?

3) Acceleration between first and second jolt.

4) What caused the jolts? Indications of column-column impact or floor-floor impact.

(I've always guessed floor-floor, but many people have calculated it as column-column with it's famous "buckling".)

5) Relative magnitudes of first and second impacts....ect (shock pattern, consistency)


Trippy's a trip.

He certainly is invited to post here if he wants to.

OneWhiteEye wrote:

No evidence of this in your data, just noise.

Then either it's noise or the definition of noise in this context includes what could be real motion.

The latter is always the case. But here we have a very good model, the B&V crush-down equation; that establishes the standard deviation. Anything less than 2sigma is just noise. Here is a "jolt" point (seconds, measurements, calculation, difference) @ 1.1607 seconds

Code: Select all

     1.0395   3.429   3.382  -0.047
     1.0679   3.712   3.567  -0.145
     1.1013   3.874   3.792  -0.082
     1.1348   4.164   4.025  -0.138
     1.1607   4.112   4.211  +0.099
     1.1643   4.092   4.238  +0.146
     1.1680   4.298   4.265  -0.033
     1.1900   4.390   4.428  +0.037
     1.2014   4.613   4.513  -0.100
     1.2348   4.789   4.769  -0.019

since the difference changed from -0.138 to +0.099, an excursion of 0.237 in 0.0335 seconds. For this run 2sigma is 0.242 so it doesn't quite meet the 2sigma threshold. Anyway, assuming random errors, this will happen about 1 in 20 times anyway. Now that purported slowdown lasts 0.0295 seconds; doesn't look like much of a speed bump to me.

... the dot tracks the target exceedingly well.

One standard deviation is about 12 cm, as I calculate it. I'd welcome your independent estimate, not using the crush-down equation to obtain it. Anyway, I doubt that even 2sigma would be "off the picture".

Look at this first difference plot:

That, as I explained before, is completely misleading. The only statistically reliable way to observe variations from a model is the way I'm doing it (AFAIK).

And, as I feared ,Tony does not have enough data to obtain a reliable estimate via the crush-down equation.

Major_Tom wrote:Whoa.

Hell yes, whoa.

I have an event I've got to leave for right now. More later.

Edit:

Major_Tom, some of your questions were the things that Dr. G and I were going to look at. But even I'm shy of doing second differences!

David B. Benson - I'll have to take some time with what you've posted but, just to be sure, do NOT look at adjacent differences, this is trending over periods > 1/3 second. There's plenty of noise about the trend, this is lower frequency, like I said earlier.