1. Mick West

    Mick West Administrator Staff Member

    On a general note, it's very easy to make math mistakes in a case like this with a variety of numbers, units, and three dimensions. Please be careful before posting analyses. Create annotated diagrams if possible. Check against what has come before. Do sanity checks by working things out in more than one way.
  2. Mick West

    Mick West Administrator Staff Member

    For spreadsheet use, to convert from CAS to TAS you can probably use TAS = CAS*368.9/254 as an accurate multiple over the small variation present.
  3. Redshift7

    Redshift7 New Member

    I'm no bird expert, but here's a tidbit of info about the Albatross that I pulled from Wikipedia:

    I also want to point out that this video isn't new. A TTSA associate had this posted on his Vimeo account as far back as 2 years ago (this has already been mentioned on various social-media platforms, but I thought it was worth stating here too).
  4. Mick West

    Mick West Administrator Staff Member

    He (Jeremy Corbell) has since clarified that this was him overwriting an older upload, so it kept the old date. This is something that happens with Vimeo, and has led to confusion before.
    • Informative Informative x 1
  5. igoddard

    igoddard Active Member

    @Justin. Yes, the more I look at it, the more I suspect the camera's motion might be the source of the object's track. I could try to stabilize the footage on something in the image. Not sure what data you mean.
  6. Kaen

    Kaen Member

    True air speed is by far not the same as ground speed. To get ground speed you must correct TAS for wind speed, and wind speed at 25000 feet can easily be 100-200 knots. Without knowing the wind speed and wind direction, you cannot rely on TAS as a reliable indicator for ground speed.
    Anyhow, the object speed is in the ballpark of a bird's speed...
  7. FWIW, I did a quick & dirty stabilization last night on the object to satisfy my own curiosity:

    • Like Like x 1
  8. Mick West

    Mick West Administrator Staff Member

    Sure, but you are more likely to get it closer if you actually use TAS (369 knots) , and not (as you did) CAS (254 knots).
  9. Justin Shaw

    Justin Shaw New Member

    @Mick West, it seems like ground speed is the most pertanent here. BTW, where did you get 369 knots? I grabbed 258 knots from the flir video.
  10. Mick West

    Mick West Administrator Staff Member


    Summary, 258 is the calibrated air speed. 369 is true air speed.
  11. igoddard

    igoddard Active Member

    Good job! That's how I was thinking of doing it. It really shows how after the object is locked on, its brightness (1) pops out and (2) stabilizes. Surely that's an artifact of the ATFLIR. It would make no sense for it to be an attribute of the object out there, that it changes its nature after a tracker is locked onto it. That implies the pre-tracked target is the best representation of the actual object, which implies that the object has a variable thermal signature, sometimes fading out and sometimes getting brighter.

    Also of note, the speed of the background seems to be constant both before and after lock-on.
  12. The version I was able to download yesterday from washingtonpost.com is 35 seconds 640x360 29.97fps 593kbps.

    Today I can't get past their pay-wall.

    Can you post/attach the 950 pixel version here or provide a link?

    I'm looking for a version with the least amount of compression, so I'm wondering if the WAPO version is an earlier generation than the version that TTSA slapped their logos an graphics on.
    Last edited: Mar 12, 2018
  13. Mick West

    Mick West Administrator Staff Member

    This is what I downloaded from WaPo, it's 1080p.

    Looking at a hex/ASCII dump it seems to have been encoded at some point with:

    I also uploaded it to YT:

    Source: https://www.youtube.com/watch?v=dMnBxI4fTac
  14. Agent K

    Agent K Active Member

    I'd guess that it's an artifact of the video compression of originally interlaced video, and the tracked object is a better representation than when it's untracked.
  15. Mick West

    Mick West Administrator Staff Member

    What are you referring to here? The object is (visually) moving faster than the background, so when the system locks onto the object the camera moves faster, so the background moves faster.
  16. igoddard

    igoddard Active Member

    Yes, I was just going to add that to the contrary of my point, the tracked object may actually be the better representation of the actual object. I think that makes more sense, as the system isn't struggling to image it once its locked-on. In that case its thermal signature looks more spherical than bird-like.

    In the stabilized footage the background's speed doesn't seem to change before or after lock-on (the stabilized views is itself locked on even when the ATLFIR wasn't). This gives me the impression that the background speed after lock-on might not be an artifact of parallax. The object itself may be moving the observed speed relative to the wavy surface.
  17. Agent K

    Agent K Active Member

    The ATFLIR system doesn't struggle to image it, it just slews the sensor and records what it sees. VideoLan's x264 video compression is what struggles with fast changes and motion.
  18. Z.W. Wolf

    Z.W. Wolf Senior Member

    Could you explain the difference between calibrated airspeed and true airspeed?
  19. Agent K

    Agent K Active Member

    Why? That's what I'd expect from parallax. When you track the background, the object moves fast. When you track the object, the background moves fast. The stabilized footage essentially does open-loop tracking, and then continues with closed-loop tracking after lock-on.
  20. Great, thank you.

    Comparing the various versions now. My initial guess is the WAPO version is the version that was used to assemble the TTSA video, i.e. the WAPO version is less compressed and has slightly more pixel detail, debatable if it's actually more useful, though. Curiously, the WAPO video is interlaced.

    Good grief, I wish TTSA would just release a link to the original video release by the DOD, not Youtube. Aren't these videos essentially property of U.S. taxpayers?

    I'd love to recreate this in 3D. But, I suppose the specific details that I'd need about the ATFLIR camera are classified - CMOS/CCD sensor size/focal length/resolution?

    (Interestingly, Luis Elizondo is on Fox News as I'm writing this post).
  21. igoddard

    igoddard Active Member

    Mick's backyard simulation might do well to be updated so as to simulate both the tracking and non-tracking segments of the footage. As it is, it only simulates the tracking portion of the footage. But just thinking about that simulation scene and fixing the camera to a background tree persuades me you're probably right, the ball will move across screen. But the camera motion needs to model the ATFLIR's.
  22. Mick West

    Mick West Administrator Staff Member

    Calibrated airspeed is the speed read by an airspeed indicator that has been calibrated for the air pressure at sea level.

    True airspeed is the actual speed of the plane relative to the air outside the plane.

    I presume that CAS is used in the display because in a fighter jet the CAS affects the maneuverability of the plane more than the TAS. i.e. you can do extreme maneuvers at a faster TAS at higher altitude, but it will be at the same CAS. So the pilot needs to know to know, for example ("is the CAS high/low enough to do a lag roll").

    In this situation, at 25,000 feet barometer altitude the TAS is very roughly 1.5x the CAS (or more accurately 1.45x)
  23. Mick West

    Mick West Administrator Staff Member

    Based on investigation in the GIMBAL thread, the NAR mode of the ATFLIR is 1.5° FOV (square) - although I think there was some small debate about that.
  24. Agent K

    Agent K Active Member

    That may be good, since raw ATFLIR video is interlaced.

    There are some details in an old brochure.

    The sensor is 640x480, but the video may only show a square 480x480 part of it. We've debated whether the narrow (NAR) FOV is 0.7 degrees, 1 degree, or 1.5 degrees based on several sources.
  25. Robert Sheaffer

    Robert Sheaffer New Member

    From Reddit:

  26. Robert Kellogg

    Robert Kellogg New Member

    So a couple things. First, have any of you guys noticed that the WSO and pilot from this video sound really familiar? Uncanny to the likes of the voices from the "Gimbal" video. The "background" radio chatter is there as a distraction imo. This stood out to me right away. I would even have a listen to some of Tommy Delong's interviews, see if you pick up anything.

    Second, I agree with Mick and the ground motion once the target is locked. This is a perfect example of illusory motion and the same effect that is happening in "Gimbal" video with the clouds moving very fast in the background. But I'm not convinced that the "said object" is not initially moving. But, hey the burden of proof is on me right.

    I'm attempting to setup in Prepar3D with the Superbug and TacPACK a scenario of of 25,000 feet and place a drone at the estimated 13k feet. Having it be static at first and see if I can simulate this, then run a scenario of it move at x speed.

    This will take me a couple days to get setup and try a couple passes at it. I will post a YT video and you guys can do what you want with it. Hopefully I will have something by the end of week or this weekend.

    Also, attached some reading material

    Mick or Ian, have either one of you guys tried calling Raytheon directly? Maybe Dr. Angelo Scotty Gilmore could be someone to interviewed.


    Attached Files:

    • Informative Informative x 1
  27. Mick West

    Mick West Administrator Staff Member

    Metabunk 2018-03-12 22-00-47.

    Interesting. The display configuration is all essentially the same. The time code at the bottom is in seconds, with a difference of 991 seconds (16.5 minutes) between the end of GO FAST (4254) and the start of GIMABAL (5245)

    This suggests is just the pilot playing around with the ATFLIR seeing what he can lock on to.
    Last edited: Mar 14, 2018
  28. Agent K

    Agent K Active Member

    Gimbal was reportedly shot off the coast of Florida in 2015. When the NY Post reported that Go Fast "was shot off the East Coast in 2015," I thought it's either the same video or they confused it with Gimbal, but I guess it is part of the same video. Are the voices similar? I guess the WSO is new to ATFLIR and doesn't identify flying objects very well.
  29. Robert Kellogg

    Robert Kellogg New Member

    Nice observation by the reddit poster, but the fact we don't have a place nor a time for the "GIMBAL" it could easily be said that it was part of the GO FAST time frame. I'm sure I will be corrected, but the GIMBAL video has never been really associated with the 2004 Nimitz video, correct?
  30. Agent K

    Agent K Active Member

    Wonder why Gimbal doesn't display the range to the target. Too far away?
  31. Robert Kellogg

    Robert Kellogg New Member

    So, this is the problem. We are seeing only a very small piece of what is going on it the jet. On the other DDI they would have their other system up, well in the case of a two seater FA-18, there is a total of eight DDI's (some call them MFD's). The pilot will typically have one DDI set to NAV and the other set for either RDR or SA. The WSO will have either ATK A/G or A/A on one DDI while the other is in FLIR mode with a configuration of TV or IR mode.

    Depending on what A/A RDR you are using VS, TWS, RWD you will not always get RNG or Vc, Typically it is TGT instead of RNG when in combat configuration. We just do not have this information unfortunately.
  32. Livefire

    Livefire New Member

    Likely to close, its designed to lock onto and track targets at long range. And yes from the first time I watch this new video I immediately thought the pilots sounded the same from the gimbol video. Listen to the "MY GOSH" part in the Gimbal video, and compare it with the "OH MY GOSH, DUDE" part in the Go Fast video, sounds like the same guy to me.
  33. Mick West

    Mick West Administrator Staff Member

    Just doing this with a TAS correction for 45° and 55°

    cos(45 degrees)*cos(27 degrees))*369-210 = 22.5 knots
    cos(55 degrees)*cos(33 degrees)*369/254*256-160 = 18.9 knots

    Indicating the object is flying away from the jet at around 20 knots. (although this all depends on the wind speed, and that's only the component of the objects motion in that direction)

    Is everyone okay with the use of TAS not CAS, as explained previously?
  34. Kaen

    Kaen Member

    Wind speed at 25000 feet kan be quite high, so ground speed will probably be in the range of TAS plus or minus 100 knots, i.e., somewhere between 269 - 496 knots. In this whole range, the object's speed is still within the range of the speed of a large bird, either flying toward or away from the jet.
  35. Justin Shaw

    Justin Shaw New Member

    The true air speed makes a pretty big difference in my simulation. Using straight and level flight (in the x direction) at 369 KTS I compute an object speed of 163 KTS.

    Taking +- 100 KTS winds at 25000 feet this translates to anywhere from 88 KTS to 254 KTS.

    Here is the update graphic with plane at 25000 feet looking down and to the left at the object.
    [Broken External Image]:https://photos.google.com/photo/AF1QipNZ9aq0-IAoyjyuUx7jBshqiypntvsyYlmTS5jp


    from numpy import *
    from pylab import *
    from scipy.interpolate import *
    #		 t sec	az deg,  el deg, range nmi  v?
    az_data = array([
    	[12 + 10./30, -43],
    	[13 + 10./30, -44],
    	[13 + 20./30, -45],
    	[14 + 19./30, -46],
    	[15 + 13./30, -47],
    	[16 + 13./30, -48],
    	[18 +  0./30, -49],
    	[21 +  6./30, -50],
    	[23 +  6./30, -51],
    	[24 + 18./30, -52],
    	[26 +  6./30, -53],
    	[27 + 18./30, -54],
    	[29 +  6./30, -55],
    	[30 + 13./30, -56],
    	[31 + 19./30, -57],
    	[32 + 25./30, -58],
    	[33 +  0./30, -58]
    el_data = array([
    	[12 + 10./30, -26],
    	[13 + 19./30, -27],
    	[16 +  6./30, -28],
    	[18 + 25./30, -29],
    	[21 +  7./30, -30],
    	[23 + 13./30, -31],
    	[25 + 25./30, -32],
    	[28 +  1./30, -33],
    	[30 +  1./30, -34],
    	[32 +  7./30, -35],
    	[33 +  0./30, -35]
    rng_data = array([
    	[12 + 10./30, 4.4],
    	[13 + 10./30, 4.3],
    	[15 +  3./30, 4.2],
    	[17 +  4./30, 4.1],
    	[19 +  1./30, 4.0],
    	[20 + 29./30, 3.9],
    	[22 + 28./30, 3.8],
    	[24 + 28./30, 3.7],
    	[27 +  1./30, 3.6],
    	[29 + 13./30, 3.5],
    	[31 + 22./30, 3.4],
    	[33 +  0./30, 3.4],
    az = interp1d(az_data[:,0], az_data[:,1])
    el = interp1d(el_data[:,0], el_data[:,1])
    rng = interp1d(rng_data[:,0], rng_data[:,1])
    def pos(t):
    	vel = array([258 * KTS, 0, 0])
    	vel = array([369 * KTS, 0, 0])
    	return array([0, 0, 25000 * FEET])[newaxis] + vel[newaxis,:] * (t[:,newaxis] - az_data[0, 0])
    NMI = 1852.
    HOUR = 3600.
    DEG = pi / 180.
    KTS = NMI / HOUR
    FEET = FOOT = .3048
    rel_speed = (diff(rng_data[:,1] * NMI) / diff(rng_data[:,0])) / KTS
    t = arange(15, 30, .1)
    azs = az(t)
    els = el(t)
    rngs = rng(t)
    x = rngs * NMI * cos(azs * DEG) * cos(els * DEG)
    y = -rngs * NMI * sin(azs * DEG) * cos(els * DEG)
    z = rngs * NMI * sin(els * DEG)
    xyz = vstack([x, y, z]).T
    p = pos(t)
    print (((linalg.norm(p[0] + xyz[0] - p[-1] - xyz[-1])) / (t[-1] - t[0])) / KTS)
    ax = subplot(4, 1, 1)
    plot(az_data[:,0], az(az_data[:,0]))
    plot(t, -arctan2(y, x) / DEG)
    ylabel('Az [deg]')
    subplot(4, 1, 2, sharex=ax)
    ylabel('El [deg]')
    plot(el_data[:,0], el(el_data[:,0]))
    plot(t, arcsin(z / linalg.norm(xyz, axis=1)) / DEG)
    subplot(4, 1, 3, sharex=ax)
    ylabel('Range [nmi]')
    plot(rng_data[:,0], rng(rng_data[:,0]))
    plot(t, linalg.norm(xyz, axis=1) / NMI)
    subplot(4, 1, 4, sharex=ax)
    ylabel('Rel vel [kts]')
    xlabel('t [sec]')
    plot(rng_data[2:-1,0], rel_speed[1:-1])
    import matplotlib as mpl
    from mpl_toolkits.mplot3d import Axes3D
    fig = plt.figure()
    ax = fig.gca(projection='3d')
    ax.plot([0], [0], [0], 'bo')
    ax.plot(x / NMI, y / NMI, z / FEET, 'r-')
    ax.plot(x[:1]/NMI, y[:1] / NMI, z[:1] / FEET, 'ro')
    xlabel('x [nmi]')
    ylabel('y [nmi]')
    ax.set_zlabel('z [feet]')
    fig = plt.figure()
    ax = fig.gca(projection='3d')
    ax.plot((p[:,0])/NMI, (p[:,1]) / NMI, (p[:,2]) / FEET, 'b-')
    ax.plot((p[:1,0])/NMI, (p[:1,1]) / NMI, (p[:1,2]) / FEET, 'bo')
    ax.plot((x + p[:,0])/NMI, (y + p[:,1]) / NMI, (z + p[:,2]) / FEET, 'r-')
    ax.plot((x[:1] + p[:1,0])/NMI, (y[:1] + p[:1,1]) / NMI, (z[:1] + p[:1,2]) / FEET, 'ro')
    xlabel('x [nmi]')
    ylabel('y [nmi]')
    ax.set_zlabel('z [feet]')
  36. Kaen

    Kaen Member

    Note that the jet is only flying straight and level with the ATFLIR locked for just 4-5 seconds (from 1:35-1:39 in the 'official' video). Most of the time it is banking left, which would considerably tilt the reference plane of the ATFLIR. I cannot immediately see whether you took this effect into account, just a heads-up...
  37. Kaen

    Kaen Member

    • Agree Agree x 1
  38. Tom Mellett

    Tom Mellett New Member

    Last edited by a moderator: Mar 13, 2018
  39. Kaen

    Kaen Member

    I redid my calculations using TAS instead of CAS and get almost the same results, though I think it’s Bernie:

    Average object altitude: 4022 m (instead of 4039)
    Average object speed: 38 knots (instead of 33 knots)

    Values from left to right in tables:
    t Left Down RNG Vc TAS ALT Vc,F18 Vc,obj Vobj ALTobj

    t = time in ‘official’ video
    Left (angle), Down (angle), RNG, Vc, ALT come from the ATFLIR display
    TAS computed from CAS on ATFLIR display and http://www.hochwarth.com/misc/AviationCalculator.html
    Vc,F18 = TAS x cos(left) x cos(down) = F18’s speed vector projected on vector towards object
    Vc,obj = Vc – Vc,F18 = object’s contribution to closure speed (negative means object is flying away from jet)
    Vobj = Vc,obj / (cos(left) x cos(down)) = estimated speed of object in direction of jet
    ALTobj = ALT – RNG x sin(Down) = computed altitude of object

    Flying level part:
    1:35 43 26 4,4 220 368 25000 242 -22 -33 4,0
    1:36 45 27 4,3 210 369 25000 232 -22 -36 4,0
    1:37 46 27 4,3 210 369 25000 228 -18 -30 4,0
    1:38 47 27 4,2 200 369 25000 224 -24 -40 4,1
    1:39 48 28 4,2 200 369 25000 218 -18 -30 4,0

    Banking left part:
    1:43 50 30 3,9 180 369 25010 205 -25 -46 4,0
    1:44 50 30 3,9 180 369 25010 205 -25 -46 4,0
    1:45 50 30 3,8 180 369 25000 205 -25 -46 4,1
    1:46 51 31 3,8 180 370 25010 200 -20 -36 4,0
    1:47 52 31 3,7 170 369 25000 195 -25 -47 4,1
    1:48 52 32 3,7 170 370 25010 193 -23 -44 4,0
    1:49 53 32 3,7 170 370 25000 189 -19 -37 4,0
    1:50 54 33 3,6 170 370 25010 182 -12 -25 4,0
    1:51 54 33 3,6 160 372 25010 183 -23 -47 4,0
    1:52 55 34 3,5 160 372 25010 177 -17 -36 4,0
    1:53 56 34 3,5 160 373 25010 173 -13 -28 4,0
    1:54 57 34 3,4 150 373 25010 168 -18 -41 4,1
    1:55 58 35 3,4 150 374 25010 162 -12 -28 4,0
  40. Mick West

    Mick West Administrator Staff Member

    We can assume it's in level flight, so /cos(down) makes sense, but isn't the actual possible range of speeds between Vc,obj / (cos(left) x cos(down)) and Vc,obj / (cos(left) x cos(down)) ? I don't think we can infer what the component of the Vobj is in the horizontal plane perpendicular to the LOS, can we?
    • Informative Informative x 1