I know a thing or two about gunshot imaging in IR, so I'll step into the fray:
On Mon, 29 Nov 1999, Robin Hanson wrote:
> Hal wrote:
> >It would also be helpful in evaluating the film to know what gunfire looks
> >like on FLIR. ...
> I agree, but I think this may be asking too much of the Waco video producers,
> whose funding and access to FLIR equipment is presumably much less than
> the FBI and allied agencies. Hopefully a reenactment will occur.
A single gunshot, fired from a weapon with a modern flash suppressor (such as what would be found on all rifles used at Waco), has a very distinctive appearance when viewed with either near or far IR imaging.
A muzzle blast coming from a typical "birdcage" flash suppressor appears as an expanding torus of light at the muzzle, oriented perpendicular to the path of the bullet. As the torus expands, the incandescence decreases until the fuel/air mixture drops below a level that will sustain combustion. In the visible spectrum, a muzzle blast from this type of suppressor generally produces a star-shaped pattern. It is interesting to note that a flash suppressor in the visible spectrum actually increases flash in the IR spectrum.
>From behind or in front, you see an expanding, glowing donut in the IR
spectrum. From the side, it looks like a pulse with a quick attack and relatively slow release. The brightness, diameter, and duration of the muzzle flash are functions of a large number of things including ammunition used, flash suppressor design, barrel length, atmospheric conditions, etc.
With respect to full-auto, the cyclic rate of most modern small arms is fast enough that muzzle blasts run together between shots, particularly if your perspective is not along the axis of the gun barrel. Automatic fire generally appears as a rapidly pulsating light, where the magnitude of the flash never quite reaches zero. This is primarily because you are seeing the summed incandescence of multiple toruses, with the rate of fade often being slower than the cyclic rate of the weapon in IR.
In the unlikely event that you have IR imaging along the axis of the gun bore (either in front or behind) you will see concentric rings of decreasing brightness.
> Hal wrote a bit later:
> > > technology. "Only an experiment makes sense. But the key is an experiment
> > > that takes into account the low probability of intercept: The way these
> > > cameras work, if 50 flashes were captured on film, there must have been
> > > a far higher number of gunshots that actually occurred."
> >This is a surprising comment, because it seemed to me that the flashes
> >were relatively obvious and well defined compared to the background noise.
> >I'm not sure what effects he is imagining which would lead to a low
> >probability of seeing them. It almost sounds like he's doing a little
> >preparatory CYA in the event that the re-creation doesn't show flashes;
> >he can explain it as due to the "low probability of interecept".
> If the "shutter" were only open for a .01sec period, and then stayed closed
> until the next opening .1sec later, and if a gun flash only lasted .1 sec,
> then you should only expect to see between 10% and 20% of the flashes.
> I don't know what the actual time numbers here are, but the concern seems
> very plausible to me.
The cyclic rate of the weapons most likely used in this instance vary from 600-900rpm.
Assuming the IR imaging was recorded at something approximating broadcast frame rates, there should be enough samples to capture a burst of automatic fire very nicely in most cases. But as you approach the Nyquist limit (frame rates vs weapon cyclic rates), you will start to get aliasing which in this case amounts to "lost" gun fire.
Individual shots fired, while probably captured (the duration of muzzle blast should exceed reasonably fast frame rates), would have wildly different magnitudes from shot to shot. In some cases, the muzzle blast signature would barely register above noise floor depending on when the frame was captured relative to when the bullet left the muzzle.
If the frame rate is relatively slow (e.g. 15 fps), you can expect to lose a large portion of the gunshots. At 30 fps, I would expect to lose almost none.
While some of the above characterizations may not apply in this case (e.g. I doubt you would be able to see the torus at the distances and resolutions involved), it should give a good baseline for discussion. Knowing the frame rates for the IR imaging would be extremely useful in determining if gun shots were "lost" and if lost, what percentage one could reasonably expect to have been captured.