Kenny Bernstein at the NHRA World Finals 11-10-01 Picture Perfect 4.49 @ 328

Some folks think nitro header flames are just hot exhaust gas, but this isn't true. The exhaust gas temperatures aren't nearly hot enough to make them incandescent. What we see as header flames is real fire, the continuing combustion of unburned cylinder charge after it is pushed out of the engine. It is COMBUSTION flame. The combustion continues and even accelerates once the exhaust charge starts expanding and dissipating into the atmosphere, where it encounters additional oxygen to help it along.

So a header flame actually has two stages: the first at the exit of the header pipe, then a second as it expands into the atmosphere. The first stage can be thought of as a "pulse" having a frequency and duration defined by the mechanics of the engine. The second stage is a flame front whose size and duration is determined by combustion dynamics.

With respect to the first stage, the "frequency" that flame is going to appear at each header pipe is of course a direct function of engine rpm. It is going to appear every exhaust stroke, or once every other engine revolution...1200 times a minute at a "nominal" 2400 rpm idle...4200 times a minute at a "nominal" 8400 rpm maximum engine speed. Expressed in terms of HZ (cycles per second), these figures correspond to 20Hz and 70HZ. (The "nominal" figures are for illustration purposes, and reason I translate this as Hz will be apparent later).

Still with respect to the first stage, the "duration" of the flame pulse at the header tube exit is also a mechanical engine function. It corresponds (more or less) to the length of time the exhaust valve is open (with some variables in flow dynamics interfering to make this an approximation). Let's arbitrarily say it's one third of the 720 degree cycle. That makes it two-thirds the duration of a single engine revolution. So at idle, that works out to 1/60th of a second (.0166 seconds) while at 8400 rpm, it's 1/210th of a second (.0047 seconds).
Remember that with this simplification the duration of each flow pulse is followed by a duration twice as long during which there IS NO flow. These durations, relative to camera shutter speeds, are significant, as will be seen later.

Moving on to the second stage, we have to consider how long it takes for the unburnt fuel in the exhaust pulse to finish burning. This will be hugely variable and dependent on many operational factors, but suffice to say that at racing rpm it is WAY longer than the duration of the exhaust pulse itself, and more on the order of a tenth of a second or even longer. In effect, then, as engine speed increases, we have more and more fresh pulses of exhaust adding to stream of still-burning mixture in the expansion area further away from the pipe. Net result: a "continuous flame" area at some distance outboard.

Thus the answer to the question is that BOTH pulsing and continuous flame areas exist in the headerflame stream of a nitro engine at speed, depending how close to the pipes you are looking.

Now on to how the human eye and cameras perceive this.

Human vision is affected by the phenomenon of "latency" that relates to the response time of our neuroreceptors. In practical terms, we can respond to changes in light intensity only so fast, and the higher the frequency and longer the duration of the light stimulus, the more uniform it appears. Light pulses at 20HZ appear to flicker. At 70Hz they appear continuous. Examples might be how a strobe light looks as you crank up the frequency (the appearance of the illumination from your timing light with the engine at idle vs. high speed) or the "flicker" on a computer monitor with too low a "refresh" rate compared to the smoothness of one that is optimum (72HZ or higher). Motion pictures rely upon this phenomenon to make sequential images presented at 30Hz appear continuous, with only moderate flicker. (They do call 'em "the flicks" after all!) From the above figures concerning header flame frequency, you can appreciate why individual pulses at the pipe tips are generally discernable at idle or off-idle rpm's but will appear to be continuous streams in the racing rpm range.

Camera film and electronic CCD's don't exhibit latency (or at least very little), and with adjustable shutter speeds are able to record events that occur too quickly for human eye response. From the above figures on exhaust pulse frequencies, it is apparent how selection of shutter speed can affect how many cylinders appear to be firing at a given engine speed. The slower the shutter the more likely multiple cylinders will be caught "in the act." The faster the shutter, the more likelihood only one or two will exhibit flame at pipe's tip. You can do the math and see precisely what I mean. Similarly, as far as the "continuous flame area" is concerned, the slower the shutter the larger the continuous flame area will likely appear to be, and the faster the shutter the more definition it is likely to have.

Now for the $64 question about the ESPN super slow-mo shots and the odd headerflame patterns revealed, especially during tire-shake. I can't be sure without knowing more specifics about the camera frame rate and shutter speeds, but the possibility exists that if the shutter speed is "in synch" or "in phase" with engine speed, you could get a strobe-like effect that catches the same cylinder several times in succession while ignoring others, creating an artifact appearance when displayed as a motion picture. Frankly though, I doubt that's the case. I more suspect you are seeing a real phenomenon...basically erratic cylinder firing due to the wilding fluctuating engine loads and component torsional response due to traction variation. Shutter speed is fast enough to catch individual cylinders in the act of misbehaving, which they may only do for a few revs. At least that's my bet. All this would be missed at slower (normal) shutter speeds, which would tend to homogenize everything together much as our eyesight would.