Rather than just list the results, I will step through
the calculations. This will give you more confidence in my
results and allow you to use the same method to do similar
calculations for other conditions.
First we need a good map. I am using the HSCA map which
has a scale of 1 inch equals 20 feet. Next we locate
Willis on the map. I place him at the curb about one foot
east of the elevation marking which is to the west of
column B. That elevation marking is 97.5 feet, which is
close enough to use for Willis. Then we have to add in his
height to camera above the curb. I estimate that at roughly
5 feet because I do not know exactly where his ear will 
be when the sound arrives and he may be leaning over slightly.
Then we have to locate the rifle. We assume it is in window
#1 of the TSBD sixth floor. The elevation of the ledge is
marked on the map as 161.1 feet. I estimate that the muzzle
would be a few inches above the ledge, so I place the
elevation at 161.3. So, we can compute the elevation difference
between the rifle and Willis as follows:
161.3 - 102.5 = 58.8 feet
Then I measure the 2-dimensional distance from the rifle
to Willis. That is 121 feet. Using the Pythagorean Theorem
I calculate the distance in air that the muzzle blast would
travel using the known speed of sound which was 1123 fps:
121 feet on the map would be 134.4 in the air.
134.4 / 1123 = .119679 sec.
We know that the Zapruder camera filmed at 18.3 frames
per second, so:
.119679 x 18.3 = 2.19 frames
So, we have found that the muzzle blast took 2.19 Zapruder
frames to reach Willis. But could the shock wave have
reached Willis before the muzzle blast?
There we have more unknowns. But let us assume, arguendo,
that the rifle was aimed at the limousine and the bullet
either hit near the limousine or passed close to the 
limousine. So, I draw a line from the rifle towards
the middle lane of Elm Street close to the position of
the limousine a couple of frames earlier. The greatest unkown
then is the speed of the bullet. Again I make an assumption
which gives us a ballpark figure. Instead of the average
muzzle velocity found from testing of 2165 fps, I round it
down to 2100 fps, as early shots tend to be slower than
the average in some cases. We also need to know what ammo
was used and how quickly it slowed down. We assume that
it was the WCC M-C ammo which had a rate of slowing down
at a first order approximation of 1 fps per foot.
Next, we drop a line from Willis position to the presumed
trajectory of the bullet so that it forms a right angle 
at the point of intersection. We then measure from the rifle
to that point of intersection. On the HSCA map it is 118 feet,
which equals 132.15 feet in the air. So, the bullet's 
instant velocity at the intersection of the dropped line
will be:
2100 - 132 = 1968 and the average speed over that distance
will be:
2100 - 66 = 2034
132.15 / 2034 = .0649705 secs.
Then we measure the distance of the dropped line
which is 27 feet. We don't need to account for elevation
since we have already accounted for all the elevation in
the calculation of how long it took the bullet to get to
the intersection with the dropped line. Then we calculate
how long the speed of sound takes for the sound of the
bullet to arrive at Willis' position from the bullet path.
27 / 1123 = .024043 secs.
Adding the two together as follows:
.0649705 + .024043 = .0890135 secs.
Then we calculate the Zapruder frames as follows:
.0890135 x 18.3 = 1.629 frames
So we see that Willis was within the shock wave cone
and could hear/feel the shock wave just before the
muzzle blast.
BTW, if you extend the dropped line past Willis and redo
the calculation, you can find the point at which the
shock wave and muzzle blast coincide, which is about 62 feet
from the trajectory. That is about at the west side of the
relfecting pool. This clearly shows why the open microphone
on Houston Street would NOT pick up the sound of the 
shock wave. In this case we would say that its position
was not privileged to be within the cone of the shock wave for
that particular shot, assuming that it was aimed at
the limousine. And indeed the shock waves of the first two
shots were not recorded on the open microphone, although
the shock wave of the grassy knoll shot was because
the open microphone was by then on Elm.
If you make different assumptions about some of the data,
you can simply plug in different numbers into these
calculations.