🚀 go-pugleaf

I'm a model rocket hobbyist, investigating the predictive
validity of rocket altitude simulation programs. These programs
use a subsonic drag model (Drag Force = -K*Vel^2) to numerically
predict altitudes of model rockets, which are fitted with (legal,
non-metallic) professionally manufactured engines of known
characteristics. No one in the hobby has done this investigation
before because K is, in general, unknown. The programs also
predict ascent time, however, and it turns out that if you know
any one of K, Altitude, or ascent time, you can find the other
two. Thus, for each rocket configuration, I graphed allowable
values of ascent time and altitude (each of which corresponded to
a unique K); I then measured altitude and ascent time directly.
The measured points fell nowhere near the curves. Actual ascent
times were, in general, much shorter than simulated ascent times
for any measured altitude. More sophisticated software, using
Runge-Kutte methods didn't do much better.

Altitude was measured with 3 elevation-only theodolites, and
since vertical trajectories were required, the only accepted
flights were those that yielded horizontal coordinates close to
the launch pad coordinates. Speeds, estimated directly from the
data, are very squarely sub-sonic (using twice average speed).
Altitudes are all below 700 feet. Rocket lengths varied between
15 cm and 1 meter. Sensitivity analysis on impulse (I have
statistical data on this from static tests), time measurement,
launch angle, and altitude demonstrate that the discrepancies are
real. Also, I have just found some data taken by another
investigator on another project that show the same, very-short
ascent times. 

Could anyone direct me to a helpful source? I'm not a physicist,
but I've had math through complex analysis and linear algebra.
Also, any helpful comments would be appreciated. Could this, for
example, be the result of viscous flow at low speeds?

Thanks in advance (Sorry it's so Newtonian),
-Larry C.