John,
No worries about the delay. I am glad to finally know where that
confidence interval came from - it's been bugging me for several years now
and for some reason I never thought to ask.
I'll try and fold this into my Excel sheets on the matter with your
permission.
Best,
-Matt L.
On Sun, Jan 23, 2022 at 2:27 PM John Schilling <
john.schilling@xxxxxxxxxxxxxx> wrote:
Sorry, I'm a bit behind on email this week.
The underlying methodology has a 95% confidence interval for
ascent-to-orbit delta-V of ~270 m/s; if I'm modeling a launcher with no
other performance benchmarks, I'll just calculate performance for the
nominal, +270 m/s, and -270 m/s cases. But for launchers where reliable
performance data is available across a broad range of missions, I can
see how closely the model matches the actual performance . In addition
to letting me throw in a constant offset to even better match flight
performance, I can usually narrow the error bands quite a bit. For the
Delta IV, all the way down to +/- 75 m/s.
On the other hand, sometimes it goes the other way and I see huge
scatter no matter how carefully I juggle the numbers. Usually those
launchers don't get included in the public version, e.g. the Electron.
The detailed math for calculating error bands, based on degree of
scatter, # of data points, and range of performance covered by those
error bands, is buried in a spreadsheet I use for calibration/
And, if the web site ever goes down, I'll try to make the model
available in some other manner. If necessary, emailing the code and
data files to anyone interested, though that would make version control
impractical so I'd rather keep it on a site I control for now. You are
of course welcome to use your reverse-engineered version as you like.
John Schilling
On 1/17/2022 5:50 PM, Matthew JL wrote:
John - Getting off-topic but one thing that eluded me with the
Silverbird methodology was the confidence interval at the end. I
wound up reverse engineering most of it into an Excel sheet in case
the site went offline (and then putting it into a forward-calculator
for building up a launch vehicle for a given payload and propellant
constraints, again for personal use though I did wind up using it in
Realism Overhaul for KSP) but I never could figure out where the 95%
CI came from. Regarding Electron - maybe it's down to greater
influence from drag rather than purely gravity losses because the BC
is smaller?
Best,
-Matt L.
