Sriram Narayan wrote:

It doesn't look that linear to me. I found a website with a similar
graph
and it appears that at sea level and at 10000ft the slope of the
curve is at
least 2x different. Your curve is quite a bit more linear (maybe 20%
increase in slope at 10k). There must some sort of mechanical
compensation
involved otherwise altimeters would be off quite a bit even at 10k
(even
with your curve). Isn't it something like 75ft accuracy requirement
for
altimeters?

http://www.atmosphere.mpg.de/enid/16h.html

His graph has more curvature than mine does because his covers a much
greater vertical distance (11000 meters is around 36000 feet). Mine
only goes to 18000 feet. But I agree that a linear assumption will
result in a worst-case error of several hundred feet. Not good enough,
so the nonlinearity must be built into altimeters even if the're only
good to 20000 ft.

Jim Rosinski