Week 4

This week was spent modeling and setting up the statistical framework for the research.

The statistical backbone of my analysis is in R, a data manipulation language. R is a PBV (Pass By Value) language; once a parameter is called up, any subsequent changes or operation do not alter it in its stored form, only its referenced form. This makes R very useful for manipulating data without integrity risk. The central object of R is the data frame, an arbitrarily large/dimensional matrix that can be referred to but not called on, allowing the user to call down an array by using its coordinate.

My physical model for the radiometer is that of a constant power, in-viscid flow, such that P=L-bv^2. L, the net thermal input term, can be measured from the surface temperature of the system at equilibrium as 65W, negating the need to know the environmental input. This setup makes use of the handy identity of v dp=p dv, to integrate subsequent terms. Ignoring thermal creep, this gives us a resting velocity of v= sqrt(P/b).

Week 3

The better part of this week was spent designing, assembling and perfecting a wooden U-mount to serve as a holder for the tachometer's emitter and diode, as pictured below. Despite slight warping in the wood and refraction by the bulb, the array is able to detect the minute interruptions in the laser perfectly.

From this setup, I received the following time series, with the anticipated results in magenta, and the actual results in blue. Although it fit the general prediction well, there is a great deal of temporal aliasing , as well as an unexplained dip near the 32 second mark. The temporal aliasing is a result of a sample rate smaller that that of the signal, leading to discretized, 'jumpy' data. It can be minimized by aggressive smoothing algorithms, which run the risk of being too reductive.