| Back | MTP Home Page |

Trouble-Shooting a Raw-Counts Discontinuity

MJ Mahoney

As should be the case when analyzing any scientific data, when you encounter something that doesn't look "right," every effort must be made to understand what is going on. This happened to me yesterday when analyzing DC-8 MTP data for a CAMEX-4 flight through Hurricane Humberto on September 23, 2001. As is shown in the Figure 1,

raw counts

Figure 1. Raw DC-8 MTP "counts" data for DC200010923.

there is a discontinuity at ~79.5 ks in both the MTP raw sky counts (C16, C26, and C36) for channels 1, 2, and 3, but also in the raw base (or reference target) counts (CB1, CB2, and CB3) for channels 1, 2 and 3. Initially I thought that the jump in the sky counts was okay, because this jump coincided with entering the eye of Hurricane Humberto -- at which time the outside air temperature increased by ~5 K. Therefore, the sky counts should increase. However, there is no reason why the base counts should increase, since the thermal time constant for the base target is very long compared to the jump time.

What aroused my interest in this discontinuity was the fact that when I attempted to use a Gain Equation to fit the outside air temperature, there was an large sinusoidal variation in the Gain Equation TB fit, which didn't make sense. This oscillation also occurred in the raw gains for each channel (G_i , i = 1, 2, 3) calculated using:

G_i = (CB_i - CS_i) / (T_target - T_sky)

where CB_i are the base counts for channel i, CS_i are the sky counts for channel i, T_target is the reference target temperture, and T_sky is the outside air temperature (see Figure 2 at ~79.5 ks).
gains

Figure 2. DC-8 MTP gains for DC20010923.

Eventually, in checking the raw counts data for the sky and base counts, I noticed the following sequence of events:

UTks	CB1	CB2	CB3	C16	C26	C36
79.436	14140	15129	15121	13873	14795	14761
79.449	14129	15120	15115	13874	14788	14756
79.463	14124	15124	15115	13864	14779	14757
79.477	14182	15179	15179	13883	14804	14775
79.491	14235	15229	15231	13959	14882	14853
79.505	14255	15260	15253	14011	14949	14907
79.518	14262	15262	15256	13984	14943	14918

Table 1. Raw base and sky counts for DC20010923.

It is clear that there is a large jump in the base counts over two cycles between 79.463 and 79491 ks (blue), but that a corresponding jump in the sky counts is delayed one cycle and occurs between 79.477 and 79.505 (orange). The reason for this is that apparently something changed beginning at the end of the 79,463 ks scan which affected the base counts for that scan, but not the sky counts. The affect was then seen in the next scan (starting at 79.477 ks) for the sky counts.

So what caused this gain change? Engineering temperatures and pitch and roll showed no correlation. The only thing that seemed to correlate was the outside air temperature, but this could not cause a gain change. In desperation, I decided to look at the accelerometer data, which is shown in Figure 3 , in the hope that perhaps the instrument has received a jolt when entering the eye of Hurricane Humberto at ~79.5 ks.
G Force

Figure 3. Raw maximum and minimun accelerometer counts (left) and derived g-force (right scale) as a function of Universal Time for DC20010923.

As is evident in this figure, the DC-8 experienced an abrupt nearly 2-g acceleration just before entering the eye of Hurricane Humberto. This apparently must have caused a slight movement of some part of the DC-8 MTP receiver -- such as a waveguide flange -- that caused a slight gain change. Note that significant g-forces are also observed during the two other transits of the eye of Hurricance Humberto: one at 74 ks and the other at 85.5 ks. There are also significant g-forces at 91 ks when convection was encountered before landing in Jacksonville, Florida.

Finally, it is useful to note that there are times (such as 81.5 ks and 83.5 ks) when the maximum and minimum g-forces are correlated, instead of anti-correlated, which is the norm. Figure 4 shows that these times coincide with aircraft turns.

Figure 4. Pitch and Roll plot for DC20010923.

The Bottom Line

When you see something unusual in any scientific data, it has an explanation! Not tracking that explanation down can lead to heartache and embarrassment. Always take the time find out what is really going on.

[Back to Home Page]