CAPÍTULO III: MARCO METODOLÓGICO
3.5 RESULTADOS
3.5.1 Análisis de los resultados
Due to alignment issues that resulted from the early UC-SCIDAR design, the cameras used for C1 and C2 were swapped. Figure 3.3 shows the physical and optical layout of UC- SCIDAR that was used from late 2004. C2 (the Thorlabs DC111 camera) was mounted on a z-translation stage to allow for fine adjustment of its positioning. The positions of L1 and L2 were fixed with respect to the corresponding camera CCD.
In 2005, the pupil-plane SCIDAR lens (L1) was replaced with a 12.7-mm focal length achromat. The shorter focal length meant that for the 1-m telescope atF/13.5 only 26.5% of the CCD was utilised with a spatial sampling, ∆r, of 0.008 m/pix. The generalised SCIDAR lens (L2) was replaced with a 10-mm focal length achromat resulting in an ideal defocus distance of ∼ 3.9 km below the telescope. The larger spatial sampling allowed for shorter exposures times to be used on bright binaries that were optimally positioned in the Southern skies during winter months. Figure 3.4 shows typical pupil-plane and generalised scintillation images captured by the system used in 2005 (henceforth termed UC-SCIDAR V2005).
An observational campaign for assessing seasonal changes in turbulence at MJUO was undertaken by Clare Worley as part of her ASTR480 project (Worley, 2005). Measure- ments were taken on a monthly basis from February through to September. Due to time constraints only a small amount of the data collected was analysed at the time. However the work provided useful parameters to optimise the collection process and assess whether
3.2. UC-SCIDAR V2005 39 Aperture Position (m) Aperture Position (m) −0.5 0 0.5 −0.5 0 0.5
(a) Pupil-plane SCIDAR
Aperture Position (m) Aperture Position (m) −0.5 0 0.5 −0.5 0 0.5 (b) Generalised SCIDAR
Figure 3.4: Typical frames from (a) pupil-plane and (b) generalised SCIDAR data taken using UC-SCIDAR V2005.
any particular run would provide a reliable estimate of the structure of turbulence. Part of this work was presented in November 2005 at the Image and Vision Computing Conference (IVCNZ’05) (Johnston et al., 2005). A significant portion of the results presented in this thesis used data taken during 2005.
3.2.1 Camera Noise Characteristics
For the configuration used in UC-SCIDAR V2005, a spatial sampling, ∆r, of 0.008 m/pix was typically used, with a frame rate of 30Hz. As such, an ideal pupil-plane SCIDAR image would be spread across 127 pixels. Due to the nature of a generalised SCIDAR image, the majority of an image taken using a binary star system with φ= 13 arcseconds would be spread across 200 pixels in the widest direction. To decrease processing time, a window of 256x256 centred around the image centroid was used for analysis. For consistency during noise removal, dark frames were sampled using the same window patch of the CCD. The following discussion uses the 256x256 windowed data.
The cameras used in UC-SCIDAR V2005 had a high level of electron noise associated with dark current. Table 3.1 tabulates the average number of photons detected per pixel over 1000 frames for both cameras used with an exposure time of 1 ms and maximum gain. For dark frames, both cameras exhibit similar characteristics with approximately 2.5 photons per pixel detected (standard deviation σ ≈0.3). When looking at images taken from neighbouring sky similar statistics are found. It was decided to use dark frames in noise removal processing, as not every run had corresponding sky frames collected and the noise statistics indicated that the level of dark current detected would have a greater effect on the noise floor of the pupil-plane and generalised SCIDAR images than the noise from the neighbouring sky.
Table 3.1: The average number of photons per pixel for the cameras used in UC-SCIDAR V2005 with an exposure time of 1 ms and maximum gain.
Dark Frames Neighbouring Sky
mean σ mean σ
Micropix M640 2.50 0.31 2.52 0.31
Thorlabs DC111 2.48 0.33 2.49 0.33
vation run using UC-SCIDAR V2005 (shown in Figure 3.5). For data collected using the Micropix M640 (typically pupil-plane SCIDAR measurements), a significant amount of the data has frequencies less than 20 Hz (Figure 3.5(c)). For data collected using the Thorlabs DC111 (typically generalised SCIDAR measurements) a significant amount of the data is less than 40 Hz (Figure 3.5(f)). For both cameras a strong peak appears at approximately 47 – 49 Hz. Upon closer examination of an averaged dark frame from both cameras (Figure 3.6), where the ensemble consists of 1000 frames, the peak seen is associated with noise. As noise of this frequency is present in both cameras it is believed that it is associated with the proximity of the power supply for the cameras to the cable connections to the computer. However it can removed from the data by way of a narrow band-pass filter.
The filter being used in the analysis of V2005 data is a twentieth order Butterworth band-pass filter design to remove only the identified frequencies. Figure 3.7 shows the scintillation frames and their corresponding frequency content after filtering has been applied for the data used in Figure 3.5. Note that the target frequency has been effectively removed while keeping relevant data intact.
Consider run #30 taken in June 2005. This run was collected using α Centauri (α Cen)∗ with an exposure time of 1 ms on both imaging channels. Figure 3.8 shows
data from the pupil-plane SCIDAR measurements. Data from the generalised SCIDAR measurements are shown in Figure 3.9.
The C2
N(h) profiles obtained for pupil-plane data (Figure 3.8(c)) and for generalised
data (Figure 3.9(c)) clearly indicate a turbulent layer at 11 – 12 km above sea level and a strong layer near ground level.†
Due to the amount of noise present in the extracted pupil-plane data (Figure 3.8(b)), the resulting C2
N(h) profile contains additional noise spikes. Using a regularisation pa-
rameter of 0.15 during pupil-plane analysis the coherence length, r0 is estimated to be 24
cm, however error on the calculated r0 is nearly 100% due to the level of noise present
in the extracted data slice. However when filtering is applied to each individual frame
∗φ= 10.2 arcseconds, ∆m= 1.36 (Source: 2005 Almanac)
3.2. UC-SCIDAR V2005 41 Aperture Position (m) Aperture Position (m) −0.5 0 0.5 −0.5 0 0.5
(a) Micropix M640 Image
−10 −0.5 0 0.5 1 50 100 150 200 250 Aperture Position (m) Number of Photons
(b) Number of photons in (a)
−50 0 50 103 104 105 106 Frequency (Hz) Magnitude
(c) Frequency content of (a)
Aperture Position (m) Aperture Position (m) −0.5 0 0.5 −0.5 0 0.5 (d) Thorlabs DC111 Image −10 −0.5 0 0.5 1 50 100 150 200 250 Aperture Position (m) Number of Photons
(e) Number of photons in (d)
−50 0 50 103 104 105 106 Frequency (Hz) Magnitude (f) Frequency content of (d)
Figure 3.5: The frequency content of typical images captured by UC-SCIDAR V2005. The strong peak at approximately 47 – 49 Hz in both cameras can be attributed to readout noise. Data captured at a frame rate of 30 Hz.
−50 0 50 102 104 106 Frequency (Hz) Magnitude (a) Micropix M640 −50 0 50 102 104 106 Frequency (Hz) Magnitude (b) Thorlabs DC111
Figure 3.6: The average frequency content of 1000 dark frames captured by the V2005 system taken at an exposure of 1 ms. Data captured at a frame rate of 30 Hz.
Aperture Position (m) Aperture Position (m) −0.5 0 0.5 −0.5 0 0.5
(a) Micropix M640 Image
−10 −0.5 0 0.5 1 50 100 150 200 250 Aperture Position (m) Number of Photons
(b) Number of photons in (a)
−50 0 50 103 104 105 106 Frequency (Hz) Magnitude
(c) Frequency content of (a)
Aperture Position (m) Aperture Position (m) −0.5 0 0.5 −0.5 0 0.5 (d) Thorlabs DC111 Image −10 −0.5 0 0.5 1 50 100 150 200 250 Aperture Position (m) Number of Photons
(e) Number of photons in (d)
−50 0 50 103 104 105 106 Frequency (Hz) Magnitude (f) Frequency content of (d)
Figure 3.7: The frequency content of the data used in Figure 3.5 after a band-pass filter removing only the signal that is at 47 – 49 Hz has been applied. Data captured at a frame rate of 30 Hz.
before its use in the determination of the covariance, termed pre-filtering, there is a clear improvement in pupil-plane SCIDAR data (Figures 3.8(d)–(f)) with the noise level dropping, resulting in data peaks becoming more readily detectable (Figure 3.8(d)). The resulting r0 estimate using the same level of regularisation becomes 23±5 cm. When
filtering is applied after the covariance has been determined, termedpost-filtering, similar improvements are seen in pupil-plane SCIDAR data (Figures 3.8(g)–(i)), the estimate for r0 becomes 22±5 cm. (Full details on slice extraction, CN2(h) determination and errors
are presented in Chapter 4.)
In generalised data (Figure 3.9) the covariance strength of the near ground layer dominates the profile and is nearly 100 times stronger than the background noise. Hence filtering data has little influence over the resulting C2
N(h) profile and the resulting r0
estimate, which was 6.7±0.7 cm for all three cases when using a regularisation parameter of 0.05. Although this shows that for generalised data filtering is not required, for consistency of analysis filtering was applied to data collected from both channels of the V2005 system. Pre-filtering is considered good-practice, however post-filtering can also be applied in its place, reducing the need to re-process data that has been processed without the filtering applied.
3.3. UC-SCIDAR V2006 43