Lock in $30 Savings on PRO—Offer Ends Soon! ⏳

A new off-point-less method for sub/mm spectros...

A new off-point-less method for sub/mm spectroscopy with FMLO: modeling and subtracting atmospheric lines with a new algorithm / FMLO 2017-07-06

@ IoA Colloquium

Akio Taniguchi

July 06, 2017
Tweet

More Decks by Akio Taniguchi

Other Decks in Research

Transcript

  1. 2017.07.06 / IoA Colloquium A new off-point-less method for sub/mm

    spectroscopy with FMLO Akio Taniguchi (D3 / Kohno Lab / IoA / UTokyo) modeling & subtracting atmospheric lines with a new algorithm
  2. / Akio Taniguchi @ 2017.07.06 FMLO: a new off-point-less method

    for sub/mm spectroscopy 2 20 Contents • New algorithm: weighed PCA by EM algorithm • Prior weight: atmospheric model of O3 Modeling & Subtracting Atmospheric lines On-site Demonstration (2014-2016) Introduction / The FMLO Method • Issues of conventional position switch (PSW) • method of Frequency Modulation Local Oscillator • Comparison of FMLO observation with PSW • An issue of remaining atmospheric lines in data
  3. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 3 20 Introduction - Scientific needs for the single-dish spectroscopy in ALMA era - ALMA (ESO/NAOJ/NRAO), R. Hills ACA 12m TP antennae +30° −30° −20° 0° +20° +10° −10° +30° −30° −20° 0° +20° +10° −10° Galactic Longitude Galactic Latitude 180° 160° 140° 120° 100° 80° 60° 40° 20° 0° 340° 320° 300° 280° 260° 240° 220° 200° 180° 170° 150° 130° 110° 90° 70° 50° 30° 10° 350° 330° 310° 290° 270° 250° 230° 210° 190° Beam S235 Per OB2 Polaris Flare Cam Cepheus Flare W3 G r e a t R i f t NGC7538 Cas A Cyg OB7 Cyg X W51 W44 Aquila Rift R CrA Ophiuchus Lupus Galactic Center G317−4 Coal Sack Carina Nebula Vela Mon R2 Maddalena’s Cloud CMa OB1 Mon OB1 Rosette Gem OB1 S147 S147 CTA-1 S212 λ O r i R g Lacerta Gum Nebula S. Ori Filament Hercules Galactic Latitude Ursa Major 0° +20° 0.0 0.5 1.0 1.5 2.0 FIG. 2.–Velocity-integrated CO map of the Milky Way. The angular resolution is 9´ over most of the map, including the entire Galactic plane, but is lower (15´ or 30´) in some regions out of the plane (see Fig. 1 & Table 1). The sensitivity varies somewhat from region to region, since each component survey was integrated individually using moment masking or clipping in order to display all statistically significant emission but little noise (see §2.2). A dotted line marks the sampling boundaries, given in more detail in Fig. 1. CO Galactic Survey (Dame+01) Figure 1. 20 GHz wide spectral scan at a velocity resolution of 200 km s−1 toward SMM J14009+0252 in the 3 mm window. A CO emission feature is seen at ∼88 GHz (see Figure 2 for a presentation of the CO line at higher spectral resolution). We first scanned the full 3 mm tuning range of EMIR with ∼2 hr of observing for each tuning. The tunings were spaced to provide 500 MHz overlap. Excellent receiver noise temperatures across the band (35–45 K) resulted in typical system temperatures of ∼100 K. The resulting spectrum had an rms noise level of 0.5 mK (≈3.5 mJy) at a velocity resolution of 200 km s−1 but did not show clear evidence for CO line emission. We then increased the integration time for the lower part (<105 GHz) of the 3 mm band until we reached an average rms noise level of 0.2 mK (1.2 mJy). The resulting spectrum, as shown in Figure 1, shows a line at ∼88 GHz. At this stage, the source redshift was still not determined as it was not clear which CO transition was detected in the 3 mm scan. We therefore used the dual-frequency 3/2 mm (E090/ E150) setup of EMIR to search for a second CO transition in the 2 mm band and to increase the signal-to-noise ratio (S/N) of the 3 mm line. In this configuration, each frequency band has an instantaneous, dual-polarization bandwidth of 4 GHz. The 2 mm mixers were tuned to 146.5 GHz, under the assumption that the 3 mm line was the CO(3–2) transition at z = 2.93. At this frequency, the receiver noise temperature was ∼30 K, yielding a system temperature of ∼120 K. SMM J14009+0252 was observed in the dual-frequency setup for ∼5 hr and we clearly detected a second line in the 2 mm band (see Figure 2). Additional 2 mm data were taken in an attempt to observe a third spectra is 160 µK (1.0 mJy) and 180 µK (1.3 mJy) at 3 mm and 2 mm, respectively. Both lines are detected at high significance (9 and 12 σ for the integrated intensities). The line profiles for both lines are very similar and well described by a single Gaussian with a FWHM of 470 km s−1. The parameters derived from Gaussian fits to both line profiles are given in Table 1. The frequencies unambiguously identify the lines as CO(3–2) and CO(5–4) (see our discussion below). Combining the centroids of both lines, we derive a variance-weighted mean redshift for SMM J14009+0252 of z = 2.9344 ± 2 × 10−4. 4. DISCUSSION At first glance, the observed frequencies cannot only be interpreted as CO(3–2) and CO(5–4) at z = 2.93 but also as CO(6–5) and CO(10–9) at z = 6.88 or even CO(9–8) and CO(15–14) at z = 10.80. The CO ladder, however, is not equidistant in frequency which results in small, but significant differences for the frequency separation of the line pairs as a function of rotational quantum number. The frequency separation is 58.577, 58.532, and 58.458 GHz for the CO line pairs at redshifts 2.93, 6.88, and 10.80, respectively. Our observations yield δν = 58.581 ± 0.017 GHz, which identifies the lines as CO(3–2) and CO(5–4) at z = 2.93. Our redshift confirms earlier photometric redshift estimates by Ivison et al. (2000, z > 2.8 based on S450/S850 and 3 < z < 5 based on the whole spectral energy distribution (SED)), Yun & Carilli (2002, z ∼ 3.5 based on the dust SED) and more recently by Hempel et al. (2008, z = 2.8–3 based on optical/IR photometry). With the precise redshift and the observed CO line lumi- nosities in hand, we can estimate the molecular gas content of Figure 2. Spectra of the CO(3–2) (left) and CO(5–4) (right) lines toward SMM J14009+0252. The spectral resolution is 60 km s−1 for both lines. See Table 1 for the fit parameters. CO Blind Redshift Survey (Weiß+09) • determining the redshift, SFR, etc of SMG candidates discovered with wide field surveys by single dish multi pixel camera • wide field spectroscopic mapping observations such as the galactic plane line surveys of molecular clouds • improving the fidelity of interferometry by obtaining total power in ALMA
  4. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 4 20 Introduction - Problems of conventional position switch method (PSW) - • low observational efficiency ηobs (=ton/ttotal) • ηobs=10-20% on high-z observation on 45m • disadvantage of Galactic plane survey (off-point is far) • baseline “wiggle” of astronomical spectrum • disadvantage of high-z line (broad line width) • additional noise of off-point • Tsys gets sqrt(2) times worse νobs ON point (target) PA νobs astronomical spectrum TA * νobs OFF point (blank sky) PA • atmosphere (H2O, O2, …) • frequency characteristics ASTE ON OFF
  5. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 4 20 Introduction - Problems of conventional position switch method (PSW) - • low observational efficiency ηobs (=ton/ttotal) • ηobs=10-20% on high-z observation on 45m • disadvantage of Galactic plane survey (off-point is far) • baseline “wiggle” of astronomical spectrum • disadvantage of high-z line (broad line width) • additional noise of off-point • Tsys gets sqrt(2) times worse νobs ON point (target) PA νobs astronomical spectrum TA * νobs OFF point (blank sky) PA • atmosphere (H2O, O2, …) • frequency characteristics OFF
 slue time
 etc 62% ON
 ʢflaggedʣ 25% ON
 ʢusedʣ 13%
  6. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 5 20 Introduction - High efficiency observation (ηobs~100%) of multi pixel camera - • High efficiency is already done with multi pixel camera • removal of correlated noises which commonly fall on the array detecters • High-rate-sampling mapping of multi-pixel camera (~10 Hz)
 with modulation of field of view (FoV) • astronomical signals are modulated at high frequency domain (~10 Hz) • correlated noises dominated at low frequency domain (1/f, 1/f2 like) • finally correlated noises are removed by high-pass filter such as PCA Power spectrum density (PSD) of timeseries data of one pixel Observing time (sec) Intensity (K) Time frequency (Hz) PSD (K/Hz) noises signal AzTEC/ASTE (Wilson et al. 2008) FFT
  7. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 5 20 Introduction - High efficiency observation (ηobs~100%) of multi pixel camera - • High efficiency is already done with multi pixel camera • removal of correlated noises which commonly fall on the array detecters • High-rate-sampling mapping of multi-pixel camera (~10 Hz)
 with modulation of field of view (FoV) • astronomical signals are modulated at high frequency domain (~10 Hz) • correlated noises dominated at low frequency domain (1/f, 1/f2 like) • finally correlated noises are removed by high-pass filter such as PCA Power spectrum density (PSD) of timeseries data of one pixel AzTEC/ASTE (Wilson et al. 2008) Time frequency (Hz) PSD (K/Hz) Observing time (sec) Intensity (K) PCA (HPF) FFT
  8. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 6 20 The FMLO Method - Frequency-Modulating Local Oscillator (Tamura, Taniguchi et al.) - demodulation and integration High-rate-sampling spectroscopy with modulation of LO frequency (FM of LO) can distinguish astronomical signal from correlated noise in Fourier domain “ ”
  9. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 6 20 The FMLO Method - Frequency-Modulating Local Oscillator (Tamura, Taniguchi et al.) - merits and limits of FMLO method comparison with FSW (FAQ!) • [o] improve observing efficiency • [o] never need off-point observation • [o] low cost implementation • [o] software-based sideband separation • [x] cannot apply to continuum observation • [x] contamination of atmospheric emission • [o] modeling of correlated noise: no additional noise by referential spectrum • [o] applicable to (sub)mm observation: where atmospheric emission dominates • [o] broadband observation: suitable to high-z blind redshift search demodulation and integration
  10. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 7 20 On-site Demonstrations (2014-2016) - Comparison of FMLO spectrum with PSW one (Nobeyama 45m) - 13CO(1-0) CH3CN O3(atmosphere) O3(atmosphere) FMLO-PSW residual spectrum Ori-KL (USB) FMLO spectrum 45m
  11. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 7 20 On-site Demonstrations (2014-2016) - Comparison of FMLO spectrum with PSW one (Nobeyama 45m) - 13CO(1-0) CH3CN O3(atmosphere) O3(atmosphere) FMLO-PSW residual spectrum Ori-KL (USB) FMLO spectrum 45m • FMLO observing efficiency is 4.6x higher than PSW one • FMLO spectrum is 1.9x deeper than PSW one for fixed observing time • no additional noise from off-point and baseline wiggle
  12. Akio Taniguchi @ 2017.07.06 / • FMLO observing efficiency is

    1.5x higher than PSW one • FMLO mapping is 1.4x deeper than OTF one for fixed observing time • no off-point observation and reducing the effect of scan pattern FMLO: a new off-point-less method for sub/mm spectroscopy 8 20 On-site Demonstrations (2014-2016) - Comparing FMLO mapping with conventional OTF (Nobeyama 45m) - Integrated intensity of FMLO mapping Integrated intensity of conventional OTF Residual S/N (residual over noise level) 45m
  13. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 9 20 On-site Demonstrations (2014-2016) - Challenge of FMLO mapping Band 8 ~ 490 GHz (ASTE) - • Verifying FMLO mapping of [CI] (tON~44min, tobs~61min) • Integrated intensity is almost consistent with conventional OTF OTF [CI] (Shimajiri+13) 10 20 30 40 50 (K km/s) FMLO [CI] (preliminary!) ASTE
  14. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 10 20 On-site Demonstrations (2014-2016) - Remaining atmospheric lines in a final spectrum is the final issue - 13CO(1-0) CH3CN O3(atmosphere) O3(atmosphere) 45m Ori-KL (USB) FMLO spectrum • Atmospheric lines (e.g. O3) must be subtracted from a final spectrum • But they have been not properly reproduced in a final spectrum because they have much wider line widths than frequency modulation
  15. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 11 20 Modeling & Subtracting Atmospheric lines - Signal processing of data reduction (late 2016) - gain correlated noises spectrum revert if estimate is not converged demodulate estimate divide estimate subtract estimated gain correlated noises estimated by PPCA raw time series data of ON point gain-corrected time series data cleaned time series data demodulated time series data time νIF
  16. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 12 20 Modeling & Subtracting Atmospheric lines - New signal processing of data reduction incl. weighted PCA - revert if estimate is not converged model fit estimate divide estimate subtract estimated gain correlated noises by weighted PCA raw time series data of ON point gain-corrected time series data cleaned time series data time νIF demodulated time series data as weight weight data by model fit demodulate
  17. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 13 20 Modeling & Subtracting Atmospheric lines - Classical PCA cannot deal with weight - D channels spectrum at t=ti • Classical PCA needs rotate D-dim data so that the 1st principal component axis has the largest variance • Cannot deal with data with weights or missing values D-dim space correlated noises by PCA cleaned time series data gain-corrected time series data transform coordinates reconstruct with top k (<D) axes
  18. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 13 20 Modeling & Subtracting Atmospheric lines - Classical PCA cannot deal with weight - PC1 PC2 D channels • Classical PCA needs rotate D-dim data so that the 1st principal component axis has the largest variance • Cannot deal with data with weights or missing values D-dim space correlated noises by PCA cleaned time series data gain-corrected time series data transform coordinates reconstruct with top k (<D) axes
  19. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 14 20 Modeling & Subtracting Atmospheric lines - Weighted PCA achieved by PCA with EM algorithm (Bailey 2012) - D-dim space D channels spectrum at t=ti • PCA with EM algorithm iteratively estimate 1st principal component axis without rotating data • Can deal with data with weights or missing values correlated noises by PCA cleaned time series data gain-corrected time series data reconstruct with top k (<D) axes Expectation & Maximization figure from Bishop's talk (2004)
  20. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 14 20 Modeling & Subtracting Atmospheric lines - Weighted PCA achieved by PCA with EM algorithm (Bailey 2012) - D-dim space D channels spectrum at t=ti • PCA with EM algorithm iteratively estimate 1st principal component axis without rotating data • Can deal with data with weights or missing values correlated noises by PCA cleaned time series data gain-corrected time series data reconstruct with top k (<D) axes Expectation & Maximization figure from Bishop's talk (2004)
  21. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 14 20 Modeling & Subtracting Atmospheric lines - Weighted PCA achieved by PCA with EM algorithm (Bailey 2012) - D-dim space D channels spectrum at t=ti • PCA with EM algorithm iteratively estimate 1st principal component axis without rotating data • Can deal with data with weights or missing values correlated noises by PCA cleaned time series data gain-corrected time series data reconstruct with top k (<D) axes Expectation & Maximization figure from Bishop's talk (2004)
  22. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 14 20 Modeling & Subtracting Atmospheric lines - Weighted PCA achieved by PCA with EM algorithm (Bailey 2012) - D-dim space D channels spectrum at t=ti • PCA with EM algorithm iteratively estimate 1st principal component axis without rotating data • Can deal with data with weights or missing values correlated noises by PCA cleaned time series data gain-corrected time series data reconstruct with top k (<D) axes Expectation & Maximization figure from Bishop's talk (2004)
  23. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 14 20 Modeling & Subtracting Atmospheric lines - Weighted PCA achieved by PCA with EM algorithm (Bailey 2012) - D-dim space D channels spectrum at t=ti • PCA with EM algorithm iteratively estimate 1st principal component axis without rotating data • Can deal with data with weights or missing values correlated noises by PCA cleaned time series data gain-corrected time series data reconstruct with top k (<D) axes Expectation & Maximization figure from Bishop's talk (2004)
  24. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 14 20 Modeling & Subtracting Atmospheric lines - Weighted PCA achieved by PCA with EM algorithm (Bailey 2012) - D-dim space D channels spectrum at t=ti • PCA with EM algorithm iteratively estimate 1st principal component axis without rotating data • Can deal with data with weights or missing values correlated noises by PCA cleaned time series data gain-corrected time series data reconstruct with top k (<D) axes Expectation & Maximization figure from Bishop's talk (2004)
  25. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 14 20 Modeling & Subtracting Atmospheric lines - Weighted PCA achieved by PCA with EM algorithm (Bailey 2012) - D-dim space D channels spectrum at t=ti • PCA with EM algorithm iteratively estimate 1st principal component axis without rotating data • Can deal with data with weights or missing values correlated noises by PCA cleaned time series data gain-corrected time series data reconstruct with top k (<D) axes Expectation & Maximization figure from Bishop's talk (2004)
  26. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 15 20 Modeling & Subtracting Atmospheric lines - Prior atmospheric (O3) model as weight - • Atmospheric model can be represented as linear combination of spectra at each altitude layer, calculated by the am (Paine 2017) • Weight is calculated to be w = exp(-Ta*) • s.t. w = 1 when Ta* = 0 (w/o line), w → 0 when Ta* >> 0 (w/ line) O3 spectra at each altitude layer around 110 GHz model O3 spectrum and its weight (=exp(-Ta*)) model weight
  27. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 16 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Atmospheric lines are successfully reproduced after the new data reduction (atmospheric model and weight are iteratively updated using data) data model weight 45m FMLO observation of 110 GHz atmosphere at Nobeyama 45m O3 (60,6-61,5)
  28. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 16 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Atmospheric lines are successfully reproduced after the new data reduction (atmospheric model and weight are iteratively updated using data) data model weight 45m FMLO observation of 110 GHz atmosphere at Nobeyama 45m O3 (60,6-61,5)
  29. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 16 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Atmospheric lines are successfully reproduced after the new data reduction (atmospheric model and weight are iteratively updated using data) data model weight 45m FMLO observation of 110 GHz atmosphere at Nobeyama 45m O3 (60,6-61,5)
  30. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 16 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Atmospheric lines are successfully reproduced after the new data reduction (atmospheric model and weight are iteratively updated using data) data model weight 45m FMLO observation of 110 GHz atmosphere at Nobeyama 45m O3 (60,6-61,5)
  31. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 16 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Atmospheric lines are successfully reproduced after the new data reduction (atmospheric model and weight are iteratively updated using data) data model weight 45m FMLO observation of 110 GHz atmosphere at Nobeyama 45m O3 (60,6-61,5)
  32. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 17 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - 45m data FMLO observation of 110 GHz atmosphere at Nobeyama 45m Atmospheric lines are not reproduced properly without weighted PCA (because line width wider than frequency modulation step is correlated) O3 (60,6-61,5) FM step
  33. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 18 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Is astronomical line successfully reproduced even overlapping O3 line ? A simulation of (artificial) line observation and reducing the data 45m data model weight FMLO observation of 110 GHz atmosphere at Nobeyama 45m + artificial line (Ta* = 5K) O3 (60,6-61,5)
  34. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 18 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Is astronomical line successfully reproduced even overlapping O3 line ? A simulation of (artificial) line observation and reducing the data 45m data model weight FMLO observation of 110 GHz atmosphere at Nobeyama 45m + artificial line (Ta* = 5K) O3 (60,6-61,5)
  35. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 19 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Artificial line is successfully reproduced even overlapping O3 line But O3 lines still remains a little (model may not be optimized) 45m data true artificial line (Ta* = 5K) residual of O3 (60,6-61,5)
  36. Akio Taniguchi @ 2017.07.06 / FMLO: a new off-point-less method

    for sub/mm spectroscopy 19 20 Modeling & Subtracting Atmospheric lines - Demonstration of modeling & subtracting O3 lines - Artificial line is successfully reproduced even overlapping O3 line But O3 lines still remains a little (model may not be optimized) 45m data true artificial line (Ta* = 5K) residual of O3 (60,6-61,5)
  37. / Akio Taniguchi @ 2017.07.06 FMLO: a new off-point-less method

    for sub/mm spectroscopy 20 20 • Weighted PCA works well to reproduce O3 lines • Atmospheric model should be optimized for better fitting (ongoing issue, but not critical) Modeling & Subtracting Atmospheric lines On-site Demonstration (2014-2016) Introduction / The FMLO Method • Issues of conventional position switch (PSW) • method of Frequency Modulation Local Oscillator • Comparison of FMLO observation with PSW • An issue of remaining atmospheric lines in data Summary