PR3 oplev's noise was improved in 3-10Hz.

In front of the QPD, an ease-made tude with black inside surface was attached against the QPD circuit. Then the duct from the viewport flange was slightly elongated so that the duct and the tube overlapped. After recentering of the beam on the QPD (actually not so much shited during the attaching of the black tube), I measured the noise spectra as shown in the attached pdf.

From 3 to 10 Hz, the noise reduces ~ 5 times at the maximum (at 10 Hz). This would mean that such a kind of easy tube can reduce the air turbulance. A reflected beam from the surface of the QPD was also dumped by the black tube, so the dump could also contribute to the improvement.

The peak structures over 20Hz are not changed from yesterday, acturally the same; in the attached figure, REF0 and REF1 are the same measurements yesterday mentioned as (iii) (Not the same REF0 and REF1 in the yesterday's figures), so the noise level improves in 3-10Hz at least. I'm not sure well but the spectra in this region now show f^{-2} characteristic.