The purppose of these investigations is to learn what channels a glitch injected into the system shows up in, and at what frequencies, and any time delays.  This knowledge should allow better tuning of glitch monitors, and hopefully a more sophisticated successor to 'donald' and 'irvana'.

First Investigation:  Resonances in the PSL Table and Periscope

The signal injection was a foot stomp by Valera outside the PSL enclosure.  The accelerometer was on the PSL table, not the periscope at this point.
The seismometers recorded signals up to  around 45 Hz, and the accelerometer recorded broadband signal to 100 Hz..   (Note that in both cases, the curves labeled 'ref' actually represent the injection.  With the data lost in a computer crash, I don't think I can change this.)  The power spectrum plots show resonances at around 18 Hz, 38, 50 Hz, and 70 Hz on ACCZ.  Smaller resonances above 100 Hz were also noted.    Checking with Robert Schofield, we learned that he has noted resonances in the PSL table legs in the 30-60 Hz band, in the table top in the 70-100 Hz band, and in the periscope in the 180-250 Hz band.  (He noted that he had never seen the table top resonances go through to AS_Q.)
 

Second Investigation:  Tracking a Seismic Injection Through the Interferometer

The second signal injection was another foot stomp by Valera - this time in the control room.  The accelerometer had been moved to the top of the periscope, to pick up vibrations of the periscope more directly.  The raw data from LVEA_SEISX and LVEA_SEISZ are shown here.   Next we looked at PSL_ACCX, and AS_Q, to see whether the signals had gone through the interferometer - as they clearly had.  A couple of other phenomena show up here, seen more clearly when the second glitch is isolated for better time resolution.  One is the splitting of the original glitch, and the other is the time delay - around 0.5 s to the peak of the first AS_Q glitch, and ~0.75 s to the peak of the second.  Although the frequencies present in the AS_Q glitches go right through the band of interest,  the power level is fortunately such that there is no danger of a stomp in the control room, or a dropped object, being mistaken for an inpiral.

Going back to a more orderly progression through the interferometer and preceding channels, we use the frames that were stored stored by the time of the analysis on delaronde:  the first glitch fell in one frame beginning at gps 742025248, and the other three in a frame beginning at gps 742025264.
 

Seismic Injection

The spectrogram of the second frame (three glitches) shows signal up to ~35 Hz, with slab resonances below 10 Hz  and at around 25 Hz.   Whether or not higher frequencies also exist and may have been transmitted to the PSL table will have to wait for another experiment.  (We plan to use one of Marcel's springs with a resonance around 80 Hz to amplify higher frequency motion.)
 

PSL

Reporting on what happened at the PSL table and periscope is complicated a little by configuration changes.  At the time of the foot stomps with the interferometer locked, the accelerometer that is normally on the table was mounted on the periscope, with X and Z directions reversed.   The accelerometer has since been returned to its normal place on the table, and a more sensitive accelerometer attached to the periscope, recording movement in the X-direction as the channel HAM3_ACCZ.  The transfer function between the table and periscope was reported by Valera Frolov and Joe Kovalik in the LLO e-log on Thursday, July 17.

The  power spectrum for the second frame is shown here.   The accelerometer was on the periscope, and remember that X <-> Z.  Huge resonances show up around 200 Hz, and between  430 and 700 Hz in the X-direction.   (The poor correspondence at low frequencies may be due to the reference spectrum being taken at a later time than the stomps.)  However, for a monitor to pick the resonances out of a time series requires careful band-passing, and the glitches almost disappear into the noise in a 100 Hz high pass filtered plot.   A  20-100 Hz band pass or a 20 Hz high pass works well for both axes.
 

MC channels

The glitches show up clearly in MC_F and MC_I, but not in MC_TRANS_DC, which is expected to indicate the power that is actually going into the interferometer.   MC time series  with 20 Hz high pass filter.   The DC channel shows no indication of the injected glitches, but they show up clearly in  MC_F and MC_L using a 40-60 Hz band pass filter.  A suggestion of the last glitch remains in MC_F after 100 Hz high pass filtering, but the others are removed above 60 Hz.
 

POB_I/POB_Q

The glitches are clearly split into components in these 40-60 Hz band-passed POB_I/POB_Q frames.
 

REFL_I/REFL_Q

The splitting of the glitches shows up in REFL_Q for the first glitch  (20-40 Hz) and to some extent in the later glitches.  The last glitch would not be noticed in REFL_Q (second REFL frame) without having its time from other channels, although all of them stand out in REFL_I.
 

MICH_CTRL

MICH_CTRL shows a very high snr signal for the first glitch in the second frame, using a 100 Hz high pass filter, and a clearly visible signal for the next glitch also.  The last glitch would not be picked out without knowing its time.  A spectrogram for this channel indicates thatthe signal goes above 500 Hz,although there is also a high level of high frequency background noise.
 

AS_Q

A sepctrogram of AS_Q shows signal from the three glitches in the second frame disappearing into the noise band around 650-750 Hz.  The signals are  possibly still present at higher frequencies, but at a level where it would be very difficult to pick them out from a time series.   A plot of AS_I and AS_Q using a 20 Hz high pass filter lets all the glitches stand out clearly, but with a 90 Hz high pass, only the first two can be seen, and only the first one (in this frame) stands out.
 

Wavelet Decompostions of AS_Q and PSL_ACCX  (courtesy of Chethan Parameswariah)

Detailed wavelet decompositions of these channels for both experiments are shown, with the frequency limits for each of the stacked bands going up in powers of 2.  These are automatically lined up in time, making it very easy to watch the progress of a glitch through the frequency bands.