Power Spectra (11/14/07)

• Again we revisit the all telling power spectra. While our results for iSeis are disappointing we are starting to see some promising results for the Null interferometer, and the iSTS1.
  
  • Figure 1: a plot from Jon Berger comparing the null interferometer to the GSN noise model
    
    
  • Figure 2: a plot from J Berger using the best fit "G" factor of the iSTS1 of 1.92 ms-2 of ground acceleration per m of mass displacement. iNull has been passed through the iSeis response

Temperature Investigation (10/25/07)

• In order for us to determing wether a correlation exists between iSeis, iNull and iSTS1 it helps to understand temperature in the vault. Currently we have three temperature sensors. Two TR-1050 data loggers and a home made temperature sensor. The home temperature sensor is located inside the bell jar and its electronics are located in the hallway. One of the TR-1050 stands outside of the insulated box on the pier (vault). The other lies on the pier within the insulated box.
  
  
  • Figure 1: A composite plot showing temperature data for a week.
    
    
  • Figure 2: A zoomed in portion of the above plot (same scales and time axis within subplots)
    
    
  • Figure 3: A second zoomed in portion of the above plot (same scales and time axis within subplots)
    
    
  • Figure 4: A plot showing a 3-day detrended record for temperature sensor. Notice the diurnal temperature fluctuations. It should be noted that the electronics for the bell jar sensor are located in the hallway; an environment more susceptible to temperature change.

iSeis, iSTS1 and Gravity Data (10/23/07 - 11/1/07)

• On October 5th, 2007 we began recording gravity data with a Scintrex relative gravity meter. Comparing these recordings with the theoretical tides, iSTS1, iSeis and a standard STS1 have yielded some interesting results. Below are the most significant findings.
  
  
  • Figure 1: J. Berger uses Matlab's cftool to perform a quadratic fit for the theoretical tides and the iSTS1 over a 90 hour period
    
    
  • Figure 2: Using Matlab's cftool, we perform a fit to the Scintrex gravity meter for 8 days of data (scintrex, iSTS1, theoretical tides)
    
    
  • Figure 3: Residual between theoretical and actual tides. Residual between best fit iSTS1 tide and Scintrex. Ambient air pressure in STF vault.
    
    
  • Data Sets: a matlab data set containing iSeis (iseis1000s), iSTS1 (ists1000s), Scintrex (gravity1000s) and theoretical (theor_gravity1000s) tidal data. All data is sampled at 1000 seconds and in units of microGal. The data range from 10/12/07 (285) 00:00:00 - 10/20/07 (292) 00:00:00.

• An interesting correlation between iSeis, iSTS1 and the null interferometer may have been found at tidal periods. It is unclear whether this is a pressure effect (perhaps cause by air pressure on the PM fiber) or a temperature effect (caused by heating and cooling of the PM fiber)
  
  
  • Image 1: comparison of iNull * 500, iSTS1 and iSeis showing the odd correlation
    
    
  • Image 2: 90 hours of iNull (blue), iSeis (red and scaled) and temperature data -- the correlation seems be be low at first glance
    
    
  • Image 3: 90 hours of iNull (blue) and IDA barometric data (green)

Null Interferometer Results (9/26/07 - 10/22/07)

• A great deal of interest lies in the null interferometer experiment being conducted at PFO. Through numerous tests we hope to obtain the power spectra results that we had in 2003. These results gave us power at about 2^-13 m at 2 Hz.
  
  
  • Run A -- Time Series and Power Spectra, 2007 (269) ellipse fitting off and modulation off, no filters
    
    
  • Run B -- Time Series and Power Spectra, 2007 (271) ellipse fitting on and modulation on, no filters
    
    
  • Run C -- Figure showing time series and spectrum from 2007 (283), ellipse fitting on, modulation on, filters installed on x and y
    
    
  • Run D -- Figure showing time series and spectrum from 2007 (284), ellipse fitting off, modulation off, filters installed on x and y (GOOD)

iSTS1 improves (9/25/07)

• The recent trip to PFO (9/21/07) in which the bell jar leak was fixed has greatly improved noise in the iSTS1. We still see the occasional pivot "blip" in the iSeis as its pivots are still wearing in.
  
  
  • Image 1: noise has improved although there is still the odd tidal phase shift and pivot jumps in the iSeis data set
    
    
  • Image 2: power spectra for 15 hours of data for both iSeis and iSTS1 in which Jon has corrected for the G factor (incorporates inertia)
    
    
  • Image 3: Observed vs theoretical fit for iSeis and iSTS1. This yields the standard Q and Wo terms we have been finding for the last few weeks. The fall off in relative response (4e-2 Hz) is most likely due to the fall of coherence (see the power spectra).
    
    
  • Image 4: Taking a power spectra from a time series which has fewer jumps in the iSeis records yields better results. The LP noise may just be in the hinges.

Transfer Functions and Relative Responses (9/17/07 - 9/25/07)

• Jon Berger has be spending a significant amount of time looking at the amplitude ratios and phase ratios (transfer functions) between iSeis and iSTS1. Through these investigations we have concluded that inertia of the seismometer (something we have been ignoring to this point) is quite important indeed. Below is a brief time line of his results .
  
  
  • Image 1: A transfer function between iSeis and iSTS for the Sumatra event. If both are SHOs with different Q and Wo terms we would expect the transfer function to be a constant for periods well below the free period . . . this is not what we see.
    
    
  • Image 2: Comparison of iSeis and iSTS1 XFR function to a theoretical XFR Function model which incorporates inertia and least squares to solve for Q and Wo for both seismometers using the Sumatra EQ (Qiseis=3.04, Qists1=1.49, Tiseis=3.66, Tists1=5.16).
    
    
  • Image 3: Fit for transfer function comparing a standard STS1 and iSeis with a coherence weighting using the Sumatra EQ (2007:255). The fit allows us to solve for Q and Wo for iSeis (Qiseis=3.13 Tiseis=3.66).
    
    
  • Image 4: Fitting iSeis and a standard STS1 to noise rather than an earthquake (as in image 3) yields similar results. In the noise fit we find: Qiseis=2.97 and Tiseis=3.61

iSTS1 woes / Null Interferometer Installed (9/10/07 - 9/21/07)

• On 2007 (253) both instruments (iSTS1 and iSeis) had recorded a nice magnitude 6.8 event from Columbia. Based on this event we thought both instruments were working fine. A bit of investigation however made us realize we were very wrong indeed.
  
  
  • Columbia Event: as recorded by both the iSTS1 and iSeis
    
    
• Image 1: seismic data from 2007 (253) comparing iSTS1 and iSeis showing the high amount of noise in iSTS1 . There also seems to be a phase shift in the tides which is not understood.
  
  
• Image 2: an image from Jon Berger shortly after the Columbia event in which data was bandpassed
  
  
• Image 3: as the noise gets continually worse we begin to suspect a leak

 

• The results for the iSTS1 are very disappointing. Not knowing the exact cause of the problem I took a trip to Piñon on 9/21/07 and indeed found that the bell jar was leaking. The bell jar was fixed and iSTS1 results have hopefully improved. During the same trip the Null Interferometer was installed at PFO.

New Pivot Install and iSTS1 install (8/10/07 - 9/07/07)

• In early July iSeis began to exhibit odd behavior. Symptoms included noisy data which decreased immediately after an earthquake, random transients, and poor data quality. It is felt that the root of these problems was worn or "tired" flex pivots. Due to extended travel in Norway, this problem was neglected until last Friday, (8/10/07) at which time new flexures were installed and iSeis was properly leveled and tuned. An optical STS1 was also installed on the same pier.
  
  
  • Figure 1: Some of the odd behavior exhibited by iSeis due to the "worn out" flexures.
    
    
    
• iSeis seemed to improve, after the new pivot installation, but "wear-in" of the flexures caused the mass to continually drifted downward until the seismometer bottomed out some thirteen hours later. On 8/16/07 the equilibrium position of the seismometer was again adjusted by turning the horizontal COM screw clockwise until the the center of mass was level with the flex pivots. As of this time, the mass has continued to drift and at the current rate should bottom out sometime late this week.
  
  
• Fortunately, after yesterday's re-adjustment (8/15/07), we were able to record the recent normal mode event off the coast of Peru. This event was captured by both the optical STS-1 (iSTS1). The main event occurred at -13.358 N, -76.522 E at 23:40 UT. The Mw 8.0 event was at a depth of 30.2 km.
  
  
  • Figure 2: Screen capture from the NRTS system showing the Peru event as recorded by iSeis and iSTS1. The continuing downward trend in the iSeis trace (top) is most likely caused by "wear-in" of the new flexures.

• Sometime in early September iSeis began acting strangely. Microseisms seemed to be suddenly attenuated and within a few days all mass displacements seemed to stop. With a trip to PFO on 9/7/07 it was found that the mass had again drifted to the lower stop (see above) and the optical paddle has moved so that it was touching the photo diodes. After a bit of readjustment and work iSeis was back to new.

STS1 Optical Feed Back Maximum Gravity Test (8/06/07)

• The optical STS1 was modified to allow feedback to be turned on and off via a couple of switches, placed on the cover of the electronics enclosure. The STS1 was then placed on an aluminum plate with two fixed feet and one micrometer translation stage. The system was translated from a negative tilt, through level, to a positive tilt position. Voltage was monitored on test point "M" (boom position) located on the "test/monitor" connector. At maximum gravity (when the sensor is level), boom position voltage reaches a minimum. This was found to be at about -0.65 Volts.
  
  
  • Figure 1: Figures from the above tests. Two trials were run with similar results. The second trial being the better of the two. Polyfit (order = 2) coefficients are as follows: Test 1 -- 164.9413 -61.6450 5.2566 and Test 2 -- 161.1167 -59.7811 4.9

Noise Spectra Comparison (6/04/07)

• Extensive noise tests have been conducted on the munki1 interferometer. During these tests a number of factors were changed including: the lasers (Melles Griot 1 mW, Melles Griot 2 mW, and the ML1), the laser modes (locked versus unlocked), environment (sealed, open, vacuum), and fitting (on or off.) The individual tests, time series and power spectra plots are detailed below.
  
  
• A preliminary composite figure (albeit a bit busy) of all the power spectra can be found here by clicking on this link. Be sure to check back later as additional figures may be added. Newer versions will hopefully be less confusing and contain "key" power spectra tests.
  
  
• Data (css and .w) files as well as the power spectra mat files can be found by clicking on the following directory: DATA
  

iSeis Tidal Response (5/21/07)

• Over the last week or so, I have been working on generating a nice tidal record for iSeis. Much work has been done and a nice tidal record has been found. Work still needs to be done, however, as there are a number of problems. The first is in our estimate of wo. In the figures, wo, has been chosen such that the variance of the residual (between iSeis and the theoretical tides) has been minimized. This occurs when wo = 1.950 or 3.2 seconds. Ring down tests, however, predict wo = 2.34
  
  The residual plots also yield some concern. At times the residual fall below 5 micro-gal yet at other times it exceeds 40 micro-gal. Future efforts needs to focus on environmental aspects such as barometric pressure and ambient air temperature. We also seem to come in and out of phase with the predicted tides. This issue also needs to be resolved.
  
  The current tidal records extends for 25 days: 4/26/07 - 5/20/07
  
  
  
  • Figure 1: time series records (filtered and unfiltered) for iSeis mass displacement covering the above period.
    
    
  • Figure 2: filtered time series records for iSeis and theoretical Earth tides at PFO
    
    
  • Figure 3: residual plots for iSeis and theoretical tides, first differences for iSeis and theoretical tides
    
    
  • iSeis Data Set: [MAT FILE] mass position in radians, 1 sps, time span: 4/26/07 00:00 - 5/20/07 23:59:59, ~6.2 Mb

Newest PFO Power Spectra (5/08/07)

• A few weeks ago we made some major changes to the DSP, the optics, iSeis, and the laser itself. Now that sufficient time has passed, allowing STF to reach thermal equilibrium, we hope to see better power spectra than those of a few weeks ago.
  
  
  
  • Figure 1: time series plots for the data covered by the power spectra, listed below.
    
    
  • Figure 2: power spectra plot using the spectrum technique in units of velocity (m/s)^2 / Hz
    
    
  • Figure 3: power spectra plot using the spectrum technique in units of acceleration (m/s^2)^2 / Hz
    
    
  • Figure 4 : power spectra plot using the spectrum technique in units of mass position m / sqrt (Hz)

Filter Timing Test (5/07/07)

• I believe there may be a problem with the filters for the fast and slow DSP femtometer channels. It seems as if an event which occurs at time t = 0 on the fast channels may occur on the slow filtered channels at time t + dt or t - dt. This should not be the case. To make matters worse, it seems declaration dependent also (i.e. 32 bit or 64 bit dependent). To further investigate this issue I have created a series of very accurately timed pulses, tied to the rising edge of a GPS clock, on munki1. Signals were injected at the following GMT times on 05/07/07 --> 17:09:00, 17:09:30, 17:10:43, 17:11:00, 17:12:15. Because of propagation delay within the circuit, we expect to see a response in the interferometer within a few milliseconds of the above times. During these tests, ellipse fitting was shut off.
  
  
  • Three impulses were chosen at random for investigation. The results can be found here: [Figure 1] [Figure 2] [Figure 3]. It appears that the timing is slightly off. On the oscilloscope we verify a circuit delay of 67 ms. In the figures we see a delay of ~100 ms.

Power Spectra Comparisons (4/27/07)

• Below, I present power spectra from three separated periods. The periods shown cover: 4/25/07 (red), 1/30/07 (black) and 8/02/06 (blue). It is felt that the period around August of 2006 yielded some of the best power spectra we have seen to date. After significant optics and DSP changes were made in November and December of 2006, power spectra results became poor. With the most recent changes made, last week, we hope to improve our spectra back to the levels seen last summer. It appears we still have some way to go.
  
  
  
  • Power Spectra Set #1 : Time Series and Power Spectra for 8 hours worth of data sampled at 1 sps.
    
    
  • Power Spectra Set #2 : Time Series and Power Spectra for 12 hours worth of data sampled at 1 sps.
    
    
  • Power Spectra Set #3 : Time Series and Power Spectra for 18 hours worth of data sampled at 1 sps.
    
    
• As stated below, a number of ring down tests were performed last Friday. Amongst the most interesting was the un-damped case, in which phase amplitudes were high enough to surpass the phase wrap threshold within the DSP. It has since been found that the phase wrap threshold is about 32,000 radians. While investigating this problem, however, a more serious problem has appeared to surface. The following figure illustrates the current issue. It seems, that each of the channels (p1f32, p1s32, p1f and p1s) have a slightly different start time for the ring down sequence (delta = 4 seconds.) This is clearly problematic and needs to be investigated.
  
  
  • Figure 1: Ring down tests from 4/20/07 exhibiting two problems. (I) phase wrapping which should be an easy fix and (II) start time delay across the fast, slow, 32 bit and 64 bit channels . This will be a more challenging fix, as filtering, decimating and timing issues can all contribute to this offset.

Ellipse Stability and Power Spectra (4/26/07)

• A new laser and polarization stability video has been created. Data in this video span from late in the day of 4/20/07 to 4/25/07. The first few days are quite noisy, but as time evolves the ellipse becomes more stable.
  
  
  • Ellipse Stability Video #1 : Quicktime Format, 103.6 Mb, 4/20/07 22:00 - 4/25/07 23:59, each frame contains 30 minutes worth of data
    
    
  • Ellipse Stability Video #2: Quicktime Format, 225 Mb, 12 hours of ellipse parameters from 4/26/07, each frame contains 60 seconds of data
    
    
    
• Power Spectra Plots were also generated. Data span from 4/25/07 00:00:00 - 4/25/07 12:00:00. The technique used to produce the power spectra -- in anticipation of adding STS1 data -- was : i) decimate iSeis data from 400 sps to 1 sps, ii) multiply by constant to convert to long period acceleration, iii) filter SIFO data through STS1 response, so that it behaves like an STS1, iv) take power spectra (pmtm or spectrum) in the velocity domain v) back out the STS1 response. If acceleration power spectra was desired step (v) data was multiplied by w^2. Power Spectra plots were normalized by dividing out the Nyquist.
  
  
  • Figure 1: time series plots for 4/25/07 00:00:00 - 4/25/07 12:00:00
    
    
  • Figure 2: velocity power spectra in units of (m/s)^2 / Hz
    
    
  • Figure 3: acceleration power spectra in units of (m/s^2)^2 / Hz
    
    
  • Figure 4: acceleration power spectra with STS1 as a reference

Newest PFO Power Spectra (4/24/07)

• Recent changes to the DSP, the optics, iSeis, and the laser will most likely have beneficial effects on power spectra plots. The following power spectra were taken from April 24th, 2007. Once the room reaches thermal equilibrium, it is expected to see more of a drop in long period noise.
  
  
  • iSeis Data Set: [MAT FILE] mass position in radians, 400 sps, time span: 4/24/07 02:00 - 22:59:59
    
    
  • Figure 1: time series plots for the data covered by the power spectra, listed below.
    
    
  • Figure 2: power spectra plot using a multi-taper technique (pmtm) <--- there is most likely an error in my technique in this plot!
    
    
  • Figure 3: power spectra plot using the Matlab spectrum technique <--- there is most likely an error in my technique in this plot!

Big Day at PFO (4/20/07)

• A number of big changes were made to iSeis on April 20th, 2007. Amongst these, were a laser change, an optics change, and a mass reentering. We also ridded ourselves of the pesky Windows machine and installed a BSD based femtometer and data logger. GPS timing was also improved with the addition of an oven controlled oscillator circuit. After the new components were installed and changes made, a number of ring down test were performed. The most interesting ring down, was the one in which damping was turned off. Here it was found that at the highest amplitudes the DSP code would phase wrap back to the zero position. This is clearly problematic.
  
  
• Temperature records over the last few months were also gathered and the plots are provided below: Figure 1 and Figure 2 show temperature records for the hallway and bell jar. Figure 3 shows a composite plot of both temperature records. Figure 4 shows bell jar temperature with the parabolic trend removed.

Ellipse Fitting Tests (April 16th - April 18th 2007)

• It is thought, that the length of the ellipse buffer (be it to long, or to short) can add noise to the interferometer power spectra. In order to test our hypothesis, a number of tests will be run to verify the effects of ellipse fitting. The rate at which new ellipse parameters are found is somewhat uncertain. While we feel that it should be directly related to the sampling rate frequency (fss), the ellipse buffer decimation rate (elldcm) and the phase buffer size (phssiz) the exact relationship has seemingly been lost over the years and numerous code revisions. (i.e. we may have a bug in the fitting algorithm).
  
  
  • Currently, it is felt that the ellipse fitting rate is: fb = fss / elldcm if fb > ff, where ff = fss / phssize. If ff > fb then ellipse fitting rates may be unpredictable.
    
    
    

    The following table contains descriptions and figures from various ellipse fitting tests. The "TS" column contains time series plots of the ellipse parameters. Fss, phssiz, phsdcm, elldcm, ellsiz and ellmin are values from the configuration file at which time the test was run. Run Number fss phssiz phsdcm elldcm ellsiz ellmin Description   TS Timing Test #1 100000 10000 5000 10 500 25 This is the first in a number of ellipse fitting tests. The configuration file is at the default setting. Analysis of the time series plots, shows that the ellipse parameters change at a rate of 1 Hz. TS1 is for a free space interferometer, and TS2 is for a PM fiber interferometer. (munki1)   TS1   TS2                         Timing Test #2 100000 10000 5000 10 500 25 April 17, 07 (10:17 PDT): In a second attempt to verify timing, the ellipse parameters were forcibly changed by tweaking the gain setting on the PDA55 photodetectors. Initially, the "x" and "y" photodetector gains were increased but voltage levels clipped the DSP. After the system was returned to baseline levels, the "x" photodetector gain was lowered one click. This changed the "x-voltage" of the ellipse from 8V to 2V. Upon inspection of figure 1, it becomes clear that something has gone awry, as we do not see the expected results. (Rather than seeing the ellipse parameters change and come to a new constant value they seem to change and then return to their previous baseline.) Perhaps it has to do with the initial clipping of the "x" and "y" photodetectors and the short amount of time in between the tests. (insufficient time to flush the ellipse buffers) A second test will need to be run. (munki1)   TS1                     Timing Test #3 100000 10000 5000 10 500 25 April 17, 07 (15:35 PDT): The gain setting on the "x" photodetector has been increased one click. X voltages have increased from 2V to 8V. If this tests runs as expected, we should see a gradual rise/fall and then near constant leveling of the ellipse parameters. Note: It appears as if the click transition occurred much to quickly and the DSP got confused as to which side of the ellipse it was tracking. A better, slower transition test needs to be found. (munki1)   TS1 TS2                       Timing Test #4 100000 10000 5000 10 500 25 April 18, 07: The mezzi0 DSP was connected to a synthetic fringe signal created by the AFG3022 function generator, on AD0 and AD1. Voltage offset along the x-axis was slowly varied from 0 to 200 mV over a period of 46 seconds. Initial ellipse parameters were roughly: a=1, b=1, x0=1, y0=0. After the transition the ellipse parameters were roughly: a=1, b=1, y0=0, x0=1.5, according the the oscilloscope The transition occurred from 22:48:30 UT - 22:49:16 UT. (mezzi0)   TS1 TS2

Simultaneous Multiple Femtometers (4/12/07)

• Over the last few days, we have investigated running simultaneous multiple femtometers on a single Datatrends DSP (munki1) and Stealth P4 computer. The first femtometer (AD0 and AD1) were connected to an interferometer modulated at 53.91 Hz. The second, third and fourth interferometers (AD2 / AD3, AD4 / AD5, and AD6 / AD7) were connected to a simulated fringe signal driven by a programmable function generator. (These femtometers were driven by a common source). The following figures show that at first glance everything seems to be working just fine.
  
  
  • Femtometer 1: modulated interferometer signal [femtometer 1]
    
    
  • Femtometers 2 - 4: synthetic fringe signal from function generator [femtometer 2] [femtometer 3] [femtometer 4]

Solomon Islands Event (4/6/07) Mw=8.0

• On Sunday April 1st, 2007 at 20:39 UT (091) a large earthquake struck near Gizo in the Solomon Islands. iSeis recorded this event. Although iSeis is not running well at this time, in the long period regime, (because of polarization issues) we still were able to record this event.
  
  
  • Figure: Mass Position
    
    
  • Velocity Figure: Clearly, we still have an issue caused by the integration term.
    
    
  • Other Figures: We still have a timing issue with the GPS and oscillator. A fix has been found, but has not yet been installed. The timing gaps create large spikes when taking the derivative of mass position. fig 1 (overall) and fig 2 (zoomed) Integration, seen in the middle of each figure is still creating some problems.

Power Spectra and Noise (2/16/07)

• Since the Kuril Event in November, it seems our noise floor has increased. A number of changes have occurred since then. The NRTS system was installed on Nov 21s, the GPS was installed on Dec 8th and the new laser was installed a few days before Christmas (Dec 21, 2006). I have taken four 72 hours segment around these days to help solve the mystery noise problem.
  
  The days are as follows: (1) Dec 14 00:00 - Dec 16 23:59, (2) Dec 18 00:00 - Dec 20 23:59, (3) Jan 09 00:00 - Jan 11 23:59 and
  (4) Jan 29 00:00 - Jan 31 23:59
  
  
  • Power Spectra Plots : Power Spectra Plots and Time Series Plots for the days in question (green = STS1, blue = SIFO)
    
    
  • Data Sets : Mass Position data at 40 sps, in radians for the above time segments.

Configuration File Woes (2/7/07)

• We are beginning to suspect that something may be wrong with the new configuration files in NRTS. According to femto.cfg the ellipse fitting time should be about 250 seconds between fits. Rather, we are seeing about 1 second (10 samples.) Instructions to compute ellipse buffer size: take fss (sample frequency) and divide by the product of the two decimation factors (phsdcm and elldcm). This gives the rate at which new points are added to the ellipse buffer. For example: 100000 per sec / (5000 * 10) = 2 per second added to the ellipse buffer. The ellipse buffer has 500 points, so it goes back into the past 500 / (2 per second) = 250 seconds.
  
  
  • Configuration File: This is the configuration file since NRTS was installed in November of 2006
    
    
  • Figure 1: Plot of ellipse parameters over 24 hours (2007009 00:00 - 23:59)
    
    
  • Figure 2: Plot showing how often the ellipse parameters change by first difference technique.

Mass Position vs Temperature (1/30/07)

• The effect of temperature on mass position has been investigated. Temperature data and mass position data for 216 continuous hours was recorded. Mass position data was converted to units of acceleration (microGal) at long periods, low pass filtered and detrended. Theoretical tides for the region were removed and the residual was plotted against low pass filtered, detrended temperature data. Below, are the results which seem to imply that there is a phase delayed mass position dependence of temperature. Note: there are some data holes in our SIFO data sets which will introduce some skews along the time axis.
  
  
  • Figure 1: Three different plots showing SIFO mass position, theoretical tidal data, and detrended temperature data
    
    
  • Figure 2: Two separate plots showing SIFO mass position with tidal correction and filtered temperature data
    
    
  • Figure 3: A single plot showing corrected mass position and smoothed temperature

Kuril Island Event (1/16/07) Mw = 8.2

• The Kuril Island Event was a normal mode event. The following power spectra plots were created with STS1 (1 sps) and SIFO (1 sps) data over 74 hours.
  
  
  • Kuril Set 1 PS : This directory contains plots showing power spectra data for 74 hours without the theoretical sphereoidal earth modes
    
    
  • Kuril Set 2 PS: This directory contains plots showing power spectra data for 74 hours with the theoretical sphereoidal earth modes
    
    
  • Non Event PS : This directory contains plots showing power spectra data with for 74 hours before the event (no teleseismic events).
    
    
  • Comparison Power Spectra Plots: smoothed power spectra plots for both seismometers before and during the Kuril Event
    
    
  • Kuril Event Power Spectra Plots: log power spectra plots for the STS1 and SIFO: [0 - 10 mHz] [0 - 50 mHz] [0 - 500 mhz]
    
    
  • Non-Event Event Power Spectra Plots: log power spectra plots for the STS1 and SIFO: [0 - 10 mHz] [0 - 50 mHz] [0 - 500 mhz]
    
    
    
    
• When one looks at the "Non-Event Power Spectra" plots it is slightly disappointing. We seem to have a large amount of noise in our spectra, in comparison to the STS1. This may be caused by temperature and pressure variations in the bell jar. Below are two plots showing pressure and temperature over 74 hours. The same 74 hours as the above "non-event" data set. Unfortunately my decimation technique for the pressure and temperature is not perfectly sound. In a perfect world, I would load the entire data set and decimate across it. (Bringing the data from 400 Hz to 1 Hz.) Our world is not perfect, however, and RAM limitations in my computer only allow me to load 24 hours worth of 400 sps data at a time. This means, rather than decimating over the entire data set, I decimate over 24 hour chunks and then append these chunks together. This will definitely introduce some noise and problems. But they should be fairly inconsequential in this case. I am just interested in the pressure and temperature trends within the bell. In the future, care more care may need to be taken. I have also created a video of the ellipse stability during this period. The laser is definitely in some sort of unstable mode!
  
  
  • Temperature Plot: temperature within the bell over the 74 hour non-event period
    
    
  • Pressure Plot: pressure within the bell over the 74 hour non-event period
    
    
  • Ellipse Stability: This video shows the laser is definitely acting strange during the "quiet" period. The time span in this video is about 96 hours. (2007009 00:00 - 2007012 23:59) <--- POSSIBLE CAUSE FOR NOISE?
    
    
    
    
• Data Sets for both time series, before the Kuril Islands Event and during the Kuril Island Event spanning 74 hours.
  
  
  • 2007009_266400_data: contains two data sets: 1) lhz_noevent -- STS1 non-event data sampled at 1 Hz, units of counts. 2) masspost_noevent -- SIFO non-event data sampled at 40 Hz, units of radians. (2007009 04:00:00 - 2007012 06:00:00)
    
    
  • 2007013_266400_data: contains two data sets: 1) lhz_event -- STS1 event data sampled at 1 Hz, units of counts. 2) masspost_event -- SIFO event data sampled at 40 Hz, units of radians. (2007013 04:00:00 - 2007016 06:00:00)
    
    
    
    
• On January 13th, 2007 (013) at approximately 04:23 GMT, the Kuril Islands (46.7N, 154.4E) experienced a large event measuring 8.2 Mw. This earthquake created a large signal and was recorded by SIFO.
  
  
  • Figure 1: A plot of the event showing mass displacement sampled at 400 sps (radians)
    
    
  • Figure 2a: A plot of the event in units of velocity with tidal signatures sampled at 400 sps (m/s)
    
    
  • Figure 2b: A plot of the event in units of velocity high pass filtered at 360 seconds sampled at 400 sps (m/s)

Ellipse Stability (1/10/07)

• Shortly before the new year a number of changes were made. The bell jar was replaced with a slower leaking jar, the laser was replaced with the frequency stabilized ML-1 and the seismometer was received and optimized. It is still unclear what benefits we may see from this. What is clear, however is that something is still peculiar with the laser. Following are two videos which show ellipse stability over a 72 hours period. The first video was from the hours immediately leaving the vault and the second video is from the last few days.
  
  
  • Video 1: Dec 22nd - Dec 24th (30.8 Mb)
    
    
  • Video 2: Jan 6th - Jan8th (30.7 Mb)
    

STF Vault Temperature (12/07/06)

• The TR-1050 Data loggers have been running in the STF vault for some time now. On a recent trip (yesterday), in which a ROVDOG gravity meter was deployed to compliment SIFO, and SIFO was re-evacuated, data records were collected from the temperature loggers. Approximately 14 days worth of data are now available. The following plots show temperature logs of both the hallway and the SIFO.
  
  
  • Figure 1: Hall Temperature
    
    
  • Figure 2: Hall Temperature Detrended
    
    
  • Figure 3: SIFO Temperature
    
    
  • Figure 4: SIFO Temperature Detrended

Kuril Island Event (11/15/06) Mw = 8.3

• The Kuril Island Event was a normal mode event! The following file (pdf) shows the normal modes as recorded by a co-located STS1.
  
  
  • Event Mass Position Data: mpdata_319a, raw phase (radians), 20 sps, length: 53 hours, 319 11:00 - 321 15:59:59
    
    
  • Event Mass Position Data: mpdata_319a, raw phase (radians), 400 sps, length: 53 hours, 319 11:00 - 321 15:59:59, file size ~190 Mb
    
    
  • Before Event Mass Position Data: mpdata_314, raw phase (radians) 20 sps, length 53: hours, 314 04:00:00 - 316 08:59:59
    
    
  • Before Event Mass Position Data: mpdata314_190800, raw phase (radians) 400 sps, length 53: hours, 314 04:00:00 - 316 08:59:59 ~185 Mb
  
  
  
• The following directories contains figures of the normal modes from and before the Kuril Island Event for both the STS1 and STS2.
  
  
  • SIFO vs STS1: normal modes as recorded by the STS1 and SIFO from 1 - 8 mHz during the event.
    
    
  • SIFO Comparisons: plots of SIFO for 53 hours before the event and 53 hours immediately following the event (key figure)
    
    
  • STS1 Comparisons: plots of STS1 for 53 hours before the event and 53 hours immediately following the event (key figure)
    
    
• On November 15th, 2006 (319) at approximately 11:14 GMT, the Kuril Islands (46.683N, 153.223E) experienced a large event measuring about 7.8 Mw. This earthquake created a large signal and was recorded by SIFO.
  
  
  • Mass Position and Velocity Data: mp_319: raw phase (radians), udot: velocity (m/s), both: 400 sps, 319 11:00:00 - 319 17:00:00 (GMT), ~83 Mb
    
    
  • Figure 1: A plot of the above data set showing mass displacement (radians)
    
    
  • Figure 2a: A plot of the above velocity data with tidal signatures (m/s)
    
    
  • Figure 2b: A plot of the above velocity data (400 sps) high pass filtered at 360 seconds. (m/s)
    
    
  • Figure 3: A screen capture of the STS-1 IDA velocity data for the above event.

TR-1050 Temperature Data Loggers (11/4/06)

• Two RBR TR-1050 temperature data loggers were placed in a deep tank for ~48 hours to determine their relative accuracies. The data loggers have a specified range of -5 °C to +35 °C, an accuracy of ±0.002 °C and a resolution < 0.00005 °C.
  
  
  • TR-1050 Data Logger #1 Data Set (.mat file)
    
    
  • TR-1050 Data Logger #2 Data Set (.mat file)
    
    
  • Figure 1: Time Series of the above data sets
    
    
  • Figure 2: Data logger temperature difference plot and linearity plot

Ocotillo Wells Event (11/3/06)

• On November 3rd, 2006 (307) at approximately 15:56 GMT, Southern California experienced a local event measuring about 4.5 Ml. SIFO recorded the event.
  
  
  • Mass Position Data: mp_307, raw phase (radians), 400 sps, 307 15:45:00 - 307 16:45:00
    
    
  • Figure 1: figure of the above data set showing ~300 radians of PP displacement

How Well is SIFO Working? (10/18/06)

• This is a difficult question to answer. However, in a recent meeting with E. Wielandt we felt things were looking quite well and progressing nicely. A number of graphics and data sets were presented in support of our case. Using data from data sets below we have been able to create a number of power spectra and earthquake comparisons. (Image credit to J. Berger)
  
  
  • 1 SPS Power Spectra: Power spectra plot of 1 sps data show SIFO and a co-located IDA STS1. Long period noise in SIFO has come a long way!
    
    
  • Power Spectrum: An 8 hour power spectrum (20 sps) comparing SIFO and a co-located STS1 and STS2.
    
    
  • Regional Earthquake: A Mb 6.1 regional event near Acapulco, Mexico showing first arrivals recorded by SIFO and a co-located IDA STS1.
    
    
  • Power Spectra of Regional Event: A power spectra plot of the above event, comparing SIFO and the IDA STS1.
    
    
  • Local Earthquake: A comparison of SIFO and a co-located IDA STS2 for a local event at broadband frequencies and band passed from 1 - 5 Hz.

Config File Change (10/16/06)

• A configuration change was made in femto.exe Ellipse fitting has been turned off to see what effects ellipse fitting has on long period noise. The old configuration file can be seen by clicking here. The new configuration file with the ellipse parameters can be read here. The change was made at 289:1929.
  
  
• Unfortunately since I have made this configuration file change, the planet has been struck by a number of large earthquakes. At least 8 events over magnitude 5.0 have occurred in the last 10 hours. The first five hours of the data set, however are uneventful and should provide a decent amount of data to look at long period noise.
  
  
• Mass Position: mp_289, 400 sps, raw mass position in radians, 8 hours, time window: 289 20:00:00 - 290 03:59:59
  

Hawaii Earthquake (10/15/06)

• On Sunday October 15 2006 at 10:07 AM (PDT) a Mw 6.7 event occurred near Hawaii. The event details are as follows:
  latitude: 19.878, longitude: -155.935, UTC Origin Time: 10/15/06 (288) 17:07.49, Depth: 38km
  
  
• SIFO recorded the event. A pdf figure of the event can be seen by clicking this link.
  
  
  • Mass Position: 400 sps, raw mass position in radians, start time of data set: 288 17:00:00

Config File Change (10/09/06)

• A new configuration file was implemented on day 227 at 22:00 hours. The change was run over the weekend, in the hopes that long period noise would be lowered. (i.e. we need to find out if long period noise is caused by the DSP). The new ellipse config file fits the ellipse more rapidly (once every second, rather than once every four seconds.) Technical: elldcm changed from 40 to 10
  
  
• Old Config File versus New Config File
  
  
• The following data set is 12 hours of mass position, sampled at 400 sps from day 280 00:00 - 280 11:59:00
  
  
  • Mass Position: 400 sps, 12 hours, raw mass position in radians, 280 00:00 - 280 11:59:00

A New DSP Problem (9/27/06)

• The last few days have been spent attempting to align the theoretical tides with the SIFO tides using least squares. Unfortunately the SIFO data seems to be skewed. (i.e. in the beginning of the data set the tides seem to be slightly ahead of phase and at the end of the data set they seem to be behind phase.) The following figure better illustrates what I mean. Figure 1.
  
  
  
• After further investigation into the individual hourly data sets, I have found some problems with the .sam data sets. The following is taken from a list of sequential sam files. The list shows the number of data points per file and the sample set size in seconds. Each file should ideally be 1440000 samples in length and 3600 seconds long. We, however, seem to be no where near this!

  1439120 samples    3597.8 secs
  1439160 samples    3597.9 secs
  1439080 samples    3597.7 secs
  1439080 samples    3597.7 secs
  1439040 samples    3597.6 secs
  1439120 samples    3597.8 secs

• And here is something even stranger. I have checked it four or five times. This is not a mistake. At least one sam file is as follows:

1463440 samples     3658.6 secs    << ---- This is over an hour long!  (maybe making up for the shorter files?)

• I have no clue as to why this is happening and I think the implications are quite bad. I will see about readfemto versions and such . . but this error in timing may explain why we can't align the tides properly.
  
  
• Even stranger, the files seem to still properly align in time, and there seems to be no data loss. The following figures show mass position across the file boundary. The red vertical line depicts where the file boundary actually occurs. The black vertical line depicts where the theoretical file boundary should occur. Figures: [1] [2] [3] [4]
  

Long Period Acceleration in SIFO (9/25/06 - current)

• We would like to examine long period accelerations in SIFO (i.e. the tides). The following data sets and figures help with this investigation. The data sets span from 213 01:00:00 to 224 20:59:59 GMT or 284 hours unless otherwise noted.
  
  
  
• Data Sets:
  
  • Mass Position: 20 sps or 1 sps -- phase (radians) (213 01:00:00 to 224 20:59:59)
  • Temperature Data: 20sps or 1 sps -- degrees C (213 01:00:00 to 224 20:59:59)
  • Mass Position Data: 400 sps for the first eight (8) hours from the above data set (radians) -- w/o EQs (213 00:00:00 - 213 07:59:59)
  • Mass Position Data: 400 sps for the teleseismic event which occurs around day 213 14:00, 1 hour, phase (radians)
  • XY Data Series: 400 sps for the first nine (9) hours (213 00:00:00 - 213 08:59:59)
  • Tidal Data Series (from D. Agnews tide programs): T = 300 seconds, microGal, (213 01:00:000 - 224 21:59:59)
  • Phi Data Series: 10 sps, phase in radians, elps(:,5) (213 01:00:00 - 224 20:59:59)
  • Pressure Data series: 1 sps, data from IDA vault, units of microbars, (213 01:00:00 224 20:59:59)
    
    
• A number of figures have been generated to see how w (the natural frequency of the seismometer), effects recorded SIFO data. In this set of figures filtering remains constant (low pass at 180s, no high pass) and w (the natural frequency of SIFO) is varied Each set of figures is as follows: A -- SIFO Mass Position Unfiltered, B -- Filtered Mass Position Data with a 2nd Order Best Fit Polynomial, C -- Filtered Mass Position Data with 2nd Order Correction.
  
  
  • W = 1.50: Figures A, B, C
  • W = 2.00: Figures A, B, C
  • W = 3.00: Figures A, B, C
    
    
• A quicktime video showing progression of ellipse stability (x and y pairs) can be downloaded here (86.3 Mb): AVI Movie
  
  
• Power Spectra:
  
  
  • Power Spectra plot of SIFO and STS1 acceleration over first 8 hours of data series (pdf). Our noise floor has dropped from where we were earlier in the year!
    
    
  • Power Spectra plot of SIFO and STS1 acceleration over first 8 hours of data series where SIFO data is computed from a static set of ellipse parameters and applied to the xy-data pairs. (pdf) Notice how the noise increases at long periods.
    
    

New Laser, Optics and UPS Installed (7/19/06)

 

• An APC 750VA Smart UPS was installed at PFO early this morning. Any power issues will be electronically mailed as well as texted to my phone. It should be noted that the PDU will no longer have an accurate power logs, as it is now connected to the battery. The new installation will hopefully prevent random reboots of the data acquisition computer.
  
  
• The existing laser and optics have also been changed. The original laser was definitely faulty as the power supply sounded as if it were arcing and the laser was haphazardly flashing on and off. We don't know if this was related to the recent electrical storms or if it was just time for the laser to die -- but a new laser was installed as well as a new optical system. A new polarization alignment was performed and the lambda/4 polarizer was set at 358 degrees and lambda/2 polarizer was set at 31 degrees.
  
  
• The seismometer bell jar is evacuated to maximum vacuum.

Laser Problems (7/13/06)

It now seems as something strange may be occurring with the laser. The following data set (1941900) may better illustrate the problem.

• Figure 1: Mass Position Data (radians)
  
  
• Figure 2: XY Data (i.e. photometers voltage in volts). Have never seen it look so one sided before. This is kind of strange.
  
  
• Figure 3: Ellipse Parameter Data (notice how much the values are changing; it would seem to suggest the config file is wrong, but this hasn't been touched for some time.)
  
  
• Figure 4: Here is a plot of "normal" XY Data.
  
  
• Figure 5: A screen capture of the real time windows also looks strange. Look at the fitted ellipse data.
  
  
• MATLAB data of the above data sets. Click
  

                             
   elps 36000x5 1440000 double array
   phfdata 1440000x1 11520000 double array
   xydata 1440000x2 23040000 double array    

   

Power Issues at PFO (7/11/06)

 

• I am noticing strange power issues at PFO. At first these problems were somewhat infrequent but as the months elapse they seem to be occurring more regularly. A number of scenarios can occur: 1) the computer reboots and data collection ceases until I notice that no new data is coming in. 2) The DSP seems to enter a strange state in which the data is irregular (figure), and 3) femto only collects data for a few hours and then stops stops collecting -- although indicates on the monitor that everything is okay.
  
  
• The following is a log file from the power strip which indicates some sort of power glitch. I have yet to hear if others are experiencing similar power issues. I will attempt to put a UPS in place, within the next week.

• Uploaded and started the new code (rev 3.07) on 172 1720510 (UTC).
  
  
• Twenty four hours worth of mass position (radians) and ellipse data from day 175. Click
  

  ellipse        863983x5             34559320  double array, 10 sps, ellipse parameters
  masspost       34559320x1           276474560 double array, 400 sps, mass position (radians)

• Twenty four hours worth of mass position (radians) and ellipse data from day 188. Click
  

    ellipse    863983x5                 34559320  double array, 10 sps, ellipse parameters
    masspost 34559320x1                 276474560  double array, 400 sps, mass position (radians)

• After closer inspection of data from day 188, I am not convinced our "spike" problem has been resolved. The following three figures may better illustrate my point. The top portion of each figure shows mass position and the bottom half show the first derivative of mass position. Notice the peculiar slope changes in the top half of the following figures. Figures (pdf)

• DSP Version 3.06 with the code changes mentioned below, seems to be working well. Previous confusion stemmed from errors within the real time display and the ellipse buffer. A simple test was ran, in which minimal ellipse buffers were used. Microseism data from a nearby STS1 was compared to microseism data on the SIFO. The results were promising and can be viewed below.

  
  Data from this hour can be downloaded here.
  

 phfdata       1439960x1       11519680 double array,  mass position data in radians at 400 sps (SIFO)
 xpf        72240x1         577920 double array,   velocity in cm/s for vertical STS1 at 20 sps (XPFO) 

• Comparisons of the microseisms between the STS1 (still noisy at times) and SIFO (highly filtered). It seems as if the peculiar problem I was seeing yesterday has gone away. Click
  
  
  
• At this time, there are no obvious large phase offsets (uberspikes) in the phfdata. This is apparent when looking at the following: Figure 1
  
  
  
• At this time, there are no obvious small phase offsets (spikes in the first differenced mass position) data. As an aside, however -- while John and I were looking at the first differenced phfdata, we noticed what seems to be a phase wrapping issue at the extreme values of the microseism curves. See the following figure. This behavior is not new. It has been around for quite some time infact. I don't know why we didn't notice it before. This definitely warrants some further investigation. Figure 2
  
  
  
• Here is a different hour of data from the SIFO. After converting it to ground velocity it looks like we may have an offset. This however can not be an offset as it occurs over many sample points, which indicates it is infact a real downward trend. Figure 3
  
  
  
• Hopefully by next week, with normal ellipse configuration parameters, and after collecting more data, I will be able to reach a conclusion about the success of the newly written code (v 3.06).

• Dave, the new programmer, has made a few code revisions which address the "uberspike" issues. The new code (v 3.05) was loaded into the DSP at approximately 12:51 on June 14th, 2006 PDT.
  
  
• As before the phase offset issue lies within the definition of tangent when equal to zero:

  Here's the workaround for the atan2f(0,x) = 0 issue:
      if (sinp == 0) {
          if (cosp >= 0)
              ph = 0;
          else
              ph = pi;
      } else
          ph=atan2f(sinp,cosp)

• After letting the code fun for a short while, it becomes immediately apparent that something seems to be wrong. It may be a lack of ellipse fitting time -- however, no changes were made to the configuration file so this seems unlikely.

Figure 1: A broad span of data. Notice the offsets and in the next figure peculiar rounded peaks (almost as if filtered) at the top of the microseism curves.

Figure 2: A zoomed in portion of Figure 1. Notice that the smaller peak to peak oscillations look like typical microseisms, but the large oscillations do not show typical microseism behavior.

Figure 3: It may be useful to compare this to an (look at the bottom plot), which has "standard" spiked data. Notice how the microseisms oscillate about a baseline. This is "normal" behavior.



• It is possible that in the above code revision something else was inadvertently changed. The new version was rolled back to an older version and then, only the above changes were made. The new code (v 3.06) was loaded and re-tested. It was started at about 14:14 PDT (21:14 UTC).
  

Figure 4: Again we see similar behavior as with revision 3.05. The entire time window looks strange. It looks as if the data is being filtered by the DSP!

Figure 5: Zooming into the microseisms even looks stranger.

• Although perhaps a futile attempt, I will power cycle the data logger and wait a couple of hours to see if the data improve. The power cycle was preformed at 15:08 PDT (22:08 UTC)
  
  
• On June 15th, 2006 at 10:08 PDT (166 17:08 UTC) the DSP code was rolled back to version 3.04 and allowed to run for a few hours. The data looked normal, so version 3.06 was re-installed on the DSP after a computer and DSP reboot. (166 22:32 UTC)
  

• Approximately five days worth of data were carefully strung together in an attempt to find normal modes of the earth. The data was dephased (by hand) and then carefully filtered/decimated so as to avoid end effects at the one hour boundaries.
  
  
• The data range from days 144 (13:00) through 149 (10:59). The 20sps mass position data (in radians) is actually a bit less than five days (4.917 to be exact or 4 days 22 hours) as the head and tail had to be chopped due technicalities in the decimation programming. (Note: the ideal time span would be from 144 12:00:00 - 149 11:59:59 but with the head and tail chopping, we get the time range stated above.)
  
  
• For a figure of the detrended data, click here. Figure 1
  
  
• For the 20 sps mass position data (not detrended) in radians: Click here (matlab) for a binary matlab file or here (ASCII) for an ASCII text file (~211 Mb).
  

  phfdata 8495972x1                  67967776  double array, mass position (radians), 20 sps        

• For STS1 data (velocity in counts) from station PFO please click here (matlab) for a binary matlab file or or here (ASCII) for an ASCII text file (~211 Mb).

• Here is an sam file containing the old style spikes. These are most apparent when you take the first difference of the phfdata and plot it. The first differenced spikes correspond to rapid an d unphysical changes in slope of the mass position. Click
  
  
• Here is a newer sam file containing the large phase offsets. (These occurred after Petar made the first code revision.) For further information about the large offsets (uberspikes) see below. Click
  

• Gravity and Tides
  
  Using SIFO mass position we investigate gravity and tidal signals at long periods. The following figures show tidal signals and corrections for SIFO (at PFO) as well as the predicted tidal loading for PFO. The tidal test ran for five days from 105 00:00 - 109 13:00 UTC. Click
  
  
  The above data set exhibits a bit of noise. Here is a quieter data set but with less noise for days 105 00:00 - 106 23:59 (48 hours). Click

• Ground Motion Comparison:
  
  We compare the recorded ground motion between SIFO and an STS1 located at PFO (about 0.5 km separation) for a teleseismic event in Tonga as well as a simultaneous local event.
  
  
  • The vertical data (udot is SIFO and xpfm is STS1 at 20 sps) set can be downloaded by clicking here: Click
    
    
  • The raw mass position data (masspost at 400 sps) for SIFO can be downloaded here: Click
    
    
• A time progression of the teleseismic and local events can be seen here: Click
  
  
• The local event in greater detail can be seen here: Click
  
  
• A nice graphic showing the local event on top of the teleseismic event can be see here: Click
  
  
  

• Power Spectra: (144-145)
  
  
• Power Spectra for a twenty-four hour period (2006144 12:00 - 2006145 12:00) for a vertical STS1 (xpfo) and SIFO have been created and compared. Figure 1 shows the power spectra and Figure 2 depicts the time series.
  
  
• The following image shows raw mass position (radians) for SIFO during the above time period. Figure 3
  
  
• Click to the right for raw mass position data of the SIFO (in radians) for the above time periods. 20 sps or 400 sps (~90 Mb)
  
  
• Click here for the ellipse parameter data at 10 sps. To see a plot of the ellipse parameters click Figure 4.
  
  
• Click here for temperature data in the SIFO vault during this period (temperature, 1sps)
  
  
• Click here for vertical velocity data sets for SIFO and XPFO

xpfm 1727040x1                  13816320  double array,  20 sps, m/s,  XPFO 
udotdec 1727994x1               13823952  double array,  20 sps, m/s,  SIFO

• Hysteresis and Creep in Flex Pivots

During the summer months a great deal of time was spent on investigating the flex pivots within the SIFO. In summary, we feel that SIFO may behave a bit better if we replace the commercial flex pivots with custom built pivots like those of the STS1. Here are a few figures which summarize the hysteresis effects.

• Figure 1: An example of a hysteresis loop with a caption explaining the experiment
  
  
• Figure 2: Multiple hysteresis loops recorded over a series of experiments
  
  
• Figure 3: A single zoomed hysteresis loop, of the above styling.
  
  

• Running the differential equation for ground velocity on SIFO data creates a long T signal over the entire time window of one cycle. This can be filtered or corrected, but it would be nice to figure out what causes this trend. Note: After a bit of analysis, it has been found that the artifact is created by a combination of the cumsum (integral) term of the differential equation and the data de-trending process. Figure 1
  
  
• Running the SIFO filter designed by J. Berger gives similar results. We still see the same trend. Perhaps this trend is real, but more than likely it comes from the numerics of the problem. A question still remains, however. Which method is better? I hope to answer that in the near future. Figure 2
  
  
• Running the same analysis on a longer time series (say 80 mins) produces the same results. This suggests the trend is not real, but rather an artifact created by the the data processing. Figure 3 and Figure 4
  
  
• The data set for the shorter time series can be found here: click

• It looks as if the uberspikes (or phase offsets) have returned! Here are a few figures showing the uberspikes: Figures 1 - 3
  
  
• The large offsets in phase are related to phase wrapping within the DSP. I have closely looked at 11 phase offsets and computed the difference in phase before and after the offsets occurs (delta offset). All of these offset deltas are integer multiples of 2*pi*N.
  
  
• Figures 1 - 11, show the phase offset (top), a zoomed in view of the phase offset with a computed delta for the offset (middle), and a zoomed in view of the first differenced phase data (bottom).
  
  
• Figure 12 shows a histogram plot of each of the phase offsets as a function of integer multiplicity of 2*Pi*N. It is clear that each of these offsets is indeed a multiple of 2*Pi.

• Day 129.00 (05/09/06 (129) 00:00:00 GMT, 05/08/06 (128) 17:00:00 PDT) -- no known vault entry or code changes; unexplained
  
  
• Day 131.00 (05/11/06 (131) 00:00:00 GMT, 05/10/06 (130) 17:00:00 PDT) -- no known vault entry or code changes; unexplained
  
  
• Day 131.75 (05/11/06 (131) 18:00:00 GMT, 05/11/06 (131) 11:00:00 PDT) -- T. Johnson and P. Davis working in IDA Vault from 11:00 to 15:00 PDT.

• I see an occasional"uber-spike" when looking at data collected over the weekend. (when sp was 3000) Figure 1 shows the overall size of the spike and Figure 2 shows the zoomed in spike and x and y voltage pairs. Here is the matlab data for these figures. Notice the lack of change in x and y. The oscillatory nature of these spikes is caused by a filter within the DSP.
  
  
• Here is another example of an uber-spike: Figure 2 It seems as if one or two of these occur an hour. Here is the matlab data set for this figure.
  
  
• And yet another example of an uber-spike as an overall view (Figure 3) and a zoomed in view (Figure 4). Notice how much larger these spikes tend to be.
  
  
• Even when sp is set to 3000, the spikes referenced below have all but disappeared. Here sp 3000 data shows how it would be hard for our ~0.8 rad spikes to be lost in the noise. Figure 5
  
  
• After setting sp back to zero, earlier this morning (5/15), it doesn't appear as if the spikes have come back. Figures 6 - 11

• Before rev to code: Here is what the phase offsets / spiking looked like before the code revision: Figure 1
  
  
• Allow new rev of code to run: After running the code for ~40 hours, I still haven't seen any phase offsets or spiking. Figure 2
  
  
• Set SP to 3000: setting the phase offset to 3000 (5/10/06) radians has not effected spiking, yet it has made the diff(phfdata) data seem noisier. Look at the 2nd plot in each of these figures (data taken hours > 12 after the sp 3000 change): Figure 3 and Figure 4
  
  
• Here is what the data looked like immediately after the SP 3000 change (Figure 5) and immediately before the SP 3000 change (Figure 6). Notice the noise introduced after the change. This is caused by least count noise within the DSP.

• A large number of pdf files showing wave forms and power spectra of an STS1 (in the old IDA vault), an STS1 (in the new STF vault) and the SIFO (in the STF vault) through 60 second windows during the Tonga EQ (mostly surface waves): click
  
  
• It seems as if the STS1 in the STF vault (green) is has a lot of high frequency noise. This is especially noticeable in the later pdf files.
  

Note: In one of the images you can see a local earthquake within the surface waves of the teleseismic Tonga event. (an earthquake within an earthquake.)

Note: J. Berger mentions that XPFO STS1 vertical has a problem that changes it's hi-f response and causes noise bursts. This will be fixed around 5/16/06 and explains the noise in the above figures.

• I have found that these seemingly "random" spikes are not that random. They do infact occur at a very particular location along the ellipse. The following folder contains a number of pdf files which show the spikes in detail: click
  
  
• Spikes above a certain threshold (pdf) were recorded and their corresponding position on the ellipse was then plotted in either red or green (pdf). These image without a doubt shows the spikes are not random.

• Mass Position: 110 hours of mass position, 1sps/20sps, units as follows: 1sps or 20sps

    masspost       394038x1                   3152304  double array  mm
    masspost_filt  394038x1                   3152304  double array  ugal  low pass filter at 1 hr
    masspost_gal   394038x1                   3152304  double array  ugal
  ***** OR *****
    masspost      7880674x1                  63045392  double array
    masspost_filt 7880674x1                  63045392  double array
    masspost_gal  7880674x1                  63045392  double array
                      

• gravity: 110 hours of gravity data, 1 sps, position in ugal click

    grav  392801x1                   3142408  double array ugal 

• xpfm: 105 hours of STS1 data (pfo), 20 sps, velocity in m/s click

    xpfm 7560000x1                  60480000  double array m/s

• tide: 110 hours of predicted tidal data for (pfo), in ugal click

     tide       1x1319                  10552  double array
                      

• tides: 110 hours of predicted tidal data for (pfo) in ugal at 1 sps : click

gdc        1x1                         8  double array
tide       1x395999              3167992  double array
time       1x395999              3167992  double array

• mat files: mat files for each hour of data (masspostion data only) click

  ellipse    71996x5                   2879840  double array
  masspost 2879840x1                  23038720  double array
  udot      143991x1                   1151928  double array
                  

• masspost: 2 hours of mass position, 400sps, position in m click
• udot: 2 hours of ground velocity data, 20 sps, m/s (same file as above)
• Ellipse Parameter Data: Ellipse Parameters, 10 sps, 2 hrs (same file as above)
• STS1 velocity data: 2 hours of STS1 velo data in m/s click
• Power Spectra Plot: click

  elps      19007x5                    760280  double array
  phfdata  760280x1                   6082240  double array
                  

• Mass Position and Ellipse Parameters: approx 1/2 hour of data click

ellipse    72000x5                   2880000  double array
masspost 2880000x1                  23040000  double array
                  

• Mass Position Data: Mass Position in mm, 400 sps, 2 hrs click
• Ellipse Parameter Data: Ellipse Parameters, 10 sps, 2 hrs click
• Code: Code to check units and compute transfer function. click
• PFO DATA: PFO Data from an STS1 for the same time period click

• Mass Displacement Data: in raw radians, 400 sps click
• Ground Acceleration Data: in m/s^2, 400 sps click
• Ground Acceleration Data: in m/s^2, 20 sps click
• Long Period Acceleration Data: in m/s^2, 400 sps click -- first two terms from seismic deqn removed

• August 10th, 2007: New flex pivots are installed in iSeis. Optical STS-1 (iSTS1) was also installed.
  
  
• April 20th, 2007: Station computer replaced with BSD Stealth machine, GPS oscillator circuit installed, iSeis re-leveled, mirrors and wave plates removed, new laser and PM fiber installed.
  
  
• December 21st, 2006: Bell Jar replaced, laser swapped from Melles Griot to ML-1
  
  
• December 8th, 2006: GPS installed onto the DSP and timing signals now arriving to NRTS system. (day 342)
  
  
• November 21st, 2006: The newest version of FEMTO NRTS is now running. Seismometer data now streams into the NRTS!
  
  
• October 18th, 2006: The first talks of installing SIFO types seismometers at BFO (Black Forest Observatory) in Germany have started.
  
  
• July 19th, 2006: UPS is installed. New optical system is installed. Old laser removed and a new laser is installed.
  
  
• Jun 14th, 2006: A new revision to the femtometer code has been made. It is our hope that this revision will address the "uberspike" issue. Dave has found that the problem lies within the definition of tangent, when it crosses the zero crossing.
  
  
• May 16th, 2006: The spike problem has not been fully resolved. The small spikes are gone but replaced by "uberspikes" with a frequency of one or two per hour.
  
  
• May 8th, 2006: A change has been made to the femtometer code. The newest version contains a higher precision, but slower, atan2f function. Hopefully within a few days I will be able to say if this was the solution to the spike problem.
  
  
• May 8th, 2006: We get a nice recording of the Tonga earthquake (Mb 8.0), as well as a simultaneous local event. Look above to find the waveforms and power spectra.
  
  
• May 5th, 2006: We find the spikes are strongly phase dependent. Look above for further information.
  
  
• May 4th 2006: We realize the spikes are not coming from the optics or the laser. They are in fact being caused within the DSP. (i.e. we don't see any spikes in the xy pairs.) Look above for further information.
  
  
• March 7th, 2006: We make a 2nd trip to PFO to check the optics, laser and seismometer to see what is causing the spikes. We re-align the femtometer, clean the optics and run some null tests. Days later we find out the spikes have returned.
  
  
• February 28th, 2006: SIFO is showing bizarre behavior. Its seems as if the mass is "stepping". This is most obvious when looking at diff(phfdata) which shows very dramatic spiking.
  
  
• February 17th 2006: We finally deploy our first prototype seismometer in the STF vault at PFO. Click here for photos!