Tempe Goes Seismic

This page is experimental and provides a place where I can post data from a homemade seismometer. In development, artifacts are expected and the data is meant for personal use - to learn.

Details: The seismometer transducer is a torsion balance consisting of two copper pennies suspended between two Nd-Fe-B disk magnets by a thin vertical wire. Motion of copper in a magnetic field produces eddy currents; thus, the magnets provide passive damping. Transverse motion causes the pennies, which are suspended at an offset from their center, to twist the line. The motion is registered by a pair of photodiodes connected differentially to an amplifier/filter. Photos of the seismometer, and recordings of a few earthquakes which this instrument has detected are posted below the plot of current data.

Version 1 of this instrument went offline on the morning of 3/13/2010 , and was replaced by Version 2.0, a mechanical re-design, on Sunday 3/14. Details and notes about V1.0's evolution can be found here. This page documents the updated V2 design. The circuit design of both versions is identical.

Seismometer current data, uploaded hourly (time is UTC-7h)

Observations:

1. 3/21/10 Sensitivity appears to be improved with the longer pendulum - 39 inches versus 9 inches in V1.0. But the sensitivity has come with a cost. It appears that any slow change in temperature sets up a thermal gradient within the enclosure, with accompanying oscillation at a frequency of about 2 cycles per minute. Peak amplitude is about 5% of full-scale. This is the effective noise floor until I can find a way to reduce this artifact. Foam insulation between the top opening and its cover helps. The artifact is not electronic in nature - the same circuit in the V1.0 enclosure never exhibited the behavior.

2. 04/06/10 M7.7 Sumatra quake recorded, distance 9400 mi., shows significantly improved sensitivity over the previous V1.0 mechanical design.

3. 01/06/11 Removed the 802.11g access node and hardwired the seismometer USB bridge to ethernet, lowering the incidence of data packet loss.

4. 02/24/11 A year's worth of data has given a better indication of the sensitivity. A M6.2 quake in Spain was at the limit of sensitivity at 5220 miles distance, while a M6.9 quake in New Guinea registered a peak response of 10% of full scale over a 6920 mile path. A M7.7 quake over a 9400 mile path made a response of 50% of full scale. Because each decrease of 0.2 in magnitude is a halving of energy, that quake would have been detectable at M7.0. Finally, the Feb 21, 2011 M6.3 quake in Christchurch, New Zealand was, after inspection, undetectable in the data - a path length of 6300 mi.

Given this, it seems that the instrument is able to detect a M7.0 quake from just about anywhere. The present limit to sensitivity is entirely mechanical - mostly the response of the pendulum frame to thermal input. Increasing sensitivity further will require improvement in the mechanical design.

Pictures

The completed seismometer board. To the right you can see the IR emitter, magnet assembly and differential photodetector. From right to left, in order:
1) Photodiode transimpedance amplifier/filter (dc-coupled, abbreviated "TZA" in the dataplots)
2) Instrumentation amplifier and HPF
3) Sallen-Key LPF
4) SE-to-balanced output buffer to DAQ board.

The response characteristic is bandpass, with passband of 0.008 to 3 Hz.

Seismometer in its enclosure. The board is mounted on top of a 10 inch circular aluminum base plate, about 1/2 inch thick. Mount point screws are tapped and threaded, providing a means for adjusting tilt.

The pendulum wire, 5-mil dia tungsten, is suspended from the top of a rigid square aluminum pipe one meter long. Height adjustment is provided by a threaded set screw with a knob attached.

Seismometer enclosure. The DAQ board, USB-IP converter and WiFi access node are on top of the cabinet to the right.

The enclosure proper is a 10 inch spiral-wound cardboard cast form for concrete, 4 feet tall. Cost: about $12 at Lowe's hardware. It is thermally isolated on both ends with foam rubber. The top opening is covered by a sheet of foam with plywood on top.

Torsion Balance
Detail
Detail of the torsion balance, showing how the copper weights are suspended between the rare earth magnets. With a meter-long pendulum, the motion is slower and the magnets provide insufficient damping alone. The small nut at the bottom of the tungsten wire in the photo, is now suspended in a vial of mineral oil for extra damping.

A long pendulum is sensitive, but the mechanical design has some flaws:

- It has multiple modes of response: lateral motion of the weight as well as twisting motion. In a long pendulum the lateral motion dominates, so it's really not a tosrion balance.
- Sensitivity is affected by the angle and height of the flag at the point where the input integrator is centered. This makes calibration impractical.
- The pennies and the end nut on the pendulum make for two separate resonances, which can be seen in FFT plots when damping is removed. Fewer natural resonances are better.
- A meter-long pendulum combined with electronics which can detect microns of motion, is hard to stabilize mechanically. There are several joints between aluminum and steel, which have different thermal coefficients of expansion. One or more of these, I suspect, creates a low-level thermal relaxation oscillation at 0.02 - 0.04 Hz which can be seen in the raw data. It tends to appear whenever there's a thermal gradient dT/dt in the garage. It's suppressed by numerical filtering, but better to suppress it at the source.

Sumatra M7.7, 04/06/2010 This is the first large quake detected by the V2 seismometer. The M7.7 quake in Sumatra on 4/6/10, distance 9400 mi. made a response peak of 60% of full-scale. The distance record for V1.0 was the M8.0 quake in American Samoa in Sept 2009, a much weaker response at 5000 mi. distance.

The surface wave frequency is about 0.07 Hz - a very low frequency typical for the large quakes.

Prior to this quake, well over 50 aftershocks have been recorded to date from the M7.2 Baja California quake of 04/04/10, some as low as M2.0.

Travel Times M7.7 Sumatra 04/06/10 USGS seismic wave travel times for the M7.7 quake, 4/6/10.

Spain (Andalusia) M6.2, 04/12/10
UTC This M6.2 quake in southern Spain on 4/11/2010 is at the limit of sensitivity for detection. The pressure and shear wave components labeled P and S can only be seen in this zoom plot, superimposed against the instrument's own twice-per-minute mechanical thermal oscillation - a flaw which I still must correct.

This is a milestone for me. This device was born because my wife Pam was in Italy when the M6.7 quake struck in L'Aquila in April 2009. My goal was to be able to detect that quake from that distance, and it appears now to have been achieved. Now on to curing the thermal/mechanical oscillation - (which wont be easy if the required thermal isolation for the enclosure is too high).

Andalusia M6.2 Travel Times,
04/12/10 UTC USGS seismic wave travel times for the M6.2 Andalusia quake, 4/12/10 (UTC). The times for P and S agree with the recorded arrivals to within seconds.

Nomenclature for the phases of seismic waves is given here. The P waves are pressure (longitudinal) waves which travel in a linear path beneath the Earth's crust. The S waves are shear transverse waves which propagate at a slower speed than the P wave. LQ and LR are surface waves which propagate along the curvature of the Earth's crust - a longer path.

Papua New Guinea M6.9 07/18/10
The M6.9 at Papua New Guinea July 18, 2010, was one of three closely spaced quakes, and had the best definition from the garage detector. There are 4 identifiable phases: P is the initial pressure wave arrival, PKiKP is the same wave bouncing from the Earth's inner core boundary, S is the shear wave, and LQ is the surface wave. The surface waves are always lower in frequency, and typically the strongest signals from a distant quake. Distance 6920 miles.

Mindanao M7.6, 072310
This large M7.6 earthquake in the Philippines on July 23, 2010 was atypical in that its epicenter was 600km below the earth's surface. Little damage was reported. Note also that the normally strong surface waves LQ and LR are not prominent.

Mindanao Mag 7.6 Quake Audio recording at 250x speed
This is an audio file of the Mindanao M7.6 quake with a superimposed image of the 22 second waveform made by playing the quake datafile at a sample rate of 3072 Hz, (about 250 times the actual sample rate). It sounds a bit like rolling thunder.

Ecuador M6.9,
08/12/10 Ecuador M6.9 quake, Aug 12, 2010, 0454 PDT, distance 3360 miles. This quake occurred in a sparsely populated area east of the Andes, facing the Amazon tropical forest.

P = initial body wave. S = transverse shear wave. sPKiKP = the same shear wave reflected off of the earth's inner core. LQ and LR are the surface waves.

An audio recording of this event, compressing 48 minutes into 12 seconds, is here.

Honshu_M8.9_031111 4 hour plot
This is the M 9.0 quake 90 mi NE of Sendai, Japan, 03/11/11. The pressure wave arrived just after 22:58:30 MST here 03/10/11. I happened to be at the computer at the time.

When it started arriving, the pressure wave was strong enough that I thought it was in one of the more local hot spots in the Gulf of California. By the time the shear waves started arriving, a check with the USGS KML overlay for Google Earth showed a M7.9 off the Honshu coast near Sendai. The instrument had picked up a M7.2 from that spot two nights earlier.

I made two postings of the plots as they were building, on Facebook. An hour after onset, the USGS amended their initial M7.9 estimate to M8.8, then to M8.9. That full point in magnitude is a 32-times increase in energy.
On 03/14/11, USGS refined its measurement to M9.0.

Tempe arrival times and a more detailed plot of this quake are shown below.

Honshu M8.9 seismic travel time to
Tempe, AZ
USGS seismic wave travel times to Tempe, AZ. The initial arrival time was accurate to within a few seconds.

One notable detail in this printout: predicted peak amplitude for the surface waves was over 1/2 centimeter. That's ground movement in Tempe from a quake 5900 miles distant. I have good reason to believe that we did indeed see that motion here.

GPS measurements showed that the land mass of Honshu Island (the main island of Japan) shifted by 8 feet after the quake.

Honshu M8.9 Quake detail
A closer look at the Honshu M9.0 quake of 03/11/11. The P group is the pressure wave. S is the slower traveling shear wave and LQ is the guided surface wave that dominates the response in powerful, distant quakes. Because the surface wave is constrained between layers of the Earth's crust, it weakens more gradually with distance.

Besides the large magnitude of this record there is one other interesting detail, highlighted by the green arrow. The blue-colored trace of the preamplifier response shows activity at the onset of surface wave arrival. The preamp is at 200-400 times lower gain than the output. I normally see a response in the blue trace only when I park my car on the same concrete slab as the seismometer - the car changes the tilt of the slab.

This makes a 1/2 centimeter peak ground motion plausible. Since it occured over a 20 second period, it normally would not have been felt.

QIP thread 1

Since it's rare to witness a piece of (geological) history as it's happening, I've posted two screen captures of Facebook threads I sent out as the seismic energy was arriving from the M 9.0 Honshu quake.

The pressure wave arrived at 10:58:35 PM MST in Tempe on 03/10/11.

The timestamps on the threads are one hour early because Facebook erroneously applied Daylight Savings Time retroactively, to a date before DST was in effect, to Arizona which doesn't observe DST.

QIP thread 2

I posted the second thread after it became apparent this was a very large event and I had a more descriptive plot of the signal arriving. Having direct access to a seismometer meant I could poll USGS as soon as I suspected amendments to their measurements, which is common as quakes are occurring. I could post news before news organizations had it.

I have friends who have friends/family in the Pacific, so I wanted to get tsunami info out on the thread for this large offshore quake. The US suffered one fatality on the Calif. coastline, and significant damage to small boats in a few harbors.

To put historical context on the human cost of this quake, the single best article I've read comes from a historian who lived in that area of Japan.

Notes:

1. Another example of a torsion balance seismometer may be found here .

2. A torsion pendulum is a special case of a compound pendulum. The two main handles on the resonant frequency are the moment of inertia of the weight, and the torsion constant of the wire (inversely proportional to the wire cross-sectional area).

3. A schematic diagram for this design is here. The seismometer is completely powered off of the 5V USB bus from the Labjack U12 12-bit data acquisition board. Supply current consumption is under 100 mA. Note: U3A should be an OP284 and not an MC33202, since low offset is needed in the feedback path. U1 and U3 have been changed to OP284, which has greatly reduced 1/f noise in the seismic frequency range down to 0.01 Hz. The IR photodiodes are Vishay BPV23NF, which are inexpensive, sensitive, and have a helpful daylight blocking filter integrated in the package.

Back to homepage

	The completed seismometer board. To the right you can see the IR emitter, magnet assembly and differential photodetector. From right to left, in order: 1) Photodiode transimpedance amplifier/filter (dc-coupled, abbreviated "TZA" in the dataplots) 2) Instrumentation amplifier and HPF 3) Sallen-Key LPF 4) SE-to-balanced output buffer to DAQ board. The response characteristic is bandpass, with passband of 0.008 to 3 Hz.
	Seismometer in its enclosure. The board is mounted on top of a 10 inch circular aluminum base plate, about 1/2 inch thick. Mount point screws are tapped and threaded, providing a means for adjusting tilt. The pendulum wire, 5-mil dia tungsten, is suspended from the top of a rigid square aluminum pipe one meter long. Height adjustment is provided by a threaded set screw with a knob attached.
	Seismometer enclosure. The DAQ board, USB-IP converter and WiFi access node are on top of the cabinet to the right. The enclosure proper is a 10 inch spiral-wound cardboard cast form for concrete, 4 feet tall. Cost: about $12 at Lowe's hardware. It is thermally isolated on both ends with foam rubber. The top opening is covered by a sheet of foam with plywood on top.
	Detail of the torsion balance, showing how the copper weights are suspended between the rare earth magnets. With a meter-long pendulum, the motion is slower and the magnets provide insufficient damping alone. The small nut at the bottom of the tungsten wire in the photo, is now suspended in a vial of mineral oil for extra damping. A long pendulum is sensitive, but the mechanical design has some flaws: - It has multiple modes of response: lateral motion of the weight as well as twisting motion. In a long pendulum the lateral motion dominates, so it's really not a tosrion balance. - Sensitivity is affected by the angle and height of the flag at the point where the input integrator is centered. This makes calibration impractical. - The pennies and the end nut on the pendulum make for two separate resonances, which can be seen in FFT plots when damping is removed. Fewer natural resonances are better. - A meter-long pendulum combined with electronics which can detect microns of motion, is hard to stabilize mechanically. There are several joints between aluminum and steel, which have different thermal coefficients of expansion. One or more of these, I suspect, creates a low-level thermal relaxation oscillation at 0.02 - 0.04 Hz which can be seen in the raw data. It tends to appear whenever there's a thermal gradient dT/dt in the garage. It's suppressed by numerical filtering, but better to suppress it at the source.
	This is the first large quake detected by the V2 seismometer. The M7.7 quake in Sumatra on 4/6/10, distance 9400 mi. made a response peak of 60% of full-scale. The distance record for V1.0 was the M8.0 quake in American Samoa in Sept 2009, a much weaker response at 5000 mi. distance. The surface wave frequency is about 0.07 Hz - a very low frequency typical for the large quakes. Prior to this quake, well over 50 aftershocks have been recorded to date from the M7.2 Baja California quake of 04/04/10, some as low as M2.0.
	USGS seismic wave travel times for the M7.7 quake, 4/6/10.
	This M6.2 quake in southern Spain on 4/11/2010 is at the limit of sensitivity for detection. The pressure and shear wave components labeled P and S can only be seen in this zoom plot, superimposed against the instrument's own twice-per-minute mechanical thermal oscillation - a flaw which I still must correct. This is a milestone for me. This device was born because my wife Pam was in Italy when the M6.7 quake struck in L'Aquila in April 2009. My goal was to be able to detect that quake from that distance, and it appears now to have been achieved. Now on to curing the thermal/mechanical oscillation - (which wont be easy if the required thermal isolation for the enclosure is too high).
	USGS seismic wave travel times for the M6.2 Andalusia quake, 4/12/10 (UTC). The times for P and S agree with the recorded arrivals to within seconds. Nomenclature for the phases of seismic waves is given here. The P waves are pressure (longitudinal) waves which travel in a linear path beneath the Earth's crust. The S waves are shear transverse waves which propagate at a slower speed than the P wave. LQ and LR are surface waves which propagate along the curvature of the Earth's crust - a longer path.
	The M6.9 at Papua New Guinea July 18, 2010, was one of three closely spaced quakes, and had the best definition from the garage detector. There are 4 identifiable phases: P is the initial pressure wave arrival, PKiKP is the same wave bouncing from the Earth's inner core boundary, S is the shear wave, and LQ is the surface wave. The surface waves are always lower in frequency, and typically the strongest signals from a distant quake. Distance 6920 miles.
	This large M7.6 earthquake in the Philippines on July 23, 2010 was atypical in that its epicenter was 600km below the earth's surface. Little damage was reported. Note also that the normally strong surface waves LQ and LR are not prominent.
Mindanao Mag 7.6 Quake Audio recording at 250x speed	This is an audio file of the Mindanao M7.6 quake with a superimposed image of the 22 second waveform made by playing the quake datafile at a sample rate of 3072 Hz, (about 250 times the actual sample rate). It sounds a bit like rolling thunder.
	Ecuador M6.9 quake, Aug 12, 2010, 0454 PDT, distance 3360 miles. This quake occurred in a sparsely populated area east of the Andes, facing the Amazon tropical forest. P = initial body wave. S = transverse shear wave. sPKiKP = the same shear wave reflected off of the earth's inner core. LQ and LR are the surface waves. An audio recording of this event, compressing 48 minutes into 12 seconds, is here.
	This is the M 9.0 quake 90 mi NE of Sendai, Japan, 03/11/11. The pressure wave arrived just after 22:58:30 MST here 03/10/11. I happened to be at the computer at the time. When it started arriving, the pressure wave was strong enough that I thought it was in one of the more local hot spots in the Gulf of California. By the time the shear waves started arriving, a check with the USGS KML overlay for Google Earth showed a M7.9 off the Honshu coast near Sendai. The instrument had picked up a M7.2 from that spot two nights earlier. I made two postings of the plots as they were building, on Facebook. An hour after onset, the USGS amended their initial M7.9 estimate to M8.8, then to M8.9. That full point in magnitude is a 32-times increase in energy. On 03/14/11, USGS refined its measurement to M9.0. Tempe arrival times and a more detailed plot of this quake are shown below.
	USGS seismic wave travel times to Tempe, AZ. The initial arrival time was accurate to within a few seconds. One notable detail in this printout: predicted peak amplitude for the surface waves was over 1/2 centimeter. That's ground movement in Tempe from a quake 5900 miles distant. I have good reason to believe that we did indeed see that motion here. GPS measurements showed that the land mass of Honshu Island (the main island of Japan) shifted by 8 feet after the quake.
	A closer look at the Honshu M9.0 quake of 03/11/11. The P group is the pressure wave. S is the slower traveling shear wave and LQ is the guided surface wave that dominates the response in powerful, distant quakes. Because the surface wave is constrained between layers of the Earth's crust, it weakens more gradually with distance. Besides the large magnitude of this record there is one other interesting detail, highlighted by the green arrow. The blue-colored trace of the preamplifier response shows activity at the onset of surface wave arrival. The preamp is at 200-400 times lower gain than the output. I normally see a response in the blue trace only when I park my car on the same concrete slab as the seismometer - the car changes the tilt of the slab. This makes a 1/2 centimeter peak ground motion plausible. Since it occured over a 20 second period, it normally would not have been felt.
	Since it's rare to witness a piece of (geological) history as it's happening, I've posted two screen captures of Facebook threads I sent out as the seismic energy was arriving from the M 9.0 Honshu quake. The pressure wave arrived at 10:58:35 PM MST in Tempe on 03/10/11. The timestamps on the threads are one hour early because Facebook erroneously applied Daylight Savings Time retroactively, to a date before DST was in effect, to Arizona which doesn't observe DST.
	I posted the second thread after it became apparent this was a very large event and I had a more descriptive plot of the signal arriving. Having direct access to a seismometer meant I could poll USGS as soon as I suspected amendments to their measurements, which is common as quakes are occurring. I could post news before news organizations had it. I have friends who have friends/family in the Pacific, so I wanted to get tsunami info out on the thread for this large offshore quake. The US suffered one fatality on the Calif. coastline, and significant damage to small boats in a few harbors. To put historical context on the human cost of this quake, the single best article I've read comes from a historian who lived in that area of Japan.