"AM1" is a home built force balanced seismometer, that detect
vertical acceleration. It is sensitive at frequencies up to one
Hertz. It could detect fifteen major earthquakes over a contimuous
operation period of one year.
Three years ago now, I began to set up an amateur seismographic station. After a two years preparation period, I finally captured some seismic data in January 2021. During the twelve month period that followed, I could detect more than a dozen major earthquakes coming from everywhere on earth. This was an interesting personal development, since the instrument is home built using modest resources, it is located in a semi urban location, and although I am experienced in a related field, I had no training in seismology or earth sciences. My closest connection to earthquakes was during precision laser spectroscopy experiments, I was trying to eliminate vibrations by using high performance anti vibration tables.
However, although design flaws are inevitable in any prototype, this one at least works. We actually met our initial objective which was to detect and identify at least one earthquake. This was not obvious, since we are located is a quiet seismic zone. For that reason, we focused our interest on major, far away earthquakes, who are always listed in public domain databases.
The other field of interest is the study of local microseisms, but we thought those are not as frequent, we had no access to a reference instrument or data, for the purpose of identification, and it would be more difficult to build a high speed digitizer.
For those reasons, the decision was made rightly or wrongly, to build a long period vertical force balanced instrument. We would try to identify one major earthquake over the period of one year by comparison with public data.
Our instrument would be located in one of two possible sites, in the basement or in the garage, and it would detect vertical acceleration with frequency up to one hertz.
Also the instrument should be of the force-balance type with electronic feedback.
It was effectively built over a period of twenty two months the
following describes its building.
Although I have some knowledge precison instrumenation, I am not a seismoligist. I had never worked with this type of equipment before. It was then necessary to document myself in order to know the most importan aspect of seismology signals. For that reason I have studied the description of many comercial and amateur instruments. Some references will be given in the text.
Another relevant factor is the station location. The only possible site is the basemenmt of the family house which is located in a suburban area. The house is located about twenty meters from the street.
Having no experience, it was not possible to know in advance if this would be adequate for the location of a seismometry station. Would the street trafic noise hide the microseism signal. Also, were were unsure about the nature of the soil under the house.
Finally the north east of america is know to be a relatively stable area where medium or large scale earthquakes do not occur often. Most people living in the region recall the earthquake about thirty years ago (1989) although it caused no damage. Based on this past experience, a "toy" type instrument would yeld some signal sometimes during the next twenty years. We would certainly wish for a more dynamic hobby.
So all of the above suggest that our best strategy would be to design a very sensitive instrument that has a low frequency response in order to detect some very distant, high magnitude earthquakes. The low frequency signal from a large earthquake would hopefully be diferent from the high frequency trafic noise and and could be more frequent.
Finally, as we all know the output signal from measuring instruments is mostly random and the hopefull metrlogist can always imagine the most unsuspected features. Wanting to avoid such fall, we would need to compare our seismograms to known results. Having no access to a reference instrument, we would use the published earthquakes from the official research organisations. These plots are often published on the internet.
There it is: We build the low frequency sensitive instrument and search on the internet for correlation with large seisms from caribeans. The goal is to detect at least ONE earthquake every year. It should be feasble.
I did in fact connect a geophone to a "toy" type amplifier! It quickly appeared that this approach is not valid. During the period I was developing this system, I saw on the internet a note from an amateur, warning us against the use of industrial vibration sensors, such as MEMS devices, as not sensitive enough. Also I became aware that the west US coast is by far more active than the north-east, so I droped the "geophone" approach. Time to "be real".
All the documentation consulted is concerned with the effect of pressure fluctuations. For obvious reasons we wanted to have a barometric enclosure. Even a vacuum enclosure would be preferable, if possible. However this choice has huge consequences. If the vessel is evacuated, then the electronic system must be built OUTSIDE the evacuated enclosure. This would imply the use of a multi-pins vacuum connector, in addition to a second enclosure since the electronics is sensitive to environment changes too and should be shielded as well (but not evacuated).
This choice is not so easy to justify. Temperature controlers are of two types. The simple ones use a heater that maintain the temperature at a level such that it will never be below the environment temperature. This is easier to say than to do. It is not easy to predict how high the temmprerature will rise.
Moreover, when there are temperature gradients in a system, these come with a set of design questions. Most of these questions can be avoided by using a pelletier cooler. Again, we chose not to use this device.
Our objective is to capture long distance signal with a period of say ten seconds. Although the temperature could fluctuate over a range of ten degrees or more, the short term fluctuation should be minimal.
However, we will monitor the temperature and study this effect. One temperature sensor is attached to the frame of the seismometer, inside the barometric enclosure. A readings is taken at every hour.
Our goal is to detect a distant large magnitude earthquake. Professonial seismologists study the intricate details of earth's crust. They can make use of the polarization information.
Traditionnaly, horizontal seismometers used the "garden gate" principle which can easily be tuned for detection of very low frequency. However it is extremely sensitive to tilt. With no more justification, we opted for the vertical type.
TAn alternating electric field is generated between two plates while a third one detects the potential that is proportional to its position. This requires a small circuit that periodically reverses the polarity.
This is similar to a standard electromagnet with a few exceptions. There is NO soft iron core. The core is a permanent magnet. It is attached to the seismometer frame and it creates a magnetic field that is PERMANENT and can NOT be modified.
The coil is attached to the beam and moves with it. The force on the beam is proportional to the current. There is NO iron on the beam or near the coil as the current in the coil would magnetize this iron resulting in a dramatic effect. Electromagnets are hundreds of times stronger than just wire coils, but are extremely non linear due to soft iron.
This is a practical choice. One can buy for a small cost an analog-to-digital converter chip already mounted to a small printed circuit board. These converters have programable sampling time and resolution. All that inside a single chip and available for a few dolars. Although the bit packet sent is up to 24 bits wide (even 32 in one case) the true resolution is much less of course. Realistically we probably have fifteen or more meaningfull bits at a sample rate of two samples per second. As explained earlier, this is adequate and a higher sampling rate would require expensive hardware, faster computers, bigger hard drives, and the surplus of data would be useless for now.
Now this choice is very easy. Magnetic shields are made of a special Nickel alloy and are not available. It is however important to protect the beam from fluctuations of the static magnetic field. For that we have used only non-feromagnetic materials, with a few exceptions. Those materials include brass, copper, aluminium, plastic, varnish, nylon.
The beam position is very sensitve and it drifts over time. As explained above, the electromagnetic system to restore the position is not very strong. When is goes out of range is necessary to readjust the beam to level position. A good feature is to have this adjustment done from remote location, without having to move or open the instrument.
Our instruement has a suspended weight that can be moved using a
small motor in order to balance the beam. This motor is made of
feromagnetic materials but it is attached to the seismometer
frame, ie. not touching the beam in normal operation. Only during
the mass reset operation, will the motor push smoothly on the
suspended mass then retract itself. The entire operation is
remotely controled from the network computers (not automatic).
The figure shows the actual arrangement of the various elements.
The frame (1) is made of four one inch square sections of aluminum extrusion, lined on the inside with thick steel plates. There is a set of eight L shaped plates that connect the sections together with a total of thirty two 10-32 screws. This forms a rigid rectangular frame, whose dimensions are five inches wide by eight inches long, on which everything is attached.
Before the two long sections were attached, they were lined inside by two thick steel plates in each section allowing to drill and tap various mounting holes to attach more structures; one is the pivot of the boom, the other one is a tall arch to which is attached the spring. Those two are mounted on the inside for the frame.
On the outside of the frame, the liners hold two removable brackets (not shown on the drawing). They are made of one inch right angle sections and can be removed and drilled with different mounting holes if needed. The mounting brackets are fitted with alignment pins to the frame. The purpose of all of this is to allow one to remove, readjust the position, remount the sensor inside the hermetic enclosure, even drill new mounting holes without taking apart the sensor.
The boom (3) and (10) is built around a two inch wide aluminum plate floating slightly above the frame and it has a "H" shape. On the pivoting end, the two hinges are placed far apart for a better stability in the horizontal plane.
At the far end, four tightly placed components perform the vertical acceleration measurement. Similarly they are thin and to optimize space and boom length.
The mass (8) is a long section of aluminum tube filled with 900 grams of lead.
Next to the mass is the electric force transducer. The elongated coil (dashed line pointed by (2) and (9)) attached to the boom, is pushed by the field produced by a permanent magnet (7), which is attached to the frame (attachment not shown). The force transducer's combined action with the position detector effectively act as a sink that absorbs the mechanical energy from the earthquake to measure.
The position detector is located far from the pivot, not in the plane where the other elements are located. But this would not be of any consequence since this element exert no force.
Two metal plates (one of them is pointed by (5)) are attached to the frame (attachment not shown). A third plate (4), attached to the boom, is floating between the other two. This detector is called "capacitive position detector". It is non-contact and force-free device, very similar in concept to "light detector".
The fourth element (11) is a bracket mounted on the bottom of the boom on which the suspension spring (6) is attached. Its purpose is to lower the angle between the spring and the horizontal plane. This reduces the natural oscillation frequency of the boom. Please note that the spring attachment goes thru two holes across the beam without touching it. Also, the black square with the thru holes is part of the holder for the position detector third plate (4) whose representation has been simplified.
Not shown on this drawing is mass positioner and the moveable mass. The mass is a short section of square tube filed with lead, attached to the bottom of the boom between the frame sections. It weight less than two hundred grams and it can be moved over a distance of fifteen millimeters. The slide mechanism on the boom has two small brass springs to force a good contact to the side of the boom.
The motor, the positioner and two limit switches are located above the boom behind the spring (not shown). As mentioned earlier, the positioner retracts to a non-contact position.
The figure shows the feedback loop configuration of the instrument.
The figure shows the Computer interface configuration.
The figure shows the computer command bus.
The figure shows the loop controller.
Three large plates closely spaced form the position detector. The two external plates are rigidly attached to the chassis. The center plate is part of the boom-mass-actuator assembly and moves relative to the other two. All the three plates are electrically insulated. Some AC voltage is applied to the outside electrodes and depending upon its position, an in-phase or out-of-phase potential will appear at the center electrode.
This can be modeled as a differential capacitor, or as one leg of a differential bridge. As the center electrode aproaches the symmetry position then the detected signal goes to zero. From there the signal grows in-phase or out-of-phase as the electrode deviates from the center position. Such null-detector allows for very high amplification gain, thus extreme sensitivity.
A good amplifier must have: i) high gain, ii) high input impedance, iii) works at low input signal level, v) is AC coupled.
We use a (electret) microphone amplifier (2SK596), mounted directly on the detector plate. a single wire (plus ground) can supply (phantom) power and send the amplified signal to the second amplification stage.
The LT1013 is a low noise bipolar input operational amplifier. The LM4871 is an audio power amplifier. Pin 8 goes to the transformer located on the mixer mixer board. The board only uses 5 Volt power supply.
The output of a crystal oscillator is sent to a series of flip-flops before it is low-pass filtered and amplified. The pure sinusiodal signal is further increased by sending it to a audio transformer wich has two balanced output windings for both the reference signal and the detector excitation. Finally a 18 volt p-p signal is applied to the detector plates.
It is difficult to obtain a distortion-less sinuosidal signal from a square wave. The MAX295 is a 8th-order, lowpass, switched-capacitor filter. To further improve the signal quality two additional analog low-pass stages are added to the input and the output buffer amplifier is also filtered. Please note the signal output from a flip-flop divider is always symmetrical.
The oscilator/divider also provides a higer frequency (307.2 KHz) signal for the SC filter clock.
The choice of 4.8 Khz as the excitation frequency was not so easy to set, as it is the the result of matching compatibility to all the chips in the system! Too low frequency result in sensitivity loss of the capacitive detector, while a higher frequency brings more sensitivity from signals phase. Most of the chips we use are frequency limited audio devices.
The entire board is powered from a unique 5 VDC supply.
This double-balanced mixer functions as a phase detector. The signal from the preamplifier is multiplied by the the reference one from the oscillator/exciter. Then the output is filtered to keep the DC output and voila! We have the mass position output.
Unfortunately this board requires four supply voltages.
The artificial ground at 2.5 V becomes the reference zero mass position.
The force element is a electromagnet coil with 17 Ohm DC resistance. This power amplifier needs to be temperature stable. The separate buffer amplifier (INA2134) provides stabilized feedback and monitoring
Please note that the "no force" reference voltage is set to 2.5 Volt.
There are 8 analog inputs to the ADC converter chip. Each can be connected of different input or to same input through a different anti-aliasing (asmpling) filter.
Input channel IN0 is the reference ZERO level at 2.5 Volt. IN1 to IN4 are connnected to the integrator output and IN1 is the current filter used at 2.5 samples per second sampling rate. The capacitors are low leakeage type.