Many years ago, after reading an Amateur Scientist article (July 1957) on swinging gate seismometers, I built a sensor based in part on the information in the article. Its design, of course, was greatly affected by what raw material I had around, and what could be obtained from friends and the local hardware store. You can access a description of it from my page on John Lahr's web site,
Originally, it was fitted out with photo-optical sensing of pendulum position. The image of a pilot lamp filament was projected through a short focal length lens on the pendulum to a pair of photocells, hooked up parallel in voltage opposition. The differential voltage output was fed to a homebuilt potentiometric recorder. It was plagued with drift problems. It never took long for the filament image to end up totally on one of the photocells. I gave up my efforts and went on to other interests.
In early 2001, my interest was revived by a former colleague and long-time amateur seismologist, Victor Aiello, who wanted me to participate in a tripartite local network he was wanting to set up. He loaned me an old vertical short period Strengnether to operate. This got me going very heavily into programming and signal processing, using Visual Basic. We also had to solve the problem of time synchronization of recordings made at separated sites, and how to set up the sensors to have the same period and damping, so that response differences would not distort our time-of-arrival estimates.
After this experience, I was inspired to take my abandoned old sensor and fit it out with a pickup coil and magnet to convert it to velocity sensing. The pendulum is short, about 8 inches to the radius of gyration, with a total mass of 100 grams. I had no intention of getting a natural period of more than 12 seconds, because in my tripartite work, I designed a digital filter that could extend the useful range of the sensor to as much as 50 seconds, and which furthermore matched the amplitude and phase response of such a long period sensor. It is presently set for a natural period of 13 seconds, and holds its centering without any attention. The data obtained from it after broadband digital filtering to emulate a 32 second period sensor, matches well what I see from the nearby LDEO PAL site, 19 miles from me, except for my higher local noise level.
My seismo buddy Victor and I were now motivated to build some more sensors of this lightweight design. Pivots are critical. The old sensor uses steel needles and sapphire bearing cups. I do not know where I could obtain more sapphire bearings, so I considered the use of mini ball pivots. From experiments with diagmagnetic supported upright pendulums, I have discovered (following a suggestion from Chris Chapman) that the ball point nib from a BIC pen makes a very good pivot. The ball is highly polished, about 1 mm diameter. You just insert the whole little brass nib into a hole in the boom, epoxy it in place, and you have your pivot. It must rest against a hard flat polished surface. Glass works, but a sapphire plate would be better. I favor a rigid strut (no wire) for mounting the upper pivot. If the upper pivot plate normal points to the center of gravity of the pendulum, there will be no lateral forces at balance on the pivots.
Actually, my second horizontal sensor is almost an exact copy of the first. For pivots, it uses sapphire cups and needle pivots salvaged by Victor from an ammeter.
My 5-second period vertical sensor was built for me by Victor, and the design is similar to the old vertical short period Strengnether. In April 2004, I had finally realized my goal of operating my own complete three-component seismic station, which, thanks to my period-extending filter, has broadband response.
I encourage others to build lightweight sensors, large or small, something like John Cole's T Max, for example. After all, we are no longer driving galvanometers, just high impedance input dc amplifiers. The only justification for weight that I can see is greater resistance to air currents. You will also find, if you use my high output magnet and coil design, that you will not need any damping other than that obtained by a shunt resistance across the coil. The resistance required will be much higher than the coil resistance, so that there will be no loss of signal.
I am willing to provide more information on how to wind the 1100-turn coil, and how to filter your data to achieve long period results. I can write post-filter programs for you for enhancing any data file, but I have to know the file format.