Well I finally got around to doing something I have always wanted to do. That was to test the frequency response of the SG (Shackleford-Gundersen) seismometer. I wanted to check if the integrator, used to make the short period sensor look like a long period one, compensates correctly for the natural decrease in response to lower frequencies of a 10 inch pendulum.
To do this test I needed to make a shake table. The heart of my shake table is a 25000 step per rev. stepping motor. By using a stepping motor I could control the speed by simple putting in a square wave signal at 25000 times the frequency I wanted. Another words, if I put in a 25000 hz signal into the stepping motor's control box I would get 1 rev per second. On the motor's shaft I super-glued a washer that was offset a little to act as a cam.
The sensor was placed on two rods that acted as rollers. A lever was made to push against the cam on the motor, and the sensor on the other end, with two springs that where connected to the sensor and base of the shake table. When the motor rotated it would move the sensor in a sine-wave motion.
For this test I made a light framed SG sensor with a 10 inch pendulum. I was not interested in sensitivity, only frequency response, so I never actually measured the distance the sensor was moving. From eye balling it I would say I was moving the sensor 1 or 2 mm. Despite this small movement I needed to reduce the gain of the electronics by a factor of 20 so it would not saturate the electronics.
After some playing around with the table I was able to get a pretty good looking sine-wave out of it. The first freq. response run was with a 130 second (4.4 uf cap & 4.7 meg feedback resistor) integrator that I have been using with my two SG sensors I have online. The original SG sensor article, in the September 1975 Scientific American Magazine, used a 62 second integrator. I'm not sure why I was running mine at the longer time constant. It was quite obvious that the 130 second integration was way too much.
So I started to deduce the integration until I got a freq. response curve that was relatively flat (+- 3 db) between 50 seconds and around .5 seconds. Above .5 seconds I ran into another problem that was related to the damping of the pendulum, but more about that later. To get the flat response out of the sensor I had to reduce the integration down to 6.2 seconds (1 uf cap & 1 meg resistor). This is 1/10 the amount that the original SG article uses.
I have now changed the integration for all of my SG sensors to the 6.2 second value. I am now waiting for a teleseismic event so I can test to see how this change effects the seismograms compared to my Lehman and the Berkeley broadband sensor I like to use for comparison.
This table also allowed me to do another test I have always wanted to do. That was to test the freq. response of a sensor and change the damping to see how it effects the response curve. As I said above, my SG testing showed an increase in sensitivity above .5 and peeking at the natural freq. of the pendulum. Before the testing I thought I had set the damping to .700. I was surprised to see so much increase in sensitivity. So, to get a flatter response curve I set the freq. of the shake table so it was moving at the period of the pendulum and then I started to add damping until I reduced the sensitivity of the sensor to the point where I was getting the same amount of signal that I got below .5 hz. After making the damping change I was able to get the sensor to have a flat (~3 db) response from ~50 seconds to ~3 hz! But to get the this flat response I really need to increase the damping by quite a lot.
Redwood City, PSN
Larry Cochrane - www.seismicnet.com/contact.html