Tilt has been the nemesis of horizontal long period seismology since its inception. Modern horizontal long period seismometers with their long natural periods are incredibly sensitive to tilt. They can sense tilts smaller than 10^-11 radians. To most readers, this is just a very very small number, so we will begin with an example, which should help to illustrate just how small 10^-11 radians is.

Suppose we have an absolutely rigid rod which is approximately 4170 kilometers long; this just happens to be the Rand McNally map scaled crow flight distance between Los Angeles and Boston. Tilting this rod 10^-11 radians corresponds to raising one end of the rod 0.0000417 meters. Alas, this is just another very very small number! However, this corresponds to slipping a little less than one third a sheet of ordinary copying paper under one end of this perfectly rigid rod. To clarify, we mean, take a sheet of paper just like the paper this report is printed on and split it a little less than one third in the thickness direction, then put it under the end of the 4170 kilometer long rod! This will tilt the rod 10^-11 radians.

Real world seismometers are nowhere near the length of this rod. A KS-54000 is about two meters long. Tilting a rod only two meters long 10~n radians corresponds to moving one end of this rod a mere 0.00000000002 meters or 0.02 millimicrons. As one of the authors old math teachers used to say, "That's PDS" (PDS = Pretty Damn Small). Unfortunately, the long period seismologist does not have the luxury of ignoring PDS numbers when it suits him as the mathematician frequently does. He must live in the real world in which tilts this small create severe contamination of long period seismic data.

At periods longer than 20 seconds, tilt noise contaminates the long period data from all instruments installed on or near the earth's surface. Many years of experimentation revealed that installing the sensors at depth in deep mines drastically reduced the level of tilt noise in long period data. However, low levels of tilt noise persisted even at great depth; this noise was caused by air convection in the vault in which the sensors were installed. Over the years, methods were developed to control the air motion with mechanical barriers (boxes) around the sensors and by stratifying (creating a situation in which the air temperature increases with height) the air in the vault near the seismometer. These methods decreased tilt noise in deep mines to very low levels. However, deep mines, that are economically and environmentally suitable and accessible to seismology, are not plentiful and are not evenly distributed over the earth's surface. Therefore, the borehole deployable Teledyne Geotech KS-36000 and later the KS-54000 sensor systems were developed to fulfill the need for instruments that could be installed at depth wherever high quality long period data was desired. Early in the development program, it became evident to the Teledyne Geotech personnel that air convection within the borehole was going to be a significant problem in KS deployments. Experimental and theoretical investigations conducted by Teledyne Geotech (see Douze and Sherwin, 1975, and Sherwin and Cook, 1976) produced a list of recommended installation procedures for reducing the effects of air convection. These procedures consisted of wrapping the sensor in a relatively thin layer of foam insulation, filling the free space volume in the vicinity of the centralizer-bail assembly with foam insulation, and the installation of styrofoam hole plugs immediately above the cable strain relief assembly at the top of the sensor package and at the top of the borehole. This technology has performed quite satisfactorily for over 20 years but evidence of tilt noise in the system output has persisted throughout the KS deployment program (the evidence was that the horizontal components were usually noisier than the vertical components) even in deep boreholes. Some deep borehole sites have been plagued by quite high levels of horizontal noise. Therefore, there has been a definite need for a new technique for controlling low level tilt noise in deep boreholes and the use of sand has been under consideration for several years.

Figure 1 contains conceptual illustrations of both the conventional holelock installed KS sensor system and the same sensor installed in sand. This figure demonstrates the major differences between the two installation methods. The curved arrows in the borehole on the left in the figure denote possible air convection cells which are believed to be the source of tilt noise in some of the conventional installations. This air motion is eliminated in a sand installation by filling most of the free air volume surrounding the seismometer with sand as shown in the right hand portion of the figure. The sand actually performs two functions; it prevents air motion and provides a remarkably ridgid clamping of the seismometer in the borehole.

This report presents the results of quantitative experimental investigations into the effectiveness of controlling low level air convection in seismic borehole installations with sand. The main body of the experimental effort consisted of installing two KS-540001 sensor systems in closely spaced shallow boreholes, allowing the sensors to reach equilibrium operation, and then pouring sand into both boreholes to observe any changes caused by pouring sand into the holes. The hypothesis of the experiment was that the sand would fill up the entire free air volume between the sensor package and the borehole walls thereby preventing movement of the air in the vicinity of the sensor package. The validity of this hypothesis had been qualitatively proven by earlier experiments at ASL and by the sand installations at the IRIS/ASL stations ANMO in 1995 and COLA in 1996. This experiment documents the degree of improved noise levels to be expected if KS instruments are installed in sand instead of in the conventional manner.