A case study of seismograph self-noise test from Trillium 120QA seismometer and Reftek 130 data logger
- 81 Downloads
Seismograph self-noise has become a de facto standard for instrument comparisons and their performance assessment and is considered as one of the most vital parameters for instrument comparison. For self-noise testing of modern force-balance feedback broadband seismometers, several factors have been thoroughly discussed and thought to be attributable to the self-noise estimate, including the data selection criteria, sensor alignment correction, timing error, correlation analysis method, and computational parameter selection during the computational process. This study focuses on some other factors, such as local site conditions, temperature insulating methods, and data logger self-noise interferences, with an aim to differentiate the self-noise contribution of these sources and their dependencies on time and frequency. A series of experiments were conducted at the Beijing National Earth Observatory using a Trillium 120QA seismometer and Reftek-130 data acquisition system at three different locations ranging from the ordinary equipment warehouse to global seismographic network level cave with a hard-rock base. Results show that noise-free site is necessary for the self-noise test in a frequency band greater than approximately 0.1 Hz. However, for a frequency band less than 0.1 Hz, the insulation method and installation procedures are far more important, although the influence of the site location cannot be neglected fully. A suitable preamp should be selected in the data logger configurations to ensure that the low-noise amplitude of the sensor signal is above the digitizer noise level.
KeywordsSeismograph self-noise Seismometer testing Broadband seismometer Seismic data acquisition system Seismometer insulation method
Seismograph self-noise defines the lower limit of the seismic noise detection in broadband seismic observations and plays an important role in seismic instrument development and seismic noise analysis. For seismometers with comparatively high self-noise, such as strong-motion accelerometers or lower-grade sensors, it is possible to obtain accurate self-noise estimates through noise power spectra estimation by testing a single senor at a relatively noise-free location when the sensor’s self-noise is well above the site noise (Ringer et al. 2015a). For high-quality low-noise broadband sensors, no locations exist where background noise levels are below that of the sensor across a wide frequency range; therefore, the two-sensor method was developed to isolate the self-noise estimates from synchronously recorded data of two collocated sensors, assuming that the internal noises between each channel pair and internal noise and common input signal are uncorrelated (Holcomb 1989; Holcomb 1990). By introducing a third collocated sensor, the three-sensor method can estimate sensor self-noise while minimizing errors during the estimation owing to the uncertainty in the transfer functions (Sleeman et al. 2006).
Practically, several factors can influence the test results, and even for the same sensor model, different test results can be obtained if these factors are not considered well (Ringler and Hutt 2010; Yin et al. 2013; Xu et al. 2017). Different coherence analytical techniques, including the two- and three-sensor methods, have been investigated well. The three-sensor method has become the preferred approach for estimating self-noise of broadband sensors owing to its relative stability and robustness in a broad range of frequencies (Ringler et al. 2015b). During the mathematical computation, the Welch method is typically adopted to evaluate the signal power spectral density (PSD), wherein specific parameters, such as sample size, sample rate, bandwidth, and windowing, need to be decided before the computation. Some optimal parameter selections have been put forward for the seismometer self-noise testing by the Guidelines for Seismometer Testing workshops to make the test results comparable in some reasonable manner (Hutt et al. 2009; Evans et al. 2010). Similarly, the effects of these parameters’ selection have been well investigated following the statistical examination of 9800 different parameter combinations by some scholars; a zone of reasonable self-noise calculation parameter combinations was identified (Li et al. 2015).
Several studies (Holcomb 1990; Sleeman and Melichar 2012; Tasič and Runovc 2013; Gerner and Bokelmann 2013; Ringler et al. 2015a, 2015b; Gerner et al. 2017) have shown that sensor misalignment is an important source of error during self-noise testing of seismometers based on collocation methods. A synthetic test was conducted to quantify the effect of sensor misalignment as a function of signal-to-noise ratio (SNR) on the self-noise estimate. Results showed that for the higher SNR, the effect of the tiny misalignment might be clarified. This implied that these types of measurements should be performed at seismically quiet locations or the misalignment error must be well corrected before the correlation analysis to avoid the misinterpretation of test result. Considering that an ideal seismically quiet location can hardly be found and misalignment error can hardly be guaranteed to be less than 1° (Ekström and Busby 2008; Ringler et al. 2013), the correction based on the trace rotation is a good option and different rotational schemes have been developed to correct the effect of sensor misalignment (Tasič and Runovc 2013; Gerner and Bokelmann 2013; Gerner et al. 2017). The timing errors between sensor records contribute to the incoherent-noise estimates and are related to the SNR of the records (Ringer et al., 2011); however, this effect is increasingly negligible because the timing clock inside modern data acquisition systems can be well synchronized once the inside global navigation satellite system (GNSS) module is normally operated.
This study aims to evaluate the performance of seismometers used in the ChinArray project, better understand the operating range of the broadband seismometers, and setup a reasonable testing environment for the instruments. Herein, we mainly focused on the effects of seismometer self-noise testing from local site conditions, temperature insulating methods, and data logger self-noise interference with an aim to differentiate the self-noise contribution of these sources and their dependencies on time and frequency. Then, a reasonable procedure and test environment for a broadband seismometer self-noise test was proposed.
2 Experimental setup
In broadband seismometer self-noise testing, the noise estimate of vertical component is typically adopted as the final result since the horizontal output often shows incoherent elevated noise levels because of the local changes in wind or pressure, which are extremely difficult to avoid in most cases. However, for a Galperin-type seismometer (Galperin 1955), since the three sensing elements are arranged in a symmetric triaxial manner, each traditional horizontal or vertical output is computed from the combination of three sensing elements using a rotational matrix; therefore, the vertical output can reflect the performance of all three sensing elements. This is one of the reasons we chose Trillium 120QA as our test seismometer. Another reason is that this model is well tested by manufacturer, offering a good reference to justify the test environment and methodology. We used the Reftek 130 data logger as the test digitizer because this model is among the most popular digitizers used in portable seismic observation in China; however, the Q330HR digitizer is mostly recommended in such experiments owing to its high resolution.
The naming of the three experimental sites is simply based on the chronological order, i.e., the tests were conducted first at site 1. However, the test result was not significantly consistent with the one provided by the manufacturer; hence, experiments at sites 2 and 3 were conducted. In the cave chamber where the site 1 experiment was completed, a granite pier built several decades ago with a ~2 m2 surface and ~70 cm in height holds a glass tank that is tightly stuck to the pier surface. The three seismometers were placed in this glass tank and a glass plate was then placed as a cover to insulate the sensors from the outside. Some small holes occur at the top of the tank wall for cables to travel between the seismometers and data loggers and they were well sealed using butter following the placement. The upper right picture in Fig. 1 shows the case before the glass cover was placed. At site 2, three seismometers were placed on a rigid granite slab 1000 × 1000 × 15 cm in size, which was supported at three points with lead pads on the floor of the instrument warehouse. As distinguished from site 1, each seismometer was individually insulated using a special cover developed by Nanometrics Inc. Afterwards, an overall cover comprising polyethylene foam was placed over all the sensors and pier; additionally, a thick blanket was placed over the equipment and down the sides of the concrete pier. At site 3, the pier was nearly the same as that of site 1 except for a slightly smaller surface size and insulation method was nearly the same as that of site 2. The lower left picture in Fig. 1 shows the case before the overall cover was placed at site 2 and lower right one shows the case after all the insulation procedures were completed at site 3.
Different self-noise test configurations used at three sites. All tests were run for at least one week to ensure sensors had settled. The serial number of data loggers and sensors are indicated in the parenthesis of each model used
Reftek 130, High gain
(B00E, B008, AF83)
(1880, 1881, 1882)
At a cave chamber with the least cultural and environmental noise interferences of three sites.
All seismometers in a well-sealed glass tank as insulation cover on an 80 cm high granite pier
Reftek 130, Unity gain
(9B13, AF80, B044)
(1880, 1881, 1882)
At the equipment warehouse of China Seismic Array Instrument Center with the largest noise interferences of three sites.
Each seismometer individually insulated with a cover designed by Nanometrics Inc. and polyethylene foam used as the whole thermal insulation for all three seismometers on a 15 cm high granite pier
Reftek 130, Unity gain
(B031, AF80, B044)
(1880, 1881, 1882)
At a cave chamber with the intermediate noise interferences of three sites.
Thermal insulations similar with site 2.
3 Analysis and results
We employed the three-sensor method to isolate the seismometer self-noise based on the continuous record of three seismometers; the methodology details can be found in a previously published paper (Sleeman et al. 2006). Although data selection had been considered as an important item to estimate self-noise, no criteria or agreement exists. Some random popcorn-like noises owing to cable stress release or semiconductor defects, which definitely belong to instrument self-noise, are difficult to describe. Most researchers claim that only quiet time periods, such as nighttime without special seismic or other events, should be used for analysis. Some scholars have attempted to combine continuous and special data selection criteria such as defining the threshold between self-noise statistics and its mode (Sleeman and Melichar 2012). Herein, instead of only analyzing the noise-free data, we obtained self-noise estimation variations during the entire test periods. Herein, we wished to see how the local environmental conditions affected the self-noise test results using different insulation methods and how these effects varied with time and frequency.
We corrected the potential misalignment angles between seismometers based on three-dimensional (3D) rotational algorithm of raw seismic traces to maximize coherence, which is similar in principle to that of previous studies (Tasič and Runovc 2013; Gerner et al. 2017). The raw data were deconvolved with instrument response of both seismometers and data loggers. For the spectrum estimation, we adopted the parameters recommended in a previous study (Evans et al. 2010) for the Welch estimation (Welch 1967), which includes 219 and 215 sampling point duration for 200 and 1 Hz stream, respectively, and a constant overlap of 87.5% of the window length chosen. Upon combining the test results of two different sample rates, the self-noise estimates were obtained as a function of frequency from 0.0005 to 50 Hz. Finally, a 25% logarithmic smoothing scheme was applied to smoothen the final results (Ringler and Hutt 2010).
4 Discussion and conclusions
Several influential factors affect the final test results of a low-noise broadband seismometer self-noise test, including the data selection criteria, sensor alignment correction, correlation analysis method, and computation parameter selection during the computation process. A seismically “quiet” site is necessary for the frequency band greater than approximately 0.1 Hz; however, for the frequency band less than 0.1 Hz, the thermal insulation method and installation procedures are more important; however, the influence of site location cannot be fully neglected. When a Reftek 130 model was used as the data logger in a seismometer self-noise test, the high gain preamp option should be adopted to ensure that the self-noise of the sensor is at a level above the digitizer noise level, thereby not interfering with the sensor’s test result. As a part of the entire observational system, data loggers should select a suitable preamp option to match the seismometers used based on different observational purposes.
Once all the aforementioned factors were fully considered, the self-noise test results of Trillium 120QA were found to be consistent with the manufacturer’s nominal curve at high- and low-frequency bands, respectively. We concluded that the elevated self-noise in the low-frequency noise at site 1 might have originated from two sources: 1) poor insulation setup that can be improved by either additional individual covers for each sensor at sites 2 and 3 or other effective insulation measures such as the whole neoprene insulation method (Sleeman and Melichar 2012) and 2) the nonnegligible contribution of the heavy insulation glass box to the underlying rock base, which made the sensor measurements uneven between each other. Test results showed that the frequency below 0.1 Hz at site 1 could be improved to a nominal value after similar thermal insulation and installation procedures were applied at sites 2 and 3, i.e., removing the current heavy glass box, leaving only the bare clean surface of the concrete pier, directly placing the sensors on the concrete pier with individual covers over them, placing an overall insulating cover to fit over this and sensors, and placing a thick blanket over everything, including the downsides of the concrete pier. However, these procedures will be implemented during our future experiment when completing such seismometer self-noise tests.
The authors would like to thank Nanometrics Inc. for useful discussions and help with the loan of their specially designed covers and an anonymous reviewer for the suggestions and comments that helped improve the manuscript. This work was mainly supported by the National Key R&D Program of China (Grant Number 2017YFC1500201) and was partially supported by the National Natural Science Foundation of China (Grant Number 41474047).
- Galperin EI (1955) Azimuthal method of seismic observations (in Russian). Gostoptechizdat 80Google Scholar
- Holcomb LG (1989) A direct method for calculating instrument noise levels in side-by-side seismometer evaluations, U.S. Geol. Surv. Open-file Rept. No. U.S. Geological Survey, Albuquerque, New Mexico, pp 89–214Google Scholar
- Holcomb LG (1990) A numerical study of some potential sources of error in side-by-side seismometer evaluations, U.S. Geol. Surv. Open-file Rept. No. U.S. Geological Survey, Albuquerque, New Mexico, pp 90–406Google Scholar
- Hutt CR, Evans JR, Followill F, Nigbor RL, Wielandt E (2009) Guidelines for standardized testing of broadband seismometers and accelerometers. US Geol Surv Open-File Rept 2009-1295:62Google Scholar
- Peterson J (1993) Observations and modeling of seismic background noise. USGS Open-File Report 93–322, 94ppsGoogle Scholar
- Sleeman R, van Wettum A, Trampert J (2006) Three-channel correlation analysis: a new technique to measure instrumental noise of digitizers and seismic sensors. Bull Seismol Soc Am 84(1):222–228Google Scholar
- Welch PD (1967) The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms, IEEE Trans Audio Electroacoust. AU-15, 2, 70–73Google Scholar
- Yin X, Chen J, Li S, Guo B (2013) The research on self-noise measurement method of moveable broadband seismometer. Seismology and Geology (in Chinese) 35(3):576–583Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.