The standard technique of drum recording is to illuminate the galvanometer mirror with light from a vertical line source, and to bring an image of the source into focus on the recording surface by means of a convex lens in front of the galvanometer mirror. A rocking prism or mirror between the light source and the galvanometer displaces the image to provide timing marks or a shutter is used to interrupt the light beam.
The system as described would make a broad streak on the photographic surface when the image was displaced horizontally. A short-focus cylindrical lens is therefore placed near the drum, in the position at which it produces a horizontal line image of each point on the galvanometer mirror (Fig. 6.1).
The combined effect of the vertical and horizontal focus ing is to produce a bright recording point, very near to the axis of the cylindrical lens.
A possible refinement of the system is to place a shaped mask in front of the cylindrical lens. By making the mask aperture narrow in the centre and wider at the ends, more light is allowed to reach the photographic surface when the light beam is strongly deflected than when it is in the central position. This ensures that at least the ends of the swings are adequately exposed when the galvanometer coil undergoes large movements, which are usually the most rapid ones. A similar effect can be obtained by setting up photoelectric cells near the undisturbed position of the galvanometer beam, and arranging for the light source to be brightened when the beam swings across them. Bourgeois and Buist (1967) have suggested an elegant system whereby an auxiliary galvanometer shines through a graticule on to a photoelectric cell. When the mirror starts to swing, an alternating e.m.f. is generated, which is amplified so as to produce a signal proportional to the angular velocity of the mirror. This signal is used to brighten the recording spot.
Standards for drum-recording were laid down by the Committee for the Standardization of Seismographs and Seismograms as follows:
Records should be made on sheets having approximately a 3:1 ratio of length to width.
The output of short-period instruments should be recorded with the drum rotating at approximately 4 revolutions/hour. For long-period instruments the speeds should be 1 or 2 revolutions/hour. Deviations from these speeds may be sufficient to cause the minute marks to progress across the seismogram, but should not be enough to prevent easy recognition of the alignment of minute marks separated by exactly one hour.
Sensitivity figures should be related to a projected image of the original sheet, on a scale such that the interval between minute marks is 60 mm for short-period records and 15 mm or 30 mm for long-period records.
In all records, time should increase from left to right and from top to bottom as in a printed page*. Galvanometers should be connected so that North, East and Upward motion of the ground moves the trace towards the top of the page.
The following information should be written or printed on the front face of the sheets:
* This recommendation has not been universally adopted.
Equivalent data (such as time signals on the record) may be substituted for these items.
Drum records are frequently copied onto microfilm for storage and exchange, and the very fine lines and minimal contrast which occur on the original records, especially in the more active parts of the trace, require very high standards of copying if the signals are to be adequately recoverable from the microfilm. Thus, although film widths down to 16 mm can be used for this purpose, 10:1 reduction (which requires 35 mm unperforated film to accommodate the image of a standard drum record) was, until recently, the smallest size which could be recommended. For the highest quality, which is widely used in world-wide data exchange, the National Oceanic and Atmospheric Administration of the United States used 8:1 reduction on film chips measuring 70 × 115 mm. Recent advances in photography now permit smaller formats, and NOAA distributes data recorded since 1 Jan 1978 only on microfiche at 32:1 reduction, giving 24 frames (4 days) per fiche.
The emergence in recent years of seismometer systems which incorporate electronic amplification instead of galvanometric detection has greatly extended the range of recording systems which can be used. Although such systems usually require more power input than a sensitive galvanometer, the difference is counterbalanced by the saving in the light-source requirement for the photographic drum, and by the avoidance of the need for darkened accommodation for recording and processing. The advantages of portability and convenience therefore lie heavily in favor of the new systems, even when power consumption (as in remote field systems) is a major consideration.
A significant distinction can also be drawn between drum recorders as a whole, and recorders which make their records in the form of a continuous strip. The great advantage of drum records is that the record of any one component can be quickly scanned for evidence of activity, and a secondary advantage is that the alignment of time signals across the record enables the hours and minutes to be counted easily. Strip records must be searched from end to end, and the difficulty of counting time marks from the beginning makes it almost essential to allocate a separate channel to record time in a coded form. The compensating advantage is that many channels can be recorded side by side. This makes for ease of comparison, and enables telemetered networks to be built around a recorder with a single timing system.
These may use wire stylus writing on smoked paper, ink pen and paper, or heated stylus on treated paper. Smoked-paper drum recorders have been used since the earliest days of seismic recording. Very compact, modern systems have been constructed for which very low power consumption is an important asset (Prothero and Brunel, 1971), and similar systems are available commercially. These recorders are particularly useful for very temporary stations such as are required for noise testing and the study of aftershock sequences, as no special support facilities are required. The same comments apply generally to pen recorders but, because of the greater size of the pen, more power is required and they are rarely as economical. Older models were liable to clog on paper dust, but some of the most recent versions are much improved. Recorders using a heated stylus on heat sensitive paper are convenient and quite widely used, but cannot be considered portable because of their bulk and power consumption. The paper must be handled carefully both before and after recording as it is both heat and pressure sensitive and cannot be 'fixed' to ensure the permanence of the record.
The advantages of drum recording are combined with the miniature scale of a film record in a number of available instruments. The mechanical and optical systems required are somewhat complex and require very careful maintenance, but they have the advantage of permitting unattended operation for periods of weeks or even months. The reduction of scale renders the necessary photographic processing less costly and burdensome than that of full-sized drum records. The viewing devices are comparable to those needed for microfilms made from standard drum records.
The most useful of these produce a strip with several channels in parallel, and process the film whilst recording. This facility is convenient for fixed-station use, but too bulky and messy for mobile installations.
The chief advantages of scratch recording on microfilm is that the equipment does not require specialized housing or very skilled operator attendance, yet is compact and economical. Good quality records can be produced at a record speed of 5 mm/min which is a twelfth of that normally employed for seismic recording. This enables the recorder to operate unattended for up to 6 days on one length of film. The records can be viewed in the field and analyzed in detail using a magnifying glass, projector or standard type 35 mm microfilm viewer. Utilization in seismology has, however, been limited, and further experience is required.
It is most useful if the first recorder purchased by a network is suitable for noise testing in remote localities, yet is adequate for the recording of both local and teleseismic events. If these conditions are met, then the equipment may be used first for noise testing of possible sites before being installed at a permanent observatory, and then remains available for mobile use during special studies of local events and aftershock sequences. It will often be found easiest to obtain housing and the assistance of an operator if a visual recorder is used, as the immediate availability of results encourages participation. The equipment most likely to conform to all these requirements is a smoked-paper recorder, a good design of ink- pen, or a 'scratch' recorder. The choice between the other recorders described above remains open when the network has settled into a permanent configuration, and will depend on individual circumstances.
The rapid development of magnetic tape recording is leading to its increasing adoption for seismic recording. The advantage of this medium is the ease with which the data can be processed after recording to obtain records corresponding to any desired pass band. The extent to which the original pass band can be modified will depend on the dynamic range of the system, for it will be impossible to recover seismic data for any frequency band within which the signal level has fallen below that of the instrumental noise.
The techniques of recording are conditioned by the fact that the wavelength of the signal on the tape must be between a lower limit, determined by the grain size and the gap in the recording head, and an upper limit determined by the length of tape which lies in contact with the head. Practical values for these limits are about 0.01 mm and 1 cm respectively.
In the so-called 'direct recording' technique, the changes in magnetization of the tape are a direct representation of particle motion in the ground, and records of this type can be made on simple adaptations of audio-frequency equipment. The ultimate range of wavelength on the tape corresponds to a frequency ratio of about 1000:1 which might, for example, extend from 10 Hz to 0.01 Hz for a tape speed of 0.1 mm/s. In fact, the fidelity of response within two or three octaves of either the upper or lower limit of frequency is very susceptible to minor variations in recording conditions, and maximum reliability is confined to a band covering a frequency ratio of about 30:1. Within this restricted range the ratio of input to playback signal can be held constant to within about 20%, so direct recording is useful as an economical way of storing data if rough approximations will suffice for final interpretation. Superior fidelity is, however, required if sophisticated processes of manipulation are to be applied to the output.
High-fidelity wide-band recording is usually accomplished by modulating the frequency of a 'carrier' signal which is recorded on the tape, and the most practicable modern standard (known as IRIG 'Intermediate band') offers response from zero frequency to 16 Hz for a tape speed of 1.2 mm/sec. Higher recording speeds can be used to extend the bandwidth, and lower speeds can economize in tape at the expense of bandwidth. The signal-noise ratio is normally about 40 db on an individual track, but can be improved, especially in the lower end of the frequency band, by recording an unmodulated carrier on a separate track. When this is played back, any variations of tape speed on recording or playback appear as modulation on the output signal, and the demodulated envelope of this 'flutter' can be used to improve the quality of output from the signal tracks. In the electronics industry, the following two types of compensation are common:
Subtractive compensation In this system, the demodulated signal from the flutter track is subtracted from the output of each seismic channel. This can give very good compensation at one particular carrier frequency, but if the carrier frequency of a seismic channel is strongly deviated, a speed variation, which changes the carrier frequency and the control frequency in the same ratio, will no longer produce the same absolute output. Compensation therefore fails to 'track' out to large deviations.
Multiplicative compensation operates by multiplying an appropriate signal in the demodulator of each seismic channel by a number which reflects the carrier deviation of the flutter track. This method works over the whole range of deviation, but fails at the upper end of the frequency band because of unavoidable phase shifts in the demodulation process.
Because seismologists are usually more concerned with the recognition of an onset from a quiet background than with the purity of large signals, subtractive compensation is preferred. With careful adjustment, signal improvements up to about 15 db can be obtained on the seismic tracks adjacent to the flutter track. With wide tapes, however, differences in 'skew' between playback and recording can introduce a phase difference between the flutter content of different tracks and this will reduce the effectiveness of compensation at high frequencies. Adequate control of skew, together with that of flutter, is therefore an essential characteristic of high quality FM recorders.
The processing of seismic data is increasingly dependent on digital computers, and for this purpose the record must be in the form of a succession of numbers, each of which describes an instantaneous condition of the seismograph output.
Paper records can be digitized by making a series of measurements by hand, or with the aid of automatic or semi-automatic curve followers. Readings taken in this way may be punched on to cards or paper tape. This process is, however, relatively slow, and the volume of data which can be handled is correspondingly limited.
If the original records are in analogue form on magnetic tape, accelerated playback can present the data to high-speed electronic digitizers. Commercial processors can now measure up to 30 000 samples per second, and pass the data either on line to a computer, or write it in computer format on digital magnetic tape. By these methods, long runs of multi-channel short-period array records can be read into a computer in a fraction of the original recording time, and the amount of processing performed will be limited by the computer capacity available rather than by the input conversion time.
In all these processes, the quality of the data is limited by the fidelity of the original analogue record and of the digitizing process. These, as we have seen, limit the overall dynamic range to about 50 db. Much better data can be obtained by digitizing the seismometer output before it is recorded. Commercially available analogue-digital converters may be used if the original output is in the form of an electric voltage or current. If the signal is in the form of a varying frequency, electronic counters are used.
By such techniques, dynamic range in excess of 120 db can be obtained, and it might be thought that the possibility of manipulating such records might drastically reduce the requirement for a variety of seismographs having specialized response characteristics. In fact, digital recording on magnetic tape requires about 10 times the volume of an FM record of equivalent bandwidth, and about 100 times the volume of a directly-recorded signal. Punched paper tape requires about 50 times as much volume as digital magnetic tape, and punched cards are much bulkier than paper tape.
For these reasons, the continuous digital recording of short-period data is almost prohibitively costly, and the modern solution to this problem is to use a minicomputer as a short-term data store, and then to use a triggering process to read out sections of data on to the magnetic tape (see, for example Peterson and others, 1976, on the Seismic Research Observatory Project). The problem is much less severe in the case of long-period data, where blocks of data can conveniently be retained for long enough to permit retrospective transcription into a permanent library.
Signals recorded on magnetic tape must be identified in the presence of background noise, and may require further processing in advance of final interpretation. Recognition may be by audio or visual search, or by analogue or digital triggering.
Audio search involves playing back at sufficient speed to transpose seismic signals into the audio range, and listening may be by loudspeaker or earphones. Speed-up ratios depend on the nature of the signals, and the recurrence interval which strikes the best balance between the need to hold the operator's attention, whilst keeping within the audio range and without an overwhelmingly high rate of event detection. Speed-up ratios between 60 and 500 are commonly in use. Even at the highest speed, the audio systems require good bass response to cover the frequencies down to 0.05 Hz, which carry most of the seismic energy of teleseisms. If the data have been collected by telemetry from an extensive network, coverage of local events, and discrimination against local noise sources, is considerably improved by using stereo earphones to listen simultaneously to outputs from two stations some distance apart in the network.
Visual search requires an initial playout in the form of a slowly-moving oscillograph trace. This method has the advantage of providing a permanent visible record of the whole content of the tape, but discrimination is not as good as that obtained by a well-trained audio operator, the records tend to be bulky and the process of recording and searching takes longer.
Analogue triggering, while simple in principle, requires quite sophisticated filtering and/or AGC if discrimination is to be kept up to the level of either audio or visual search.
Digital searching requires A/D conversion if the original record is in analogue form, but then enables quite sophisticated triggering processes to be applied by small computers (see, for example, Crampin and others, 1974). Main storage requirements range upwards from about 8K, and speed-up ratios are commonly in the range from 8:1 to 32:1. The advantages of the method are that the triggering criteria are precisely defined in the program, and that the records can easily be accumulated on tape or disc for storage and subsequent processing.
The only primary standards worth considering for new installations are those which are built around electronic oscillators controlled by quartz crystals and these can yield, at quite low initial cost, constancy of rate better than 0.1 s per day. Such clocks can be used in conjunction with an electronic comparator or stroboscope to allow the error referred to a radio signal to be accurately determined. In addition, these can be made to supply sufficient frequency stabilized power to drive recorder motors at synchronous speed, which allows accurate interpolations of the time scale between time-marks. Clocks should be arranged to operate directly from accumulators to ensure record continuity in the event of power failure.
On drum records, the time mark is usually generated in the form of a contact closure lasting one or two seconds each minute, with a longer closure or omission to distinguish the hour. These marks should be made to deflect the seismic indicator, without interrupting the flow of superposed seismic data. It is also very desirable for radio time signals to be included on the record and, if the clock is to indicate almost correct time, it is necessary to overcome the problem of mutual interference between clock and radio signals. In some types of recorders the throw of the time-marking relay varies in amplitude, dependent on the current passed through it. If resistive elements are placed in the two control lines, then simultaneous radio and chronometer marks will be indicated by a summation of amplitudes. It may be possible to reverse the direction of recorded radio time pips by reversing the polarity of the current supply to the time-mark deflector. If the connections to the deflector are isolated from the power supply, then this can be easily achieved by interconnecting the change-over contacts on the radio and chronometer relays.
If, however, one connection of the time-mark deflector is bonded to the recorder frame which is then connected to one terminal of the power supply, then an additional reverse polarity supply must be used.
If the amplitude and direction of time-marks are fixed, then the radio and chronometer relays can still be connected to result in no deflection of the time-mark relay when they operate simultaneously.
Because of the tedium of counting back to the start of the record, time marks on magnetic tape should be recorded on a separate channel, and encoded in such a way as to enable the absolute time to be determined from any short section of the record. The need is not quite so pressing for visibly-recorded tapes, but coded systems are still applicable. The recommended code is that developed for the 'VELA Uniform' project of the United States, and assumes the signal is presented as a succession of marks every second, some of which are lengthened in a pattern which varies every minute to provide the desired information (Fig. 6.4.2). Some commercial encoders generate the more complicated 'IRIG' code which uses half-second marks, stores the VELA-Uniform data in the first half of each minute, and leaves the second half of each minute free for additional data. As the IRIG code makes greater demands on the resolving power of the final record, it is harder to read at the playback speeds which are most commonly used in seismology and is not recommended for this purpose.
Station-clock corrections should be referred to standard time-signals broadcast by stations conforming to Co-ordinated Universal Time (UTC) (see OP 2.1.1), and a short-wave radio receiver is required for the reception of these signals. Better quality radios sold for personal broadcast reception are sometimes adequate, but such radios often lack selectivity, that is, they are difficult to tune precisely. Simple communications receivers, which have a tuning scale fairly well spread at the frequencies to be used, and a sensitive fine-tuning control, are quite inexpensive and much better. The best communications receivers and the specialized time-signal receivers are usually equipped with a band-spread facility, crystal locked tuning on selected frequencies, or very precise tuning controls. These will be required if the reception is found to be poor.
The operation of the time-signal receiver is much enhanced by the use of a good antenna. If a simple indoor whip proves to be inadequate, the options are to use a long horizontal wire or a dipole tuned to a particular wavelength: the latter has a directional characteristic which helps to exclude unwanted signals. To make a tuned dipole, a length of wire equal to half of the wavelength of the most important radio signal to be received is divided in the middle with an insulator, and the separate sides terminated to the conductors of a length of 70-ohm twin line. In the absence of such cable, flat twin flex as used for reading lamps may be substituted in temporary installations. The length L (meters), required to tune the antenna for a radio signal of frequency f (Hz) or wavelength lambda (meters) is approximately given by L = lambda/2 = C/2f, where the velocity of propagation C~ 3 × 108 m/s. For example, for 15 MHz,
The length given by this formula should be reduced by about 5% to allow for capacitative end effects.
The antenna should be erected at a height of about lambda/2 above the ground and at right angles to the direction of the transmitting station. The ends of the twin lead-in should be connected to balanced antenna terminals if these are available on the receiver, and any shield on the cable should be connected to the chassis. If only one antenna terminal is available, the lead-in should be connected to it and to the chassis.
When the receiver input is unbalanced, coaxial cable used with a 'balun' transformer at the antenna, yields the best protection against interference from nearby electrical equipment when the core and screen are connected to the antenna and Earth connections of the receiver (a co- axial connector joined to the core and screen of coaxial cable will do this automatically). Any cable should run as directly and straight as possible to the radio receiver. If interference remains troublesome the receiver itself should be earthed to a metal stake or sheet of wire mesh buried in the ground.
Seismograph stations are sometimes located at or near universities, meteorological observatories or other scientific establishments. These are frequently in seismically noisy areas, and hence the need arises to put the seismometers in a quieter location and telemeter the signals back to a convenient recording site. In other cases, such as for the study of local earthquakes, it is desirable to have a number of remote seismometer locations all recording at a central point.
Seismic signals can be transmitted over lines up to some kilometers in length by simply amplifying the seismometer output so that the signal level is raised well above the noise due to induction and other causes in the line. However, for satisfactory telemetering over many kilometers or for telemetering by radio links, it is desirable to use the seismometer output to modulate the frequency of a carrier wave that is recorded or demodulated at the receiving point.
Many of the telephone companies of the world provide a commercial telemetering service over their lines that is completely adequate for seismological purposes. Commercial equipment for telemetering by high frequency radio links is also available and can be used where telephone lines are unobtainable.
The capital or rental cost of a telemetering channel is quite substantial, but is largely offset by the fact that a single chronometer provides the time service for all the channels which enter a common recorder, and that a single multi-channel recorder is usually less expensive than a number of separate units of equivalent capacity. The cost and difficulty of the telemetering link, and the degradation of the signal which passes along it, all increase with distance. Convenience in data processing is enhanced by having all the relevant information on a single record. For distances up to about 10 km the balance of advantage lies strongly in favor of the telemetered system. For most types of operation, the use of separate recorders is superior when the range exceeds 100 km. The choice between these limits will be determined by technical and environmental factors.
Date created: 1/7/97 Last modified: 9/9/97 Copyright © 1997, Global Seismological Services Maintained by: Eric Bergman firstname.lastname@example.org