 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
When will SEG Y be replaced or revised?
Background
Back in the 1960's, the seismic industry showed signs of developing many competing and incompatible formats for digital seismic data, repeating what had happened with analog recording.
Early digital seismic recording systems used either 1" magnetic tape or ½" magnetic tape. The 1" tape was only used by one significant instrument builder, Texas Instruments, and so was standardized. But the narrower tape started to show up in a different format for every manufacturer. Data processing contractors and major oil companies started to face the problem of supporting many different formats. In September 1966, the SEG appointed two subcommittees to recommend standard formats for 9-track, ½" digital tape. One subcommittee was to make recommendations for a field format, the other for an exchange format.
The field tape format committee produced two formats A and B, both based on IBM tape specifications, and both using ingenious techniques for recording a maximum amount of seismic data in a minimum amount of tape, using a minimum amount of memory in the field system.
The exchange format committee defined SEG Exchange Tape Format (which became known as SEG "Ex") also based on IBM specifications. The committee included specifications for a variant of the format to be used on 7-track tapes. These were in common use in processing centers which used non-IBM computers such as the Control Data machines.
In early 1971 the field tape subcommittee realized that the existing formats were inadequate. The new Format C recorded data in a multiplexed format, using IBM System/360 floating point numbers, requiring 4 bytes per sample per trace. The committee noted that even with the availability of 1600 bpi tape transports, the new format could not record more than 62 channels of seismic data with 2 ms sampling. This format allowed an effectively unlimited dynamic range, and dynamic resolution of 144 dB.
In 1973 the SEG formed a new subcommittee to develop a new demultiplexed seismic exchange tape format. As a successor to the "Ex" format, the new format was called "SEG Y Format".
Seismic exploration in 1973 was two dimensional. The GeoSpace GS2000 recording system promised a "new dimension in data acquisition", but it did not mean 3D exploration. 3D seismic exploration was implicit in Harry Mayne's patent for the CDP seismic method (U. S. Patent 2,732,906 Jan. 31, 1956) but was not explicitly spelled out, and in any case, the technology needed to implement it did not exist for another two decades. Most explicitly 3D seismic patents did not appear until later in the decade. The new format did not have to accommodate 3D seismic surveys.
Seismic interpretation in 1973 was on paper. The first commercial sales of seismic interpretation workstations were not for another decade. So the exchange format was for moving data between processing systems.
The SEG Y format contains three types of file block:
The physical arrangement of bits on the tape, together with BOT (beginning of tape) and EOT (end of tape) marks, encoding details, EOF (end of file) and IBG (interblock gap) are all specified.
When the format was first defined, the EBCDIC block was intended as a standardized text page which could be printed out easily to identify the data following on tape.
Part 2 of the Reel Identification Header is generally called the "binary header". It contains, in a form easily read by a computer, the data which are constant for a whole 2D seismic line: the line number, data format code, the sample interval, etc. All values are 16-bit or 32-bit integer.
Each seismic trace is written as a 240 byte header followed by a series of data sample values, in one of four formats as specified in the binary header. Meanings are assigned for the first 180 bytes of the trace header, with values in 16-bit or 32-bit integer format.
SEG Y was defined at a time when mainstream computers were IBM System/360 mainframes, enormously expensive machines occupying specially air-conditioned rooms with a dedicated staff to coddle them. IBM used proprietary formats for dat representation, while the rest of the computer industry supported standards such as ASCII for text representation, and IEEE for floating point numbers. So today, when almost all handling of seismic data uses Unix workstations or personal computers, reading and writing SEG Y data is, for the computer, like conversing in a foreign language.
The original SEG Y specification required the use of ½" tape with 9 tracks, recorded at either 1600 or 800 bits per inch. The logical arrangement of the data on the tape can easily be done on a more modern tape format such as 8mm or DLT tape, but there is a more subtle problem. In 1974, a typical tape was 2400 ft long. At 1600 bits per inch, it could hold about 45 megabytes of data (without interblock gaps). A DLT tape might hold 50,000 megabytes of data, and an 8mm tape 10,000 megabytes. If interblock gaps are taken into account, a typical seismic trace of 2000 samples would use 8,240 bytes, equivalent to 5.15 inches, and the interblock gap 0.6 inches, for a total length of 5.75 inches. A reel of tape might hold, at most, about 5000 traces. In 1974 a typical land seismic line might consist of 100 records of 24 traces each, for a total of 2400 traces, and a marine seismic line might be 1000 records of 48 traces, for a total of 48,000 traces. It is easy to see that individual seismic lines were likely to take more than one tape, and putting more than one line on a tape would not be needed. Today, it seems a waste to put a 20 megabyte seismic file by itself on a tape capable of holding 5,000 megabytes or more.
Originally, SEG Y was mainly used for unstacked seismic data. Today, it is widely used for stacked seismic data. There are three problems arise from this:
The widespread use today of SEG Y format for processed 3D seismic data to be loaded onto workstations has introduced a new set of questions:
With 25 years hindsight, there are a few things we can say that the Technical Standards Committee should have thought of in the first place:
Probably never. Actually, it has been: SEG D format (1980) was intended as a replacement for all of the older formats, but is so complicated and contains so many options that hardly anyone uses it as an exchange data format (it is used as a field recording format). Furthermore, as far as I can figure it out, it does not address in a meaningful way any of the shortcomings in SEG Y noted here.
|
|
|
|
|
|
|
|
|
|
|
|