Guralp Systems CMG-SAM - Manual (Page 21)

Operator’s Guide CMG SAM & CMG CRM
5. DATA STORAGE
4.1 SILICON FILE
The solid state memory provides the SAM and CRM units with a large amount of low-
power storage and data buffering so that systems can utilise data transmission links on a
time shared basis or where the data link is unreliable providing local storage and re-
transmission. It is also used in conjunction with the hard disk storage option to buffer
data accesses to the disk allowing the disk to be powered down for extended periods to
reduce overall system power consumption.
The module can be configured with from 1 to 8 Mbytes equivalent to 1000 to 8000 data
blocks and upto 2 modules can be fitted to one system.
This unit uses low-power dynamic RAM (uPD42644 1Mb*4 Silicon file) accessed as
1k blocks in the processor address space. The processor provides the low frequency
‘self-refresh’ clock and control timing in conjunction with an FPGA to provide a rapid
access, large capacity circular data buffer without moving parts and the associated high
power consumption.
4.2 S.C.S.I. DISK
For systems that require long term storage of continuous and event triggered data, the
SAM unit features an interchangeable SCSI disk drive (>1 Gbyte). This allows the
acquired data to be retrieved simply by removing the disk drive and fitting a
replacement unit.
To minimise the system power consumption this module is usually used in conjunction
with a large RAM buffer (either system 512k static RAM or silicon file) so that the disk
drive can be powered down for long periods while data is accumulated in the buffer.
This intermediate buffering also allows the drive to be removed from the system for
data retrieval.
This module like all others occupies 1 page of the processor address space and includes
32k of RAM for read/write buffering in addition to the SCSI controller chip (53C94).
The disk filing system and data format have been designed especially for efficient
operation in a continuous real-time system. All data are stored in a 1k block with a
header which identifies the data stream (1 stream per ADC channel per sample rate) and
the time of the first sample (by definition on an integer UTC second). On the disk all
data blocks for a particular stream are linked together by disk address, the address on
each block pointing to the next block - the current end of data has a null pointer. Disk
space is allocated chronologically to the next ‘cluster’ (8 *1k blocks) of data available.
When the end of disk is reached the system can simply wrap round to the beginning of
the disk and automatically overwrites the oldest data. The data block format is described
in more detail in the next section.
Issue D March 2001 19