Guralp Systems CMG-DCM build <10,000 User Manual
Page 106
Acquisition Modules and Platinum Firmware
Digitiser Configuration
Once the digitiser has rebooted, a new section appears on the screen, as
shown below:
The “Source” drop-down menu allows you to select the tap that will be used
as the input to the causal filter (which need not, therefore, be the final filter in
the decimation chain). The “Decimate by” drop-down menu allows you to
select the divisor implemented by the filter. The “Output” check-boxes allow
you to select which components to output via the causal filter.
The causal filter may be disabled by clicking the
button,
then scrolling to the bottom of the page and clicking
.
8.1.1.6 Compression mode
Compression mode: Controls how samples are packed, affecting both data
latency and line utilisation. GCF packets contain a 32-bit starting value and
then a series of differences between consecutive samples. When the input
signal is relatively quiet, these differences can often be expressed as 8-bit
quantities. When the input signal includes large transients, the differences
are transmitted as 32-bit quantities. For intermediate level signals, 16-bit
values can be used. This is known as compression as it can “compress” four
samples into the space otherwise occupied by a single value. The digitiser
can be configured to limit the amount of compression used, decreasing
latency.
The difference values are stored in records, which are four bytes long and, so,
may contain four 8-bit differences, two 16-bit differences or one 32-bit
difference. A GCF packet can contain up to 250 records so the maximum
number of samples in a packet is between 250 (when 32-bit differences are
used) and 1000 (for eight-bit differences).
Packets must start on whole-second boundaries, so they are not always filled.
In addition, it is possible to configure the digitiser to further restrict the
number of records in a packet in order to decrease latency.
The drop-down menu controls both of these settings and normally offers the
following choices:
•
Recommended for 200sps or less (8 bit 20 records, maximum 80
samples): - Represents the best compromise for throughput: allowing
8-bit compression potentially increases the number of samples per
106
Issue E - February 2014