SonTek/YSI
10
ADVField Operation Manual (September 1, 2001)
1.3. ADV Sampling Strategies
During data collection, the ADV pings as rapidly as possible and averages pings together to out-
put velocity data at the user-specified sampling rate. Each sample is calculated as the block aver-
age of pings during the sampling interval. No filtering or averaging is done on adjacent samples.
The ADV uses two clocks: a real-time clock to mark the start of data collection, and a high fre-
quency sampling clock for the timing of each sample. The separate sampling clock is required
because of the digital processing requirements of the ADV. The high-speed sampling clock is ac-
curate to 1 part in 50,000 (about 2 seconds per day) and thus can introduce drift over long peri-
ods. The ADV data acquisition software uses the controlling computer for the real-time clock,
while the ADVField internal real-time clock is used during autonomous deployments.
The ADVField can include internal memory and battery power for autonomous operation. Since
the sampling clock may drift over long periods, the ADV uses burst sampling during autonomous
deployments to provide a periodic reference to its real-time clock. In burst mode, the ADV col-
lects a fixed number of samples at the user-specified rate (a burst) at an interval specified by the
user. Data analysis, memory requirements, and battery limitations also make continuous sam-
pling impractical for many deployments; thus the autonomous ADV has been designed to ac-
commodate a variety of sampling strategies as described below.
Figure 7 shows ADV sampling strategies for real-time data collection and autonomous deploy-
ments. In the first example (real-time data collection), the ADV outputs data continuously at the
user-specified sampling rate until terminated by the user. Data is stored on an external device
(computer or data logger). The ADV relies on its sampling clock for the relative timing of re-
maining samples – recording the real-time clock only at the beginning of data collection.
Continuous sampling for an autonomous system is not practical for two reasons. First, the quan-
tity of data generated is, in most cases, prohibitively large for storage and analysis. Second, the
sampling clock may drift with time, and the data needs to be periodically referenced to the real-
time clock. Therefore, all autonomous deployments use burst sampling. In burst sampling, you
must specify three values: the sampling rate, the number of samples per burst, and the time be-
tween the start of successive bursts (called the burst interval). These values must meet the fol-
lowing criteria.
Burst Interval ≤ Overhead + (Number of Samples / Sampling Rate)
Overhead is the combination of the time required to prepare the system for data collection at the
start of each burst and to calculate statistics at the end of each burst. An overhead time of about
five seconds is required for each burst.
To Next Page
Loading...
Need help?
General
