Comprehensive Manual60
© 2018 Nortek AS
not to an orthogonal coordinate direction such as X, and any screening should take this into
account. Therefore, we recommend looking at each beam individually when looking at correlation
and SNR rather than an average across two or more beams. Problems can occur in only one
beam and averaging the data across beams may obscure potential problems.
Check each beam individually. It does occur that just one beam is bad, and this can be indicated
by a signal level near the noise floor, for example. The data may still usable, but if using the
Vector one has to process the data with the two other beams only; a solution that is known as a
two-beam calculation. This will result in 2D currents only (since there is only data from two beams
available. Note that the processing makes one important assumption; the vertical currents are
zero. This is a reasonable assumption in the vast majority of current flow. For the 4-beam
systems, one can discard data from one of the beams and still get 3D velocities out.
Removing low correlation measurements is a good idea because correlation is a strong indicator of
data quality in the sense of a valid Doppler phase shift determination.
Lower correlation means more noise in the data.
The SNR is calculated by subtracting the noise level in counts from the amplitude levels measured
in counts during velocity measurement, and it tells you the rate of signal over noise. The noise
level is measured at the beginning of each burst (and once at the beginning of continuous
measurements). A rule of thumb is that SNR should be >15 dB. The noise level should be around
50 counts for the Vector.
Correlations above ~70% is considered to be generating good quality data. The value of correlation
threshold of 70% is actually fairly stringent and are may be throwing away good data. A close
examination of the dataset is the best way to set a correlation threshold for discarding bad data
points.
Screening data: There is no point using the SNR values to screen data, use the correlation value
instead. When you discard data with a low correlation, you are screening the data with respect to
SNR as well.
Typical processing tasks:
1. Assess data quality (QA/QC)
2. Data screening
3. Statistics to describe the flow (mean, variance)
4. Spectral Analysis (turbulence, waves)
2.8 Common Data Analysis Scenarios
Correlation looked good during data collection, but there are several range cells with poor
correlation
Low correlation in only a few range cells is symptomatic of weak spots. Unfortunately, there is not a
way to correct for this in post processing. Please see the Configuration Guide for how to eliminate
weak spots before recording data.
SNR looked good during data collection, but there are several range cells with much
different SNR (higher or lower)
This is also symptomatic of weak spots or an obstruction (e.g. an object) within the beam. Weak
spots will generally see higher SNR because the strong boundary echo is causing problems, but
may be associated with lower SNR depending on the nature of interference.
A beam obstruction will typically produce a strong echo and high SNR. It may create an acoustic
shadow resulting in low amplitude and SNR in bins behind the object as well.
Weak spots can not be corrected in post processing. Removing biased velocities associated with an
object in the beam is possible by determining which range cells are affected and discarding them
from analysis.
Mean velocity profiles do not look good even though correlation and SNR are high
It is important to assess data quality and perform some basic quality control on a dataset before
attempting to interpret the results. At minimum, discarding very low correlation measurements
(<40%, adjusted as needed based on data quality) and very low SNR values (<15 dB, again adjusted
To Next Page
Loading...
