YSI SonTek ADVField, Техническая документация инструкция

Страница: 212 / 228

SonTek/YSI
Acoustic Doppler Velocimeter Principles of Operation (September 1, 2001) 6
When operating in highly turbulent flows, the ADV may show low correlation values that indi-
cate increased noise in velocity measurements. For turbulent flow, the noise level may be re-
duced by increasing the velocity range. The magnitude of the turbulent fluctuations can be ap-
proximated by the standard deviation of velocity data displayed by the ADV data acquisition
software. If these are 5% or more of the velocity range setting (i.e., greater than 5 cm/s when us-
ing the
±
100 cm/s velocity range), performance should improve by changing to the next higher
velocity range. This change should increase the correlation coefficient and reduce the noise in the
velocity data.
6.1.2. Sampling
The ADV is designed to measure velocity as rapidly as possible. A single estimate of the 3D ve-
locity field is referred to as a
ping
. The ADV pings 150-250 times per second (the rate varies
with velocity range setting). As the noise in a single ping is too high for practical use, the ADV
averages a number of pings to reduce the noise level in each output velocity sample. The number
of pings averaged is set to meet the user-specified
sampling rate
, within the range of 0.1 to 25
Hz. For example, when sampling at 25 Hz the ADV will collect as many pings as possible over a
40-ms period, average these values together, and output the average as one
sample
.
An important result of the ADV sampling scheme is that reducing the sampling rate decreases the
noise in each sample (by increasing the number of pings averaged per sample). The instrument-
generated noise, also referred to as Doppler noise, is random; averaging multiple points will con-
verge towards the true value without introducing bias. The noise level decreases with the square
root of the number of samples averaged; thus, data output at 1 Hz has about one-fifth the noise of
data output at 25 Hz.
6.1.3. Short Term Uncertainty (Noise)
All Doppler systems have an inherent measurement noise that is a result of the physical process
by which the sound waves are scattered from particles in the water. This is referred to as Doppler
noise. Doppler noise is purely random and can be assumed to follow a Gaussian distribution. Av-
eraging multiple data points converges to the true value without introducing bias.
Under good operating conditions (i.e., SNR > 15 dB, correlation greater than 70%), the noise in
ADV horizontal velocity data is estimated at 1% of the velocity range when outputting data at
25 Hz. For example, individual samples at 25 Hz will have horizontal velocity noise of about
±
1 cm/s if using the
±
100 cm/s velocity range (or about
±
0.3 cm/s if using the 30 cm/s velocity
range). As mentioned in §6.1.2, noise decreases with the square root of the number of pings av-
eraged per sample; thus, individual samples at 1 Hz would have noise of about
±
0.2 cm/s when
using the 100 cm/s velocity range.
Note that the above noise estimates are for horizontal velocity. Since the bistatic axes of the
ADV receivers are 15
°
off the vertical axis, noise in horizontal velocity measurements is larger
than in vertical velocity measurements by approximately a factor of four.
6.1.4. Accuracy
Accuracy for ADV velocity data refers to the presence of a bias in mean velocity measurements
after removing instrument-generated noise. Two main factors influence the accuracy of ADV ve-
locity measurements: sound speed and probe geometry. Sound speed is a user input parameter;
errors in user input sound speed will cause errors in velocity measurements that can be corrected