In general, the correlation values should be upwards of 90% and not widely scattered. In a time series signal “correlation” is simply a measure of “self-similarity”. The correlation is affected not only by the strength of the acoustic echo that comes back from the water but also by other processes such as pulse-to-pulse interference.
The SNR, or Signal-to-Noise Ratio, is specified in dB, and it is a measure of the strength of the acoustic signal relative to the background noise level. SNR (in dB) = 20*log10 (Amplitude with signal/Amplitude with no signal). Note that using Amplitude as a quality-control parameter is a bit trickier than using SNR since it is just a number.
If the SNR is very low:
If a low SNR is encountered, try adding scatterers to the water column such as the seeding material provided with the velocimeter. Add a little at a time and see whether the SNR improves. If there are enough scatterers in the water, check the Power Level in the Configuration dialog and make sure it is on high. The SNR is strongly dependent on the amount of scatterers in the water. Higher velocities will lead to more scatterers remaining in suspension, so most users will tend to see an increase in SNR with velocity.
Vector: Typical SNR values should be about 30+ dB.
Vectrino: SNR should be around 10 dB. Data improves as SNR increases, up to an SNR of around 15 or 20. Beyond that, the SNR plays little role in the data quality. Depending on your objectives, you can also get good results even when the SNR falls below 10.
Vectrino Profiler: SNR should be around 30 at the center of the profile and above 20 at the ends, with a parabolic shape peaking at the sweet spot of 0.05 m range. SNR at ranges greater than 0.06 m will drop sharply so attention should be paid to this region of the profile if used. You’ll want to be a little more careful at the ends of the profile, especially the far limit, because the geometry of the head causes a decrease in SNR there.
If the SNR values are zeros at some depths: The first data record that the Vectrino Profiler produces is a noise floor record. This is set up and collected in the same way as the other velocity records except that the transmitter is turned off. The amplitude data in the noise floor record is subtracted from the amplitude data of each velocity record to produce the SNR data. If the noise floor amplitude is greater than the velocity record amplitude, then the SNR is set to 0. If there is very little scattering material in the profiling region, then it is quite possible for the noise floor to have amplitude cells larger than the velocity cells, with the result being a 0 SNR. Interference from other instrumentation (or other noise sources, electrical or otherwise) can also influence the noise record and cause this sort of behavior.
The correlation is low or varies widely:
Vector and Vectrino: Check your Nominal Velocity Range setting in the Deployment Planning dialog and ensure it is high enough for the flow you are measuring. In the same dialog, on the right-hand side, two velocity ranges are reported, one each for horizontal and vertical velocities. The Nominal Velocity Range setting will cause changes in the reported horizontal and vertical ranges. Make sure these are above the expected flow velocities or you will not measure usable, high-quality data.
Here is an example of a spiky velocity time series and correlation from one range cell:
Vectrino Profiler: Correlation profiles should have the same parabolic shape as SNR. Try to achieve correlations greater than 90% throughout the profile. The highest correlations will be at the sweet spot at a range of 0.05 m, with correlations at ranges greater than 0.06 m dropping off sharply.
To improve correlation, first make sure SNR is high (see above). Then check the Velocity Range and Ping Algorithm in the Configuration dialog. The actual Horizontal range and Vertical range are reported in the configuration dialog just below where the Velocity Range is set. These need to be appropriate for your flow. These ranges are also reported in the Text Data area in the leftmost pane of the Vectrino Profiler software window. Click the Text Data button if this pane is not visible.
If your velocity range is not what you expected, check which ping algorithm is selected. Reported velocity ranges will depend on the algorithm. Here are the descriptions from page 20 of the Vectrino Profiler SW User Guide:
Max interval: Produces the longest possible ping interval to achieve desired ambiguity velocity and sampling range. This is best used for smooth flow conditions.
Min interval: Produces the smallest possible ping interval to achieve the sampling range and, at minimum, the desired ambiguity velocity. This is best used for turbulent measurements.
Adaptive (If licensed): Collects a long profile and examines the return echoes to determine where acoustic interference is occurring. The ping rate is then calculated to achieve the desired ambiguity velocity and sampling range while minimizing/removing acoustic interference. The adaptive ping interval selection is carried out with every configuration performed so the ping intervals selected may vary when the beam return signals change. This is the best general choice.
An appropriate Velocity Range for a flow will typically produce a Horizontal Range (m/s) and Vertical Range (m/s) that are larger, but not significantly so, than the expected velocities. Horizontal refers to the x and y instrument velocity components, Vertical to the z instrument component. Most users will want to match the Horizontal Range to flow conditions. If you expect large velocities aligned with the Vectrino Profiler’s z velocity component, make sure the reported Vertical Range is large enough.
When adjusting the Velocity Range to improve correlation, try moving to a slightly higher value first. In most circumstances a Velocity Range that is too large is better than a Velocity Range that is too small.
In a turbulent flow:
Highly turbulent flows (e.g. a turbulent jet) will typically need a larger Velocity Range than mean velocities suggest. The best way to increase correlation in a highly turbulent flow is to adjust your velocity range. For some conditions, when you will be working with a larger velocity, it will mean dual-PRF operation for the Vectrino Profiler. You will need to ensure there is adequate SNR to make good measurements in this type of flow as well.
Can we use data with low correlation to calculate turbulent features, such as Reynolds stress and turbulence spectra?
If the reduced correlation comes from a lack of scatterers in the water (i.e. low sediment concentration), you can generally perform turbulence analyses. However, you will find that the spectra are quite noisy in the high frequency range. If the reduced correlation comes from other processes (for example, SNR > 20 dB and Correlation < 90), then you have to be more careful since the velocity data may be biased.
If the Velocity Range is set appropriately, but there are low correlation regions in the profile:
These are due to pulse interference, also called weak spots, where a past acoustic pulse’s boundary echo is interfering with the current pulse’s processing. Read more about how to detect this in the data by clicking the “WeakSpot” tag below. Here is an example of how it may look in the velocity data: