Re: Model of 2MHz Aquadopp Noise
Up to Current profilers and current meters
I am trying to develop a numerical model of the error and noise present in the 2 MHz Aquadopp's xyz data output for varying bin sizes in an open ocean environment. Once I have the model, I want to add the characterized error and noise to theoretical current conditions to generate a simulated sensor output for varying bin sizes, but I have a few questions:
- I found the equation for sig(vmean) in the Functional Description of the user manual. Is this the process used by the software to calculate the horizontal and vertical precision estimate shown in the "Deployment Planning" window of the AquaPro software? If so, is the velocity precision the expected standard deviation of the output data?
- Do you guys have a publically available numerical model or equation for calculating the measurement noise as a function of bin size.
- Is it ok to assume that the primary source of error in the output of the 2MHz Aquadopp (excluding flow turbulence) is Gaussian-distributed Doppler noise that is uncorrelated between adjacent bins and beams?
- The manual states that the noise floor can be found by collecting data in air. As I understand it, the noise floor is the overall sum of all the noise sources in the instrument. But, this measurement would exclude Doppler noise right? I collected some noise floor data (1 Hz, ENU, 0.10m bins) with the sensor in air. I meant to collect in xyz, but forgot to change the setting. I attached the plotted data. The noise pk to pk was a lot greater than I would have thought.
Thank you so much for looking into my questions. I look forward to hearing back from you soon.
Previously Matt Young wrote:
This is just the expected reduction in variance of a sample due to averaging. The software uses this to calculate the precision based on single ping uncertainty as the averaging interval changes in software.
I found the equation for sig(vmean) in the Functional Description of the user manual. Is this the process used by the software to calculate the horizontal and vertical precision estimate shown in the "Deployment Planning" window of the AquaPro software? If so, is the velocity precision the expected standard deviation of the output data?
I know there are equations for the pulse coherent and Broadband processing, so I think there's probably one for the narrowband processing the Aquadopp uses. I would check the literature for this type of thing.
- Do you guys have a publically available numerical model or equation for calculating the measurement noise as a function of bin size.
This is a pretty reasonable assumption except for the correlation between adjacent bins. The bins overlap in a standard Aquadopp profiler (meaning they aren't truly independent), so there will be some correlation between them. I think people generally assume the noise is uncorrelated within one bin separation, but I would verify this in actual data to be sure. The noise will be largely uncorrelated between independent beams.
- Is it ok to assume that the primary source of error in the output of the 2MHz Aquadopp (excluding flow turbulence) is Gaussian-distributed Doppler noise that is uncorrelated between adjacent bins and beams?
- The manual states that the noise floor can be found by collecting data in air. As I understand it, the noise floor is the overall sum of all the noise sources in the instrument. But, this measurement would exclude Doppler noise right? I collected some noise floor data (1 Hz, ENU, 0.10m bins) with the sensor in air. I meant to collect in xyz, but forgot to change the setting. I attached the plotted data. The noise pk to pk was a lot greater than I would have thought.
Hmmm, I've never tried to estimate the noise floor this way. I generally prefer to do it with actual data I'm looking at using a velocity spectrum and identifying the noise floor there. I think the manual is actually referring to the electronic noise in the receiver circuits with the in air testing. This is different than the Doppler noise. It's what is used to calculate SNR. I think all of our instruments now measure briefly without transmitting at the start of data collection to obtain a noise amplitude in counts which is used for SNR calculations.
Good luck with the model, but I would be sure to do a good literature search. I think a lot of this is probably covered in papers already. There should be quite a few from the IEEE Ocean Engineering conferences and journals worth reading.
P.J.
Powered by
Ploneboard
Forgot your password?
New user?
