1. Real-time Uncertainty Output for MBES Systems
Eric Maillard, George Yufit, Pawel Pocwiardowski RESON, Inc

2. Introduction What uncertainty? How is it measured?
Imperfect Sonar in a perfect world
Sound speed assumed perfectly known
No refraction correction …
Quantify random error in range or depth measurement
How is it measured?
Using classical formulas published by Simrad and IFREMER
Adding our sonar specifics into them

3. Methodology
Start with a set of formulas

4. Methodology Adapt to Sonar specific
e.g. bottom detection with blending

5. Methodology Use Monte-Carlo simulation to check the adaptation:
Simulate acoustic signal from bottom
Apply beamforming and bottom detection
Measure specific parameters:
Number of points used in phase processing
Blending coefficient

6. Output of Monte Carlo simulation

7. Adapting error model

8. Published models Simrad’s model IFREMER’s model
A priori model based on simple sonar characteristics
Can be tuned to matched observed performances
IFREMER’s model
Using in-depth sonar modeling
Accurate bottom detection characterization
Environment dependant
Increased level of complexity

9. Comparison on a simple case
30 meter depth, sandy bottom
400kHz
1.0 x 0.5 degree system
220dB source level

10. Comparison on a simple case
No baseline decorrelation
No shifting footprint

11. Revisiting IFREMER’s model
Zero phase difference instant

12. Phase bottom detection random error
Linear regression
Sample parameters from measured phase difference time series

13. Depth error

14. Increased model accuracy
Amplitude of signal varies according to Rayleigh law
New estimation of phase noise
Filtering of phase before regression
No exact derivation
Least Mean Square modeling

15. Phase measurement error
When N >> 1
With Rayleigh distribution modeling
With phase filtering

16. Monte Carlo validation
Beam angle, degree. 45
52.2 60
Depth error, equation
2.58*10-4
2.94*10-4
3.39*10-4
Depth error, Monte - Carlo
3.23*10-4
2.87*10-4
4.41*10-4
SNR at array output 20 dB, depth 25 m

17. Experimental validations
SeaBat 400kHz
Three environments
Tank
Harbor (boat and sonar static)
In open water
Boat drifting
Sonar mounted over-the-side

18. Tank test Set the sonar on a rigid frame
Try to maximize coverage given confined space
Collect series of pings at high ping rate
Statistical analysis is not concluding
Need to get more realistic environment

19. Harbor test Set-up Very shallow water
Increase incident angle range by tilting and rotating sonar

20. Typical data

21. Bottom topology
Average depth computed over 60 pings

22. Uncertainty

23. Real life versus models
Actual results better than prediction
Too small number of pings
IFREMER and RESON models match experimental data on phase detection
Too pessimistic on amplitude detection

24. Effect of pulse length
Amplitude detection proportional to pulse length
Check validity of models

25. Open sea test
Bottom topology (without refraction correction)

26. Uncertainty results

27. Conclusions
Better match between model and experimental data at larger depths
Still Shallow Water 
Some more tuning of model is required
Real-time output measures environmental conditions
Validation for other SeaBat will follow