Simrad EY500, Инструкция по эксплуатации

Страница: 131 / 262

Theory of operation
P3403E/A
11
3 EY 500 DB FORMAT
Simple conversion between dB and linear scale is obtained in a computer by
using 10 dB x log(2) = 3.0103··· dB as a reference value in the dB domain. The
EY 500 algorithms use 16-bit words to represent dB quantities.
A B
XXXXXXXX XXXXXXXX
The eight most significant bits (A) correspond to the integer part relative
3.0103··· dB and the eight least significant bits (B) correspond to the fractional
part. Thus, the least significant bit corresponds to an increase/decrease of
3.0103··· dB/256
.
0.01 dB. Assuming as an example the linear decimal number
178.125 the value of A and B becomes
10 dB x log(178.125) = 22.507··· dB
A B
22.507···/3.0103··· = 7.4767··· = 00000111.01111010
Conversion to linear scale is based on the relation
10
0.30103··· x A.B
= 2
A.B
= 2
A+0.B
= 2
A
x 2
0.B
Evidently, the upper byte A is simply the exponent in binary floating point
format, and 2 to the power 0.B is the mantissa. Thus, the mantissa can be
obtained by using B as the address in an antilog lookup table containing 256
elements, and a similar technique can be used for the inverse conversion from
linear to dB scale.
4 BOTTOM DETECTION
The bottom detection algorithm is implemented solely in software. The
algorithm is designed with emphasis on reliability in the sense that erroneous
depth detections are never output. Whenever uncertainty is associated with a
detection the algorithm outputs zero depth to indicate that no reliable detection
was obtained. The algorithm is designed to maintain bottom lock for a
discontinuous jump in bottom depth, and special features have been included
in order to avoid false bottom detection on schools of fish. Operational
experience has shown that the algorithm indeed is quite robust; erroneous
bottom detections are virtually absent, a dense school of fish does not confuse
the algorithm, rough bottom contours cause only a few dropouts to occur.
Basically the algorithm is implemented as a fourfold tracking algorithm. For
each ping up to four candidate bottom returns are identified, and their
association with previous bottom candidates is determined in order to perform