Skip to content

Latest commit

 

History

History
103 lines (76 loc) · 4.72 KB

trex_fish_notes_2017.md

File metadata and controls

103 lines (76 loc) · 4.72 KB
Raw

TREX 13 fish echo analysis @APL

Part II (2017)

2017/01/13

Spent this past Mon/Wed/Fri trying to figure out the offset between my beamforming routine and Jie's output. Finally figured out what's going on...

There are 4 compensations that need to be taken into account:

  1. Gain associated with match filtering (gain_pc in code).
  2. Gain associated with beamforming (gain_beamform in code). This is se to 20*log10(48) because there are 48 elements when used in the linear configuration, so total energy is enhanced by 48 times. Note that since cardioid beamforming was normalized to the level of linear beamforming output, this factor is the same under cardioid beamforming.
  3. Differences of load when driving the FORA array (gain_load in code). The gain is 46.95 dB when the array is driven as triplet, and 42.35 dB when the array is driven as linear array. My confusion before was that I thought the gain would be different when different combinations of elements were used.
  4. FORA system gain (gain_sys in code). This gain is 12 dB for all runs starting from run 41, and 18 dB before that.

2017/01/23

Calibrate for source level

Going back to original code get_SL. The transmitted voltage is recorded at the output of the power amplifier (L10) before feeding into the transducer (see figure below). Data are called HAARI. HAARI files are sorted according to dates and transmit start time. In the data, ch0 is voltage and ch1 is current.

Transducer used is ITC-2015. Use get_TVR_ITC2015 to get the TVR curve.

Unit for TVR is dB//uPa/V @1m. Unit of voltage recorded on HAARI (Vh) is dB//V. So source level (SL) is TVR*Vh. Note there is a scaling factor A.ch_0_voltage_scale_factor=100 that needs to be applied to get the current transmit signal voltage.

Updated code get_SL:

  • Note Jie uses rms values for calculating the source level, but I have to use the actual spectral level for potential resonance fitting using fish scattering models. In get_SL, the rms value is tx_rms_dB, which can be recovered by calculating 10*log10(sum(10.^(tx_psd_dB/10))*df) (integrating over all spectrum).

  • In the output of get_SL, SL=SL_psd/fswhere fs is the sampling frequency, i.e., SL_psd is the spectral level.

    { width=70%}

2017/01/25, 27

  • Check to make sure subset_beamform is getting the right set of data --> Yes
  • Revise get_pie_outline to work with degree input and simplied its output to only x,y coordinate points
  • Sorting out the fft/psd issues: Within get_spectrum_mtm, the following are equal: 
     [10*log10(sum(abs(sig).^2,1))' 10*log10(sum(10.^(sig_fft_dB/10),1))']
 ans =
 	155.8193  155.1548
 	155.7985  155.2243
 	155.8041  155.3125
 	155.7904  155.3492
 	155.7719  155.3426
 	155.7733  155.3355
 	155.7757  155.3366
 	155.7324  155.3077
 	155.6346  155.2192
 	155.5392  155.0943
 	155.5087  154.9790

 [10*log10(sum(abs(sig).^2,1)/fs)' 10*log10(sum(10.^(sig_psd_dB/10),1))']
 ans =
 	114.8502  114.1857
 	114.8294  114.2552
 	114.8350  114.3434
 	114.8213  114.3801
 	114.8028  114.3735
 	114.8042  114.3664
 	114.8066  114.3675
 	114.7633  114.3386
 	114.6655  114.2501
 	114.5701  114.1252
 	114.5396  114.0099

 [10*log10(sum(10.^(sig_psd_dB(idx_sig,1)/10))*mean(diff(sig_freq_vec)));...
 10*log10(sum(pxx(idx_pxx,1))*mean(diff(f)))]
 ans =
 	117.9556
 	118.2376

2017/02/01-08

  • Started re-writing the manuscript
  • Updating split-window and energy normalizer code to work with new data format:
    • normalizer_split_window
    • normalizer_energy
  • For reasons currently unknown, the energy normalizer has "rings" of elevated levels throughout range, corresponding to the locations of nulls in the non-pulse compressed but beamformed time series. LOOK INTO THIS WHEN HAVE TIME!
  • The energy normalizer also requires using all data instead of just the subset. This makes processing a lot slower... for now only use the split-window normalizer for the paper.

2017/02/15

  • Functions to plot echogram:
    • plot_small_echogram: can be called to plot small scale echogram on a given axes handle
    • plot_large_echogram: not a function, can be used to plot full scale echogram
  • Try echo_info on multiple overnight sessions
    • run 103: ping 155 and 160 seem stronger than other pings by a few dB??
    • run 79: fish spread toward southern side of the shipwreck
  • Remember that run 94 pings after 573 have not been beamformed
  • Remember to change the part re. ping_num swicth in get_SL

2017/09/07

  • Recap to-do's for TREX fish echo paper:
    • Coherent averaging over night-time frames to determine the actual strength of shipwreck (by assuming water column variability is negligible and that fish influence averages out)
    • re-do all analysis with matching sizes of AW1 and AW2