---draft---

Nathaniel B. Palmer:
Ocean Surveyor 38KHz Data Report

Jules Hummon, Eric Firing
University of Hawaii
http://currents.soest.hawaii.edu

Introduction

An RD Instruments Ocean Surveyor ADCP (OS-38) operating at 38 kHz was installed on the RVIB NB Palmer in drydock in Auckland in September, 2004. It was first used on the short transit and test cruise from Auckland to Lyttelton, Oct 4 to 8. This was sufficient to establish that the instrument was working correctly, and to begin to discover the characteristics of the installation. Here we give a short description of the OS-38 system and its performance as seen on the test cruise; the reader is cautioned that only a small range of conditions was encountered, and much more will be learned from future cruises. As a point of comparison, we also discuss the results from the RD Instruments narrowband ADCP (NB-150), which has been in operation on the NB Palmer for many years, and which is now operated in parallel with the OS-38.

The motivation for installing the OS-38 is that profiling range varies inversely with frequency, to a first approximation. Operating at 150 kHz, the NB-150 on the Palmer can profile to about 300 m under good conditions; without the thick acoustic window required to protect the transducer from ice, it might reach 450 m in Antarctic waters. The OS-38, however, is capable of reaching 1600 m under ideal conditions: bubble-free water, low noise, good scattering. Although conditions encountered by the Palmer may rarely if ever be good enough to yield this maximum range, even the short test cruise has shown that 1200-m range is achievable. It is clear that the OS-38 provides a major improvement to the scientific capability of the NB Palmer.

Although it cannot approach the range of the OS-38, the NB-150 remains an important part of the NB Palmer's current profiling system. The inherent tradeoffs for range are vertical resolution, ping rate, and short-term accuracy. Hence the NB-150 complements the OS-38 by yielding better resolution and accuracy in the top 300 m of the water column, and by profiling somewhat closer to the ship's hull. As it turns out, the NB-150 on the NB Palmer also seems to be less affected by aerated water than the OS-38, so there may be adverse conditions under which it alone will yield valid velocity profiles. This remains to be seen, however.

Instruments: Installation and setup

The OS-38 and NB-150 are both installed in transducer wells filled with a mixture of propylene glycol and freshwater. A thick (about an inch) acoustically transparent window protects the transducers from the ice. Neither instrument is accessible unless the ship is in drydock, since instruments are considered to be on the ocean side of the well's watertight seal. More details of the Palmer OS-38 installation will be available later (Raytheon report).

ADCPs use a Doppler frequency shift along each of 2 pairs of opposing beams to determine horizontal and vertical velocities. The OS-38 is a phased array transducer, generating these beams from a grid of much smaller transducer elements. The NB-150 has 4 fixed transducers. Beams from both instruments are oriented at 30deg up from the vertical, and rotated about 45deg starboard of the ship's bow. The much newer phased array electronics are capable of transmitting the 38KHz pulse in one of two modes: "broadband", a slightly wider bandwidth signal in which the carrier frequency is modulated with a pseudo-random code, and "narrowband", which is a more tightly focused and uncoded beam. The broadband setting results in higher short-term accuracy (or better vertical resolution) at the expense of vertical penetration; the narrowband setting requires greater vertical averaging (hence lower resolution) but appears to penetrate deeper into the water column. The NB-150 can only produce narrowband pings.

In both instruments, the following parameters can be set to configure data acquistion:

blank (meters)	how long to wait for reverberations to die down before interpreting the signals as velocity
bin size (meters)	thickness of each averaging interval (meters in the vertical)
pulse size (meters)	how long the signal lasts before starting to interpret the return
bottom track (on or off)	devote one profile per cycle to finding the bottom. good for calibration if the instrument is in shallow enough water; undermines statistical accuracy of an averaging interval
ping type	broadband (OS-38) narrowband (OS-38 or NB-150) interleaved (OS-38)

Because the characteristics of the OS-38 broadband and narrowband pings may be quite different, the "interleaved mode" was chosen for testing and inital logging. Blanking of 16m was used for both NB-150 and OS-38. Data were collected with 8m, 12m, and 16m bins for NB-150, OS-38 (broadband mode) and OS-38 (narrowband mode), respectively.

Data: Logging and Processing

Data are logged using software written at the University of Hawaii ("UHDAS") on a 1-U rackmount Dell server with a Digi Neo 8-port serial card, running the most recent version of the Mandrake Linux distribution. UHDAS controls all tasks related to data acquisition, archiving, processing, display, and reporting. It is made up of routines written in Python, C, and Matlab. UHDAS provides a gui interface in which a technician can start or stop a cruise, start or stop data acquisition, and change ADCP acquisition parameters. By communicating directly with both ADCPs and logging only one set of ancillary data for use with both instruments, raw data volume is kept to a minimum. Daily emails are sent with a sample of recent data, as well as the status of logging, system, processing, and display. Raw (ADCP and ancillary) data and processed ocean velocities are available on a web site on the ship's network, which is hosted by the UHDAS computer. UHDAS provides regularly updated figures and data, available through this web site.

Ancillary data are needed in processing to correct the timestamps of the raw ADCP data and to provide position and heading. All raw data components must be reliable and of high quality or the final dataset will suffer. These ancillary serial inputs are made available from Black Box splitters direct from the instruments, and have been reliable. The serial inputs used for both instruments are a Trimble GPS reciever, the ship's gyro, and the Seapath. In addition a soundspeed probe in the NB-150 transducer well allows us to correct for speed of sound at the transducer (a correction that is unnecessary for the OS-38).

Reliable heading is obtained from the gyro, but accurate headings are necessary for an ADCP dataset to be of sufficient quality to be used for science. An Ashtech ADU2 unit was in use through this cruise but we have switched over to using the Seapath exclusively for a heading correction to the gyro. This move was prompted by three factors: (1) the Seapath has worked well for the last year, (2) a complete set of Seapath spares will be present on the Palmer, and (3) the L.M.Gould ADU2 unit was in need of replacement parts, so the ADU2 from the Palmer was moved the Gould.

Cruise Summary

The ship left Auckland Oct 4 and headed SE with strong and gusty tailwinds. After rounding the East Cape corner, the local winds decreased. The bow, which had been trimmed about 2' high, was lowered to neutral by moving ballast at about the same time as the ship moved into the lee of the South Island. OS-38 data quality was worse prior to this transition, but it is unclear whether the bow trim was related to the improvement. During the remainder of the trip to Lyttelton, ADCP and Dynamic Positioning tests were undertaken. The cruise track and test locations are shown in Figure 1.

The weather and the cruise track allowed ample opportunity to collect data under limited sea states and varying depths. Four additional ADCP tests were conducted: (1) a speed test, to investigate the extent to which ship's speed affected data quality (2) a heading test (conducted at two speeds), to see if heading made a difference in data quality, (3) an acoustic interference test (to observe the effect of various other sonar devices on the ADCP data), and (4) a reciprocal track over 320-350m to obtain calibration offsets for both NB-150 and OS-38. The calibration run was a simple matter of data processing to obtain calibration values. The other three tests were more exploratory in nature, providing valuable information but no quantitative results.

Data Quality

Overview:

Single ping data are affected by beam width, ringing (acoustic reverberation inside the transducer well), background noise (eg. from mechanical vibration or flow conditions), acoustic interference (from other sonars), and acoustic blocking (eg. bubbles, aerated water, or ice). Because the beam width is a characteristic of the instrument design, it will not be discussed in this report.

Ringing can be caused by reverberation within the well, which is exacerbated if there is an air bubble within the well or if the window is not horizontal relative to the instrument. A blanking range of 16m was needed to remove the symptoms of ringing from both instruments. This value is the default for the OS-38. Although 16m is large for an NB150 it is the value we have used for the last several years. (Prior to the addition of acoustic tiles in the NB-150 transducer well several years ago, it was even greater).

Acoustic interference was visible in the signal return strength but in most cases, because the frequencies are so different, they did not seem to adversely affect the velocity data. The 3.5 KHz Bathy chirp was most visible but was automatically edited out and so caused no problem. There is the potential for the 38KHz setting of the EK500 to disrupt the OS-38 data: in general it should be secured.

Acoustic blocking and flow noise are generated through a complex interaction of factors, including hull design, ship speed, sea state, heading relative to seas, wind strength and direction (relative to ship, and to seas), tranducer well design (fluid-filled well, mechanical coupling to the hull, presence of a window, rough edges), transducer placement (the bow gets more bubbles), underhull protrusions and roughness, nearby gratings, holes, or flows (eg. bowthruster), to name a few.

Background or flow noise while underway resulted in a decrease in range of the affected profile, but under the conditions of this cruise, did not affect the remaining data in the profiles. On station, the predominate issue was the bow-thruster, which is located less than 10m from the OS-38 and frequently disturbs the water beneath the ship. On-station data, especially from the OS-38, are likely to be spotty.

The largest underway effect on single-ping data came from acoustic blocking (presumably from bubble sweepdown or aerated water). Acoustic blocking renders useless all returns from the outgoing signal. It results in a loss of data at all depths, with the critical exception of a few highly biased bins at the top. (These must be edited out prior to averaging or the final data quality can suffer greatly). The OS-38 was severely affected by this issue; the NB-150 less so (on this cruise). This will be discussed below.

Once the single-ping data have been merged with the ancillary serial streams and the bad values edited out, the data are averaged and undergo various post-processing routines. A preliminary version of this ocean velocity dataset is available in the form of plots and data files generated by UHDAS processing and is accessible through the ship's web. Two examples of the final ocean data are shown in figures 2 and 3 (at 40 S and 42.5 S), showing the best depth coverage for each , and good agreement between the instruments where they overlap.

Bubbles and Data quality

When acoustic blocking from bubbles or ice occurs, most of the profile is thrown away by the instrument, but a few bins near the top may be left with measured velocities near zero.
When bubbles blocked the signal in the OS-38, data from the top two bins (in narrowband mode) or the top 4 12-m bins (in broadband mode) were affected. If the ship is underway, these short profiles can severely bias the processed data if they are not removed prior to averaging. Therefore, editing criteria require OS-38 profiles to have valid velocities below the top 48m or the profiles will be removed. From earlier experience in the ice, the NB-150 is known to show the same behavior, also requiring that short profiles be excluded.

The presence of bubbles, the "short biased profiles" phemomenon, and its effect on calculated ocean velocity can be seen in figure 4. The top panel is the NB-150 return signal strength, the middle panels are the OS-38 (broadband mode) and the bottom panels are the OS-38 (narrowband mode). The time range is the same, and shows a short period during which the heading changed by 90deg. The two lower left panels are signal return strength; the two lower right side panels are the beam 1 measured velocity.

The NB-150 return signal strength shows a small bubble event at about 4.5 minutes (a vertical blue stripe) and several even smaller events (eg. at 6 minutes). The vertical rainbow slash marks during this time are acoustic interference from another sonar.

The OS-38 signal strength shows the ocean scattering strength during the first 3.5 minutes; high near the surface, decreasing with depth, with a subsurface scattering layer at about 400-450m, and a reflection off the bottom at about 2000m. From about 4-6.5 minutes the signal is alomst entirely blocked (dark blue). There is a transition during minute #3 in which a decreased signal strength is dicernible. The beam velocities (right panels) break down in to two time periods which correspond to the "clean" versus "blocked" times seen in the return signal strength. During the first 3 minutes, the profiles reach to the bottom of the range and are vertically coherent. During the second period (4-6.5 minutes) the profiles are divided into two regions: the top few bins, and the rest. The top few bins are more coherent in time, with nearly uniform values near zero, and are decoupled from the deeper parts of the profiles. These bins will result in a bias in the calculated ocean velocity and must be removed prior to calculating ocean velocity. Below the top few biased bins, almost no profiles reach full depth, and some are missing entirely. This figure suggests that acoustic blocking is correlated with (1) bias in the top few bins and (2) shorter profiles.

Experience with two other OS-38 instruments shows that acoustic blocking is not inherent in the instrument. A cruise ship with a flat hull and a fast underway speed (>20kts) suffers badly from underway acoustic blocking, whereas a swath ship with deep draft is almost unaffected by underway bubbles.

Between Auckland and East Cape, about 5% NB-150 profiles and about 20% of OS-38 profiles were short, biased profiles. Later in the cruise, the number of short biased profiles was under 5% for the OS-38 and was negligible for the NB- 150. It is not clear to what extent this shift was due to a change in sea state, winds, and heading as opposed to the neutral ballast of the bow which also occurred at about the same time. Failure to adequately screen out short profiles west of East Cape leads to an error in OS-38 velocities of over 40cm/s.

Two bubble tests

Fundamentally, for a given installation, the relationship between ship's heading, ship's speed, wind speed and direction, and sea state play a roll in generating bubbles, but quantifying those effects is not trivial. Two tests were undertaken to gain experience with the instrument under different environmental conditions. The conclusions are of limited use.

During the transit from Auckland and East Cape, while OS-38 broadband profiles suffered from bubbles, a speed test was undertake to see how slow we would have to go to eliminate the presence of short biased profiles. Slowing to 9kts and subsequently to 6 kts for 10 minutes each had no effect. Only at 3kts were OS-38 data clean. In this example, a drastic slowing of the ship was required to get unbiased OS-38 data.

A test of heading relative to wind was also performed. The winds were moderate to low, and the ship was asked to steam in a box pattern, 10 minutes on each side, starting by going downwind. Immediately after that box, the test was repeated with the wind starting off the port quarter. The results of this test are shown in figure 5. The top panel shows ship heading and speed, with downwind marked. The second panel shows the signal return strength from deep water. The unblocked value in this example should be about 80, but when the signal is blocked, it decreased to 20. The bottom panel shows the number of bins in each profile, color coded by the maximum deviation from ocean velocity. It is clear that shorter profiles are more likely to contain high deviations (red and green dots). Full-depth profiles are more likely to contain no high deviations. By compararing the second and third panels, one can see that high signal return is correlated with good velocity data (low deviation). For this test, it also seems that at 4kts, downwind headings are more likely to result in clean velocity data; upwind (surprisingly) also gave clean data. Wind off the bow impacted the data negtively, and wind on the beam was detrimental. There is no quantitative result or recommendation, but that heading and seastate are obviously going to impact the data, even at slow speeds.

The second part of the test was to repeat the first box at full speed. Figure 6 shows that (under the conditions of this test), at cruising speed (10 kts) down wind was worse than 4kts downwind, and at any other heading the data are adversely affected. Some headings are worse than others, but even in only moderate wind and seas, at full speed, no heading is good.