Read an ADV data file, producing an object of type adv. This
function works by transferring control to a more specialized function,
e.g. read.adp.nortek() and read.adp.sontek(),
and in many cases users will find it preferable to either use these or
the several even more specialized functions, if the file type is
known.
read.adv.sontek.text(
file,
from = 1,
to,
by = 1,
tz = getOption("oceTz"),
originalCoordinate = "xyz",
transformationMatrix,
longitude = NA,
latitude = NA,
encoding = "latin1",
monitor = FALSE,
debug = getOption("oceDebug"),
processingLog = NULL
)An adv object that contains measurements made with an ADV device.
The metadata contains information as given in the following table.
The Nortek name'' is the name used in the Nortek System Integrator Guide (reference 1) and the Sontek name'' is the name used in the relevant
Sontek documentation. References are given in square brackets.
metadata name | Nortek name | Sontek name | Meaning |
manufacturer | - | - | Either "nortek" or "sontek" |
instrumentType | - | - | Either "vector" or "adv" |
filename | - | - | Name of data file(s) |
latitude | - | - | Latitude of mooring (if applicable) |
longitude | - | - | Longitude of mooring (if applicable) |
numberOfSamples | - | - | Number of data samples in file |
numberOfBeams | NBeams (reference 1, p18) | - | Number of beams (always 3) |
numberOfBeamSequencesPerBurst | NPings | - | number of beam sequences per burst |
measurementInterval | MeasInterval (reference 1 p31) | - | |
samplingRate | 512/(AvgInterval) (reference 1 p30; reference 4) | - #' |
The data list contains items with names corresponding to adp
objects, with an exception for Nortek data. Nortek instruments report some
things at a time interval that is longer than the velocity sampling, and
these are stored in data as timeSlow, headingSlow,
pitchSlow, rollSlow, and temperatureSlow; if burst
sampling was used, there will also be items recordsBurst and
timeBurst.
The processingLog is in the standard format.
a connection or a character string giving the name of the file
to load. It is also possible to give file as a vector of filenames,
to handle the case of data split into files by a data logger. In the
multi-file case, header must be FALSE, start must be a
vector of times, and deltat must be provided.
index number of the first profile to be read, or the time of
that profile, as created with as.POSIXct() (hint: use
tz="UTC"). This argument is ignored if header==FALSE. See
“Examples”.
indication of the last profile to read, in a format matching that
of from. This is ignored if header==FALSE.
an indication of the stride length to use while walking through
the file. This is ignored if header==FALSE. Otherwise, if this is
an integer, then by-1 profiles are skipped between each pair of
profiles that is read. This may not make much sense, if the data are not
equi-spaced in time. If by is a string representing a time interval,
in colon-separated format, then this interval is divided by the sampling
interval, to get the stride length. BUG: by only partially
works; see “Bugs”.
character string indicating time zone to be assumed in the data.
character string indicating coordinate system, one
of "beam", "xyz", "enu" or "other". (This is
needed for the case of multiple files that were created by a data logger,
because the header information is normally lost in such instances.)
transformation matrix to use in converting beam
coordinates to xyz coordinates. This will over-ride the matrix in the file
header, if there is one. An example is rbind(c(2.710, -1.409,
-1.299), c(0.071, 2.372, -2.442), c(0.344, 0.344, 0.344)).
optional signed number indicating the longitude in degrees East.
optional signed number indicating the latitude in degrees North.
a character value that indicates the encoding to be used for
this data file, if it is textual. The default value for most functions is
"latin1", which seems to be suitable for files containing text written in
English and French.
boolean value indicating whether to indicate the progress
of reading the file, by using txtProgressBar() or otherwise. The value
of monitor is changed to FALSE automatically, for non-interactive
sessions.
a flag that turns on debugging. The value indicates the depth
within the call stack to which debugging applies. For example,
read.adv.nortek() calls read.header.nortek(), so that
read.adv.nortek(...,debug=2) provides information about not just the
main body of the data file, but also the details of the header.
if provided, the action item to be stored in the log. This parameter is typically only provided for internal calls; the default that it provides is better for normal calls by a user.
Sampling-rate and similar issues
The data format is inferred from the System Integrator Guide (reference 1A) and System
Integrator Manual (reference 1B). These document lacks clarity in spots, and so
read.adv.nortek contains some assumptions that are noted here, so
that users will be aware of possible problems.
A prominent example is the specification of the sampling rate, stored in
metadata$sampingRate in the return value. Repeated examination of
the System Integrator Guide (reference 1) failed to indicate where this value is
stored in the various headers contained in Vector datasets. After some
experimentation with a few data files, read.adv.nortek was set up to
calculate metadata$samplingRate as 512/AvgInterval where
AvgInterval is a part of the User Configuration'' header (reference 1 p30), where the explanation is average interval in seconds''). This formula was
developed through trial and error, but it was confirmed later on the Nortek
discussion group, and it should appear in upcoming versions of (reference 1).
Another basic issue is the determination of whether an instrument had
recorded in continuous mode or burst mode. One might infer that
TimCtrlReg in the User Configuration'' header (reference 1 p30) determines this, in bits 1 and 2. However, this was the case in test files available to the author. For this reason, `read.adv.nortek` infers the mode by reverse engineering of data files of known configuration. The present version of `read.adv.nortek` determines the sampling mode from the ```NRecords`'' item of the Vector Velocity Data'' header, which seems
to be 0 for data collected continuously, and non-zero for data collected in
bursts.
Taking these things together, we come upon the issue of how to infer
sampling times for Nortek instruments. There do not seem to be definitive
documents on this, and so read.adv.nortek is based partly on
information (of unknown quality) found on Nortek discussion boards. The
present version of read.adv.nortek infers the times of velocity
observations differently, depending on whether the instrument was set to
record in burst mode or continuous mode. For burst mode, times stated in
the burst headers are used, but for continuous mode, times stated in the
``vector system data'' are used. On the advice found on a Nortek discussion
board, the burst-mode times are offset by 2 seconds to allow for the
instrument warm-up period.
Handling IMU (inertial measurement unit) data
Starting in March
2016, read.adv.nortek has offered some support for handling IMU
(inertial measurement unit) data incorporated into Nortek binary files. This
is not described in the Nortek document named System Integrator Guide'' (reference 1A) but it appeared in System Integrator Manual'' (reference 1B;
reference 1C). Confusingly, 1B described 3 varieties of data, whereas 1C does not
describe any of these, but describes instead a fourth variety. As of March
2016, read.adv.nortek handles all 4 varieties, because files in the
various schemes appear to exist. In oce, the varieties are named
after the byte code that flags them. (Variety c3 is the one
described in (reference 1C); the others were described in (reference 1B).)
The variety is stored in the metadata slot of the returned object as
a string named IMUtype.
For each variety, the reader is cautioned that strong tests have not been
performed on the code. One way to test the code is to compare with textual
data files produced by the Nortek software. In March 2016, an oce
user shared a dataset of the c3 variety, and this permitted detailed
comparison between the text file and the values inferred by
read.adv.nortek. The test suggested agreement (to within the
resolution printed in the text file) for velocity (v in the
data slot), signal amplitude (a), correlation (q),
pressure (p), the three components of IMU delta angle
(IMUdeltaAngleX etc), and all components of the rotation matrix
(IMUrotation). However, the delta velocity signals did not match,
with IMUdeltaVelocityX disagreeing in the second decimal place,
IMUdeltaVelocityY component disagreeing in the first, and
IMUdeltaVelocityZ being out by a factor of about 10. This is github
issue 893 (https://github.com/dankelley/oce/issues/893).
Variety c3 (signalled by byte 5 of a sequence being
0xc3) provides information on what Nortek calls DeltaAngle,
DeltaVelocity and Orientation Matrix. (Apart from the orientation matrix,
Nortek provides no documentation on what these quantities mean.) In the
object returned by read.adv.nortek, these are stored in the
data slot as IMUdeltaAngleX, IMUdeltaAngleY,
IMUdeltaAngleZ, IMUdeltaVelocityX, IMUdeltaVelocityY,
IMUdeltaVelocityZ, and IMUrotation, all vectors except the
last, which is a 3D array. In addition to these, IMUtimestamp is a
time-stamp, which is not defined in the Nortek documents but seems, from IMU
documents (reference 5), to be defined based on a clock that ticks once per 16
microseconds. Caution may be required in dealing with this timestamp, since
it seemed sensible in one test case (variety d3) but kept reseting to
zero in another (variety c3). The lack of Nortek documentation on
most of these quantities is a roadblock to implementing oce functions
dealing with IMU-enabled datasets
Variety cc (signalled by byte 5 of a sequence being
0xcc) provides information on acceleration, angular rotation rate,
magnetic vector and orientation matrix. Each is a timeseries. Acceleration
is stored in the data slot as IMUaccelX, IMUaccelY,
IMUaccelz. The angular rotation components are IMUngrtX,
IMUngrtY and IMUngrtz. The magnetic data are in
IMUmagrtx, IMUmagrty and IMUmagrtz. Finally,
IMUmatrix is a rotation matrix made up from elements named
M11, M12, etc in the Nortek documentation. In addition to all
of these, IMUtime stores time in seconds, with an origin whose
definition is not stated in reference 1B.
Variety d2 (signalled by byte 5 being 0xd2) provides
information on gyro-stabilized acceleration, angular rate and magnetometer
vectors. The data stored MUaccelX, IMUangrtX,
IMUmagrtX, with similar for Y and Z. Again, time is in
IMUtime. This data type has not been tested as of mid-March 2016,
because the developers do not have a test file with which to test.
Variety d3 (signalled by byte 5 being 0xd3) provides
information on DeltaAngle, DeltaVelocity and magnetometer vectors, stored in
IMUdeltaAngleX, IMUdeltaVelocityX, and
IMUdeltaMagVectorX, with similar for Y and Z. Again,
time is in IMUtime. This data type has not been tested as of
mid-March 2016, because the developers do not have a test file with which to
test.
The file argument does not actually name a file. It names a basename
for a file. The actual file names are created by appending suffix
.hd1 for one file and .ts1 for another.
This file type, like other acoustic-Doppler types, is read with a hybrid R/C++ system, for efficiency. The processing steps are sketched below, for users who want to inspect the code or build upon it.
In R, readBin() is used to insert the file contents into
a vector of type raw.
In C++, this raw vector is scanned byte by byte,
to find the starting indices of data "chunks", or subsections of
the data that correspond to individual sampling times.
Checksum computations are also done at this stage, to detect
possible data corruption. Warnings are issued for any bad chunks,
and they are skipped in further processing. The valid
starting points are then passed back to R as a vector of
type integer.
In R, readBin() is used to read the components of each chunk.
For speed, this is done in a vectorized fashion. For example,
all the velocities in the whole file are read in a single call
to readBin(). This process is done for each of the data
fields that are to be handled. Importantly, these readBin()
calls are tailored to the data, using values of
the size, endian and signed parameters that are tailored
to the structure of the given component.
Scaling factors are then applied as required, to convert the
components to physical units.
Finally, in R, the acquired items are inserted into the data or
metadata slot of the return value, according to oce convention.
Dan Kelley
Files without headers may be created in experiments in which a data logger was set up to monitor the serial data stream from an instrument. The lack of header information places a burden on the user, who must supply such basic information as the times of observations, the instrument orientation, the instrument coordinate system, etc. Example 3 below shows how to deal with such files. Three things should be noted.
The user must choose the appropriate read.adv variant
corresponding to the instrument in question. (This is necessary because
oceMagic(), which is used by the generic read.oce()
routine, cannot determine the type of instrument by examining a file that
lacks a header.)
The call to the read function must include a start time
(start) and the number of seconds between data (deltat),
again, because the instrument data stream may lack those things when the
device is set to a serial mode. Also, of course, it is necessary to set
header=FALSE in the function call.
Once the file has been read in, the user will be obliged to specify
other information, for the object to be well-formed. For example, the
read function will have no way of knowing the instrument orientation,
the coordinate system being used, the transformation matrix to go from
"beam" to "xyz" coordinates, or the instrument heading, pitch,
and roll, to go from "xyz" coordinates to "enu" coordinates.
Such things are illustrated in example 3 below.
In ADV data files, velocities are coded to signed 2-byte integers, with a
scale factor being used to convert to velocity in metres per second. These
two facts control the maximum recordable velocity and the velocity
resolution, values that may be retrieved for an ADV object name d
with d[["velocityMaximum"]] and d[["velocityResolution"]].
1A. Nortek AS. System Integrator Guide (paradopp family of products). June 2008. (Doc No: PSI00-0101-0608). (Users may find it helpful to also examine newer versions of the guide.)
1B. Nortek AS. System Integrator Manual. Dec 2014.
(system-integrator-manual_Dec2014_jan.pdf)
1C. Nortek AS. System Integrator Manual. March 2016.
(system-integrator-manual_Mar2016.pdf)
SonTek/YSI ADVField/Hydra Acoustic Doppler Velocimeter (Field) Technical Documentation (Sept 1, 2001).
Appendix 2.2.3 of the Sontek ADV operation Manual, Firmware Version 4.0 (Oct 1997).
Nortek Knowledge Center (http://www.nortekusa.com/en/knowledge-center)
A document describing an IMU unit that seems to be close to the one named
in (references 1B and C) as being an adjunct to Nortek Vector systems is at
http://files.microstrain.com/3DM-GX3-35-Data-Communications-Protocol.pdf
Other things related to adv data:
[[,adv-method,
[[<-,adv-method,
adv,
adv-class,
advSontekAdrFileTrim(),
applyMagneticDeclination,adv-method,
beamName(),
beamToXyz(),
enuToOther(),
enuToOtherAdv(),
plot,adv-method,
read.adv(),
read.adv.nortek(),
read.adv.sontek.adr(),
read.adv.sontek.serial(),
rotateAboutZ(),
subset,adv-method,
summary,adv-method,
toEnu(),
toEnuAdv(),
velocityStatistics(),
xyzToEnu(),
xyzToEnuAdv()
if (FALSE) {
library(oce)
# A nortek Vector file
d <- read.oce("/data/archive/sleiwex/2008/moorings/m05/adv/nortek_1943/raw/adv_nortek_1943.vec",
from=as.POSIXct("2008-06-26 00:00:00", tz="UTC"),
to=as.POSIXct("2008-06-26 00:00:10", tz="UTC"))
plot(d, which=c(1:3,15))
}
Run the code above in your browser using DataLab