%~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ % $Id: ProfileCycleModel.tex,v 1.2 2008/07/14 17:02:19 swift Exp $ %~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ % RCS Log: % % $Log: ProfileCycleModel.tex,v $ % Revision 1.2 2008/07/14 17:02:19 swift % Added compensator hyper-retraction feature to allow floats to be % parked at mid-water pressures. % % Revision 1.1 2006/11/03 19:08:57 swift % Added user manual to CVS control. % % Revision 1.3 2006/07/10 22:24:49 swift % Modifications to bring the manual up to date with % changes to the SeaBird CTD firmware (v1.1c). % % Revision 1.2 2006/04/10 16:11:51 swift % Added the subsection describing the deep-profile cycle. % % Revision 1.1 2005/12/27 23:29:06 swift % Initial revision % %~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ \newcommand{\Pn}{P$_{N_2}$} \section{Controlling \iridium\ \apex\ behavior: A parametric model.} \label{sec:ProfileCycleModel} The \iridium\ \apex\ firmware is highly configurable so that the user can control float behavior by adjusting the values of more than 20 parameters and by selecting several optional modes and features. \subsection{Sample missions.} \label{sec:SampleMissions} The ability to configure the float within a 20(plus) dimensional parameter space means that that range of possible float behaviors is practically infinite. However, some general characteristics span the whole parameter space while many potentially useful kinds of missions are excluded entirely. Figures~\ref{fig:P4P}, \ref{fig:P1P}, and \ref{fig:P254P} represent common mission cycles within the usable parameter space. Figure~\ref{fig:P4P} represents the most general kind of mission cycle and is referred to as \emph{Park-n-Profile} (PnP). The original motivation for PnP was as a mechanism to balance the competing objectives of energy savings versus direct measurement of salinity drift in the deeper water. The basic idea was to collect most profiles from the park level but occasionally execute a deep profile to facilitate evaluation of CTD performance. The ``$n$'' in PnP refers to the cycle length of the PnP mechanism; every $n^{th}$ profile is a deep profile. \begin{figure}[htbp] \begin{center} \leavevmode \mbox{\includegraphics[scale=0.5,clip,draft=false]{P4P.ps}} \caption{Schematic of a PnP mission with cycle length $n=4$. The park level is the same for all profiles. Every fourth profile is a deep profile. The shallow blip prior to the (special) first profile represents pressure activation.} \label{fig:P4P} \end{center} \end{figure} The first profile is special because it is executed immediately after the mission prelude, does not drift at the park level, and is also a deep profile. The first profile will be telemetered within 24~hours after the mission is activated. The exact timing will depend on the user's specific parameter selections. This feature was implemented to satisfy the often-heard request for a profile to be executed immediately after deployment. Figure~\ref{fig:P1P} represents a PnP mission with $n=1$ (ie., a P1P mission). In this way, PnP firmware can be programmed to collect lagrangian data from a shallower level while still being able to collect deep profiles. As with the P4P mission in Figure~\ref{fig:P4P}, the first profile is executed immediately after the mission prelude. \begin{figure}[htbp] \begin{center} \leavevmode \mbox{\includegraphics[scale=0.5,clip,draft=false]{P1P.ps}} \caption{Schematic of a PnP mission with cycle length $n=1$. Every profile parks shallow but profiles deep. The shallow blip prior to the (special) first profile represents pressure activation.} \label{fig:P1P} \end{center} \end{figure} Figure~\ref{fig:P254P} represents a degenerate case of the PnP model where $n$ is large and the park level has been chosen to be deep. This mission cycle is so common amongst \apex\ users that it was implemented as a special case. The value $n=254$ is a special sentinel value that disables the PnP feature so that only park-level mission parameters (ie., park pressure and park piston position) are used for controlling the profile cycle; the profile-level parameters (ie., profile pressure and profile piston position) are ignored. \begin{figure}[htbp] \begin{center} \leavevmode \mbox{\includegraphics[scale=0.5,clip,draft=false]{P254P.ps}} \caption{Schematic of a degenerate PnP mission with cycle length n=254. Every cycle parks and profiles from the park level. The shallow blip prior to the (special) first profile represents pressure activation.} \label{fig:P254P} \end{center} \end{figure} As with the previous two examples, the first profile is executed immediately after the mission prelude. \subsection{Deconstructing the profile cycle.} \label{sec:DeconstructProfileCycle} Details of the firmware architecture and design are outside the scope of this user manual. However, deconstructing the profile cycle into its constituent elements will give meaning to many of the configuration parameters. The \apf\ firmware design makes fundamental use of the concept of ``sequence points'' for controlling the flow of the profile cycle. A sequence point is defined to be a point where one phase of the mission cycle transitions to the next phase. Most of the sequence points are based on time but there are several sequence points that are event-based. Given a properly functioning \apf\ controller, the firmware guarantees the phase transition at each sequence point regardless of the health of any other float component. The schematics below illustrate different parts of the mission cycle: \begin{enumerate} \item The pressure-activation phase. \item The mission prelude. \item A profile from the park level. \item A deep profile. \end{enumerate} \subsubsection{Pressure-activation phase (\emph{optional}).} The pressure-activation feature is an optional phase of the mission that preceeds the mission prelude. It was designed to accomodate requests from ship's crew to be able to deploy the float without being required to start it with a magnet. One event-based sequence point is implemented that activates the mission prelude if the pressure exceeds the activation threshold (ie., 25~decibars). \begin{minipage}{6in} \begin{verbatim} |----------------- L ----------------------| -+----------------+-----+-------+----------------------------------+- Time P| . | . . r| . | . Sequence Points . e| . . ----------------------- . s| B . L = Mission prelude . s| B = Pressure-activation u| r| e| \end{verbatim} \end{minipage} Pressure activation mode is not ``on'' by default. The user must enable pressure activation mode via the interactive user interface (see Section~\ref{sec:MissionConfiguration}). Enabling the pressure activation mode immediately induces the firmware to perform a self-test of the float. In order for pressure activation to work the float has to be able to sink from the surface down to the activation pressure (ie., 25~decibars). Obviously, if the float is too buoyant to sink then it can never self-activate. Enabling pressure activation mode causes the firmware to put the float into a state of minimum buoyancy; the buoyancy pump is fully retracted to deflate the oil bladder and the air solenoid valve is opened to deflate the air bladder. Then the firmware enters a nonterminating loop where the the CTD is queried for pressure every two hours. If the pressure is less than the activation pressure then the \apf\ controller puts itself back to sleep for another two hours. However, if the pressure exceeds the activation threshold then the firmware launches the mission and enters the mission prelude. \subsubsection{The mission prelude.} The purpose of the mission prelude is mainly to allow the float to transmit a fix of its deployment location and to telemeter its mission programming parameters. The mission prelude is the time period between mission activation and the first descent. The sequence point 'L' is time-based and is the transition between the mission prelude and the first descent. The period of the mission prelude is user-defined (see Section~\ref{sec:MissionConfiguration}). \begin{minipage}{6in} \begin{verbatim} |--------------------------- L -----------------------------------| -+-----------------------------------------------------------------+- Time P| . r| Sequence Points . e| ------------------- . s| L = Mission prelude . s| u| r| e| \end{verbatim} \end{minipage} When the mission is launched either manually or else by the pressure activation mechanism, the firmware puts the float into a state of maximum buoyancy by fully extending the buoyancy pump to inflate the oil bladder and then inflating the air bladder. \subsubsection{Profile from park depth.} \label{sec:ParkProfileCycle} The profile cycle for a shallow profile consists of four phases (each associated with a time-based sequence point): descent (F), park (K), profile (P), and telemetry (C). Two additional event-based sequence points (S,T) ordinarily cause phase transition before their associated time-outs force the phase transition. \begin{minipage}{6in} \begin{verbatim} |--------------------------------------------------- C ----------------| |------------------------------ P ---------------| | |----------------- K -------------| | | |--- F -----| | S | T | -+-----------+---------------------+-----------+-------------------+----+- Time P|. | | . . r| . | | . Sequence Points . e| . | | . --------------- . s| . | | . F = Descent . s| . | | . K = Park u| . . . . . . . . . . . . S = Surface detect r| P = Profile e| T = Telemetry | C = Cycle \end{verbatim} \end{minipage} This is the simplest kind of profile cycle which \apex\ floats have executed since their initial development. The float sinks to its park depth, drifts for a period of time, profiles to the surface, and then telemeters its data. However, unlike \apex\ with ARGOS telemetry, the length of time for each profile cycle is not fixed. The profile cycle for \iridium\ \apex\ ends as soon as the telemetry is successfully completed or the profile cycle time-out (C) expires, whichever happens first. Typically, an \iridium\ float is on the surface for only 15 minutes or so before the next profile cycle begins. For this kind of profile cycle, the end of the park phase (K) coincides with the end of the ``down-time'' and the beginning of the ``up-time''. The sequence point C coincides with the end of the up-time. The (maximum) length of the profile cycle is the down-time plus the up-time. \begin{description} \item[Descent phase\label{DescentPhase}]---The profile cycle starts with the descent phase. The float queries the CTD for the surface pressure and then the buoyancy pump is retracted to the park position. The float sinks until the descent period expires (ie., the firmware forces a phase transition at the sequence-point~F in the schematic above). Hourly pressure measurements are logged as well as one at the completion of the buoyancy pump retraction. These pressure measurements are referred to as \emph{descent marks} and are telemetered as engineering data. \textbf{Special notes for floats with \NComp:\label{DescentPhaseNComp}} The gas pressure (\Pn) in the \NComp\ plays an important role during the descent phase. While the float is at the surface, the \NComp\ piston is fully extended which adds $\sim$80~grams of buoyancy to the float. In order to descend back down to the park level, the buoyancy pump must reduce the buoyancy enough to descend deeper than the gas pressure \Pn. To accomplish this, the piston is retracted beyond the park piston position. This is referred to as \emph{compensator hyper-retraction}. Near the end of the descent phase, the piston is extended back to the park position. This allows floats with \NComp\ to park at any depth greater than 850~dbars. \textbf{Warning:} The \NComp\ renders the float buoyantly unstable in the range of pressures $\sim$400--700~decibars. Therefore, such floats must be parked outside this range. \item[Park phase\label{ParkPhase}]---Active ballasting is accomplished and park-level PT measurements are collected during the park phase. The \apf\ wakes once each hour to accomplish these tasks. A PTS sample is collected at the end of the park phase. \begin{description} \item[Active ballasting:] The float wakes each hour to monitor the pressure and make buoyancy adjustments if three \textbf{consecutive} measurements violate a 10~decibar dead-band on either side of the user-specified park pressure. Measurements that are within the dead-band or that completely cross the dead-band reset the violation counter and will not induce buoyancy adjustments. \item[Park-level PT samples:] The float collects hourly low-power PT measurements and telemeters them as hydrographic data. Refer to Section~\ref{sec:PtSample} for their data format. Hourly salinity data are not measured due to energy considerations. \item[Park-level PTS sample:] The float collects one PTS sample at the end of the park phase (K). Refer to Section~\ref{sec:LowResPtsSample} for its data format. \end{description} \item[Profile phase\label{ProfilePhase}]---As might be expected, the profile phase is the most complicated of the profile cycle. Three asynchronous processes are active during the profile cycle: Ascent rate control, hydrographic sampling, and surface detection. These processes operate on a 10~second heartbeat; the \apf\ controller sleeps for 10~seconds then wakes to attend to these processes before going back to sleep. \begin{description} \item[Ascent rate control:] As an initialization step of the profile phase, the buoyancy engine adds a user-specified initial increment of buoyancy to start the float ascending toward the surface. Thereafter, the firmware monitors the pressure at 5~minute intervals to determine if the average ascent rate has been maintained above 0.08~decibars/sec. If the ascent rate falls below this threshold then the buoyancy engine adds a user-specified increment of buoyancy. \item[Hydrographic sampling:] \iridium\ \apex\ is designed to collect hydrographic profiles with relatively high vertical resolution. The \sbe\ \ctd\ has a continuous profiling (CP) mode that runs asynchronously and autonomously from the float's \apf\ controller. When CP-mode is active, the \ctd\ collects 1-Hz samples and stores them internally in nonvolatile memory. The \apf\ shuts down CP-mode 4~decibars below the surface to avoid contaminating the conductivity cell with ingested surface scum. To protect against pressure-sensor drift, the \apf\ commands the \sbe\ to shut down 4~decibars deeper than the most recent surface pressure measurement. As a fail-safe measure, the \sbe\ will shut itself down when its pressure sensor reaches 2~decibars (but no attempt is made to account for drift of the pressure sensor). After the float reaches the surface the \apf\ commands the \sbe\ to sort the 1-Hz samples into 2~decibar bins and compute the arithmetic mean of the samples in each bin. The resulting bin-averaged profiles have 2~decibar resolution though we often refer to them as ``high resolution'' or ``continuous profiles''. These high resolution profiles are telemetered using the format defined in Section~\ref{sec:HiResPtsSample}. The \sbe\ also has a spot-sampled mode that is roughly simlar to the Sbe41 used on ARGOS \apex\ floats. This \iridium\ firmware implements an optional mixed-mode sampling strategy in order to save energy in the deep water where gradients are small. At the user's discretion, the float can be programmed to collect spot samples in the deep water and automatically transistion to continuous profiling when the float ascends to the \emph{CP Activation Pressure}. To disable this feature and force continuous profiling from top to bottom then the user should specify the activation pressure to be deeper than the float's operating range. Spot-samples are formated according to Section~\ref{sec:LowResPtsSample} and collected according to a pressure table that is hard-coded in firmware. \item[Surface detection:] The surface detection algorithm terminates true when the float ascends to a pressure that is within 4~decibars of the most recent surface pressure measurement\footnote{Actually, the surface detection algorithm is more complicated than this but the complications handle pathological situations. Refer to the C source code (src/profile.c) in your distribution.}. Surface detection causes the profile to be terminated, another increment of buoyancy to be added by the buoyancy pump, and transition to the telemetry phase. \end{description} \item[Telemetry phase\label{TelemetryPhase}]---If the end of the antenna is even a centimeter below the water's surface then telemetry will not be possible. Therefore, the telemetry phase starts with precise surface detection using its SkySearch algorithm. The heart of this algorithm involves attempting to register the LBT with the \iridium\ system. If the algorithm terminates true then GPS acquisition and telemetry can proceed; otherwise, the buoyancy engine adds another increment of buoyancy and sleeps for one telemetry-retry period before repeating the attempt. After determining that the float can see the sky then the LBT is shut down and the GPS engine is used to acquire the float's location. After acquiring the fix then the GPS is shut down and the LBT is reregistered with the \iridium\ system. The float places a call via the \iridium\ system to the remote host and logs in using the float's username and password. Once logged into the remote host, the float downloads its new configuration from the remote host and uploads its hydrographic and engineering data to the remote host. Finally, the float logs out of the remote host and reprograms its mission paramters according to the new configuration file that it just received from the remote host. \end{description} \subsubsection{Deep profile.} \label{sec:DeepProfileCycle} For PnP cycle lengths less than 254, the first profile cycle will be a deep one as will profile cycles for which the internal profile counter (PrfId) is an integral multiple of the profile cycle length. For example, a P4P mission will execute a deep profile cycle when the internal profile counter is 1, 4, 8, 12, and so on. The following C source code represents the test executed by firmware to determine if the current profile cycle is a deep one: \begin{verbatim} (PnpCycleLength<254 && ((!(PrfId%PnpCycleLength)) || PrfId==1) ? Yes : No; \end{verbatim} The profile cycle for a deep profile consists of five phases (each associated with a time-based sequence point): descent (F), park (K), deep-descent (D), profile (P), and cycle (C). Three additional event-based sequence points (Q,S,T) ordinarily cause phase transition before their associated time-outs force the phase transition. \begin{minipage}{6in} \begin{verbatim} |------------------------------------------------------ C -------------| |--------------------------------- P ------------| | |----------------- D --------| | | |------- K -------------| | | | |--- F -----| | | S | T | -+-----------+-----------+----+----------------+-------------------+------- Time P|. | | | . . r| . | | | . Sequence Points . e| . | | | . --------------- . s| . | | | . F = Descent . s| . | | | . K = Park u| . . . . . . . | . Q = DeepProfile r| | . D = DeepDescent e| . |. S = SurfaceDetect | . P = Profile | . T = Telemetry | Q C = Cycle \end{verbatim} \end{minipage} For this kind of profile cycle, the end of the deep-descent phase (D) coincides with the end of the ``down-time'' and the beginning of the ``up-time''. The sequence point C coincides with the end of the up-time. The (maximum) length of the profile cycle is the down-time plus the up-time. \begin{description} \item[Descent phase]---Refer to description on page~\pageref{DescentPhase}. \item[Park phase]---Refer to description on page~\pageref{ParkPhase}. The park phase is shortened for the deep profile cycle in order to allow time to descend from the park level to the deep target pressure. \item[Deep descent phase]---The purpose of the deep descent phase is to allow the float to descend from the park level (eg., 1000~dbars) to the pressure where the deep profile should begin (eg., 2000~dbars). The maximum time allowed for the deep descent phase is user specified (see Section~\ref{sec:MissionConfiguration}). The deep descent phase begins by retracting the piston from the park piston position to the profile piston position. During the descent, the pressure is monitored every five minutes to determine of the target pressure has been reached (ie., sequence point Q). \begin{description} \item[Sequence point Q]---If the float descends to its deep target pressure before the deep descent phase times out then the profile piston position is incremented by one count (but no piston extension occurs). The intent is to reduce the descent speed during the next deep profile so that the float will reach the target pressure closer to the end of the down time (ie., sequence point D). \item[Sequence point D]---If the deep descent period times out before the target pressure is reached then the profile piston position is decremented by one count (but no piston retraction occurs). The intent is to increase the descent speed during the next deep profile so that the float will reach the target pressure before the end of the down time (ie., sequence point D). \end{description} Transition to the profile phase is forced at either sequence point Q or D. \item[Profile phase]---Refer to description on page~\pageref{ProfilePhase}. \item[Telemetry phase]---Refer to description on page~\pageref{TelemetryPhase}. \end{description} %%% Local Variables: %%% mode: latex %%% TeX-master: "IridiumApex" et %%% End: