Sensor Configuration¶

A Seaglider talks to its science sensors in one of three ways. Each needs configuring on the glider and on the basestation, and most "sensor failures" in the first days of a mission are really configuration problems on one side or the other.

Three ways in¶

On Rev E, none of the serial sensors are built in: everything goes through serdev or logdev with a .cnf file. A self-test that lists only the CT after a board change is the typical symptom of missing .cnf files. The manufacturer's factory Rev E builds don't use SciCon.

For sensors too complex for a .cnf (odd protocols, hex output, lots of data), operators have used a small interface board between the sensor and the glider (e.g. a "Smart Cable"). The glider then talks to a simple serial device and the board handles the sensor.

The library and the slots¶

Two separate things have to be set up, and they're independent:

1. The library — which .cnf files the glider knows about. Add, list and remove with seradd / serlib / serdel (serial sensors) and logadd / loglib / logdel (loggers), under param/config.
2. The slots — which sensor is attached to which port. These are the $DEVICEn and $LOGGERDEVICEn parameters, set with Configure sensor / Configure logger sensor in param/config.

A sensor only appears as a choice in step 2 once its .cnf is in the library. If Configure sensor offers only the CT and "not installed", the library is empty, not the firmware. Setting $DEVICEn or $LOGGERDEVICEn directly (by hand or in a cmdfile) runs the same setup as the menu, but only works if the .cnf is already in the library. Otherwise the glider rejects or resets the value.

Getting files onto the glider¶

• On the cable: from PicoDOS, transfer with XMODEM (xr) and then run strip1a to remove the transfer padding. Make sure the target isn't read-only.
• If XMODEM keeps failing ("0 files transferred"), try YMODEM (yr). In one case repeated xr attempts failed and yr worked first time.
• Rename the old file before overwriting, and check that the new one arrived before you delete anything. One failed transfer after a rename left the glider with no calibration file at all.
• Remotely, the same menu actions can be put in pdoscmds.bat, for example menu param/config/loglib or menu param/config/logadd device=0 file=gpctd.cnf. Send one command per call and read the result before sending the next.

.cnf gotchas¶

• Trailing blank lines matter. An optode .cnf with two blank lines at the end returned "got 1 of 5 columns", then "got 0 of 5 columns", even though the sensor streamed good data in direct comms. The fix was exactly one blank line at the end.
• Start-up chatter. A sensor that prints a banner or mode line on power-up (e.g. "MODE RS232") can confuse the parser. Check the sensor's own configuration in direct comms, not just the .cnf.
• The prompt must match. The glider waits for the prompt string in the .cnf. If the sensor's settings have changed (for example, a GPCTD with its "executed" tag turned on), the glider never sees the prompt and reports "no prompt detected". Either change the sensor setting or change prompt= in the .cnf.
• Warm-up and timeouts. Too short a warm-up gives empty first samples. You can adjust it without editing the file (edit warmup=… in the sensor's hardware menu).
• Don't ask a logger for depth on early Rev E firmware. On one Rev E build, a logdev .cnf that sent the depth (%D) in its start/stop strings made the glider crash and reboot (a bus fault) the moment the string was sent. Removing %D fixed it.
• Power cycling. Some sensors must not be powered off between samples (the RBR legato is one). Newer serdev drivers have .cnf options for this (power-policy, and cycles for frequency-counting instruments). They have been in APL-UW's Rev E firmware since about 2020 but not in Rev B builds. On Rev B, run such a sensor as a logdev instead, or slow sampling right down.
• voltage= and current= are bookkeeping. They tell the fuel gauge which battery to charge the sensor to, and how much. They don't change the supply. See Batteries. A CURRENTS file entry is optional if the .cnf has a current value.

The basestation side¶

The glider can log a sensor perfectly and the basestation can still drop it on the floor.

• The CTD is special. The basestation needs to find one of the CTD types it knows, in the units it expects. Otherwise most processing stops ("No CT data found"), including the flight model and the derived quantities. Basestation2 knows the Seabird CT sail on the glider, and the GPCTD and legato only via SciCon. An RBR legato as the main CTD on a serial port needs basestation3, sg_ct_type = 4 and a legato_sealevel value (the sea-level pressure reading from a self-test) in sg_calib_constants.m.
• Other sensors need a matching .cnf in the basestation's Sensors directory (listed in .sensors) to reach the NetCDF files with proper names and metadata. Without it they may appear in the .eng files but not in the .nc.
• Column names differ between sensor builds. A WET Labs puck whose columns didn't match the calibration names was fixed with a column remap in sg_calib_constants.m (remap_wetlabs_eng_cols = "…"). Another team corrected a mislabelled column directly in the .eng files before reprocessing.
• Send the basestation log with any processing question. The error lines ("No handler found for columns…", "Unknown nc metadata…") usually say exactly what's missing.

Per-sensor notes¶

CT sail (SBE)¶

• A step change at apogee (for example, +10 °C and −10 PSU from one dive on, with unrealistic density afterwards) has been traced to water entering the thermistor: micro-leaks where the cap is welded onto the thermistor tube. Sea-Bird resumed leak-testing thermistor tubes after these cases. A thermistor wire broken off by over-twisting the C and T wiring was found on another sail.
• Zero or erratic counts on a CT that tests fine elsewhere point to the main board. See Electronics.
• Coefficients on the glider differ slightly from the cal sheet. That's normal. The glider's floating-point precision can't hold every digit, and it only uses the values for onboard density. Processing uses sg_calib_constants.m.
• The plug seals. The CT plug uses a -012 o-ring on the bore (with a backup ring) and a -016 on the face.

GPCTD (Sea-Bird pumped CTD)¶

• Prompt and output format. If the GPCTD's "executed" tag has been turned on (e.g. by using Sea-Term), the glider can't find the prompt. Turn the tag off, or change the prompt in the .cnf. The output format also has to be what the .cnf expects (hex).
• Garbage at the end of the first half-profile. A known GPCTD behaviour puts about 20 samples of garbage at the end of the dive (a) file. On a short deck dive the whole dive file can be nonsense while the climb looks fine. Don't judge a GPCTD on a deck dive.
• Configuration can silently disappear. On one glider $LOGGERDEVICE2 was found disabled after a firmware swap, so no CTD files were made ("No pumped CT data found"). Setting the parameter again, via the menu or the cmdfile, restored it. Always save and restore parameters around firmware changes.
• Clock-sync errors during a self-test have been intermittent. Timing changes in the .cnf can help.
• Dive and climb temperatures disagree by degrees → suspect the pump. The pump's energy use can be checked roughly by comparing the BATTERY file between two dives.
• Connectors. Corroded IE55 pins on a GPCTD had to go back to Sea-Bird. The small IE55 bulkheads are not a field repair.

Oxygen optodes (Aanderaa)¶

• 4330F vs 4831F. The two are essentially the same sensor with a different interface. The 4831F has a standard wet-mateable bulkhead (and an analog output option). The 4330F's sensor foot needs the manufacturer's own cable plug, so a home-made cable is much harder. For a Seaglider, the 4831F is the simpler choice.
• Storage. Keep the foil wet and dark, with water in the cap. If it has been stored dry, hydrate it for about 24 h before calibrating or deploying.
• Flooded cables and connectors are a common cause of optodes dying mid-mission.

WET Labs / Sea-Bird ECO sensors (BBFL2, SeaOWL)¶

• Channel wavelengths and names differ between models and builds. Check the characterisation sheet and set the column mapping to match.
• Calibration sheets go missing with second-hand and refurbished gliders. Sea-Bird can supply them by serial number, but it takes time.
• For a deck check, leave the sensor on the glider and run it through the glider (a bucket underneath works). Taking it off needs its own cable and software.

RBR legato¶

• Can run as a serdev or a logdev, depending on firmware and integration. The .cnf for one won't work for the other.
• Needs basestation3 to be processed as the main CTD (see above).
• Reported issues on manufacturer-integrated gliders:
• power use not being counted by the fuel gauge;
• CTD data repeated several times in the NetCDF files;

PAR sensors (Biospherical)¶

• The older QSP2150 has a simple serdev .cnf (one value per sample).
• The newer MPE-PAR outputs hexadecimal at 115200 baud, free-running at about 1 Hz once powered, and applies no calibration itself. Log the raw hex and convert afterwards: subtract a dark reading, divide by the calibration coefficient. Also log the internal temperature, which is useful for correcting the dark offset.
• A cosine (flat) collector lets in more light than a spherical one, but is more sensitive to the glider's pitch and roll.
• Don't try to convert a digital unit to analog by flipping its internal switch. The analog mode is logarithmic, needs different firmware settings, and needs recalibration.

AD2CP (Nortek)¶

• Connected either through SciCon (APL-UW's usual way) or as a logdev. The logdev approach uses a .cnf plus an ncp_go command file, runs at 38400 baud, and returns the AD2CP's own averaged telemetry file as the real-time data.
• Beam switching. Normally the instrument switches beams by pitch when configured for glider use. One manufacturer-integrated glider sampled the wrong three beams on dive and climb. The workaround was to sample all four beams and sort them out afterwards.
• Converting the .ad2cp files needs Nortek's tools, and beam geometry has to be checked against the configuration.

SciCon¶

• Start-up handshake. The glider sends a "log start" command and expects a reply and the SciCon prompt within about 1.5 s. It tries three times. On one glider, after 400+ good dives, SciCon began missing the handshake: dive data went missing and climb data was labelled as the dive. Turning on SciCon debug capture (capvec HSCICON DEBUG BOTH in pdoscmds.bat) changed the timing enough to make it work. The team left debug on for the rest of the mission.
• A 0 interval in scicon.sch skips that sensor in that depth bin.
• SENSOR_SECS in the log is SciCon's total on-time, not per sensor. SciCon measures per-sensor power itself and reports it to the glider's BATTERY file.
• The auxiliary compass on SciCon provides pitch and roll only. See Compass Calibration.

Removing a sensor¶

When a sensor comes off the glider, remove it from the configuration too (set its slot to "not installed"). Otherwise the glider keeps trying to talk to it, and the self-test and the fuel gauge are wrong.

See also¶

• Electronics — board swaps, Rev E, and why sensor configuration has to be redone.
• Compass Calibration.