Nestbox Field Test 3

This field test was carried out with the prototype 2.0 in the time frame between 12th of July 2018 and 14th of July 2018. The test was carried out using a outside Nestbox installed at a farm next to EPFL.

  • 12th July 2018: Installation of the Prototype and the IR-Camera
  • 17 July 2018: Data download + Correctly re-setting the logging (raspberry pi) time.
  • 27th July 2018: Try to recalibrate and reflash the MSP430. Removal of the IR camera. Data download.
  • 03th August 2018: Data download.
  • 14th August 2018: Data download. Take down of the prototype.

Valid data could be recorded during the time period of July 12 - July 19 with the exception of the night from July 15, 21h20 to July 16, 9h02, most likely due to a mains power cut, which was supplying both the raspberry pi (datalogging) and the cameras.

Subsequently, the timestamps during the 16-17 July are offset by the time of the “black-out” (lagging behind by 11h40min). Thanks to a data download on the 17th, the time is synched correctly again thereafter.

Video offset

  • Raspberry Pi Timestamp: UTC (Swiss daylight saving time is UTC+2)
  • Video Timestamp: UTC + 2 h 4 min 15 sec

Camera timestamps would resume automatically to a correct value after the black-out.

Reliability

While Raspberry Pi logging was paused during power-cut, the main nestbox device continued to reliably make measurements. There was no firmware reset or failure until the device was shut down automatically to prevent battery draining.

Test conditions

We reported very high temperature of the electronics (over 45°C), as it was directly exposed in the sun and measurements were done during an extreme heat season (air temperature reached 35°C).

Prototype Material

  • Nestbox-V2.0 prototype, powered by battery
  • RFID reader installed on the perch
  • Load-Cell installed on the perch
  • PIR sensor next to the perch
  • PIR sensor inside the Nestbox

Test Material

  • Raspberry Pi for data logging (connected to mains power supply)
  • MSP430-Launchpad as a Serial-to-USB converter (powered by raspberry pi)
  • IR camera recording continuous video (connected to mains power supply)

Aggregated events captured

The following gallery shows all owl appearances captured on camera, together with information on whether weight measurements, RFID tag detection or PIR detection is available. Some entries may not contain an image, because it was not an interesting picture (owl just staiding on the entrance directly). A table view can be consulted under this link.

search?q=https%3A%2F%2Fairtable.com%2Fembed%2FshrGbEXjjSdicJKhe&btnI=lucky

Presence detection (PIR)

General

The PIR detection worked very reliably and detected the owl approaching or leaving the nestbox with the external sensor with 85 to 100% success rate. Some of the events that were only captured by video may not have been detected by the PIR sensor due to the orientation of the lens or because the owl was too quick.

The internally mounted detector triggered much more often, since the baby owls may move around inside the nestbox without leaving it.

Timing of the PIR trigger

Most trigger events by the external PIR sensor happened before the owl got detected by the balance. In some cases it was however only when the owl left the perch, due to the orientation of the PIR lens.

TODO: check individual events if PIR trigger was early enough

RFID detection

Owl identification via its RFID tag worked very well and within milliseconds after the owl touched down on the antenna pad. Nearly every occurence that was recorded on video and when the owl landed on the pad resulted in a correct RFID identification (over 65%). Below is a graph displaying both the external presence detector and successful RFID detections.

In the table below, we display each video event when the owl was standing at least shortly on the antenna pad of the perch, and should therefore have been detected. On 9 out of 26 such events, the owl did not get identified by the RFID reader. Note that each touch-down of the owl was very short (max. 5 seconds) and sometimes moving a lot.

search?q=https%3A%2F%2Fairtable.com%2Fembed%2FshrRQPkzgttBjuCKT&btnI=lucky

No consistent weight measurements could be recorded, due to following reasons:

  • The owl stayed only a short time on the perch (often less than 5 seconds).
  • It moved while standing on the perch, making a stable measurement very difficult.
  • When on the way to enter the nestbox, the owl would very often carry a mouse.

Recorded Data

Measurements without mouse

Average weight without mouse: 432 [can be scaled to grams] search?q=https%3A%2F%2Fairtable.com%2Fembed%2FshrOtMlsKEBYxG47q&btnI=lucky

Measurements with mouse

Average weight with mouse: 470 [can be scaled to grams] search?q=https%3A%2F%2Fairtable.com%2Fembed%2Fshr2XsQyRuRNvOy2a&btnI=lucky

Oscillating measurements

What is displayed in the night of the 16th-17th in the above graph is so far unexplained. (!! This was actually during the daytime due to the nightly power-cut and timestamp de-synch !!). The sensor readings show oscillations. This behaviour is also observed after the 27th of July where cameras were not used anymore.

and zoomed in:

Unconclusive calibration attempt

The calibration of the 27th failed. Due to unknown reasons, the weight measurement was not stable (a strong creep of the value is apparent). Possible reasons could be the high temperature and degradation of the load cell (which is within specs only between -10°C and 40°C).

Power

The first 6 days of the field test contain the only relevant power measurements. Afterwards, the test firmware switched to RFID-on-mode at all times.

During 6 days, the battery voltage decreased from 5.28 to 5.23 V, a very satisfying performance. Note the high