# HDF5 File Structure
### Directory Structure
Sensors are grouped according to their sensor name. Integrations are stored as datasets in a subgroup called Rawdata. The name of a dataset is a combination of its timestamp and integration counter.
HDF5 directory structure.
Each dataset (integration) is a 6 x 256 array. Each dataset contains six 256-channel spectra for cal\_HH, cal\_VV, data\_HH, data\_VV, data\_U, and data\_V:
* cal\_HH, cal\_VV: auto-correlation products of the calibration load
* data\_HH, data\_VV: auto-correlations products of the soil signal
* data\_U, data\_V: cross-correlations products of the soil signal
Data array.
### Datasets
Each integration is stored with a set of metadata. Most of the metadata fields are self-explanatory, except for the following ones:
* accelerometer: A 3-element array containing the output of the accelerometer for the x, y, and z-axis. The look angle of the sensor is $$\tan\alpha = \dfrac{a\[2]}{a\[1]}$$.
* bw\_mhz: Sensor bandwidth in MHz.
* course\_deg: GPS course (azimuth) in degrees.
* flight\_counter: A counter that increases when the drone returns to its "home" position. This can be used to identify battery swaps.
* freq\_mhz: Sensor centre frequency in MHz.
* gyro: A 3-element array containing the gyroscope output for the x, y, and z-axis.
* integration: Counter increases for each new integration containing cal-load and soil signal.
* lna\_temperature\_degC: The physical temperature of the calibration load in degrees Celsius.
* runtime: Sensor runtime since powerup in milliseconds.
* sats: 2-element array containing the number of GPS satellites used and visible.
* sf\_mhz: Sampling frequency in MHz. The frequency of each channel is: $$f(\mathrm{MHz}) = \mathrm{freq\_mhz} - \mathrm{sf\_mhz}/4 + \mathrm{channel} \cdot (\mathrm{sf\_mhz}/2)/256$$
* speed\_kmh: GPS speed in km/h.
