The Research Portal allows for a maximum of 50 GB of compressed data to be downloaded for each single data export request submitted. Moreover, each data request (sensor or summary raw data) can cover a period of a maximum 31 days of data collection.
It is therefore important to understand the expected size of downloaded data files, in order to make sure that all your requests can be properly processed by the Research Portal.
Sensor raw data
A full 24h day of raw data for a single study participant equals about 35 MB of compressed data downloaded from the Research Portal. Therefore the 50 GB limit allows for a total of about 1500 full days of sensor raw data download.
For example, if you want to download sensor raw data for 200 subjects and a month of data collection, your request cannot be fulfilled (200 x 31 x 35 = 217 GB of data). You will need to submit separate requests, reducing the data collection window to 7 days. A good rule of thumb to determine the optimal data interval is to consider 1500 / n = days, where n is the number of enrolled subjects. If your study has 100 enrolled subjects, you should adjust your sensor raw data export window to a maximum of 15 days at a time.
Summary raw data
This dataset is considerably smaller in size compared to raw data. A full 24h day of summary data equals about 35 KB of compressed data downloaded from the Research Portal. For this reason, you should not experience any issue when downloading summary raw data for all subjects.
Data format
Embrace records physiological data in binary format. Binary data is sent to the Mate App and then to the Empatica cloud where it is processed. When a download request is made from the Research Portal, data is converted into text-csv format, organized in folders, compressed in a zip file, and made available for download.
The 50 GB download limit per single request refers to the size of compressed data. Please note that when the file is opened uncompressed, the size can increase significantly (up to 5 times). Make sure your computer or server has enough space to accommodate the files.
Note: The values and the calculations indicated in this article have been approximated for demonstration purposes. Real dataset sizes might vary due to multiple factors (including, but not limited to subject compliance, connectivity interruptions, and full Embrace2 memory).