Clipping Issue due to wrong scaling of Raw Data

First of all: thank you for this nice package!

I am running into weird behavior when processing some external files, recorded by Ax6 over 7 days with 100 Hz and ±8g.

Apparently, it occurs while reading the binary data (.cwa) and is related to some wrong scaling of the data. Virtually all values get clipped and are set to ±8g even though they should lay well below the threshold.

With the provided sample.cwa I do not reproduce this error and everything, including the calibration and activity classification works as expected. I also compared my raw data to the sample file loading them externally. The data is very similar (range, mean, std, ...) and has very few to no clipped values, indicating that the error is not hiding there but rather in the de-binarization.

What I noticed is the different .cwa file headers/data_fromats extracted from the binary files using the external software:

for sample.cwa: {'deviceFractional': 55355, 'sessionId': 1, 'sequenceId': 0, 'light': 646, 'temperature': 29.1015625, 'events': 1, 'battery': 4.1953125, 'rateCode': 74, 'numAxesBPS': 48, 'sampleCount': 120, 'frequency': 100.0, 'timestamp': 1399469391.6893005, 'timestampOffset': 125, 'timestampTime': '2014-05-07 13:29:51.689', 'estimatedFirstSampleTime': 1399469390.4393005, 'channels': 3, 'bytesPerAxis': 0, 'bytesPerSample': 4, 'samplesPerSector': 120, 'accelAxis': 0, 'accelRange': 8, 'accelUnit': 256}
for my_file.cwa: {'deviceFractional': 40055, 'sessionId': 12940, 'sequenceId': 0, 'light': 38, 'temperature': 34.375, 'events': 1, 'battery': 4.04296875, 'rateCode': 74, 'numAxesBPS': 50, 'sampleCount': 80, 'frequency': 100.0, 'timestamp': 1660149007.2223816, 'timestampOffset': 80, 'timestampTime': '2022-08-10 16:30:07.222', 'estimatedFirstSampleTime': 1660149006.4223816, 'channels': 3, 'bytesPerAxis': 2, 'bytesPerSample': 6, 'samplesPerSector': 80, 'accelAxis': 0, 'accelRange': 8, 'accelUnit': 4096}

could the error lie in the different accelUnit = 256 and accelUnit = 4096 values? resulting in way too high scaling during de-binarization of the raw data?

Note: I'm using openjdk=11 instead of the suggested openjdk=8, due to apple silicon architecture.

the .json file of produced by accProcess shows the wrong amplitude scaling and thus falsely classified clipped values:

{
    "file-name": "path/to/my_file.cwa",
    "file-size": 390550016,
    "file-deviceID": 55191,
    "calibration-xOffset(g)": 0.0,
    "calibration-yOffset(g)": 0.0,
    "calibration-zOffset(g)": 0.0,
    "calibration-xSlope": 1.0,
    "calibration-ySlope": 1.0,
    "calibration-zSlope": 1.0,
    "calibration-xSlopeTemp": 0.0,
    "calibration-ySlopeTemp": 0.0,
    "calibration-zSlopeTemp": 0.0,
    "calibration-errsBefore(mg)": 12856.130487908302,   <-----------------------
    "calibration-errsAfter(mg)": 12856.130487908302,    <-----------------------
    "calibration-numStaticPoints": 1,
    "quality-calibratedOnOwnData": 1,
    "quality-goodCalibration": 0,
    "file-startTime": "2022-09-20 15:00:07.011000+0100 [Europe/London]",
    "file-endTime": "2022-09-27 14:59:07.011000+0100 [Europe/London]",
    "file-firstDay(0=mon,6=sun)": 1,
    "totalReads": 61061181,
    "quality-daylightSavingsCrossover": 0,
    "clipsBeforeCalibration": 60842031,   <--------------------------------------
    "clipsAfterCalibration": 0,
....
}

OxWearables / biobankAccelerometerAnalysis

Clipping Issue due to wrong scaling of Raw Data #247