kebasaa
commented
1 year ago Wow, thanks a lot for this. It is still highly relevant
I've been trying to apply it, using Python's struct library (I'm using the little-endian encoding, i.e. "<I". But I ended up with a few questions, and I'm wondering if you can help me:
- An integer is 4 bytes, i.e. when I convert sample and sampleDark (each is 1800 bytes total) as you suggested, I end up with many more values than 400, even after removing the header (which is 8 bytes according to your instructions). Basically, I would have an 8 byte header, followed by 400 integers, and then 196 bytes that are unknown. For sampleGradient, I end up with too many values as well (1656 bytes total, i.e. 8 byte header, 400 integers and 52 bytes left over). But if instead all the values after the header were integers, then I get 449 integers for sample and sampleDark, and 413 for sampleGradient. That makes it difficult to do the reflectance math. Is it simply that the remaining data is not used? Do you know what's going on?
- I assume that the best way to create a reflectance spectrum from the units of "counts per second" is to normalise it to 0-1? But there must be a fixed maximum and minimum for that? The values that I'm currently getting are huge
- Do you have any idea how to apply the calibration? I guess it's based simply on a scan inside the calibration box, but what is the math behind it?
- Here is some example data:
- sample first few hex bytes:
ba0208070000000018c9a124fc3f2e57aa716475154067b118384dd9407d30dda544906a9753... - sample, resulting in: -70 (protocol) 2 (signal type: scan) 1800 (length in bytes), then the following integers
614582552 1462648828 1969516970 - sampleDark first few integers:
4269192318 1192275574 2032346525fromba020807000000007eb476fe76ae10479d252379e65579bccb56d70f85bfecfd6729eb3ebd... - sampleGradient first few integers:
405697630 247858264 128659535fromba0278066e0000005e742e185804c60e4f30ab0756c7b26e52e1672bba13a1898f76f0c0b647922... - Resulting reflectance:
9.45933685e-01 -2.86285788e-01 3.30041416e-02
- sample first few hex bytes:
hbsagen
Nice to see you're still tinkering with this. In the readme it says: "Every SCIO bluetooth LE message contains 3 parts: sample, sampleDark and sampleGradient (No clue so far what that those mean or how to convert them)." Not sure if that's up to date, but hope the below is helpful.
sample: This is the raw spectral data from the sample. It represents the light that is reflected off the sample and detected by the SCIO.
sampleDark: This is the raw spectral data from the SCIO's internal dark current reference. It represents the background signal that is detected when there is no light present.
sampleGradient: This is the raw spectral data from the SCIO's internal white reference. It represents the signal detected when the SCIO is measuring a known white reference
To calculate the reflectance values of the sample, you need to subtract the sampleDark data from the sample data and divide the result by the difference between the sampleGradient data and the sampleDark data. This is expressed by the equation R = (S - D) / (G - D), where R is the reflectance value, S is the sample data, D is the sampleDark data, and G is the sampleGradient data.
Let's take -bark.txt sample:
Each packet consists of a header and three data sections: sample, sampleDark, and sampleGradient. The header contains information about the packet, including a packet identifier (01 or 02), the protocol identifier (ba), and the length of the data sections (in this case, 02 90).The sample, sampleDark, and sampleGradient data sections are each 400 bytes long and contain spectral data measured by the SCIO spectrometer.
To calculate reflectance values from the sample, sampleDark, and sampleGradient data, you need to perform the following steps:
The example code below should extract the data from the log, converts these arrays to numpy arrays of integers and performs the reflectance calculation using numpy array operations.
The reflectance values will be in units of "counts per second"...