For example, let's say you specify 10 seconds for the CL acquisition time. The software divides that into intervals based on the RealTimeInterval specified in your Probewin.ini file which is usually around 400 milliseconds or so. It then adds those 400 ms integrations together for the total spectra.
So summing or averaging fractional second integrations, one should get the same results I would think!
John,
I think there are small difference between those two methods. On my method, the the random noise is cancelling each other, while your implementation seem to be not.
On my methods, I've subtracted the light spectrum intensities with the dark references before summing it up. In this way, I will end up with some pixels to have positive and some negative value. This is useful to remove some random noise, particularly for some non-emitted wavelengths. Averaging those positive and negative value should give me a zero average. I've also make sure to take the dark ref with the same acquisition time with the light spectrum to get the same noise characteristic for both.
On the other hand, on your implementation, it seem that you are summing all the short acquisition. Since all the pixels are having a positive value, we will end up accumulate the random noise if we just simply sum them up.
Hi Riko,
I don't quite get why your method of performing the net intensity subtraction for each integration would provide better data than performing the net intensity subtraction after summing them. Mathematically they are exactly equivalent...
Unless there is some non-linearity with the spectromotor that changes with each sub second interval acquisition?
But still I'd like to try and figure out how to make your net (light - dark) method an option in my software. The problem is, I want to keep both the dark and the light spectra stored separately, not just the net intensities. Storing the dark and light spectra for each sub second interval would be an excessively large amount of data. Also doing a light - dark net subtraction would require inserting the faraday cup for each integration! We don't want to interrupt the WDS/EDS acquisition!
Maybe I could just add a third storage array for the net spectra. That way I could perform a net subtraction for each sub second acquisition and store that as an accumulation, instead of calculating it on the fly after all acquisitions are complete... I would have to add it as a third column in the EMSA file.
I guess I don't quite understand why your net method would work better. Random noise should still average out over time. Can you provide some examples here of the two methods compared?
john