It turns out that some commercial spectrometers are not very linear in their response. And that means intensities, even relative intensities, reported with the expensive commercial instruments might not be very reproducible.
But, we can do better.
The commercial instrument:
The following are fluorescent lamp spectra collected with one of the most popular commercial CCD spectrometers at different exposure times. We divide intensity by exposure time. So, all of those curves should lay on top of each other. To make things more clear, the third graph shows the ratios of the heights of three of the peaks from the spectra.
What we see with this $4,000 instrument, is that the peaks heights are not proportional to the exposure time, and even the height of one peak compared to another changes when you change the exposure time.
That is really not very good. How do you rely on something like that to publish your research?
The new instrument:
And now, here is the data from the instrument that I just recently designed. BOM is about $150, optics around $200 from Thorlabs or ebay. The processor is a Teensy 4.0
As you can see, the curves do lay over each other, except where it saturates at longer exposure times. And the relative peak heights are pretty flat, again until one of the peaks saturates. That is what you want if you want to be report spectra when you publish your research.
The how and why of this is very interesting, and I plan to return to the topic and add some explanation in due time.
The design files and firmware should be on github soon.
Meanwhile, here is a picture of the first board set (sensor and controller), and assembly of the spectrometer. The resolution you see in the above is with a 200um slit. That was part of the spectrometer design, but that is a different subject. There are some minor updates since the first boards. Those will be in the github repo.
