Hello everyone
I have developed a compact spectrometer (photo attached) and working towards developing it into a milk analyzer. I will be using 600-1100nm, based on literature, where researchers have used this range to quantify the fat% in milk.
Before beginning, I would like to make sure how the calibration could be transferred from one device to another. There are some products available already in the market like Foss Milkoscan, but they use mid-infrared wavelength range, making it less challenging to transfer calibration. Others like Fatscan, haven't yet made it to the market.
Is it impossible? Can someone please elucidate the actual challenge in scale up while transferring calibration? Any leads would be appreciated.
Thanks.
What are you going to calibrate, wavelength or transmission (absorption)?
Dr. Jan
Can you please elucidate what do you mean by Quality scattering.
Dr. Taran
We are going to calibrate the transmission.
If components aber well specified anschauen Property assambled, you aber able to use the calibration of one instrument in another instrument of the same model,especially , if there is no need to have really small uncertainties ( in the range of about 1% in every part of the spectrum). If you use file of different models it could be worse than using no correction, or better- you can not know.
You have to distinguish between 2 different kinds of 'calibration':
There is no formal need to separate these two, but it's wise to do so: while the basic calibration is applicable to all devices of this kind, the device-dependent calibration is specific for a single device and not applicable to another. Omitting device-dependent calibration typically results in lower-precision devices.
Re your question: basic calibration could be transfered, device-dependent data shouldn't be transfered.
Regards
We have faced such problematic and the best way is to use a "transfer function".
Yhis is achieved by recording the vey same typical sample on both devices (consider repetitions to get a narrow standard deviation) and to calculate the transfer function as the ratio of spectra collected on both devices. Any new spectrum is then multiplied by the TF to genrate a spectrum which can be analysed by the statistical model you developed.
This MUST be done for each new spectro used.
Hope it help
Very well explained Dr. Dreher. Thanks.
Dr. Olivier
Thanks a lot. To bring it down to laymen's terms, we have to perform one-time calibration on a large sample set (say 1000) on one 'mother device', and perform a small sample set (say 5) calibration on each of the successive 'daughter' device and then based on that comparison between mother and daughter, there will be a transfer function which will let us 'fit' the conclusion of large sample set calibration of mother device onto the daughter device?
- Badges
- Science topic