Calibration for cereals and others on Antaris II

MagdaFka's picture

I’m a New user of Thermo Scientific Antaris II FT-NIR with Ingot calibrations for Animal Feed and Ingredients. We want to use the equipment mostly to test cereals. But the problem is that we’ve got no instructions on how to prepare samples. Firstly the seller told us that we can test whole cereals without milling it. This is what we wanted to have as we can’t imagine milling for example rape samples. But then it turned out that we have to ground them. I know that the level of milling and the amount of the sample does matter but I’ve got no idea how to prepare the sample so the results are true. What should we do?
And also before buying the instrument we weren’t aware that we have to test our samples by reference methods to determine BIAS. It is a little problematic for us as we haven’t got all of the samples in a number that is required for determine BIAS. We were sure that if we buy calibrations the instrument should be ready to use. What is your experience about this instrument and calibrations?

Hammink's picture

Dear Magda,
Basically no real sample prep is required. Grounding or milling makes your sample better homogenius which should give more accurate result and less scattering. The Antaris can have this setup part to create a litle rotating table, ideal for larger grains and particles. This would also have a cup, which gives you an idea on the amount of sample needed.
Ideally, your sample prep should be similar to the sample prep that was done on the samples that are used making the calibration. However I assume the Ingot calibrations contain loads of data that take sample size and physical changes into account to some extend.
A bought calibration made using a different instrument is never ready to use as it is. Also your cereal migth be different from the cereal used for the calibration. It is  needed to determine BIAS because no NIR system is the exact same.
You should always be doubtfull if a sales rep claims that if you buy a premade NIR calibration your instrument is ready to use. You should always set a BIAS and or add samples of your own.
Good luck!

shileyda's picture

Dear Magda,
I agree with Hammink, your best result would come from using a rotating sample cup. Bias adjustment (actually SLOPE and Intercept) adjustment should be evaluated on every calibration even if you develop this on your own data. This standardization is a critical step often overlooked. This is extremely important if you have multiple instruments or if you are using a transferred calibration. Yes, always be skeptical when anyone tells you they can transfer perfectly with no work on your part. Someone will have to do work, whether this is the person that transfers the model or you. The standardization set does not have to be very large, 12 to 20 samples is what I recommend. But the set should have as wide a range of constituent values as possible.  Also don't forget to use to correct moisture basis when you evaluate for slope/intercept. Labs typically report using As-Is basis and models may produce As-is or dry matter basis predictions. Of course these are not the same so it is very important that you understand the moisture basis of any predictions and of any reference data that you obtain before you calculate the adjustments.
Best regards,

dwhopkins's picture

Hi Magda,
The Ingot calibrations you mention are supplied by Aunir, I presume.  They are a good company, and you should get complete instructions from them on how to prepare the samples for use with your Antaris.  They may have calibrations for whole or ground samples, and you need to know what is required to use the calibrations.  For whole, unground samples, you may get better results by averaging multiple readings.  For ground samples of grains such as rapeseed, the exact grinder type, sample size and grind time should be specified. 
I agree with the others, it is wise to test your results with reference lab tests.  You may be able to participate in a test sample program with Aunir, to give you confidence in your NIR results.  NIR is a great technology, but you certainly should not use it as a "black box".  You would be smart to use as many independent samples as you can to test each calibration, whether that is just a couple samples that would give an indication of bias, or 20 samples over a range of values that would give you an estimate of slope and bias.  If you can start with only a few samples, then you can send out subsamples as you find a better range of analyses.
You should expect aid from Thermo and Aunir, just ask for it.
Best wishes,
Dave Hopkins