Range Equipment Online measurement

Analyst's picture

Dear all 
We are gonna purchase a Online NIR, with a DAD detector. it will be placed above a conveyor belt as process controll. 
We got 2 NIRs in mind both DAD but the first 1 had a range from 300nm-1700nm , the second has an extended range from 300nm to 2100nm. 
Which range would be needed for online, or is 1700-2100 usefull?
also , if a visual detector is included does it need other pretreatments then a NIR detector
Best regard
Tom Van de Walle

Venheads's picture

Dear Tom
I am sure what is meant by online Detector and not sure that 300nm could be described as NIR
But anyway, it depends of consituents you want to analyze and accuracy you need to maintain
As far as i know by exampble of Foss Nirs 2500 which is 400-2500, there is no special treatment for 400-700 nm detector

Analyst's picture

Dear Venheads,
The amount of constituents are "in scope" 3 and out scope 7.
The accuracy they want is Moisture: 200ppm or 0.2% absolute, Byproduct 300ppm.
The colorcoating , no specs yet.
The 300nm to 700nm will be used for he color coating.
Now only question does the extended range wll have a  beneficial effect or just makes it more complex.
At the moment we use a FOSS Infraxact NIR with a range of 570-1850, and there we meet the requested specs.
Also to the FOSS users: Weighted MSC how you can translate this in Unscambler.
Best Regards
Tom Van de Walle


ptillmann's picture

since you wrote, that you have a model based on Infraxact data that meets the specs, you ought to be able to test whether you need the range 1700-1850 nm with these data. So the answer to your qustion lies within our data.
Generally a detector with an upper limit of 2100 nm compared to a detector with 1700 nm will either have less pixels per nm across the whole range - larger spacing - or less pixel size - more noise - or a higher lower limit of detection range. So you will always loose something. What is benificial can only be answered by your intended application.
Our customer do use online NIR instruments with an upper limit of 1700 nm as well as 2100 nm, depending on the application and targeted constituent. It is not true that 1700 nm is always sufficient. (Sorry, Phil)
The penetration depth of light decreases with increasing wavelengths, so 2100 nm might not be ueful anymore if you have granular objects with a coating or husks or ...
The ISI "Weighted MSC" does not translate into any Unscrambler data treatment as far as I know.

tricour's picture

dear Peter, i agree with you concerning the explanations for resolution and other stuff on the needs and request for online analysis. This is also one of the reasons that AOTF is superior to diode array for such tasks...
concerning the weightzd MSC as described in Winisi, i would like to check within Unscrambler to compare by using some MSC combined with other pretreaments... as you, i never found such weighted MSC in unscrambler but until i didn't try to compare with some combination i would not prononce myself definitely...

Analyst's picture

Dear Peter
Thx for the reply, in the meanwhile we founded a comparison result with unscrambler. 
BTW does AOTF have no problems with accoustic noise? i'm not familiar with this technique. 
best regards 

tricour's picture

Dear tom, as explained before, most of the time, visible part is not relevant to use combined with Nir. Mostly providing noise, so a good range would be 900-2500nm.except if you have specific parameters in visible...
But based on your parameters you don't need to use such a wide range. Maybe 1100 to 1900nm.
Diode array is a good techno but could be limited in accuracy depending on resolution and range. Try aotf techno which would provide ft quality with speed of diode array...
Concerning weighted msc, it is not so complex to translate into unscrambler. As unscrambler official trainer for France, don't hesitate to ask for more help by MP.
[email protected]

Analyst's picture

Hi Ricour
Thanks for the information. The visual data will be needed, as it  is possible to detect the color with NIR (which is only 50ppm) but the accuracy is not so good, around 10ppm. Therefore the Vis detector would take this over and predict maybe more accurate. 
We will maybe take the AOtf but at the moment we did contact some suppliers but none spoke about this technic. 
Concerning the unscrambler help, it would be helpfull.
today we proven that it is possible with NIR to predict the 3 in scope constituents, with limited calibration samples ;D

ianm's picture

The InGaAs detector range from about 900-1700 is fine for on-line, with a diode array instrument.

(posted on behalf of Phil Williams by Ian Michael)

Analyst's picture

Thank you Phil for the reply, 

venkynir's picture

Hi  Venheads,
Can  I know the vital constututes to measure and control in on-line .Then I can suggest the DAD and other deatiails  I can also provide solution foe Low  cost and good measurment system
. Venkynir ,India

NIRonCV's picture

Dear Tom, to answer your question more correct, you may need to say a little more about the product and parameters that you want to analyze on the belt. And to go in detail , more about the speed and surface. All this can be of influence of the selection of your device. 
please let us know so we can help you

Dennis Zijderveld's picture

Dear Tom,
Metrohm Belgium does offer lab and process NIR analyzers that are made by Foss.
There are two direct light processs NIR anayzers: one InGaAs 1100-1650 nm and one XDS direct light NIR analizer with a 850-2200 nm range. There are also a couple of "inbetween" lab and process NIRS. These analyzers do work with fibers and a reflection probe that covers the full 400-2500nm range. This gives some possibilties with using  automated sampler system.
When Vison software is used the visable part can be used to do qual- or qantification analysis. There are also possibilties to export to 3rd party software programms such as unscrambler.

Is you application food related?
With kind regards,


ptillmann's picture

if you think of a data pretreatment as a tool fo spectroscopy you might very well find another tool that performs equally well. SNV detrend and MSC are in this meaning equivalent for many applications.
If you think of "inverse MSC" as an algorithm that treats data mathematically, then there is obviously no matching algorithm in Unscrambler. You need not try other data petreatment options or combinations of options.
Three annotations to this:
- many derivatives are not equal across chemometric packages, they differ mathematically strongly
- SNV and detrend in ISI and many other chemometrc package differ mathematically
- Inverse MSC is a tool designed especially for transmission measurement in the Infratec.In my view there is no valid reson to use it on reflectance measurements.

tricour's picture

Peter, i agree with most of your comment. I never used winisi so far and made comparison with unscrambler and others as eigenvector, mathlab....
i am not also a great fan of msc, but I need to have spectra before taking any decision on treatments...,
concerning aotf, the techno doesn't care about environmental conditions like acoustics vibration so far. I could tom send you in PM a presentation of the techno.

Analyst's picture

Dear Thomas 
I would like to have this presentation, to present to the Board as new innovating lab equipment.
[email protected]
best regards
Tom Van de Walle