Performance Optimization of DLP NIR Scan EVM

Hello Derek Ness,

We are pleased to know  your interest in evaluating the DLP NIRscan system.

Our laboratory measurements shown better S/N compared to the one in the market, but our suggestion at this point would be to try yourself first for your confidence and satisfaction. 

Inside the DLPNIRscan system uses TI Sitara processor AM335x processor for ADC sampling and computation of final absorption spectrum of the sample scan. You can setup AM335x application software to further speedup and improve S:N parameters by adjustment such as -

1. S/N: There is Programmable Gain Amplifier in the system - you can adjust the gain of the signal.

2. Speed: The pattern exposure time can be reduced upto ~700us per pattern as compared to presented hard coded value of ~1.4ms time. This can improve the scan time.

Regards,

Sanjeev

Sanjeev,

thanks, that was really helpful. In my experience it seems to be a balance between sensitivity and resolution with spectrometers if you want good sensitivity ( low signal to noise ) you can suffer with poorer resolution, unless you up the scan time or number of scans.  I will have to have a look at the diagram but is this module dispersive where it splits the frequencies and scans them through the sample, or FTIR where all the radiation passes through the sample and an interferogram is generated and then converted to the spectra?

regards,

Derek.

Hello Derek,

You are correct, the S:N improvement is trade-off to the resolution parameter in the system.

It is FTIR, all the radiation passes through the sample.

Regards,

Sanjeev

Hello Derek,

I made a mistake about FTIR part. My apologies.

DLP NIRscan spectrometer uses dispersive approach to be more precise it is post-dispersive i.e., after passing entire spectrum over the sample at the single point detector side is follows dispersive approach.

Regards,

Sanjeev

Sanjeev,

yes, I realised that when I checked on the set up and saw the grating splitting up the spectra. In mid IR the instruments have moved away from the dispersive approach and use the FTIR process exclusively, where there is no splitting up of the wavelengths and all the radiation passes at the same time through the sample. In NIR and Raman spectroscopy they still use the dispersive method for some instruments although some do use FTIR.

regards,

Derek

Sanjeev,

I think that the concept is great and I would love to get my hands on one of these to try one out in ernest, ( the smart scanning would be great for keeping a track on our production materials mix quality) but sadly for just now we haven't got the  funds  in the technical  budget. So maybe I will try to get one into next years programme . I just hope that they haven't all gone by then .  Good luck with the DLP NIR I'm sure that it will be a great success.

Derek.

Hi Derek,

Thank you for your comments. Do let us know what you feel when you get a chance to work the Kit.

Regards,
Sanjeev

Sanjeev,

I was wondering how to improve the resolution of the NIR. If the slit was decreased from 25 microns to 10 microns, in most spectrometers this would improve the resolution at the expense of signal to noise. With the DLP is this the case or does the resolution of the DLP limit the resolution of the spectrometer? If not it would be quite easy to have a scan at 1sec. rather than 300msec. to get more signal at the lower slit size and if these were averaged out over 30 scans you should get an improved resolution with good signal to noise with a total scan time of 30 seconds which for a bench machine is fine. How easy would it be to change the slit? I have looked at the BOM but I cant see wheee the slit comes from.

regards,

Derek.

Hello Derek,

I have to get in touch with our expert. Here is the response to your query.

Decreasing the slit with does indeed typically increase the resolution of the spectrometer at the expense of SNR.  This SNR loss could potentially be compensated for by increasing the integration time (averaging a larger number of samples).  However, there are three items which enter into the final resolution of a spectrometer such as this:

 1 – slit width

2 – DMD pattern width

3 – optical point spread function

 The optics should be designed to support the smallest resolution anticipated for the instrument.  At some point, decreasing the slit width and DMD pattern width has minimal impact due to the optical design. This is always a trade between resolution (optical performance) and optics costs.  Of course, the resulting spectrometer resolution only needs to be as good as required by the specific application.

Feel free to revert back if you need any further information.

Regards,

Sanjeev

Hi,

Maybe I am missing something on the info (on the web) and this forum, but what is the highest resolution of the NIR Scan EVM ?

Regards,

Hans

Hello Sanjeev,

Do you have a document that describes the experimentally derived parameters that you have discussed? For example, 10 micron slit plus averaging of 50 scans equals ??? signal to noise and resolution.

What is the minimum spectral resolution (measured in full width half max) that is attainable with a system using 700-2500 nm?

Hi again,

Mh, no fast response this time.

Regards,

Hans

Hello Hans and Others,

Very good questions are being asked regarding design considerations of the NIRscan.  Our White Paper released in February discussed some of these topics, but clearly there is a need for a deeper discussion.

We are writing another application note, to publish in early August, to answer these types of design considerations.  Please post other questions in this tread that you would like us to address.  We will do our best to provide everyone with the additional insight needed to make good development decisions.

Thanks,

Eric