RoadTest: DLP® NIRscan™ Nano Evaluation Module
Evaluation Type: Independent Products
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: n/a
What were the biggest problems encountered?: software setup, firmware updates
I reviewed the Texas Instruments DLP NIR Scan Evaluator Module (EVM)
This Eval module is meant to get you up and running and familiar with the capabilities of this nifty device.
Near IR (NIR) refers to the 700nm to 2500 nm range of the light spectrum. Outside of our visible spectrum, this range of light is used for spectroscopy in a variety of industries. Spectroscopy can allow identification of different materials based on the unique "fingerprint" they have when illuminated by a light source and having optical properties measured such as transmittance, reflectance, or absorbance. The degree of transmission or absorbance at each wavelength can create a fairly unique "fingerprint" to tell different materials and compounds apart.
NIR Spectroscopy has applications in several industries, including medical, pharmaceutical, agriculture, food production and QA, non-destructive evaluation (NDE), chemical production, chemistry, and atmospheric science. I dialed into the TI Sales call where they made the case for this device, and provided great explanations on its competitive advantages.As with all spectrometer, a sample is illuminated by a broadband (multi spectral) light source. The light reflecting off the material is directed, focused, split by a diffraction grating, and then received on some sort of sensing surface. Convention spectrometers frequently employ a linear array to sense each wavelength at a different point on the sensing surface.
This DLP NIRScan spectrometer can sense in the typical 700-2500 nm wavelength range, but while built on this EVM, it can only sense in the 900-1700nm range. The manual talks about many ways to interface this hardware differently. I'm a little foggy in this area, but it appears you can interface it with your own logic controllers or microcontrollers and make complete custom use of this device. The manual and files talk about some RTOS and other softwares that enable you to get started. Reading deeper in the manual, the EVM board has a microcontroller running a Texas Instruments OS called Tiva, which runs on RTOS. The EVM board also contains micro SD card for saving data, Bluetooth adapter for interfacing via apps in iOS or Android devices, and a USB port for computer based interfacing.
The actual DLP detector board contains 6 main components. The moneymaker is a 1mm non-cooled Hamamatsu G12180 Indium Gallium Arsenide InGaAs detector. A TMP006 Thermopile will measure temperature of the detector and so that temperature compensation can be performed. Signal from the InGaAs sensor is amplifed by an OPA2376 low noise amplifier, and then converted into a 24 bit value by a ADS1255 ADC. Two other precision ICs provide precision voltage buffering and supply to the transimpedance amplifier. These are a REF5025 and OPA350. In short summary, this EVM module consists of 4 boards. The EVM microcontroller board (the one connect to via bluetooth or USB), DLP controller board (logic board that controls detector and contains lamp driver), detector board (board that does sensing and ADC), and the DMD board (contains DLP NIR mirror). From a brief look at the manual, it looks like you'd want to interface the J500 port directly on the DLP controller board if you wanted to replace/bypass the EVM microcontroller and take full control of the sensing hardware. More conventionally, you could send commands to the TI EVM microcontroller board via UART4 on the J3 expansion connector or the J4 JTAG ARM Cortex 10-pin. The EVM contains 3 hardware buttons that are frankly a pain to access. All buttons are at the edge of the PCB but basically require you to cradle this little device and jam your fingernails in. Trying to get in there with a plastic tool probably still leaves you at a risk of jabbing the device. Luckily you really only need the scan/bluetooth button when you want to get up and running on Bluetooth.
Installing the software is easy enough. Go to http://www.ti.com/tool/dlpnirnanoevm and look around for a good list of manuals, install files, etc. The page could really use a clean-up, as many older files are mixed in with new files, and clutter up the web page. Overall, it easy to gain access to a lot of material, including the hefty 100 PDF user's guide. But the website is a bit overwhelming if you want to just plug in the device, install the GUI, and start scanning items.
In not reading this page properly the first time, I installed an out of date GUI.
The Update GUI has a lot more bells and whistles when you install the correct one!
Installing the software goes well enough, except for some out of date Firmware files.
If you really care, you can go back into the website, download more install files, go into the extract folders, and then point to the ISO and BIN files to update your firmware. This part is not very user friendly and poorly thought out.
When you open the software, you find a pretty well laid out application. There is a quick start page that arrives in the package for those eager engineers (me) who hate manuals and just want to play with the tech. It does a good job of getting you rolling and doing a first scan.
The manual and webpage mentions that there is an iOS app available for reading from this device. There is also an Android App available, but it must be quite new because it is not mentioned anywhere at all. You can find it here: https://play.google.com/store/apps/details?id=com.kstechnologies.NanoScan&hl=en. To setup with the app: Simply open the app on phone, turn on bluetooth, and then hold the Bluetooth/Scan button on the device for 3 seconds. When you see the blue light flashing, you can press the little icon in upper right that looks like an electrical cord connecting. Connection should happen pretty quickly. When you load, you notice they have pre loaded some different NIR signatures, to show you what the different spectra can look like, and to give you some ideas on things to go scan.
Let's get to actually scanning some goodies! The quick start recommends you just use the stock scanning parameters. All you need to configure is the save location and filename. Click Scan, and the device takes care of the rest. As mentioned above, the device will bath the mystery object in broad spectrum light (probably like an incandescent type light source), and the measures the light reflected by the object in each wavelength. The intensity of this received light is registered. From the intensity of this received light the device determines the light reflected by the object, and the corresponding light transmission through the sample can be determined.
My first scan was of an IR remote. It's a cheap little one with an IR LED, and I honestly didn't know what the beam profile would look like. I see the expected peak in the 900 nm range, but see several peaks, which I didn't really expect. At first I wondered if the light beam is coming out "choppy" due to the diffraction grating of the NIR DLP device. But I expect this is the true output, and that this is perhaps the noisy IR output of the very cheap IR remote.
I hooked up some other IR LEDs in a basic regulated circuit to verify the emission spectra. Below is a TSAL6200 IR LED. It should have a peak wavelength at 940nm.
I do the scan and get the expected intensity peak at 940 nm. This device is working! The software shows a cool display of the detector senses light reflected or absorbed, and also the light intensity received at each wavelength. What isn't working is this GUI software. The stock GUI software outputs a CSV of only the absorbance values, which are not interesting at all for an IR LED.
Testing some other materials on hand, we can see some interesting patterns.
I set up a basic testing setup in my kitchen and basically tested all the organic components of a charcuterie board, and different alcohols, for good measure. Because the device runs off a standard 5V USB, you can easily power it with a battery pack, and then control scans on your phone via Bluetooth connection. The app allows me to configure saving to the DLP micro SD card port or the phone. Very handy out of the box! Below I am scanning salt water.
This was definitely a very fun device to really dive into. After my software / firmware concerns, I was really able to get up and running, and start testing different fruits, vegetables, cheeses, and wines in my kitchen. This is definitely science. Delicious science.
Looking at wet and dried Basil leaves, we can clearly see a difference. It's likely that the water content (and possibly other plant health factors) can be differentiated by the different NIR signature.
Not being much of a spectroscopy expert, I was a little lost on the difference between comparing intensity and absorbance. From looking at the manual, I think intensity might essentially be the raw power levels recording with the DLP. I think the absorbance (and correspondingly the reflectance) are a calculated, and better normalized value. With that in mind let's use absorbance values moving forward.
My wet basil has a peak absorbance at 1400 nm, while the dry basil has overall much lower absorbance values (look at y axis), and peaks are poorly pronounced, but visible at 1400nm and 900 nm. We can see a clear difference.
Below is an orange. Honestly is visually hard to tell all of the produce apart just by looking at these spectra, but with a mathematical analysis I'm sure you could develop a decent finger printing method. In the image below I can see that there is a small local max just past 1580 nm, probably around 1650nm. I also see a deeper trough around 1085nm, or higher peaks around 1150nm.
Here is a fresh apricot. Right below it is apricot jam. Note that the signal appears noisier for the jam, but this is likely due to me scanning it through the glass jar (not wanting to contaminate my EVM). We can see that the 1400nm peak is relatively reduced, and there are higher readings across 1100-15nm. Part of this may be due to the glass (borosilicate glass should have low absorbance up past 2000nm), but a this should mainly be caused additives to the jam, such as sugar, pectin, and citric acid preservative.
Let's review a couple other materials, and look at some liquids. I tried to take a NIR scan of regular water in a glass as a good baseline, but again ran into issues with my sample jar. I guess a well designed flat sample container would help mitigate these issues.
Looking at salt water, we see following absorbance. Slight peak around 1400-1500nm, but generally quite flat. Noise is probably an issue.
Possibly because the bottle was quite clear, we got a great reading from our scotch sample.
Absorption is highest at 1085nm, and has a low point around 1150 nm.
I took a sample of black coffee, and got a different spectra. Their app has a spectra for coffee available, which ended up being very different from mine. This may be due to my extremely small sample size, as I was just emptying the remnants of my coffee maker.
A couple other items with markedly different absorption spectra.
Honey shows very different absorbance, with peaks at 950nm and 1200nm.
Flour shows strong absorbance only around 1400nm.
Let's look at some sugar samples
White sugar has a sharp peak at 1700nm ish and a noticeable peak around 1450nm.
For Brown sugar we notice a very different peak profile, with a broad peak that spans 1450 to 1600nm.
In general I am seeing that most materials have a noticeable absorbance around 1400nm. Looking at spectral absorbance of light in our atmosphere I notice that almost all light in 1400nm range is blocked by water moisture in the air. This seems to be our explanation. The 1400 nm peak is attributed to water. In the dry basil, we see the remaining leftover water as a small peak but it is a relatively small peak compared to the baseline signal.
In summary, this was a great piece of hardware to get exposed to the hands on science of taking NIR spectroscopy measurements. This is a well designed piece of hardware, and the user is able to get the system up and running quickly, and take their first spectra readings within the hour. Most of the confusion comes from their online website, which confusingly serves up multiple versions of files, and there isn't even a newest version that contains all software and firmware together with the GUI. This one roadblock is enough to waste time of the purchasing engineer, who may be left questioning the device. But once you get the device up and running, this is a slick little product. Future updates to this product should include updates to smooth out the install process, the GUI layout itself, and the CSV export in windows, which at the point of writing still only outputs absorbance values only. The EVM pack would be even more useful if the sensor probes were included in the EVM kit. These probes make liquid probing a lot easier, removing the need for a special glass test vials or slides for testing.
A couple lessons learned from myself: I need a better method of investigating wet samples. You don't want to damage the module, so you need a safe and reliable way of getting a wet sample right up against sensing window without contaminating or damaging the unit. Most flasks and test tubes are not ideal, since they are curved and I noticed that this really messed up readings. The ideal scenario is to get an external probe so that you can in close without risking damage to the unit (just like they would be doing in industry). Alternatively for small batch tests of wet materials, some sort of flat glass container or a test slide could work. I guess you could even put the wet sample in plastic wrap, but I wouldn't trust that. Below is an example of my folly in scanning my red wine in a curved glass. Even if I get in close to the sample, there is so much light scattering occurring, which seems to generally lead to a very noisy signal output, and depression of the real peaks. My only guess is that this geometry is really affecting the path length of the beams, and that different amounts of transmission in the glass medium are really distorting results.
It is a very good review in my opinion. The first part is nice but I have appreciated the test of the different products and the analytical approach.
The problem with wet objects is the water absorption and reflectance issues.
There are some processes and algorithms you can use to run a baseline water scan and then you have to subtract that…
As for the choppy spectra from IR remote - I also noticed this, but in my opinion that does not depend on the quality of IR LED. I noticed that this device does not work well with pulsed radiation…
I did not interpret the data of Hadamard scan yet, but column scan gives me results I kind of expected.. I cant upload images here for some reason, website crashes..
That could account for the gaps.
If you only have energy being collected over a short time, then the device would be off while the spectrometer is expecting input.
Does the spectrometer have a setting where you can have it dwell longer on each frequency? That would give you time to push the buttons multiple times while it is looking at each wavelength.
If you have a spare remote, you might just rewire it to turn it on continuously while you scan.
I will eagerly look into this issue when I can get back into my lab and get my pulsed IR source.
I have a hunch that the pulsed source is simply affecting the spectrometer acquisition, which takes place over 1-2 seconds depending on the scan config. I read that the default "Column Scan" method "selects one wavelength" at a time. The manual doesn't really specify the exact acquisition procedure and how the scheduling works, but I imagine the pulsed source wreaks havoc with this, and this is likely gives my result which looks like spectral chopping.
I believe that this IR LED should have a fairly regular (and smooth) Guassian peak, but it looks chopped simply by virtue of this spectrometer's default acquisition scheme. The other scan scheme they offer is the Hadamard, which "creates a set with several wavelengths multiplexed at a time and then decodes the individual wavelengths."
I'm still not crystal clear on what the Hadamard proposes, but it sounds like the acquisition might be less affected by a modulating source. However my first scan at home with a wimpy TV remote doesn't not seem yield a much better result.
I'll have to do some more acquisitions with my better IR remote before I can confirm this with more confidence.
I would check on the various codes used in IR remotes.
The serial code is standard across devices, so you should be able to resolve the output.
I am still kind of puzzled as to why that would cause any spectral pattern unless they have added multiple doped materials in the substrate to cover a wider range of input filters for better spectral response.
It could be worth contacting the manufacturer of the device to see if they will provide some information.
Maybe, depending upon the wavelengths, you get water reflection and absorption effects that can cause dead bands in your reflectivity and absorption measurements.
As I said, water is tricky stuff and the bane of spectroscopy measurements. Though in the right situations, very useful at the same time.
You get interface results at each material interface and with each substance.
So you get effects of the glass, liquid, plus the substances in the liquid.
Yes, you want as flat a surface as you can get to reduce the light paths. The curvature can also introduce lensing effects, which also cause issues.
Yes, spectroscopy is a fascinating field. You never stop learning new things.
You guys have me intrigued.
I wonder if these IR LED's can be adjusted over the 940 to 1600 nm range by PWM or frequency changes?
Could be a really cool experiment if you have the time. It would greatly increase a use for these simple LED devices for other applications.
A puzzle to be solved.