|Product Performed to Expectations:||10|
|Specifications were sufficient to design with:||10|
|Demo Software was of good quality:||9|
|Demo was easy to use:||10|
|Support materials were available:||9|
|The price to performance ratio was good:||10|
|TotalScore:||58 / 60|
This is a RoadTest review of a TI's DLP NIRscan™ Nano evaluation kit - a DLP based spectrophotometer for 900-1700nm range. It is important to disclose that I'm not a chemist/biologist or a physicist specialising in optics by any stretch of imagination, so some things written here should be taken with a pinch of salt. My background is in robotics and computer vision so initially I wanted to use this spectrophotometer mostly to evaluate IR sources in common RGB-D cameras, such as Kinect, KinectOne or Intel's RealSense. But quickly I realised that these cameras, of course, employ normal CMOS sensors therefore their IR emitters are in the range where CMOS is still sensitive (<900 nm), while the NIRscan spectrophotometer is specified for 900-1700nm range. Therefore I had to devise other experiments for this RoadTest.
Due to my background I've used the device mostly as a monochromator, not a reflectance spectrometer as seems to be intended by TI. I guess other reviewers, more inclined to bio-chem fields would cover these scenarios.
It would be good to discuss typical spectrophotometer designs, because TI's novel design took me a bit by surprise.
This is a classical one, following the original manual spectroscope design to large extent. This particular unit is Jasco V-630 spectrophotometer, used for bio-chem transmissive spectrommetry. It covers UV and visible wavelengths and the diffraction grating is moved mechanically (stepper motor in the middle) to select the wavelength of choice.
The source of light is either a deuterium lamp (UV) or a halogen bulb (visible + IR), this is switchable with the mirror in the top right corner. The source light goes through a filter wheel through an entrance slit to reach the reflective diffraction grating. This splits the light into various components, and by rotating this diffraction grating we can select which one makes it through the exit slit. This wavelength, in this particular spectrophotometer, bounces off a mirror and then via beamsplitter to create two rays to examine two samples at once and after exiting the samples the signal reaches the photodiode detectors (in some spectrometers photo-multiplier tubes are used).
This is another example from my collection, Ocean Optics S2000 spectrophotometer, visible spectrum range. It is way more compact and lightweight than the one described above. Instead of mechanically moving the diffraction grating it uses a linear CCD sensor, like TCD1304AP, similarly to the flatbed scanners. Here this sensor captures the entire spectra at once. This results in a much more compact design, but has some downsides. First of all such CCDs are not very sensitive for wavelengths beyond 1100nm and other materials, such as InGaAs are necessary. Such linear (or even 2D camera) sensors for far infrared exist (e.g. Hamamatsu makes them), but are very expensive to manufacture. Another problem is that the CCD array has finite resolution (e.g. 3648 pixels) to capture the entire spectrum (e.g. 200-1000nm).
(Image ©Texas Instruments)
This is where TI's design surprised me. It somehow never occurred to me to use a DLP mirror chip for spectrommetry. It is a very neat idea. This allows to use a single point InGaAs detector, which is massively cheaper (and possibly has better SNR) than linear/image array and still have the compactness similar to the linear CCD design described above. Here the optical path, after reaching the diffraction grating reaches the DLP chip and now the magic happens! Thousands of mirrors either direct the selected part of the spectrum to the detector or elsewhere to a beam stop. Nice thing about DLP chips is that they can flip these mirrors very rapidly (2880 Hz) - that's in contrast to a mechanically rotated diffraction grating in the mechanical spectrophotometer.
While over the years I amassed few useful things for optical/laser experiments from eBay or waste recyclers, I'm not in the possession of a comparable spectrometer (NIR/SWIR range), as these things are worth more than my entire home lab. For example Ocean Optics NIRQuest 512-1.9 is listed online at $25900. Therefore instead of comparisons I've decided to devise some interesting experiments that could put the NIRScan Nano through its paces. For most experiments I use the device in a monochromator mode - the halogen bulbs are disconnected and often I've made relative measurements - that is the reference scan was taken with the light source under test switched off to capture and substract background & stray light levels in my setup.
Let's start with a simple experiment. I've found a 940nm LED in my junkbox and pushed 20mA through the diode. The spectra is cleanly visible, and as it is an LED it is quite wide.
Unfortunately, as I've mentioned before, most of my IR light sources are under 900nm, therefore unsuitable for this spectrometer. So I've decided to get a cheap SFP transceiver off eBay to pull out a 1310nm laser - FTLF1318P3BTL from Finisar. I also wanted to keep things safe and this laser is Class 1 with less than 0.5mW optical power. Working with higher power lasers requires sturdy optical bench, beamstops, careful experiment planning and wearing safety goggles at all times - too much hassle and risk. Additionally, I think I went through most of TI's documentation related to NIRScan Nano (quite a lot! impressive in general), but I haven't found specification of maximum allowable optical power on the input. So staying below 1mW was not only to keep my eyeballs safe, but the spectrometer as well (I hope!).
I couldn't find any info on the laser extracted from the SFP module, but the closest thing I've came across online was FP-1310-4I-56A laser diode from the same manufacturer (Finisar). It helped me with figuring out pinout and the configuration of the laser diode and the internal photodiode.
Here's the spectra of the laser, it runs at 10mA, start lasing around 7-8mA. I would rather use the embedded feedback photodiode and put LDC-3722 into optical power control mode, but I couldn't calibrate it as this would require an optical power meter for such wavelengths and low power levels. The only one I have is a Coherent OEM sensor head that is good for very high power lasers (up to 100W), outputting 0.4mV/W and shining this tiny laser into it doesn't really bring it out of the noise, even on a 7.5 digit meter.
Here's the same setup, but with NIRScan Nano software configured to scan narrow part of the spectra, with very high resolution and averaging 10 scans as well. The steps in the dataset are only 1.15nm wide! The specs for the laser diode range from 1290nm to 1330nm and we can clearly see that even without much time to stabilise, running probably not at the 100% correct current (unknown part, just guess) the laser frequency is almost spot on (1306.4mm vs 1310nm).
Another experiment I came up with was measuring the pass-band characteristics of fibre optic cables. To do that I've used a cold light source - a glorified halogen lamp, with some DIY adapter for SMA905 connector (but no collimating lens, sorry). Halogen is a broadband source, covering wavelengths from visible light to far, thermal, infrared.
On the other side I've encountered a bit of a problem. The only attachment NIRNano came with was the one for reflective measurements (built-in 2 halogen bulbs). TI did a great job and on the website they've published the designs for few other adapters, including SMA905 fibre optic coupling, proper one with necessary lenses etc. However, these are only designs and I don't have such sophisticated mechanical engineering capabilities in my home lab to manufacture them myself. It would be good if TI could offer them through their network of distributors as optional accessories. If the price would be OK I would surely buy one. Unfortunately, for this RoadTest I had to improvise with a bit of PCB, SMA905 connector and soldered pins to keep the connector covering the sapphire glass input.
I think the sloppy optical setup, without collimating lenses and dodgy adapters contributed to poor and strange results of this experiment. The reference spectrum was taken with SMA connectors (of the light source and spectrometer) looking at each other. Then the measurement was taken with a fibre optic cable in between.
First subject was a telecom fibre patch cable - 50/125um OM2 multimode. Second one is a lab grade UV-VIS 200um core singlemode fibre cable. The plots below employ either Column or Hadamard scanning methods and the relative measurement is on the Y1 axis, while the halogen reference is on the Y2 axis. The strangeness is in the scale and noise. These were relative measurements and the values are tiny compared to the reference and look like being in the noise, no particular trend. So either these are excellent fibre cables with 99.9% transmission across the entire spectrum or there was something wrong with my measurements - I suspect the latter.
Another idea that came to my mind was to check various laser safety goggles. Possibly not to precisely measure their OD rating, as this is often massive for power lasers and probably exceeds the dynamic range. It was just to see how it affects the spectrum.
Here below we have Bolle safety goggles for YAG lasers, specified for 980-1100nm and OD6 attenuation. The plot on the right clearly shows that wavelengths in this part of the spectrum vanished completely.
Similar to the previous one, with the telecom 1310nm laser described above, but this time with some blue + IR laser safety goggles (190-450nm and 800-2000nm, OD4). With glasses on the laser frequency is gone.
Some of the readers might find this experiment interesting. Here I shined a green (532nm) laser pointer module (also <1mW optical power, stay safe!) at the spectrophotometer. Should be entirely out of the range of the device, right? Wrong. Most green lasers are using frequency doubling technique, therefore in addition to the desired colour (green - 532nm), we also have a component at half the frequency (1064nm - IR) and this we can observe well with NIRScan Nano.
So I wanted to also test the spectrophotometer in its intended configuration - to measure the reflectance spectrum. I wanted to check if Teflon (AKA PTFE) is really almost as good as Spectralon when it comes to reflectance. I've found a piece of PTFE, stuck NIRScan Nano to it and there was something wrong with the reflectance spectrum. So I thought the block wasn't PTFE but maybe HDPE, as I bought pieces of HDPE off eBay as well long time ago. But I was pretty sure I had a PTFE rod (also from eBay). I've checked the rod - the same non-uniform spectrum. So I've checked purchase history and it clearly said in the auction titles for both items PTFE. So, as a last resort, I took the plumbers teflon tape, few turns, stuck that onto the spectrometer and got the spectrum I anticipated from the start. Are my PTFE bits and pieces from eBay fake?
When the spectrometer arrived it required firmware update which went flawlessly. The software stack (on the PC side) is composed of multiple parts, such as USB communications library (wonder why it's using HID class and not something else like CDC ACM or vendor specific), DLP Spectrum library above that and finally a Qt based GUI. While I don't think it is officially supported, with some tweaks it is possible to compile all the software on Linux and it works identically to the Windows version.
I've only used USB mode, as this powers the device as well. I suppose Bluetooth mode might be useful as well, but for that to make sense I need to buy proper battery and connectors and integrate that nicely. The third mode of operation possible is completely standalone, with the spectrometer configured beforehand and running a scan with a press of a button and saving the results onto the SD card. I cannot see this mode being very useful - the device doesn't have an LCD display to check if the scan is OK, just saves everything I presume and if one is working in the field one has a laptop or a smartphone so the Bluetooth mode would be more appropriate here.
The device has a connector for trigger I/O so possibly the standalone mode with SD card can be used in conjunction with that in some sort of automated test system.
As you can see the cons are rather minor and of cosmetic nature. I was really positively surprised by this little device. From the novel (to me) DLP based design, to a ton of support material available (not always that common with dev. kits nowadays) to software good even for unskilled user. I guess professional users of spectrophotometers will stay with specialised companies providing complete solutions with support, accessories & certifications - the TI kit, however excellent is just an evaluation kit in the end, but I guess it might attract semi-professional and hobbyist customers.