RamanPi – The 3D Printable Raspberry Pi Raman Spectrometer | Hacker News Reader

RamanPi – The 3D Printable Raspberry Pi Raman Spectrometer

publiclab.org

63 pointsMichaelAO23 comments

Love the PublicLab's enthusiastic embrace of open source hardware and software licenses. If you don't want to 3d-print your own, you can get an awesome bluray laser spectrometer + webcam at their Kickstarter. https://www.kickstarter.com/projects/publiclab/the-homebrew-...

Participatory citizen-science FTW!

This is neat, but I wonder what the actual resolution of this device is (and how stable are the measurements)? Is it useful for anything other than just novelty?

Not sure that I really see the point of 3d printing the parts (although scratching many geek itches, yes) when your optics are going to be a much larger cost than the mechanical mounts, and the quality of the mechanical mount is so very important.

"when your optics are going to be a much larger cost than the mechanical mounts"

It would be nice if that were always the case. I suspect this is very much like my electronics stuff, a $2 MMIC amp mounted in a $30 aluminum diecast box. You can blow a lot of money on chassis and component mounting.

Also, speed. Could print something where an exact model just drops in place with superglue and a perfect fit. I could probably replicate their chassis in aluminum by hand in a days work, but I'd much rather go lazy and print one. Also each engineering revision drops from "days work" to "hit print and come back later"

A custom housing could certainly be on the order or more expensive than the optics, but if you were using catalog parts the mechanics should not be the largest cost.

For example, a "cheap" 1" notch filter from Thor would run about $500, whereas a kinematic mount would be somewhere closer to $40.

For Raman spectroscopy, a "cheap" notch filter won't even work. a 0.5" long pass filter made for Raman can easily cost $2000. When you have a $2000 filter, you have to use temperature stabilized laser, otherwise the laser wavelength shifts, thus, spectrum shifts. Grating and concave mirror can also cost a fortune.

Also, Raman typically is very weak, roughly 10^-6 of the laser intensity. The CCD/CMOS has to be low dark count.

The things is, if you have spent $5000 on components, why should you save couple hundreds of dollar on good optical mounts?

This 3D printed plastic base/structure will never have good stability and precision for any serious scientific application.

Nonsense, you can run a perfectly good student lab with a cheap notch filter and pull a usable spectra off something like polystyrene or Tylenol. What I'm wondering is if this thing could even do that. Looking at the abuse the diffraction grating has taken suggests the answer is no.

As far as a decent scientific bench instrument goes, we're mostly in agreement. I'd love to know if you could make a decent optomechanic system like this on a SLS with a metal substrate.

Curious: what are you doing with microwaves?

Radio's always interested me but I haven't done anything with it since high school.

Very good point, I would think before making a fancy case and adding useless colorful LED rings, you would try to prove the actual detection concept, which hasn't been shown by the maker.

Hey, I just want to thank everyone for the kind words I've been getting..! I'd love to hear any suggestions or comments anyone has on my project..!

Take a look if you're interested at the project page on hackaday.io... hackaday.io/project/1279

From the article:

Ordinarily, an expensive notch filter would be used which is cost prohibitive for most average people. My system avoids this cost by using two less expensive edge filters which when combined in the correct manner provide the same benefit as the notch filter...at the minimal cost of a little extra computing time.

If the goal is to cancel the incident light "carrier" frequency, couldn't this also be done interferometrically, just using mirrors and beam splitters?

That's exactly what expensive notch filters do.

Cool. I guess it was naive of me to visualize a piece of colored glass. :)

I'm not a physics grad, but I'm curious: can one predict the Raman response of a molecule? For example: suppose I'm interested in detecting molecule X. I have a laser of frequency f. Given these two, can I predict that the reflected light (after Raman effect) will definitely be of frequency f' ?

In principle, yes. The relevant equations can't be solved exactly for any non-trivial molecules. But there are a large number of different approximation methods, which boil down to a numerical optimization problem that converges on a bounding value.

These methods generally don't scale well, however, so as the size of the system grows, the approximations that can be successfully solved quickly grow too coarse to be useful.

There are also a lot of simple heuristic methods that a chemist can employ that don't require collaboration with a theorist, e.g. particlar chemical "motifs"/"functional groups" generally have a Raman resonance at a consistent frequency that is affected only slightly by the surrounding parts of the molecule (and the general direction and order of magnitude of that shift can also be approximated heuristically).

Finally, there are enormous catalogs of recorded spectra for a huge range of molecules at e.g. the NIST webbook (http://webbook.nist.gov/chemistry/).

Thanks. Followup question: does the light source have to be a laser? I read on the Wiki that CV Raman demonstrated the phenomenon with sunlight. So would it be possible to just observe the reflected light from an object and determine the molecules on it that are lighting up? If so, can (someday) someone come with a Star Trek-style tricorder that you look through and figure out the molecular composition of an object? Maybe at a distance of 100 meters?

More or less, you start with a quantum chemical DFT calculation, the Raman frequencies are associated with the vibrations of the molecule (second derivatives of the energy, once you're at the relaxed potential energy state). Raman frequencies are shifts (Stokes and anti-Stokes) relative to the laser frequency.

For medium sized molecule, a hybrid DFT calculation (which scales in computational time with N_eletrons^3 ) would cost a few CPU days of time, giving pretty accurate frequencies (there are known correction factors of ~0.95x to compensate for systematic failings in the theories). What we are very bad at is predicting the Raman intensity from theory. But you get the frequencies in the correct order, and can use this to assign observed peaks to particular vibrations within the molecule.

http://www.gaussian.com/g_tech/g_ur/k_freq.htm

How does this compare to vessyl[0]? Is building one of these harder than building a radio kit?

[0] https://www.myvessyl.com/

They are opposites, in that the vessyl site strictly discusses apps and salesmanship without mentioning how it works whereas this project is strictly how it works and what it does and you figure out what to do with it.

I enjoyed reading the vessyl site because it explained nothing at all about how it works, so I immediately applied my engineering gut sense to trying to figure out how I'd make one if I wanted to. Small scale calorimetry heating and cooling with a peltier device a couple degrees at a time? That works for metals and phase transitions but probably not enough to tell diet sodas apart. Some kind of high res ultrasound to analyze waves and thus viscosity and thus density/composition? Personally I'd go all EE on it and shove a modest AC signal thru a big capacitor and get the dielectric constant of the "stuff" in the cup. A really accurate capacitive sensor to tell exact volume with a really accurate force sensor on the base to read weight/mass and a decent thermistor to calibrate for temp and you've got density at a specific temp. Once the raw data is in there, I'm guessing its all just lookups and best fits and the like.

I'd do all of it, and then pull it apart with machine learning.

it differs in that it's not an obvious scam

.. or how about a Spectruino:

    http://myspectral.com/

My big question is - what can these be used for? What are they actually useful for? I guess .. gardening? Medicine?

spectruino's ADC isn't fast enough to capture all the data from a CCD data frame, thus to get an entire spectrum one needs to sample multiple times and shift the ADC sample clock... or deal with not having all the pixels. This method would be fine if your samples aren't changing, but for something like enzyme kinetics this non-simultaneous method will likely yield poor results. For applications that can use a monochromator, this method should be fine (albeit with lower spectral resolution since monochromators usually have quite high spectral resolution)

I'm still yet to find an actual use for the thing, regardless of constraints/limitations. The fact is, until we have a use of the hardware which allows average-person to gain some advantage, we have little progress.

Can I check for Ebola with the thing?

Reply on news.ycombinator.com