DIY High Resolution Spectrometer (2018) | Hacker News Reader

DIY High Resolution Spectrometer (2018)

publiclab.org

101 pointsdisago32 comments

Anyone know of active DIY higher-er resolution spectrometer projects using image sensors? Is it possible to create a sharp image of 940-1240nm 3000 pixels across with less then $3000 in parts? I want to learn how to make a broader(uv to ir) but crappy-er version of this: https://solarsystem.nasa.gov/resources/390/the-solar-spectru...

I will add that I am willing to do some advanced image processing and 3D geometry to make this happen. I am unimpressed by projects I have seen so far, but it's not fair assessment because I don't know the field very well.

That's a beautiful link, thanks for that.

I'm probably missing something obvious, but could it be done by using a larger number of inexpensive setups?

There's no reason that each camera has to be focused on the same part of the grating, right? 4x cameras, point each at a different region and stitch it in software.

Yes you can stitch: have 4x cameras, or move the camera and take 16 images, or a single detector are all going to work. My question is what is the lens/optics set up? I think the answer in understanding optics and or know the right terms and or understanding all the different types of spectroscopy. It's a bit to look into. It's seems someone would be publishing DIY things like this already, but I don't know the right terms to search for.

It's easier to rotate the grating. That's how we did this in labs in undergraduate (we did a session on imaging the sun with a high res spectrometer). Or put it on a stepper or servo motor and sweep it in several positions. We used an SBIG camera and a micrometer-adjustable spectrometer. It seems like SBIG got bought up by the folks that make MaximDL. They used to sell these instruments off the shelf for $$$$.

It's more complicated to use multiple cameras on one grating because you'd need far more light shielding.

You might be able to achieve it by placing the diffraction grating at the focal point of the optical path. Given the incident angle would vary across the field you would probably have to calibrate it and realign the samples, but it seems at least plausible.

I don't know how long most commercial imaging sensors typically go, but you'd also have to do some work to remove the IR filter that's typically added by the manufacturer. There are a number of tutorials for this.

> Anyone know of active DIY higher-er resolution spectrometer projects using image sensors? Is it possible to create a sharp image of 940-1240nm 3000 pixels across with less then $3000 in parts? I want to learn how to make a broader(uv to ir) but crappy-er version of this: https://solarsystem.nasa.gov/resources/390/the-solar-spectru...

The linked project uses a 4000px wide image sensor and the second example in the article is a solar spectrum like what you want. It's not as sharp, but it's remarkably good given that the diffraction grating was a fragment of a DVD and the opening was a 3D printed slit.

940-1240nm isn't "uv to ir" like you described, though. That's just pure deep infrared. You'd have to test to see if the camera's image sensor was sensitive in those regions with any filters removed, but I doubt you'd get much signal in the deep IR region. Good sensitivity that far outside of the visible range requires a specialized sensor. Some security cameras designed for "invisible IR" will have such a sensor inside, so you could start by looking there.

Not going to happen!

Photons of different wavelength are absorbed and converted at different depth in silicon sensors. Blue photons mostly convert at the surface, while red go much deeper into the sensor. The pixel electronics of a regular cmos image sensor however will only collect the resulting electrons in a certain region at the surface. If an electron is generated somewhere else, it will recombine and be 'lost', not contributing to the signal.

Depending on the sensor, at 1000nm only <10% of the photons will result in electrons that are captured by the pixel electronics. The sensor does not give you any signal any more.

Why? Silicon has a band gap of 1.12eV - photons with lower energy will not interact. This corresponds to ~1100nm - you will need sensors with different materials

Thanks, this was very helpful(and interesting) for the IR part of the question.

You'd better start saving for a cryo setup as well then, because even if your sensor has the right sensitivity at those wavelengths it will be hard to avoid contaminating the results if the gear itself is not chilled, unless the signal you are sampling is overwhelmingly stronger than the stray photons produced by the gear.

Worth checking out: https://hackaday.com/2021/04/23/pi-based-spectrometer-puts-t...

Have a look at the work by Christian Buil [0]. He's been building incredible DIY astronomy widgets for years.

Off the shelf you're looking at around $8k for a spectrometer from Ocean Optics. Just point the fibre at sun on a white PTFE plate outside (teflon has pretty uniform reflectivity) and it should work. You generally need an InGaAs sensor to go beyond 1000 nm with any appreciable sensitivity. Silicon is still sensitive to around 1200, but performance drops significantly. The solar image from NASA is only in the visible anyway, but as I mentioned in another post the simplest way to do this is to use a high lines per mm grating (e.g. 1200) and rotate it so you capture multiple spectra in shorter ranges.

[0] http://www.astrosurf.com/buil/spectrographs.html

[1] https://www.oceaninsight.com/products/spectrometers/general-...

Do you need to collect all the wavelengths at the same time? If not could you simply rotate a grating/prism and use a single photodiode? Lots of spectrometers work like this.

When I worked for an imaging company, we used to have diffraction grating spectrographic plates (the DVD is a great idea) embedded in 35mm slide mounts.

We’d stick a standard CCD (like in any DSLR) behind them, and voila! You have a spectrometer! This would also give you a good optical path, for focusing light.

You’d still need to calibrate the driver, but that’s just paperwork.

One thing that nobody mentioned yet is that many things absorb a lot in the uv, so focusing optics etc will be tricky. As other said getting an ir to uv range on a single detector is tricky to impossible depending on what you call ir and uv.

As mentioned below, the main problem is the detector. Most detectors are Si CCD. meaning the Si is fundamentally limited to 1100nm due to bandgap. Going beyond that means you need a different material such as InGaAs or Ge, unfortunately CCDs are not as common or are expensive for these materials. Which then gives you two choices:

- an array detector, which is pricey.

- a single detector such as a photodiode behind a second slit. a detector such as these https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=12... With a moving(rotating) grating, so that the wavelengths can be reached by moving the grating.

The latter is quite common in lab spectrometers.

"Photodiode behind a second slit": Thanks for the link. This is looks like a practical way to go, and in price range.

almost completely off topic but btw the o in spectrometer, speedometer, worldometer, whatever is called an interfix. you've heard of suffixes and prefixes, well interfixes unlike those two have no meaning other than to join two root words (as in words that can't be used as suffixes or prefixes).

You may be doubting if your comment was actually useful: thanks for posting it, for not discarding it.

I've been wanting a cheap, no-cloud raman spectrometer in my pocket for a while now.

Like something that can plug into something as universal as an audio jack on a phone, reflecting the laser back into its built in camera. I know there was some options out there that were cloud-based but I want one simply lets us interpret the samples and their readings ourselves with lab device quality output.

DIY stuff like this is ultra cool but it's like miniaturize-able for sure, and something you can run with open source software.

I know it's around the corner, I've read about it. It's like the next killer feature for new cell phones already. [1] Making it open source so some corporations who are doing cloudy things with your spectrometer data is definitely an international research security imperative.

[1] https://aip.scitation.org/doi/10.1063/5.0046281

I don't think smartphone manufacturers are incentivized to put better, cooler sensors into them; unless those sensors can be used to better spy on you, which I don't think a spectrometer can be.

Ever hear of drugs before?

Really cool! I wonder if this could be combined with a digital micromirror and compressed sensing to inexpensively recover both spectral and spatial resolution. (see: "single pixel camera")

I guess it makes sense if you already have a DSLR, but you can get a proper used spectrometer for about $1k which is surely cheaper and better than a DSLR.

How does this compare to Les' RPI photospectrometer?

https://www.hackster.io/news/les-wright-s-diy-raspberry-pi-s...

Does it just have a much larger spectrum?

A lower resolution but easier-to-construct variant: http://scheeline.scs.illinois.edu/~asweb/CPS/

Without an enclosure to get rid of ambient light, you are not likely to get a lot of useful data out of that.

Interesting. I garden and have thought it would be cool to build a spectrometer that I could feed soil samples and get NPK and other nutrient levels quickly without chemical processes.

The pdf mentions using a cfl to calibrate the position -> wavelength mapping, but wouldn't the peaks depend on the particular phosphors used for that lamp?

... and soon a small x-ray spectrometer? (even though this is illegal in most countries since x-rays are really dangerous).

It's easy enough to generate x-rays, for better or worse, but the devil's in the details. The x-rays apparently need to be monochromatic, or nearly so, so you can't just apply HV to any old random radio tube that's lying around. And then you need a beryllium exit window on the tube and another one at the detector entrance.

Fascinating teardown on one of these things at Ben Krasnow's channel https://www.youtube.com/watch?v=KdfHVcU8U7U .

Not the same thing, but I’d love to see an open-source XRF gun.

Reply on news.ycombinator.com