Comment by burfog | Hacker News Reader

Comment by burfog | Hacker News Reader

It hurts me to know that I can't cheaply buy a spectrometer, colorimeter, spectrophotometer, densitometer, spectroradiometer, turbidimeter, or similar. There are toys sold with essentially the same level of complexity that go for roughly 1% of the price. It seems that silly games are far more popular than chemistry labs.

Consider a 3-channel colorimeter. You need a pair of tri-color LEDs, a squareish glass test tube, a power source (a USB cable or battery holder), a chunk of black plastic, and a multimeter. In place of the multimeter, a USB interface can be had via a very cheap chip. Let the user move the tri-color LED, and you get a turbidimeter too.

The whole OLPC XO laptop, with that custom screen and support ASIC, cost only about $188 to make.

The problem is popularity. Almost nobody wants to study science. Meanwhile, we can get Pokemon Power Action Pikachu for $19.95, or Nerf Laser Ops Pro (2-Pack) for $41.25, Boxer (Interactive A.I. Robot) for $64.99, and so many other useless things.

I designed, and my old company sells, a NIR/VIS spectrometer at around the £800 mark (I forget if that's Inc vat). It's easy to use, has a WiFi interface, a simple python API etc. It's a scientific grade instrument and comes calibrated.

The opinions here are mine, and I've left out any commercially sensitive stuff.

Why are these things expensive? The optical design is not trivial. Yes, it's fundamentally two mirros, a grating and a detector. But you also need a good optical bench inside which means baffling, multiple prototypes to figure out how to reduce stray light, making it strong enough to resist handling, etc. Mostly this means a CNC base.

The electronics for a precision instrument aren't cheap either. The ADC has to be fast and low noise which is expensive. The linear detector itself is expensive for what it is. The big companies put stuff like order sorting filters and coatings on the sensor - that's expensive. The OPLC is largely commodity components in a common configuration. They're also selling more units than we are!

Also don't forget labour. Although we went through several iterations to make assembly as easy for the engineers as possible, they're still all hand made. That takes time and so does calibration. You're paying for that, and so you don't have to buy a line emission lamp yourself.

So yes, Ocean Optics has a monopoly and they probably overcharge. We're about 0.3-0.5x the price with slightly lower performance accordingly. Going lower would be possible, but to be honest without the volume, BOM cost kills you.

http://www.is-instruments.com/phone/miniature-spectrometer.h...

I wonder what kind of experiments/projects you will do with given instrument.

My company tried to sell "cheap" (as in ~1000/1500 USD) VIS/NIR spectrometers, but it was very hard to actually find some viable market.

You can actually find modules[0] that will give you quite a good performance for given price. Electronics needed to run those devices are quite complex, but I even seen some open hardware boards available that might be half decent. Still, you are going to be in territory of 500 USD even in fully DIY build.

[0] https://www.hamamatsu.com/eu/en/product/photometry-systems/m...

"My company tried to sell "cheap" (as in ~1000/1500 USD) VIS/NIR spectrometers, but it was very hard to actually find some viable market."

You'd have done well with the mining market if you had your resolution down to 5nm precision. 15nm is just a bit too much of a drift range to allow for reasonably accurate mineral identification.

Interesting - the system we sell is better than 5nm (depending on the fibre) and is in that price bracket. What's the application in mining?

Soil sample analysis so you can see if there's specific ores present.

Specifically, direct on-site, without the need to go transporting to a lab. IR spectometry is a wonderful thing, look up ASTER and LANDSAT 7/8.

I'd teach AP Chemistry. It's useful for teaching AP Biology too. The equipment, with some application-specific packaging and instructions, might also work for: aquarium keepers, gardeners, pool owners, beer brewers, people on well water.

Full DIY builds are cheap for a colorimeter, photometer, and turbidimeter. One can make all three for about $30, not counting labor and a computer.

The trouble is manufacturing volume. Sadly, not every student takes AP Chemistry. I realize that the manufacturing processes with low per-unit cost (casting, forging, blow molding, extrusion...) all have high set-up costs. This really is my complaint: if people cared about learning science as much as they cared about pointless toys, the production volume would let us have these things at prices ranging from a few dollars (simple colorimeter) to a hundred dollars (stuff with decent lenses and moving parts). The affordability of low-end cell phones makes it clear how low prices could go. I know you're having trouble due to the costs associated with low volume, but to an ordinary purchaser your price looks like a rip-off.

FYI, there are about 320000 people learning AP Chemistry, of which about 9600 are homeschooled and thus unlikely to reuse equipment from prior years. Assuming the equipment in schools needs replacement every 20 years, perhaps 25000 could be sold per year if the price were low enough to not be a budget problem. I'm thinking you'd have to knock about 90% off the price, and I suspect that 25000 per year is nowhere near the volume you'd need to justify the set-up costs of high-volume manufacturing processes. I may be wrong though... maybe you can get a good deal on plastic molds or custom aluminum extrusion dies or molds for some sort of casting process.

>I know you're having trouble due to the costs associated with low volume, but to an ordinary purchaser your price looks like a rip-off.

Yup, being a hardware engineer makes you appreciate just how cheap some things are. Bear in mind that our motivation to develop our instrument was shock at how much the incumbents were charging! I think getting to O($500) for a quality instrument without high volume would be fairly doable with some clever design work. Getting to O($100) yes, if you could make the entire optical bench as a MEMS product and you have scale on your side (this has been done partially by SCiO).

A big limiting factor is that there aren't many perceived use-cases beyond "science". The big consumer one that everyone touts is food inspection, but outside a few very specific uses, it's a gimmick to most people. Maybe for avocado ripeness... It's also not ready for primetime yet. The real markets are places like astronomy where people don't hesitate spending thousands on glass and silicon. Brewing is another. You want hobbies which are taken up by data nerds.

The other problem is that spectrometers are overkill for a lot of stuff. Colour measurement for instance - (calibrated) colorimiter ICs are incredibly cheap. For applications where you need the resolution you either need an experienced user or software to automatically interpret the results.

There are some other off the shelf solutions: Hamamatsu make a low resolution spectrometer that's good enough for general demonstration - costs a few hundred[1]. AMS make an 18-channel "spectral" sensor which costs $10 (ish?). There have been a few Kickstarters which are trying to address this e.g. the SCiO [2]. That's about the closest we've got so far to a mass market spectrometer and it costs $299.

The problem as you say is: how many schools can afford even $100 on a device which may only get used once or twice a year? Holding a transmission grating up to your eye will solve maybe 90% of the needs of most teachers. It's enough to show different line emission from a bunch of lamps. The public labs spectrometer would do for the other %10.

[1]https://www.consumerphysics.com/scio-for-consumers/

[2]https://groupgets.com/manufacturers/hamamatsu-photonics/prod...

I don't think MEMS is required. For $60 you can get a 2 TB hard drive. It contains platters that are several inches in diameter, and there is a swinging arm that is a couple inches long. All these parts move with incredibly tight tolerances. DVD writers are $26, and Blu-Ray writers are $120. Those are an even better comparison, with moving optics.

Food inspection is silly unless that is your job. People who keep aquariums, particularly saltwater ones, are a more likely market.

18 channels sounds great for most school labs. It's enough for a crude graph of chlorophyll's absorption and/or florescence. It might even be enough to distinctly identify the four main sources of color in spinach leaves. Repeated sampling at about 10 Hz would be plenty for all the labs involving reaction rates; most could be done with 0.1 Hz. For example, a popular lab involves measuring the changing rate at which Crystal Violet is destroyed by lye. Some less-demanding popular labs involve measuring the amount of copper in brass (as copper nitrate via nitric acid) or the molarity of blue food coloring in drinks.

Do you fancy chatting further about this? I've been toying with writing a hacker's guide to spectroscopy, and it would be interesting to know what sort of experiments schools would find useful. Feel free to get in touch at josh at iayc dot org.

> Food inspection is silly unless that is your job.

Nevertheless, that's what the majority of the consumer devices are going for. Ripeness detection, "quality" measurements (eg measuring brix) and so on. SCiO also mention things like pill identification and body fat measurements. This is the problem: these are all gimmicky things that get VC funding but aren't all that useful in the real world.

This is the AMS sensor btw https://www.digikey.co.uk/product-detail/en/ams/AS72651-BLGT...

I can speak about turbidimeters as this is instrument I've designed few of (as in 2 commercial products). Getting into 30 USD would be hard if you want any performance out of it.

Just source of light is going to be some kind of NIR laser, VCSELs are getting cheaper, but it is still nowhere cheap as LEDs. Then you are going to need some form of amplification for your light sensor and most probably two of those. Then some ADC chip and related circuitry, power section and so on.

But thats just a board, you will need enclosure and you want something better than just a generic box as you are going to deal with stray light. Also add some vials so users have something to put the liquid in.

We are talking about DIY builds now, but if you want to sell those: you need to budget for packaging and sending products. But there are also going to be problems: some devices are not going to work or die soon, you will get returns, some products will get lost and you will have to pay for it.

It is dollar here, few cents there but those will add up quickly.

I'm not familiar with how to do any sort of spectral analysis with a few discrete LED elements. Any pointers for further reading on that?

A good way to get your hands on a true optical spectrum analyzer is to look for one of the Verity units as shown in Kerry Wong's video ( http://www.kerrywong.com/2015/08/16/yet-another-scanning-mon... ) and the Signal Path blog ( http://thesignalpath.com/blogs/2014/11/21/teardown-upgrade-a... ). I bought one and wrote some software for it after seeing Shahriar's video. Very cool toy. Unfortunately, they have become somewhat harder to find on eBay since those videos were posted, and quite a bit more expensive when you do see them for sale.

This would be an absorption photometer or something similar. You shine light of a known wavelength through your sample and measure the intensity that's transmitted. If you have enough wavelengths and they're useful (ie you're looking for some feature in your sample that is either present or absent, or is a proxy for something you want to measure), you can do some basic analysis of your sample. It might be enough for QA in some industries.

You can also do fluorescence spectroscopy where you illuminate with multiple wavelengths and image in multiple filters (or a single filter). You end up with a grid of measurements that can be used to fingerprint samples.

Monochromators are a cheap way of measuring spectra, but since you need to measure each wavelength individually they're pretty slow. Most lab spectrometers are real-time.

There's a lesson to learn, that people will still gravitate toward simple feelings and notions even if tech gives you all science on your desk for 10$

Reply on news.ycombinator.com