PDA

View Full Version : Spectrometer in MAKE:Magazine #14



Adrenalynn
05-21-2008, 04:51 PM
So did anyone else have their mind blown by the incredibly simple arduino-driven spectrometer in MAKE:14? [Page 134-135] by Eric Rosenthal?

For those that are living in a cave, you can check the whole project out here on Mr. Rosenthal's site: http://creative-technology.net/MAKE.html Really - if this doesn't make you subscribe to this magazine please stop reading right now...

It's psycho-crazy-simple, but man is it cool. If you already have an Arduino board, it's probably $15 in components at its simplest, sky's the limit from there...

Why am I posting this here? Think about the applications of your robot being able to do spectranalysis! Meta - how about allowing your robot bartender to positively identify the alcohol it's grabbing? Building its own index of what it has available in the bar?

This is a transmittive device as its written today, but there's no reason it can't be made reflective. Want to build your own "mars" rover? Pick up rocks and soil samples and figure out what it's holding? How about identifying what room its in based on the floor sample or paint sample? It's tremendously cheaper than using a camera both in hard cost and in processing power. I have enough stuff in junk boxes to build an improved version of this device...

How would you improve it? Ultraviolet LED? Of course. Multiple LEDs of each color and a couple photosensors and do differential analysis? Duh. How about refrigeration of the LEDs to keep their color more constant?

The mind boggles (at least mine does) at how simple but powerful this is. I guess I just never considered what it'd really take...

metaform3d
05-21-2008, 05:53 PM
That is pretty awesome. It would be interesting to see if it had the sensitivity to distinguish gin from vodka. Perhaps you could use an incandescent source and a moving prism or diffraction grating to improve the frequency range.

Adrenalynn
05-21-2008, 06:00 PM
That's also a possibility, but wouldn't that take us back to a full-fledged camera?

I'd be willing to test and see if it has the sensitivity. I'm going to build one in a few days...

I'm betting that it would, especially with the addition of at least longwave UV. The photosensor is IR-sensitive, and it has an IR diode by default. I'm betting those two would differentiate, but only one way to find out! Now, if it can distinguish brands and tell me if they really set a top-shelf tequila down on the table - that'd be cool (although I can tell by odor anyway ;) )

LinuxGuy
05-21-2008, 06:32 PM
For those that are living in a cave, you can check the whole project out here on Mr. Rosenthal's site: http://creative-technology.net/MAKE.html Really - if this doesn't make you subscribe to this magazine please stop reading right now...
I don't live in a cave. I just can't afford to travel to all the neat events and conferences I'd like to go to. I read the project pages and can see definite possibilities here.


This is a transmittive device as its written today, but there's no reason it can't be made reflective. Want to build your own "mars" rover? Pick up rocks and soil samples and figure out what it's holding? How about identifying what room its in based on the floor sample or paint sample? It's tremendously cheaper than using a camera both in hard cost and in processing power.
When W.A.L.T.E.R. v2.0 goes active, he will have an arm that can easily reach to the ground and pick stuff up, as well as store it on his cargo deck for transport. :happy: The arm is designed and mostly assembled now (5DOF total) and I just need one more SES bracket so I can attach the 2DOF Little Grip to it. I couldn't get the flexibility I wanted with just 3DOF for the arm.

This arm is very flexible and has one double joint. There is an IR and Ultrasonic ranger on the end of the arm with the gripper, and it would be very cool to be able to add a reflective spectrometer there also. Hmmm, identifying colors - this has whole new possibilities. SFE just came out with a cool digital color sensor also - I want one. :happy: I'd definitely like to experiment with this spectrometer stuff.


I have enough stuff in junk boxes to build an improved version of this device...
Are you going to design and make a reflective version?


How would you improve it? Ultraviolet LED? Of course. Multiple LEDs of each color and a couple photosensors and do differential analysis? Duh.
For myself, I would want to concentrate on the reflective spectrometer. I'd want to be able to locate the actual sensor assembly away from the processing node so it could be attached to the end of a robot controlled arm. I'd want to get as much accuracy as possible to reduce error.


The mind boggles (at least mine does) at how simple but powerful this is. I guess I just never considered what it'd really take...
I am getting pretty boggled over this myself, at the possibilities. I'd like to test the reflective spectroscopy in RoboMagellen environment one day or something similar. Most of the RoboMagellan folks are depending on vision and image processing to detect the orange cones. I've got ideas now!

8-Dale

metaform3d
05-21-2008, 06:33 PM
That's also a possibility, but wouldn't that take us back to a full-fledged camera?Dunno. I'm a bit uncertain how the photodiode "sees" the entire range of wavelengths. If it's only sensitive to IR is it depending on florescence effects?


Now, if it can distinguish brands and tell me if they really set a top-shelf tequila down on the table - that'd be cool (although I can tell by odor anyway ;) )Ha! A robot with taste would be an excellent roboexotica project. A robot that turned up it's nose at tequila mixto would be cool.

Adrenalynn
05-22-2008, 01:59 AM
Meta,

I'm hypothesizing here, as always, this is open to debate. [always the scientist...] I'm sure the majority of this isn't new to you, I'm just setting the stage for the answer - and it might be new to others...

I've done quite a bit of spectranalysis using prisms in extension tubes for astrophotography. The light coming into the prism isn't really "white", so when the spectrum splits you can measure each "band" (wavelength) and analyse the composition. The key is to look at each color band's width.

A photodetector is just a photovoltaic. The overall intensity of light hitting the detector is measured. No spectrum, just intensity.

So when each color of light hits the sample in question a certain amount will be absorbed and some amount will pass through the sample and hit the detector. We're measuring just what actually impacts the detector.

That's my reasoning behind stating that a prism or diffraction grate would bring us back to requiring a camera.

Make sense?

------

Wouldn't it be neat to have a 'bot that could sort something like Tequila by its grade? That'd be a hoot - yes, it'd have "taste"! Awesome!

Adrenalynn
05-22-2008, 02:24 AM
Dale,

>> I don't live in a cave. I just can't afford to travel to all the neat events and conferences I'd like to go to. I read the project pages and can see definite possibilities here.

I'm not suggesting traveling around to the events, rather subscribing to MAKE:Magazine. http://www.makezine.com Some issues are better than others, but overall it always gets my mind racing...

>> and it would be very cool to be able to add a reflective spectrometer there also.

That's my plan, put it on the claw of the CrustCrawler arm!

>>Are you going to design and make a reflective version?

It's really just a small extension. Positioning the detector(s) for reflected light measurement, and some code change. But yes, that's the plan!

>> I'd want to be able to locate the actual sensor assembly away from the processing node so it could be attached to the end of a robot controlled arm. I'd want to get as much accuracy as possible to reduce error.

Yes, exactly. I'd whack the display, no use for it in my application (and adds expense, weight and power-draw) The detector and diodes would be on a wiring harness with a "cup" to block the light. The unprocessed data will end-up at the Make Controller and then the combined data packetized and shipped back to a PC for more advanced processing.

>> test the reflective spectroscopy in RoboMagellen environment one day or something similar. Most of the RoboMagellan folks are depending on vision and image processing to detect the orange cones.

Probably not the right application. You'd need to enclose the cone in total darkness surrounded by the LED and photovoltaic, and then measure the light reflected off it from each LED. That's not really all that sensible I don't believe.

This is really for analyzing a small sample in a semi-controlled environment.

Adrenalynn
05-22-2008, 11:13 AM
Dunno. I'm a bit uncertain how the photodiode "sees" the entire range of wavelengths. If it's only sensitive to IR is it depending on florescence effects?


DOH! Wait. I get it!

You propose moving the sensor through the entire light path at some known rate and sampling the intensity of the entire spectrum generated at any given point in time defining a known path, right?

That just might be genius. Have to give that some thought!

Sorry I misunderstood upon first reading!

metaform3d
05-22-2008, 01:42 PM
You propose moving the sensor through the entire light path at some known rate and sampling the intensity of the entire spectrum generated at any given point in time defining a known path, right?More or less. If you used a prism rotating it would shift the spectrum across the sample. You'd need a pretty bright incandescent light source -- something like a projector bulb.

Adrenalynn
05-22-2008, 01:52 PM
That would work too. Some compounds might not appreciate the heat a projector bulb is going to generate...

"Oh, I guess that was nitro sample, huh?" ;)

Doctor Robotnik
07-27-2008, 03:51 AM
N0ob here with stupid questions.

1. How is it going with the spectrometer?

2. Could use use a beamspliter to split the image from the sample so one part of the image goes to a camera and the second goes to the spectrometer, giving the bot..smision? It would be neat if it could determine the composition of something just by looking at it.

Adrenalynn
07-27-2008, 05:53 AM
Sorry I didn't follow up more. I'll dig it out and shoot some photos at some point.

I ended-up using a combination of sensors - CS photocell for longwave UV flourescense and IR sensitive photocell for absorbtion. I didn't explore reflectivity.

I think what you're suggesting is similar to what we do in astrophotography. If you have something that emits light or absorbs certain wavelengths of reflected light, you can use a diffraction grating or a prism in front of a camera and sample the width of the "rainbow" the prism splits out. If you have narrow color bands or some bands missing altogether, it becomes pretty obvious what the chemical makeup is.

Generally, though, most sources don't emit light that way, so you need to set fire to them to figure out what they are. I guess you could build a robot with an adequately large laser and just fry anything you're curious about. ;) ;)

The applications for this spectrometer as-built that I immediately see: It's great if you want to build a robot that searches for blood stains or pet spots. :) There's an application for installing it on a robotic arm and having it sort liquids that it grabs, great science-faire project. You could probably put "unknown" beverages on Metaform3D's robotic bartender and still get a perfect drink. I just kinda ran out of applications that excited me, so it got put in the "dead project" drawers and I moved on. I have a short attention span once the initial problem has been explored - it's a personal failing...

Electricity
07-27-2008, 11:43 AM
Well that IS pretty cool. I actually just ordered an arduino, maybe i'll order up the rest, and try it out!

LinuxGuy
07-29-2008, 02:34 AM
I'm not suggesting traveling around to the events, rather subscribing to MAKE:Magazine. http://www.makezine.com Some issues are better than others, but overall it always gets my mind racing...
Right now, I have more than enough ideas of my own to explore. I don't need a lot of additional outside stimulation of my brain at present. :happy:


It's really just a small extension. Positioning the detector(s) for reflected light measurement, and some code change. But yes, that's the plan!
Very cool. I am extremely interested in anything I might be able to use on W.A.L.T.E.R. for environmental or warm body (animals, Humans, etc) detection and mapping. At some point, I need to work on the "L" and make the "E" more proactive, and I am excited about the possibilities. :veryhappy:


Yes, exactly. I'd whack the display, no use for it in my application (and adds expense, weight and power-draw) The detector and diodes would be on a wiring harness with a "cup" to block the light. The unprocessed data will end-up at the Make Controller and then the combined data packetized and shipped back to a PC for more advanced processing.
I am also interested in potentially using a MAKE controller as a peripheral processor to Hammer. I SO wanted to get both an Arduino and Proto Shield kit as well as a MAKE controller at OSCON this past week. Just no money to do it. :sad: I want to put a display on W.A.L.T.E.R. also.

8-Dale