spectroscopes; the end is near?
Page 4 of 4

Author:  Brian [ Fri Jan 16, 2009 2:20 am ]
Post subject: 

Whoa JB, perhaps you read something in my statement that isn't there. The comment was a specific example of my general lack of understanding of all commercial activity. As for me, I haven't looked at my paystub in years.

gemscientist wrote:
The cost-benefit of a digital spectrometer is not clear to me given the existing body of applicable information.

This seems to be the general consensus of opinion. Even I agree.

gemscientist wrote:
However, for it to be beneficial, the best approach would be for several members of the forum to put together the exact same set-up ...

Sounds like communism to me! Or is it elitism, I get the two confused. um... seriously... given the same stone, its relative transmission (or absorption) spectrum should be reproducible using different equipment. Not everyone uses the same spectroscope; why should everyone use the same spectrometer et. al.?

Author:  MacGyver [ Fri Jan 16, 2009 10:52 am ]
Post subject: 

Brian wrote:
Whoa JB, perhaps you read something in my statement that isn't there. The comment was a specific example of my general lack of understanding of all commercial activity. As for me, I haven't looked at my paystub in years.

OK Brian, please don't think I'm picking on you, but this reminds me of a line from the movie Ghostbusters:

Dr. Ray Stantz when Dr. Venkman was fired from the university wrote:
Personally, I liked the university. They gave us money and facilities, we didn't have to produce anything! You've never been out of college! You don't know what it's like out there! I've *worked* in the private sector. They expect *results*.

Here's a soundbite I found if anyone wants to set the Way Back Machine to 1984. :) ... esults.mp3

Author:  Brian [ Fri Jan 16, 2009 11:13 am ]
Post subject: 

I'm sure many people would question why I am paid for what I do.

Author:  MacGyver [ Fri Jan 16, 2009 11:16 am ]
Post subject: 

Brian wrote:
I'm sure many people would question why I am paid for what I do.

I don't care why they pay you. I'm just glad that they pay your way so we get to reap the benefits of your intelligence. It's a self serving attitude, but what the heck, I'm a private sector capitalist. Image

Author:  gemscientist [ Fri Jan 16, 2009 11:59 am ]
Post subject: 


I find that commonality of equipment really helps with communication and productivity, whether it's elitism or communism, no matter. In the companies I run, all the execs and senior management get issued a MacBook Pro with the same suite of software on day one. Since we are spread out all over the globe there are no incompatibility problems and trouble shooting is easy.

If two or three of us had the same spectrometer set-up we could help each other with techniques and resources and I think be more productive.

I know that Gene (G4Lab) could fix us all up with some cost effective units. Gene, how about telling the group what you have to offer?

Author:  jleb [ Sat Jan 17, 2009 5:20 am ]
Post subject: 

Bruce, it's not that bad.. that link I posted will generate a color spectrum and the software will de-bayer your sensor, generating a color image like the books. Nice thing is, it's free...

For pattern matching and validation, if you run the data through the software, you'll be looking at something much like an opl. If you wanna geek out ( many of us do) then there is lots of fun to be had w/ checking your stones.

while I'm dying here from this long term hacking cough I seemed to have gotten over the holidays, and laying in bed doing absolutely nothing while my sinuses and lungs decide they belong to me again, I picked up two connectrix b&w web cams off of ebay. Got one for $1.00 and the other for $5.00. (make sure the wine glass you pick up at the new years party was the one you set down.. and that whoever is going to plant a wet one on you.. is healthy!)

Hopefully i'll be better soon and can play with them after I clear my cutting backlog.

Author:  G4Lab [ Sat Jan 17, 2009 7:54 pm ]
Post subject: 

jleb has used the term ghetto spectroscopy in this thread. It could have various meanings but I am presuming that it means clever results obtained on a shoestring.

But in your last post you mention debayering and other conversions that software intended to display spectra taken from color cameras needs to use to end up with the spectrum graph we are familiar with.

I for one don't want data taken that way. There are too many things done to the data that you don't know about unless you are the architect of the camera. The newer the camera the more complicated the digital architecture. You simply can't publish data taken that way.

I hope Brian will discuss this matter. I am working on some Ocean Optics gear with Brian and Alberto and Brian does not even want to use Ocean Optics software for anything except recording the rawest form of data. He won't even let it subtract dark current or a reference spectrum which should be a fairly easy calculation. But if he is going present the data at a meeting and have somebody from a competing team ask him a question about how the data was handled he doesn't seem to be willing to trust the software engineers at Ocean Optics and actually why should he. And this is software written solely for recording spectra produced by a company whose main product and distinctive competence is the design and manufacture of modern miniature spectrometers. They are a market leader and have been at it since the early nineties.

One of the reasons I have not bothered with the astro software is you can get the same wiggly line out of Photoshop. IMO the only reasonable way to do that is with a black and white camera and the minimum amount of processing. But even the simplest cameras have gamma corrections built in that change the response of photocurrent to light incident on the sensor.
Unless designed to be used as a linear sensor( Rare , cameras built by Hamamatsu and a very few others and way more expensive than any of the gear we are talking about just for the camera alone) they have black level and white level adjustments (which in the old days of TV used to be called AGC automatic gain control) unless you can disable this and calibrate against some kind of repeatable (even if not absolute nor traceable) brightness standard then faint bands may or may not be represented in your results and and strong bands may show up as different than they actually are. There are some video cameras designed to have this level of control but they are not webcams and so do not have a built in frame grabber to digitize the signal. And the frame grabber ALSO has to be very linear and accurate for the results to mean anything. Once again this means not a piece of crap designed to retail for twenty dollars.

Another difference between what we do and the astro guys is that they usually use spectrometers that are bigger than we would want to have on our desktops. And they are more expensive than even Ocean Optics units.
They usually look at narrow spectral bands and spread them out really widely. And the emissive stars they are looking at also have lots of known lines that allow a degree of self calibration with respect to wavelength.

If the camera has a ccd or cmos sensor there is built in software to correct for dead or hot pixels. This also needs to be taken into account. The expensive cameras from Hamamatsu and others use so called Grade Zero sensor chips which have Zero defects and are designed to be used the way we are discussing. The same just isn't true for a camera designed to retail for $20.

If you can't calibrate against something, then your curves are worthless except for some pattern recognition as described by Dr. Hanneman. Even then your camera may be fooling you.

The above all applies to the Y axis , the intensity of light transmitted at a particular wavelength after going through the stone.

Calibrating for wavelength is a whole nother smoke as one of the tobacco commercials used to say.

In the case of "real" spectrophotometers like Ocean Optics there is an algorithm which you can apply to compare the output of the unit (in terms of wavelength) with known wavelength standards. This is frequently done with an an argon mercury lamp which has enough very accurately know lines in the UV VIS NIR range that you can compare the output of your unit under test to the table of lines published in the early twentieth century and you can generate a correction equation. Even for a simple unit like the Ocean optics this is a complex looking equation (though not actually that complex) whose coefficients (the numbers that effect the correction) are determined to many decimal places.

To talk about databases when using either the eye or an imaging sensor
(ie. any Bayer pattern color camera, digital camera, or any non spectroscopic imaging sensor) is akin to a kid who is a paintball wizard saying he has combat experience.

The data for both axes needs to be unmolested if it is to serve any useful purpose. That requires a real spectrophotometer. The entry level for that is Ocean Optics and some of its similar competitors such as Avantes or Stellarnet and a few others.

When he brilliantly proposed video spectroscopy some twenty years ago
Dr. Hanneman described a black and white camera plugged right into a monitor. Both the camera and the monitor had a few controls (contrast=gain and brightness=black level) and cameras at that time sometimes had AGC off switches. He proposed to place a transparency on the screen and to mark the location of calibration lines.
This would allow a pretty good estimate of the wavelength of "newly discovered" features if one would take the effort to place enough calibration lines. This essentially does graphically what a correction equation does arithmetically. It all stays simple and that is why it works.
(PS occasionally modern digital cameras will have an AGC control switch usually a tiny "DIP" switch) Perhaps Dr. Bill would grace this discussion with some of his wisdom. He was a professional spectroscopist in his real job.

With a black and white ccd camera you could have an advantage in that you could use the X axis pixel number in a similar fashion and even calibrate with respect to wavelength. But to see the pixel number you need a decent frame grabber. (The gadget that digitizes the image so you can view it in the computer) Just like cameras, spectroscopic imagers, and spectrometers, frame grabbers (when used in a real scientific context) is a whole entire subject unto itself. The same applies to data acquisition and spectrophotometric software.

Author:  jleb [ Sat Jan 17, 2009 8:51 pm ]
Post subject: 

All good points!

My canon digital rebel xt (350D) allows me to capture the image in RAW format and black and white. The manual says that if I want to apply bayer processing to the image stream, I'll have to do it in software. That makes me think that I can get B&W greyscale data out w/out internal camera processing.

I just spent a lot of time looking at all the spectra processing freeware out there, and I think, for our purpose of generating a spectra digitally rather than drawing it free hand, it would serve.

Now as to identifying the exact line that it the diffraction grating puts on the sensor, is that not a linear function across the wavelength? In otherwords wouldn't an A - D - G set of lines be proprotionally spread across the sensor?

The accepted astro method for subtracting dark current is to shoot an image at the same room temp as the spectra and then use that ref image to subtract the pixel values from the image under observation. Wouldn't that work on the Digital SLR's?

The connectrix camera's are apparently fairly simple w/ out a bayer pixel filter and some rudimentary controls. A lot of the astro spectrograhers are bying the NexImage cam and swapping out the color
CCD array for a B&W greyscale one.. They seem to have no problem finding the H3 lines in the star spectra and computing accurate red-shift from the image that matches the reference value. what do you think?

I appreciate all the wisdom here, I"m trying to find the sweet spot between publication level examination and something I can use in the field to tell the difference between pieces of rough I'd be buying...

Author:  G4Lab [ Sat Jan 17, 2009 9:08 pm ]
Post subject: 

I meant to mention the RAW format which hard core digital photographers are using so as to avoid some of the processing which most people prefer to have taking place in the camera. It is analogous to what Brian does with the output of an OO spectrometer. But you can't get a black and white picture from a Bayer pattern chip without deBayering it somewhere. The pixels have
filters in front of them. (Red Green and Blue with twice as many greens since the human eye peaks its sensitivity in the green.)

I am a pretty good electronics hand but am not terribly interested in swapping chips on anything already built unless they are in a socket.
CCD chips of the quality I would want are too expensive to risk. If you want a b/w camera get a Dage-MTI CCD 72 when one comes up on ebay. They have a very compact camera head and lots of controls so you will know exactly what you are doing to the signal. They come up pretty regularly at moderate prices , less than $100 and they are a precision scientific camera.
And since you are an electronics hand you can just display the signal on an oscilloscope. The rawest and most unprocessed.

The use you mention, ie out in the field might well benefit from having a B/W webcam grafted to an OPL spectroscope and displayed on your laptop. But you probably would not want to publish the results from it.
Most of these webcams use DSP techniques that are inherent to their electronic design and not very able to be disabled. This especially applies to digital SLRs. The professionals demanded and got access to the raw signals but exactly how raw they really are I don't think anyone can actually say and the manufacturer's who should be saying don't want to because most of the processing are trade secrets. (DSP =Digital Signal Processor a chip like an intel computer chip designed specifically to handle things like audio video and high data rate signal acquisition and process it in real time. Serious engineering there)

If a lap top is going to be part of the equation and you want more than an OPL or other hand spec then the Ocean Optics Red Tide which would be about $1500 delivered with software and has USP connectivity and would
do a better job than anything you could gin together. Remember that standard video has about the same number of horizontal pixels as regular old VGA which is 640 across and 480 top to bottom.

You mention the Canon Rebel 350D. Here is a link to what the
astro types do to there cameras. Some of these guys claim to be hobbyists but some of them are actually astrophysicists that can' switch it off when they get home. ( I DO love it!)

The first thing they do is remove the IR filter that almost all consumer cameras and video gear have. They also often remove so called anti aliasing filters that may be present. These are low pass optical filters designed to prevent Moire' patterns from occuring in the output signal.

Here is the home page of the above which has TONS of stuff to look at.

And here is the homepage for VisualSpec which can eat .pic files and
spit out spectrum curves.

Author:  jleb [ Sun Jan 18, 2009 2:16 am ]
Post subject: 

Great site astroturf... The full canon 350D/Rebel XT IR conversion instructions can be found here: Canon Digital Rebel/350D IR instructions

They also sell the clear glass drop in for the filter for $180. The body can be purchased on ebay for $100-200 or so.

Here's a better link on diy w/a edmund filter much cheaper and more details.
here and here's more

Those frenchmen are really dedicated.. So it does look like calibration ( they supply the curves) and the Canon camera turns out to be the right one.. According to their test of Nikon vs Canon, Canon RAW does not have any internal processing mods of the image (which I guess is why they have them mounted on all those pictures of 12 - 24 inch scopes).

Way cool.. those links are a goldmine G4!

Author:  G4Lab [ Sun Jan 18, 2009 4:58 pm ]
Post subject: 

You are welcome John! :D

I did not know that the 350 raw mode circumvents processing but if those guys say so I believe it. And I presume they can give an accurate assessment of the effect of de Bayering the signals.

I also ignored most of the Canon stuff because I use Nikon and am not that interested in doing some of the things they do (like major surgery on a relatively recent digicam).

I do love fanatic science hobbyists no matter what they are looking at. :twisted:

Author:  jleb [ Sun Jan 18, 2009 6:44 pm ]
Post subject: 

They compared the Nikon to the Canon. The Nikon lives in a state of Sin, as it does process the RAW data and they don't tell you. Got caught out on some comparisions between the two, when they noticed artifacts in the star field of the Nikon vs the Canon. Further investigations nailed them to the cross.

Canon bodies go for as little as $109.00 on fleece bay, w/ detailed instructions on how to remove the IR filter and replace it (due to R.I. expectations of a 1.567 index sitting there) with clear plate so autofocus still works.

I have collected all instructionables and sources for those interested in extended range IR photography.

Author:  G4Lab [ Sun Jan 18, 2009 7:08 pm ]
Post subject: 

All the Nikon's I have are film cameras :oops:
Except for some microscope cameras that have Nikon mounts.

Author:  Brian [ Mon Jan 19, 2009 7:06 pm ]
Post subject: 

jleb wrote:
Now as to identifying the exact line that it the diffraction grating puts on the sensor, is that not a linear function across the wavelength? In otherwords wouldn't an A - D - G set of lines be proprotionally spread across the sensor?

No. Wavelength spread across the CCD is only approximately linear. But don't let that hold you back. :wink:

Page 4 of 4 All times are UTC - 4 hours
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group