Re: [dorkbotpdx-blabber] OpenSpectometer

[Tom Frisch]

I've been following another spectrometer project for a while- the public labs one- but haven't jumped in because they are using a webcam and I don't see how it could produce good enough results to bother with.  

My needs for a spectrometer are to measure color temp, CRI and CQS of a light source.  I have a spreadsheet w/ macros from the NIS that can generate these from a spectrograph, but the Public Labs spectrometer that uses a webcam suffers because the webcam has unknown white balance presets, and often has auto exposure enabled, which makes it kinda useless to me.  Also, the AtoD response curve of different colors and brightnesses is completely unknown.

Have you seen the latest hand-held device from UPRTek? 

http://cinescopophilia.com/led-light-meter-mk350-spectrometer-now-at-ikan/

Think it would be possible to do this kind of thing with your device?

and make it portable?

-Tom

[Jerry Biehler]

I looked at the Public Labs one when it was on kickstarted. It is good for someone who wants to look at something like a lamp and tell what gas mixture is in it or something like that. No good for the kinds of stuff I would like to do.

The UPRTek one is not a whole lot better in terms of spectral bandwidth. At 12nm it is pretty awful. For the stuff I want to do it needs to be sub-nm. I know I am asking a lot but I think the openSpectrometer could do it.

A couple things I not on the website, 3648 pixel CCD. This may be kind of high, the pixel size is pretty tiny at that point and you can run into problems. The one I had (Gamma Scientific Radoma) used a 1024nm sensor which gave sub-nm resolution in the 290-900nm range it was set up for.

Ditch toslink and use an industry standard SMA905 connection. They are very, very easy to make. Then you have a massive amount of cables out there you can use like silica ones which will allow you to get down into UV. I can take some pics of the Verity monochrometer that I have that has a SMA and a slit. Very simple to reproduce.

Add a aperture slit. You still need to get one, especially if you want to take advantage of the resolution of the sensor. Slits also allow you to attenuate light levels. Smaller slit the less light, the higher the resolution, and the greater integration time needed. A shutter would be cool too as some CCDs have additional noise when reading out with light on the sensor. Slits are not a big deal for me to make if you want some. My laser cutter can cut the thin metal for these.

Add an option for cooling the sensor. It does not take much and you can eliminate a lot of noise in the sensor and increase repeatability between uses. There are chip solutions from Analog that would make it pretty easy to implement. You can even do it with a couple opamps. Leave a hole in the circuit board below the sensor for a cold finger.

-Jerry

> _______________________________________________
> dorkbotpdx-blabber mailing list
> dorkbotpd...@dorkbot.org
> http://music.columbia.edu/mailman/listinfo/dorkbotpdx-blabber

_______________________________________________
dorkbotpdx-blabber mailing list
dorkbotpd...@dorkbot.org
http://music.columbia.edu/mailman/listinfo/dorkbotpdx-blabber

[Mathew Lippincott]

The public lab spectrometer is really a piece of software--spectral workbench-- that we sell webcam hardware to use with. The software is hardware agnostic.  You could also take your spectra using either a commercial diffraction grating or DVD diffraction grating with a good camera (I know Tom has a lot of these), and set the exposure consistently. I don't see why white balance matters on a spectrometer though.  the exact color of the color bands in a spectra isn't important if you have the thing calibrated, and it won't effect intensity. The response curve of a CCD is not "completely unknown" either.  You can calibrate around full spectrum light (the sun) to fix the dip in the green range that all commercial CCDs have. Alternatively, you could get a monochrome security camera that lacks a Bayer filter for your image source.  

[Nathan Bergey]

RE White balance, I think it matters if you are looking at normalized
flux rather than rectified intensity spectra.

Normalized flux being where you try and say x brightness of a pixel
maps to x flux, but this is hard when you have unknown white balance
between exposures. I'm not sure I believe the claim that it doesn't
effect intensity, imagine in the extreme case where, say the blue
channel is ignored completely, then all the pixels on the blue side of
the spectra would be much dimmer than they should be. Most white
balance settings are more subtle than that, but clearly it must do
something to the output.

Rectified intensity, on the other hand is where you try to fit the
continuum part of the spectra as close as possible to a flat line
(y=1) and let the dips and peaks stand relative to that. This is not
nearly as sensitive to overall brightness bias since you are fitting
the continuum.


Either way I suspect calibrating wavelength is far more important than
calibrating intensity.

[Nathan McCorkle]

Just found these replies!

On Fri, May 17, 2013 at 9:43 PM, Jerry Biehler <jerry....@gmail.com> wrote:

> I looked at the Public Labs one when it was on kickstarted. It is good for someone who wants to look at something like a lamp and tell what gas mixture is in it or something like that. No good for the kinds of stuff I would like to do.
>
> The UPRTek one is not a whole lot better in terms of spectral bandwidth. At 12nm it is pretty awful. For the stuff I want to do it needs to be sub-nm. I know I am asking a lot but I think the openSpectrometer could do it.
>
> A couple things I not on the website, 3648 pixel CCD. This may be kind of high, the pixel size is pretty tiny at that point and you can run into problems. The one I had (Gamma Scientific Radoma) used a 1024nm sensor which gave sub-nm resolution in the 290-900nm range it was set up for.
>
> Ditch toslink and use an industry standard SMA905 connection. They are very, very easy to make.

Good to know, thanks. Have any instructions or parts numbers that are
easy/cheap to make/buy?

PCH cables here in Hillsboro had TOSLINK wall jack connectors for
around $2 or $3, if SMA905 are 3X the price, they're still worth
considering if a user DOESN'T need UV. Personally I'm aiming for UV,
so again, thanks for the tip.

>Then you have a massive amount of cables out there you can use like silica ones which will allow you to get down into UV. I can take some pics of the Verity monochrometer that I have that has a SMA and a slit. Very simple to reproduce.
> Add a aperture slit. You still need to get one,

Can't you just buy a thinner fiber, or don't they come in in such a
variety of diameters?

>especially if you want to take advantage of the resolution of the sensor. Slits also allow you to attenuate light levels. Smaller slit the less light, the higher the resolution, and the greater integration time needed. A shutter would be cool too as some CCDs have additional noise when reading out with light on the sensor. Slits are not a big deal for me to make if you want some. My laser cutter can cut the thin metal for these.

With household aluminum foil or what? I wonder if the stepper motor
pulses would cause undulating edges and thus diffraction... for
production lining up two snapped crystal edges parallel to each other
could be used to make an mold for injection molding.

>
> Add an option for cooling the sensor. It does not take much and you can eliminate a lot of noise in the sensor and increase repeatability between uses. There are chip solutions from Analog that would make it pretty easy to implement. You can even do it with a couple opamps. Leave a hole in the circuit board below the sensor for a cold finger.

It's something we've talked about before, could you send a link to the
Analog solution? I'd be worried that a coldfinger spot would lead to a
non-uniform dark noise and possibly photo response. What about a metal
bar with gel running below the sensor, and small TECs on either end,
or on the ends and in the middle like you said. My point is that I
think the cooling should be as uniform as possible.



I've also been talking about using the PCB in a slotted fashion, to
take advantage of the mechanical stability of PCBs for the optical
system. Not quite press-fit, since some adjustment would be needed,
but maybe some pivot points and a curved slot at the radius. (to mount
the sensor perpendicular to the optical plane, also where the grating
would be mounted)

[Jerry Biehler]

On May 25, 2013, at 6:06 PM, Nathan McCorkle <nmz...@gmail.com> wrote:

> Just found these replies!
>
> On Fri, May 17, 2013 at 9:43 PM, Jerry Biehler <jerry....@gmail.com> wrote:
>> I looked at the Public Labs one when it was on kickstarted. It is good for someone who wants to look at something like a lamp and tell what gas mixture is in it or something like that. No good for the kinds of stuff I would like to do.
>>
>> The UPRTek one is not a whole lot better in terms of spectral bandwidth. At 12nm it is pretty awful. For the stuff I want to do it needs to be sub-nm. I know I am asking a lot but I think the openSpectrometer could do it.
>>
>> A couple things I not on the website, 3648 pixel CCD. This may be kind of high, the pixel size is pretty tiny at that point and you can run into problems. The one I had (Gamma Scientific Radoma) used a 1024nm sensor which gave sub-nm resolution in the 290-900nm range it was set up for.
>>
>> Ditch toslink and use an industry standard SMA905 connection. They are very, very easy to make.
>
> Good to know, thanks. Have any instructions or parts numbers that are
> easy/cheap to make/buy?
>
> PCH cables here in Hillsboro had TOSLINK wall jack connectors for
> around $2 or $3, if SMA905 are 3X the price, they're still worth
> considering if a user DOESN'T need UV. Personally I'm aiming for UV,
> so again, thanks for the tip.

Thorlabs has a bulkhead mount SMA fitting for $17.30. Toslink cables are just really crappy quality cables.

http://www.thorlabs.com/thorProduct.cfm?partNumber=ADASMA

>
>> Then you have a massive amount of cables out there you can use like silica ones which will allow you to get down into UV. I can take some pics of the Verity monochrometer that I have that has a SMA and a slit. Very simple to reproduce.
>> Add a aperture slit. You still need to get one,
>
> Can't you just buy a thinner fiber, or don't they come in in such a
> variety of diameters?
>

Not really. The light coming out of a fiber is pretty crumby. It spreads into a wide cone when it exits. You really need a slit.

>> especially if you want to take advantage of the resolution of the sensor. Slits also allow you to attenuate light levels. Smaller slit the less light, the higher the resolution, and the greater integration time needed. A shutter would be cool too as some CCDs have additional noise when reading out with light on the sensor. Slits are not a big deal for me to make if you want some. My laser cutter can cut the thin metal for these.
>
> With household aluminum foil or what? I wonder if the stepper motor
> pulses would cause undulating edges and thus diffraction... for
> production lining up two snapped crystal edges parallel to each other
> could be used to make an mold for injection molding.
>

My laser is a but different than anyone else's, my cutter is servo driven. None of that stepper stuff for me! And mine is a frequency tripled yag (355nm UV). I have cut BeCu up to at least .010" thick and stainless as well. I have some thicker BeCu I am going to try later. I might be getting an old laser welder from work. If I can get it working I can cut up to 5mm thick metal according to the manual.

http://www.youtube.com/watch?v=W071IQhpCFA

>>
>> Add an option for cooling the sensor. It does not take much and you can eliminate a lot of noise in the sensor and increase repeatability between uses. There are chip solutions from Analog that would make it pretty easy to implement. You can even do it with a couple opamps. Leave a hole in the circuit board below the sensor for a cold finger.
>
> It's something we've talked about before, could you send a link to the
> Analog solution? I'd be worried that a coldfinger spot would lead to a
> non-uniform dark noise and possibly photo response. What about a metal
> bar with gel running below the sensor, and small TECs on either end,
> or on the ends and in the middle like you said. My point is that I
> think the cooling should be as uniform as possible.

What you would do is have a rectangular cutout in the circuit board for the finger and the tec through that with a heat sink outside of the case. Something like this CCD mount for a SITe CCD sensor I linked below. It doesn't have to cover the entirety of the sensor housing, just approximate the size of the sensor itself. The sensor does generate heat as it runs, but not a whole lot.

http://www.flickr.com/photos/67292116@N00/8835366248/

>
>
>
> I've also been talking about using the PCB in a slotted fashion, to
> take advantage of the mechanical stability of PCBs for the optical
> system. Not quite press-fit, since some adjustment would be needed,
> but maybe some pivot points and a curved slot at the radius. (to mount
> the sensor perpendicular to the optical plane, also where the grating
> would be mounted)

I think I would just mount the sensor to a sub-board connected via a flex cable.

-Jerry

[Jerry Biehler]

>
> It's something we've talked about before, could you send a link to the
> Analog solution? I'd be worried that a coldfinger spot would lead to a
> non-uniform dark noise and possibly photo response. What about a metal
> bar with gel running below the sensor, and small TECs on either end,
> or on the ends and in the middle like you said. My point is that I
> think the cooling should be as uniform as possible.

http://www.analog.com/en/fiberoptic/laser-drivers/adn8831/products/product.html