Open source Hamamatsu-based spectrometer

Started Apr 17, 2017 | Discussions
Iliah Borg Forum Pro • Posts: 27,546
Open source Hamamatsu-based spectrometer
5
Jack Hogan Veteran Member • Posts: 7,471
Re: Open source Hamamatsu-based spectrometer

Cool, what is an estimate of the cost of the parts?

Erik Kaffehr
Erik Kaffehr Veteran Member • Posts: 4,412
Re: Open source Hamamatsu-based spectrometer

Iliah Borg wrote:

https://github.com/Alexey-Danilchenko/Spectron

Sounds like an excellent project when I go into retirement, December this year.

Best regards

Erik

-- hide signature --

Erik Kaffehr
Website: http://echophoto.dnsalias.net
Magic uses to disappear in controlled experiments…
Gallery: http://echophoto.smugmug.com
Articles: http://echophoto.dnsalias.net/ekr/index.php/photoarticles

Alexey.Danilchenko Senior Member • Posts: 1,264
Re: Open source Hamamatsu-based spectrometer
2

Jack Hogan wrote:

Cool, what is an estimate of the cost of the parts?

For the Spectron 2 board (spectrum measuring one) - around $75-80 without Hamamatsu sensor and Photon. The stepper motor board - around $25 without Photon.

Jack Hogan Veteran Member • Posts: 7,471
Re: Open source Hamamatsu-based spectrometer

Thanks.

OP Iliah Borg Forum Pro • Posts: 27,546
Re: Open source Hamamatsu-based spectrometer
1

Jack Hogan wrote:

Cool, what is an estimate of the cost of the parts?

Light sphere is a rather easy DIY project by the way. It simplifies many tests, like I use it to check digital vignetting.

-- hide signature --
rf-design
rf-design Contributing Member • Posts: 606
Re: Open source Hamamatsu-based spectrometer

I was also interested in a photo-spectrometer. Technical line sensors combined with a reflective and focusing diffraction grating I guess are a better approach.

My interested is to measure the spectral gain of foveon sensors. A grating then deliver monochromatic light to a specific spatial position. The relative layer response comes then for free. To get startet I ordered an USB-low cost spectrometer:

http://www.ebay.de/itm/282424977612

Later I consider to built DSLR as M42 adapter with the grating and slit in a box to compare different cams.

I do not understand if the very compact monochromator/line sensor combination need a stepper motor?

"The Spectron Motors board is a hardware for controlling monochromator positioning and moving through the selected spectral range"

 rf-design's gear list:rf-design's gear list
Sigma DP1 Panasonic LX100 Sigma dp0 Quattro Sigma SD9 Sigma SD10 +68 more
OP Iliah Borg Forum Pro • Posts: 27,546
Re: Open source Hamamatsu-based spectrometer

I do not understand if the very compact monochromator/line sensor combination need a stepper motor?

No, it does not need it, but it is a convenience.

-- hide signature --
Alexey.Danilchenko Senior Member • Posts: 1,264
Re: Open source Hamamatsu-based spectrometer
1
I do not understand if the very compact monochromator/line sensor combination need a stepper motor?

"The Spectron Motors board is a hardware for controlling monochromator positioning and moving through the selected spectral range"

Unless you plan to scan the whole spectral range in question manually by setting monochromator to each wavelength then yes it is not needed.

I thought the approach was described fairly clearly - monochromator selects a wavelength (one at a time), light is passed through it resulting in narrow band light source, that light is measured (spectrally) and photographed (to get raw RGB channel responses at that wavelength). This process repeated for the whole spectral range - automatically with selected step (hence stepper) and at the end we will have the data to obtain spectral sensitivity curves for the sensor.

rf-design
rf-design Contributing Member • Posts: 606
Re: Open source Hamamatsu-based spectrometer

AlexeyD wrote:

I do not understand if the very compact monochromator/line sensor combination need a stepper motor?

"The Spectron Motors board is a hardware for controlling monochromator positioning and moving through the selected spectral range"

Unless you plan to scan the whole spectral range in question manually by setting monochromator to each wavelength then yes it is not needed.

I thought the approach was described fairly clearly - monochromator selects a wavelength (one at a time), light is passed through it resulting in narrow band light source, that light is measured (spectrally) and photographed (to get raw RGB channel responses at that wavelength). This process repeated for the whole spectral range - automatically with selected step (hence stepper) and at the end we will have the data to obtain spectral sensitivity curves for the sensor.

Sorry, I did not read the description carefully. The intention is to bring up a narrowband spectral light source with the help of a spectrometer. So the spectrometer is only operating within the mechanical feedback loop. That was the misunderstanding regarding the usage of a stepper motor.

To measure spectral sensitivity of sensors a direct usage of a grating seems to be more useful because you get all wavelengths at one time. The tricky point is that the grating themselve have a spectral shaping which is needed to calibrated with a known receiver first.

 rf-design's gear list:rf-design's gear list
Sigma DP1 Panasonic LX100 Sigma dp0 Quattro Sigma SD9 Sigma SD10 +68 more
Jack Hogan Veteran Member • Posts: 7,471
Re: Open source Hamamatsu-based spectrometer

rf-design wrote: To measure spectral sensitivity of sensors a direct usage of a grating seems to be more useful because you get all wavelengths at one time. The tricky point is that the grating themselve have a spectral shaping which is needed to calibrated with a known receiver first.

Not just the spectral shaping.  I did something similar and on the cheap here though I was more interested in the process and my results were never meant to be used quantitatively.   It would be a nightmare to calibrate out all the distortions in my setup (device, lens and sensor).  For reliable results I would use Iliah and Alexey's approach.

Jack

rf-design
rf-design Contributing Member • Posts: 606
Re: Open source Hamamatsu-based spectrometer

Jack Hogan wrote:

rf-design wrote: To measure spectral sensitivity of sensors a direct usage of a grating seems to be more useful because you get all wavelengths at one time. The tricky point is that the grating themselve have a spectral shaping which is needed to calibrated with a known receiver first.

Not just the spectral shaping. I did something similar and on the cheap here though I was more interested in the process and my results were never meant to be used quantitatively. It would be a nightmare to calibrate out all the distortions in my setup (device, lens and sensor). For reliable results I would use Iliah and Alexey's approach.

Jack

Jack,

I did read your results and think it is for me easier to follow this route. Instead of having another sensor and built up with all the electronics and code I could direct use the DSLR sensor. The ordered USB-Spectroscope is using a small grating and a webcam. In principle I could replicate this concept with an bigger and more resolving grating and put his with the slit into a M42 adapter box. So in between the broadband light source and the sensor are only the focusing and refecting grating making calibration more easy.

Then I found that the grating themselve have a significant spectral function. To calibrate them you need a second monochromator based on prism dispersion where the transfer function could be calculated.

Furtheron the flash sources have also a spectral function because of the glass coating. Using outside light could be also a source if the air is clean. Otherwise you "see" the pollution.

But anyway a lot to learn.

Reiner

 rf-design's gear list:rf-design's gear list
Sigma DP1 Panasonic LX100 Sigma dp0 Quattro Sigma SD9 Sigma SD10 +68 more
Alexey.Danilchenko Senior Member • Posts: 1,264
Re: Open source Hamamatsu-based spectrometer

rf-design wrote:

To measure spectral sensitivity of sensors a direct usage of a grating seems to be more useful because you get all wavelengths at one time.

Really? And just how you will ensure the same light levels for each band at each location projected to the sensor? Calibration won't help you much there. Besides if you do it without lens (to take the lens out of the path and measure sensor spectral response directly) you will need a very precise alignment of the grating or calibration of sensor response to account for unevenness. It is impossible to use integrating sphere to diffuse the light and make illumination being photographed even without yet another calibration that is fairly tricky. And the last, I am not sure how exactly in that setup you will determine precise location of each band - using just grating without controlling output slit is quite imprecise (very dependent on the falling angle) so yet another alignment and calibration is needed.

So all of this adding up, you will end up with a very tricky setup - it will be very hard to maintain consistency of the results.

The setup with monochromator is very simple and precise, does not require any complicated calibrations to achieve precise and repeatable results and has very few variables to maintain (the most important one - unevenness of light is catered for by spectral measurements).

This is an example of the spectral sensitivity curves taken with this setup

Alexey.Danilchenko Senior Member • Posts: 1,264
Re: Open source Hamamatsu-based spectrometer

rf-design wrote:

I did read your results and think it is for me easier to follow this route. Instead of having another sensor and built up with all the electronics and code I could direct use the DSLR sensor.

The thing is - you get the output of proportionate usefulness with this. I.e. if you want to apply those spectral response curves to build camera profiles then setup without controlled environment will have questionable accuracy. Why do you think I ended up building all that electronics (which is not that complicated by the way) and software (I started also with a simple setup)?

rf-design
rf-design Contributing Member • Posts: 606
Re: Open source Hamamatsu-based spectrometer

AlexeyD wrote:

rf-design wrote:

To measure spectral sensitivity of sensors a direct usage of a grating seems to be more useful because you get all wavelengths at one time.

Really? And just how you will ensure the same light levels for each band at each location projected to the sensor?

First of all I am not an expert and learn on the base of my education.

Calibration won't help you much there.

Calibration is the key to put an accuracy number on every nm of the spectrum. The calibration is a chain of uncertainity. It start for shure with the light source.

Light source themself is complicated because flash temperature is time dependend. Next issue is the filtering of the encapsulation.

Besides if you do it without lens (to take the lens out of the path and measure sensor spectral response directly) you will need a very precise alignment of the grating or calibration of sensor response to account for unevenness.

Plane focus accuracy of the colliminating grating reduce the spectral resolution. A 100mm focal length provide enough pixels.

It is impossible to use integrating sphere to diffuse the light and make illumination being photographed even without yet another calibration that is fairly tricky. And the last, I am not sure how exactly in that setup you will determine precise location of each band

There are atomic lines which are used as calibration.

- using just grating without controlling output slit is quite imprecise (very dependent on the falling angle) so yet another alignment and calibration is needed.

So all of this adding up, you will end up with a very tricky setup - it will be very hard to maintain consistency of the results.

There are other tricky solutions outside:

https://en.wikipedia.org/wiki/Grism

http://www.ebay.com/itm/132132416869

The setup with monochromator is very simple and precise, does not require any complicated calibrations to achieve precise and repeatable results and has very few variables to maintain (the most important one - unevenness of light is catered for by spectral measurements).

This is an example of the spectral sensitivity curves taken with this setup

Do you derive an error enclosure?

 rf-design's gear list:rf-design's gear list
Sigma DP1 Panasonic LX100 Sigma dp0 Quattro Sigma SD9 Sigma SD10 +68 more
Alexey.Danilchenko Senior Member • Posts: 1,264
Re: Open source Hamamatsu-based spectrometer

rf-design wrote:

Calibration is the key to put an accuracy number on every nm of the spectrum. The calibration is a chain of uncertainity. It start for shure with the light source.

Of course it does. Now please do some experiments and see how consistent readings of the flash or tungsten lights are over time to be reliable for you to normalise camera raw readings.

Light source themself is complicated because flash temperature is time dependend. Next issue is the filtering of the encapsulation.

Because of this complexity, eliminating uncertainty of the measurements is exactly what monochromator setup does.

Plane focus accuracy of the colliminating grating reduce the spectral resolution. A 100mm focal length provide enough pixels.

You are measuring sensor + lens spectral response here. Based on this you would need to account lens transmittance curves and also unevenness across the surface. to eliminate lens influence on those spectral responses. I take photos directly without lens - no such problem or extra measurements to be done. Even with the lens in setup with monochtromator+sphere, camera values could be taken from narrow central area and lens characteristics (transmittance and flat field could be much easier calculated there).

There are atomic lines which are used as calibration.

In every shot with flash as a light source?

There are other tricky solutions outside:

https://en.wikipedia.org/wiki/Grism

http://www.ebay.com/itm/132132416869

I am not interested in tricky or complicated solutions - Iliah and mine is easy and gives stable results. But to each its own - you don't need to take my or anyone word for it. Try yours, use the results compare it with profiling using targets and see where you can get from there.

Do you derive an error enclosure?

???

rf-design
rf-design Contributing Member • Posts: 606
Re: Open source Hamamatsu-based spectrometer

AlexeyD wrote:

rf-design wrote:

Calibration is the key to put an accuracy number on every nm of the spectrum. The calibration is a chain of uncertainity. It start for shure with the light source.

Of course it does. Now please do some experiments and see how consistent readings of the flash or tungsten lights are over time to be reliable for you to normalise camera raw readings.

Stability and accuracy are different terms.

Light source themself is complicated because flash temperature is time dependend. Next issue is the filtering of the encapsulation.

Because of this complexity, eliminating uncertainty of the measurements is exactly what monochromator setup does.

But the monochromator have to be calibrated. So the uncertainity chain start somewhere.

Plane focus accuracy of the colliminating grating reduce the spectral resolution. A 100mm focal length provide enough pixels.

You are measuring sensor + lens spectral response here.

A focusing reflective grating avoid quartz glass to extend into UV/IR. There is nothing other than the grating and the slit.

Based on this you would need to account lens transmittance curves and also unevenness across the surface. to eliminate lens influence on those spectral responses. I take photos directly without lens - no such problem or extra measurements to be done. Even with the lens in setup with monochtromator+sphere, camera values could be taken from narrow central area and lens characteristics (transmittance and flat field could be much easier calculated there).

There are atomic lines which are used as calibration.

In every shot with flash as a light source?

There are other tricky solutions outside:

https://en.wikipedia.org/wiki/Grism

http://www.ebay.com/itm/132132416869

I am not interested in tricky or complicated solutions - Iliah and mine is easy and gives stable results. But to each its own - you don't need to take my or anyone word for it. Try yours, use the results compare it with profiling using targets and see where you can get from there.

For shure but have patience I am not an expert. My interest is to get insight into the accuracy of the disclosed foveon processing.

Do you derive an error enclosure?

???

Estimation of the relative accuracy of the spectral function. For instance % spectral dependend derivation to the total spectral energy of a ideal boltzmann light source.

 rf-design's gear list:rf-design's gear list
Sigma DP1 Panasonic LX100 Sigma dp0 Quattro Sigma SD9 Sigma SD10 +68 more
Jack Hogan Veteran Member • Posts: 7,471
Re: Open source Hamamatsu-based spectrometer

AlexeyD wrote: This is an example of the spectral sensitivity curves taken with this setup

Very nice Alexey. May I ask:

1) What are the units on the vertical axis?
2) Any reason why the x-axis is limited to 400-700 instead of 340-780nm?
3) Are these the original curves or those obtained after deconvolving the spectrometer's impulse response?

Thanks,

Jack

alanr0 Senior Member • Posts: 2,323
Re: broad band spectrometer

rf-design wrote:

AlexeyD wrote:

rf-design wrote:

Light source themself is complicated because flash temperature is time dependend. Next issue is the filtering of the encapsulation.

Because of this complexity, eliminating uncertainty of the measurements is exactly what monochromator setup does.

But the monochromator have to be calibrated. So the uncertainity chain start somewhere.

Plane focus accuracy of the colliminating grating reduce the spectral resolution. A 100mm focal length provide enough pixels.

You are measuring sensor + lens spectral response here.

A focusing reflective grating avoid quartz glass to extend into UV/IR. There is nothing other than the grating and the slit.

For broad band UV/IR operation, won't you need an order sorting filter (higher order blocking filter) to prevent, e.g. 800 nm detector responding to 400 nm light?

https://www.newport.com/t/grating-physics

-- hide signature --

Alan Robinson

rf-design
rf-design Contributing Member • Posts: 606
Re: broad band spectrometer

alanr0 wrote:

rf-design wrote:

A focusing reflective grating avoid quartz glass to extend into UV/IR. There is nothing other than the grating and the slit.

For broad band UV/IR operation, won't you need an order sorting filter (higher order blocking filter) to prevent, e.g. 800 nm detector responding to 400 nm light?

Alan,

shure.

The "Free Spectral Range" for order m=-1 (reflective grating) is 2 expressed as ratio of the max to min non-overlapping range. So typical 380nm-760nm have to be filtered before the grating. Anything outside alies into the data.

Otherwise two half band filters, one UV and one IR, after the grating would allow a broadband application. Do you know of a commercial solution?

Reiner

 rf-design's gear list:rf-design's gear list
Sigma DP1 Panasonic LX100 Sigma dp0 Quattro Sigma SD9 Sigma SD10 +68 more
Keyboard shortcuts:
FForum MMy threads