Halogen Lamp Spectrum 12V, 50W MR16: A vs B

[phy_optics]

12V, 50W, MR16 halogen lamp spectrum is measured as shown ('B') with our newly developed mini-spectrometer having resolution ~1nm. But in internet it was similar to 'A' (We don't know how much resolution spectrometer they've used). Could anybody let me know what exactly the spectrum look like when measured with ~1nm resolution spectrometer (A or B) and why?

just check the below link for more understanding
https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_ID=3482

 

[Charles Link]

It's a little hard to determine from just looking at the data provided. Suggestion would be to try measuring the spectral transmission curves $\tau(\lambda)$ vs. $\lambda$ of some optical filters by taking $\tau(\lambda)=I(\lambda)_{source \, with \, filter}/I(\lambda)_{ source \, without \, filter}$ and comparing to the manufacturer's curves. If you get good results with that, it would demonstrate to some degree the quality of your spectral measurement apparatus.

 

[Tom.G]

Overlaying "OURS" and "OTHERS" images indicates someone is either compensating for optical fiber transmission when they shouldn't be or not compensating when they should be. Or maybe you got the sign of the correction wrong.

Interesting in any case. Please let us know what you find.

 

[phy_optics]

HI Tom!

Thanks for the info. Following is more accurate image data for comparison.

The waviness pattern is exactly same, we felt that the performance of our spectrometer is matching to that of our competitor,
but in internet the spectrum would be something like the following.

I still didn't why is this waviness?

I inquired about it to concerned spectrometer manufacturer, they replied that it could be due to etalon effects caused by the glass window placed on to the linear CCD array. But the thing is we don't have any glass window on to the CCD array. I didn't understand what is that etalon effect.

Pls help.

 

[Tom.G]

For a quick tutorial on the etalon effect see:

https://www.thorlabs.com/images/TabImages/NDFilter_Etalon_Lab.pdf

The CCD probably does have a window over the sensor chip itself for mechanical and environmental protection. It is rare to purchase a packaged chip (one with pins or solder bumps) that is not sealed. Bare chips are available directly from the chip foundries for use in very high volume products, They are then mounted directly to a circuit board and covered with a spot of epoxy. The total cost of the assembled product is a few pennies lower that way. But mounting and connecting is not usually a hand operation and the machinery is a bit pricey.

edit: A somewhat closer examination of the above link shows etalon effect is NOT likely your problem. Your peaks are too far apart. So the below would not be applicable. end edit

Take a close look at your sensor and talk to the manufacturer to get details. If they acknowledge a flat window, ask for the thickness, then, using the formula in the above link, see if the calculations match your effects.

If it is a flat window, maybe they can supply a wedge window on the sensor.

 

[Tom.G]

I still didn't why is this waviness?

The two plots you supplied for "OURS" in messages #1 and #3 are different. Message #1 shows a peak at 618nm but in message #3 it has shifted to 590nm. What changed?

You could ask the supplier of the "OTHER" plot what caused the ripples and if they are not real, can they be filtered out in software or removable with a calibration. I suspect the ripples are a common instrumentation artifact and published halogen lamp spectra have been corrected, probably automatically in the spectrometer.

 

[phy_optics]

Plot 'OURS' in message #3 is correct, pls ignore plot in #1 msg. I just calibrated the spectrometer.

I asked the 'OTHERS' i.e., Thorlabs, they just said it was due to etalon effects. I too feel it might not not be the reason. We are very sure that we are using CCD without glass window.

Pls check I'm just sharing the other spectra too.

 

[sophiecentaur]

Those graphs are so like what you get with a mismatched RF transmission line. The peaks would suggest that the standing wave is formed between two interfaces, separated by an order of magnitude of about a wavelength (?). That, presumably is your 'etalon effect'. The mismatch reflections are not too strong as the 'standing wave ratio' is only about 20%. I guess the situation could be improved by coating (blooming) each face. A cheaper solution would be to cancel out the variations of the data. numerically

 

[Andy Resnick]

12V, 50W, MR16 halogen lamp spectrum is measured as shown ('B') with our newly developed mini-spectrometer having resolution ~1nm. But in internet it was similar to 'A' (We don't know how much resolution spectrometer they've used). Could anybody let me know what exactly the spectrum look like when measured with ~1nm resolution spectrometer (A or B) and why?

Why not use your spectrometer to examine a standard- for example, the sun? Or an incandescent bulb... Then you will know if your instrument has artifacts or not.

 

[Charles Link]

The wavy lines in he spectrum do look like some kind of "thin film"=etalon effect. Perhaps there is some kind of thin film coating layered onto your detector. @Andy Resnick has a good suggestion to test the spectrometer with an incadescent bulb.

 

[Tom.G]

So far, consensus points toward the etalon effect. Is there anything other than air or vacuum between the grating and the sensor? To test the sensor , tilt the sensor so its active surface is not perpendicular to the light coming from the grating. Rotate the sensor around its long axis, keeping that axis fixed in relation to the grating. After re-calibrating, if the peak wavelengths are shifted and/or drastically changed in amplitude, that would confirm a sensor etalon effect. (If, by chance, they disappear, you have found a fix.)

As @Charles Link said: "Perhaps there is some kind of thin film coating layered onto your detector."
Likely. For protection from the environment (principally water vapor and mechanical damage), semiconductor chips are generally passivated near the end of the manufacturing process. Passivation is done by coating the surface with a thin layer of Silicon Dioxide (SiO₂), also known as Silica or Quartz. That could be the etalon source.

 

[phy_optics]

@Charles Link the sensor is TCD1304 (with glass window removed), which I don't this it has some coating on it. Incandescent spectrum is found very similar to that of halogen lamp.The competitors (others except Thorlabs) claim that they use the same TCD1304, but their halogen spectrum seem to be flat i.e., without waviness.

Following is are the spectra with different light sources.

@Tom.G The Detector (CCD) tilting experiment, I'm going to do it now. I'll share the observations

Spectrometer is of Czerny Turner configuration. I doubt the reflected light (very small amount) from front and back surfaces of the plane grating/colimating mirror/
focusing mirrors would have crated an interference pattern and appended on the flat smooth curve (of course this is just like etalon effect, but from other optical components, not from coating surface). What are your thoughts on this? I'm trying to find a way to test this.

 

[Tom.G]

Spectrometer is of Czerny Turner configuration. I doubt the reflected light (very small amount) from front and back surfaces of the plane grating/colimating mirror/
focusing mirrors would have crated an interference pattern and appended on the flat smooth curve (of course this is just like etalon effect, but from other optical components, not from coating surface). What are your thoughts on this? I'm trying to find a way to test this.

I would have expected front-surface mirrors without any hard-coating. If you are using second surface mirrors or first surface with a hard-coat, I think you have found your etalon(s)! The only way I can think of at the moment to test this is try some non-coated, front-surface mirrors.

@sophiecentaur noticed that the ripple amplitude was about 20%. That matches rather well with the rule-of-thumb of 4% reflectance from a glass-air interface.

edit:
To get an idea of surface reflectance, shine a LASER beam on a second surface mirror at a shallow angle. Aim the reflected beam on a white surface several feet away. You will see two spots next to each other with one of them very dim. That dim spot is from the front surface of the mirror. In addition to creating the etalon effect, this smears the intended spot to the next few pixels of your sensor, degrading the SNR (Signal-to-Noise-Ratio) of your system. Be aware that even the metal coating on a mirror has a spectral sensitivity +-5% over the visible band so calibration is needed at a few wavelengths. And of course it changes as the exposed metal surface oxidizes from exposure to the air. (several years?) That's why large astronomical observatories have their own on-site mirror coating facilities.
end edit:

 

[phy_optics]

@Tom.G Mirrors are having hard coating on its front surface.

 

[sophiecentaur]

Without paper n pencil, I can't estimate the thickness of the layer to produce that ripple spacing but it can't be more than 1micron. Does that not give a clue as to what / where it is?
The logic has to indicate that, if the same effect happens for more than one source, it has to be the sensor. Doesn't the spec sheet give a clue?

 

[Tom.G]

I doubt the reflected light (very small amount) from front and back surfaces of the plane grating/colimating mirror/
focusing mirrors

Perhaps I mis-interpreted that statement. I took it to mean that they are back-surface mirrors.

However this statement implies that they are front-surface mirrors.

Mirrors are having hard coating on its front surface.

A clarification please?

I finally came up with a way to test the mirrors:
Use some of them to reflect the light source several times before the light beam enters the spectrometer. If the ripple amplitude shows a significient change, they are a problem source.

 

[Andy Resnick]

Spectrometer is of Czerny Turner configuration.

What happens to the fringe visibility when you adjust the slit width? (Can you adjust the slit width?)

The reason I ask is shown here (last figure):

http://www.shimadzu.com/an/uv/support/fundamentals/monochromators.html

 

[phy_optics]

@Andy Resnick When I adjust the slit width, the waviness is reduced at the cost of resolution. But the resolution is too worst that we can not use this technique.

@Tom.G I tilted the CCD and found variation in spectrum pattern (waviness at other regions, amplitude differences..etc), it may be due to combined effect of 'etalon (we suspect) and out of focus).

I found following paper pls have a look.

I just pointed the CCD surface with laser pointer and observed as below. (just for ur info)

This particular CCD have glass on its surface.
1) If I point the light on glass plate --pattern is not forming
2) Pattern is forming only, when I point light on to the pixels (CCD pixel size 8um, beam dia ~2-3mm)
3)Pattern is circular
4) Patten have fringes with increasing gap

From all the above I feel the the waviness is due to "diffraction caused by the CCD pixels randomly reflected/scattered into the spectrometer/detector section and that made this". But still need to confirm this, most importantly Solve this.

 

[Charles Link]

What happens to the fringe visibility when you adjust the slit width? (Can you adjust the slit width?)

The reason I ask is shown here (last figure):

http://www.shimadzu.com/an/uv/support/fundamentals/monochromators.html

In this paper the author discusses slit width and he says you adjust the slit width to "1 nm". He is actually referring to the resolution $\Delta \lambda=d \frac{\Delta x}{f }$ where $d$ is the distance between grooves on the grating, and $\Delta x$ is the slit width. $f$ is the focal length of the focusing mirror in the spectrometer. A good slit width $\Delta x$ is normally about 100 microns.

 

[Tom.G]

@Tom.G I tilted the CCD and found variation in spectrum pattern (waviness at other regions, amplitude differences..etc), it may be due to combined effect of 'etalon (we suspect) and out of focus).

It is unclear if the second image is the result of the 'tilt' test. Testing with the tilted sensor needs to be done with a continuous spectrum, as from a halogen lamp. This will show if the ripple peaks shift wavelength when the light goes thru a different thickness of any layer on or in the sensor. Do you have that data?

From all the above I feel the the waviness is due to "diffraction caused by the CCD pixels randomly reflected/scattered into the spectrometer/detector section and that made this". But still need to confirm this, most importantly Solve this.

It looks like we are down to etalon effect somewhere in the optical chain and/or diffraction in the sensor. If it is in the sensor the only way I see to correct it is in a final calibration with the correction done in software. That would probably mean each a correction file included with each machine.

If the cause is etalon somewhere in the optical chain, then we are back to finding and eliminating the effect, or using a correction file as above.

I have only had a chance for a briefl look at the articles mentioned above. I will read them in the next few days and see if anything else comes to mind.

 

[Charles Link]

It is unclear if the second image is the result of the 'tilt' test. Testing with the tilted sensor needs to be done with a continuous spectrum, as from a halogen lamp. This will show if the ripple peaks shift wavelength when the light goes thru a different thickness of any layer on or in the sensor. Do you have that data?It looks like we are down to etalon effect somewhere in the optical chain and/or diffraction in the sensor. If it is in the sensor the only way I see to correct it is in a final calibration with the correction done in software. That would probably mean each a correction file included with each machine.

If the cause is etalon somewhere in the optical chain, then we are back to finding and eliminating the effect, or using a correction file as above.

I have only had a chance for a briefl look at the articles mentioned above. I will read them in the next few days and see if anything else comes to mind.

Normally, a spectrometer requires a calibration type source, either a blackbody or calibration lamp, because even if the detector doesn't exhibit the etalon effect, it, along with the optics, are not usually spectrally flat in their response. There are power sensors, typically pyroelectric type detectors, that are spectrally flat, but photodiode type detectors, in general, are not spectrally flat. The calibration equation assumes a linear response at a given wavelength $\frac{V_s(\lambda)}{V_{cal}(\lambda)}=\frac{I_s(\lambda)}{I_{cal}(\lambda)}$. $\\$ $I_s(\lambda)$ is the spectrum of the source that you are trying to determine.

 

[Andy Resnick]

@Andy Resnick When I adjust the slit width, the waviness is reduced at the cost of resolution. But the resolution is too worst that we can not use this technique.

Thanks for checking.

I tilted the CCD and found variation in spectrum pattern (waviness at other regions, amplitude differences..etc), it may be due to combined effect of 'etalon (we suspect) and out of focus).

What do you mean 'out of focus'? Is there a lens in the optical path?

 

[Tom.G]

Over the weekend I managed to do some reading/research and I think sophiecentaur is onto something:

Without paper n pencil, I can't estimate the thickness of the layer to produce that ripple spacing but it can't be more than 1micron. Does that not give a clue as to what / where it is?
The logic has to indicate that, if the same effect happens for more than one source, it has to be the sensor. Doesn't the spec sheet give a clue?

The calculation for the ripple spacing yields an interference (etalon) layer thickness around 2.2μm. That seems a bit thick for the CCD but still possible. It could be the sum of passivation layer thickness on the 2nd mirror + CCD.

 

[Andy Resnick]

12V, 50W, MR16 halogen lamp spectrum is measured as shown ('B') with our newly developed mini-spectrometer having resolution ~1nm.

This problem has been bugging me for a few days now...if it is an etalon-type effect (which is the case, according to ThorLabs), then using your supplied graphs to estimate a free spectral range of 50 nm and a center wavelength of 540nm, the cavity path length is 2.9 um (air) or 1.9 um (glass). This doesn't obviously correspond with ThorLabs' claim of a 'window' causing the effect- unless the window is in part of the beam path where there is focusing or converging rays (which you mentioned in another post). Then, the cone angle of the focused beam allows for path length differences, analogous to Newton's rings. In this case, the 1.9 um path length difference is trivial- a 1mm thick window acquires this for a f/32 system. So, if it is indeed an etalon effect, it can be explained by path differences acquired as converging/diverging light passes through a window.

But, there a problem with this simple result: Czerny-Turner spectrometers are fast: f/4 or so. That's too fast to provide nice fringes like you see- they would be washed out. So, the window needs to be located at the entrance coupler rather then within the system (e.g. the CCD), which is where (I think) the ThorLabs window is located.

And, of course, there's another problem- you didn't specify if there's a window present! Your OP says it's a 'newly developed' spectrometer: developed by your group? What is the optical design? Or, is it actually the ThorLabs spectrometer?

In the end, if you guys are simply using a ThorLabs spectrometer and don't like the fringes, then correct the measured spectrum against a known ('standard') source and use that to correct other measurements.

 

[phy_optics]

Thank you very much friends for all your postings and discussion.
Specially @Tom.G and @Andy Resnick I'm very much grateful to you.

I could successfully convey the most probable and possible root cause for this issue to my client. Thanks :)
Could anybody suggest any mathematical algorithm to process this spectrum(Waviness) and make it smooth one? I'm trying FFT and will share you the observations..

 

[sophiecentaur]

Could anybody suggest any mathematical algorithm to process this spectrum

I would suggest a 'look-up table' approach would probably be best. Start with a good flat source (sunlight?) and measure the waviness over the visible band(ref a data filtered spectral curve. Then modify all results by the (1/A) factors in the table. I don't think there would be any advantage in trying to characterise the waviness with an analytic function (algorithm), which would probably involve a number of coefficients that would need to be measured and stored for each different set up.

 

[Tom.G]

I agree that a lookup table of gain corrections would be best. Its size of course depends on the number of ripples and the required accuracy, then interpolate between entries. Perhaps adaptive sampling could be used for denser sampling around inflection points and sparser sampling in linear regions.

An analytic function would need a polynomial of degree one more than the number of inflection points, around 17 or 18 just for the part of the spectrum shown in your posts.

Probably the best bet would be a calibrated light source as a standard. Using Sunlight on the Earths surface involves too many unknowns re the atmosphere. Also Sunlight is not even close to spectrally flat, it has a strong peak in the middle of the visible spectrum. (Apparently we evolved to make maximum use of what was available.)

http://www.greenrhinoenergy.com/solar/radiation/images/solar%20radiance-07.jpg

 

[Tom.G]

If you really need an analytic solution, some use could perhaps be made of cepstrum processing as described in an article by
Merla, et al. "Portable System for Practical Permittivity Measurements Improved by Homomorphic Deconvolution"
published in IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, Vol. 66, No. 3, March 2017, pgs. 514-521, (DOI:10.1109/TIM.2016.2644859) .
It is rather compute intensive, a couple FFT, IFFT, and Complex Log and AntiLog operations over the dataset. Even at that it may need some changes because your ripple periodicity varies across the spectrum.

 

[Ajz]

I am sure that the problem is with the CCD sensor. I had the same problem but solved after replacing the CCD.