Transmission measurement of an UV-Pass filter

[ulf]

I would like to share the process I use to analyse an UV-Pass filter in a series of posts.

 

#1 The Equipment.

 

Spectrum analyser:

I use a Ocean Optics Flame-S XR1, USB-type analyser with a wavelength range of ca 200-1000nm.

It is like an optical building block to be used in different setups together with other optical components.

https://oceanoptics....e-spectrometer/

 

This array type of analyser is small and flexible and have its strengths and weaknesses.

The main weakness for this type of measurement is that it only has have a signal noise ratio of 1:250 with full signal input.

There are some things to do to improve this during data gathering and analysis, but it is far from what you can obtain with more advanced instruments.

 

Light sources:

My Main Light Source is an OEM unit from Heraeus Noblelight FiberLight DTM 6/50.

http://www.alphaligh...Fiber_Light.pdf

It has an output spectrum matching the one of the spectrometer.

The light is generated by two sources, one Tungsten Lamp and one Deuterium Lamp.

The unit is builtin an aluminium box with switches to control the two lamps and the output shutter.

 

Beside the Main Light Source I improvise and use other Light Sources if they are more suitable for the task

 

Light Fibres:

To guide the light between source and analyser I have several different fibres with different lengths and core diameters, 200um-1000um

All are UV-VIS capable.

 

Collimators:

I use collimators from Avantres, COL-UV/VIS and COL-UV/VIS-25.

https://www.avantes....ollimating-lens

 

Optical Bench:

I use a home built system based o in Arca Rail components, M42 extension tubes and step rings.

More details about this another time.

 

To be continued...

[ulf]

#2 The Basic Transmission Measurement.

 

The Transmission curve ( T ) is a result of calculations with three measurement data vectors (curves) using values for each examined wavelength-section.

 

The source curves are.

 

** Lamp ( L )

** Background ( B )

** DUT^* ( D )

 

* ( Device Under Test )

 

The Transmission is calculated. T = ( D - B ) / ( L - B )

 

The quality of the measurement is dependent on the stability of L and B.

 

The Lamp intensity become reasonable stable after enough warmup time if everything including the fibres are kept mechanically stable.

This is most important for the transmission result of the higher values not close to zero.

 

The background will drift and for long sampling periods this can become a problem.

This is most important for leakage measurements.

 

It is also important that the Lamp has enough intensity within the examined wavelength range.

Unfortunately the lamp intensity, as seen by the spectrometer, often has a lot of variation over the wavelength range.

 

For different wavelengths this is caused by the combination of

** Lamp Spectrum.

** Transmission variation in fibres and collimators.

** Spectrometer Response.

 

The Tungsten Lamp as My Spectrometer sees it:

Graph#1

 

The Deuterium Lamp as My Spectrometer sees it:

Graph#2

 

And the combination:

Graph#3

 

This looks rather terrible as a measurement signal source, but it is rather normal.

It works well for wavelengths where the intensity is at least 15-20% of the maximum value, as long as the intensity is stable.

 

The combination is the Light Source I used for the first overview measurement of the filter.

 

To be continued...

 

Edit: Numbered the graphs.

[Andrea B.]

Ulf -- thank you so much for your contribution here about measuring transmission of filters. It is great to see some of our members begin to take this on and provide us all with data about lens, filter and/or camera transmission.

 

I will be back to read in detail. (And I expect we will have comments from other folks too.) Just wanted to post a thanks and a "reading with interest" comment. :D

[ulf]

#3 The Overview Measurement of the Filter

 

I connected and aligned the measurement components for maximum signal response at the Spectrometer:

 

Light Source =-= UV-VIS Fibre =-= Collimator = = = Collimator =-= UV-VIS Fibre =-= Spectrometer

 

When ready I had a graph similar to the last one in previous post.

Then I adjusted the exposure time to get the highest peak as high as possible without overloading the spectrometer.

( => 0.14s)

( This is just as exposing to the right with a camera)

 

An averaging of 10(?) measurements was chosen

A Lamp ( L ) curve and a Background ( B ) curve was obtained and loaded into the spectrometer software.

The function to calculate the transmission from ( L ), ( B ) and ( D ) was activated.

 

The resulting graph was a straight line at most of the wavelengths with a trumpet of noise beginning around 900nm.

 

After placing the filter between the two collimators the response changed to this:

Graph #4

 

Changing to Log scale

Graph #5

 

The Log view clearly show noise due to dynamic range limitations.

 

As we are not looking for narrow wavelength peaks or very steep transmission changes the graph can be improved.

The spectral resolution of the detector array is ca 0.4nm per pixel.

The program can do a moving average smoothing of the graph and with 9-pixels soothing the graph looks like this instead:

Graph #6

 

However I would sill not trust the values below 0.1% very much.

 

The leakage response between 400nm and 700nm to be examined further.

How to do?

 

A real cliff hanger isn't it? :)

 

To be continued...

 

Edit: Numbered the graphs.

[ulf]

Ulf -- thank you so much for your contribution here about measuring transmission of filters. It is great to see some of our members begin to take this on and provide us all with data about lens, filter and/or camera transmission.

 

I will be back to read in detail. (And I expect we will have comments from other folks too.) Just wanted to post a thanks and a "reading with interest" comment. :D

 

Hi Andrea,

 

Thank you for the interest and positive words.

This series of posts has been in the pipes for some time now, but I have not been able to write them until now.

The measurements were done several days ago when there was a debate about if an infamous cheap UV-Pass filter was OK to use on The Convoy UV-lamp.

 

Here the filter type is irrelevant.

I want to describe my methods to measure it.

 

I really hope that people here with more experience with spectrometer measurements add their comments and help me improve my methods and correct any error in my reasoning.

 

What I am writing here is what I have learned when working with the spectrometer, trying to understand the results and sort valid and not valid data. It is a very hands on method.

I hope that my conclusions so far have been reasonably correct.

[JMC]

Thanks for sharing Ulf.

 

I've found with my setup I get issues with the baseline drifting slightly during my measurements, so sometimes I get % transmission results which go below zero. I'm guessing there is either an issue with light source variability or the spectrometer response changing as it warms up during use. Do you get artifacts like this as well?

 

It'd be great to see the bench you are using - I'm using an Ocean Optics RTL-T stage, but I love seeing home built solutions for technical problems.

 

I wonder if there is a filter we both have, and then we could put up a direct comparison?

[ulf]

I have my spectrometer connected to USB all the time.

The overall temperature should be reasonably stable due to that, but then there is a strong possibility that the power consumption of the electronics vary depending on how active it is.

I still se some baseline drift and are very careful obtaining fresh background and lamp values close in time to the measurement sampling.

The background is also sensitive to the light in environment. Optimal would be to cover the setup when taking that sample to avoid stray light.

 

To get optimal stability both the environment and the sampling activity should run for some time before the sensitive measurement is done.

Thermal stabilisation time could be rather long as I suspect the electronics is not very well coupled thermally.

 

I will eventually post about my homemade optical bench setup.

 

Jonathan - Lets find a solution for a filter comparison. TBD

 

/U

[Shane]

Heraeus Noblelight FiberLight DTM 6/50

I've found with my setup I get issues with the baseline drifting slightly during my measurements, so sometimes I get % transmission results which go below zero. I'm guessing there is either an issue with light source variability or the spectrometer response changing as it warms up during use.

 

Both contribute to baseline drift. Although the Noblelight is cheaper and a small convenient package, I tested it in the past and found it to be unsuitable (at that time) due to limited weak output and output stability. Weak output requires longer collection times which increases the probability of source and detector drift.

 

I always ran calibration immediately before starting data collection and then checked after data collection for drift, then reran the data collection if necessary, depending on the importance of accurate results.

[Shane]

For comparison.....

http://www.beyondvisible.com/BV3-filter.html these spectra were mostly collected using an expensive (massive) dual beam UV-Vis Spectrophotometer which used a 50W quartz halogen and 30W D2.

and these http://www.beyondvisible.com/BV3a-ICF.html with an Ocean Optics spectrometer but using a 300W UV enhanced xenon source (very short collection times).

[Shane]

The Noblelight must be pretty successful ..... design doesn't appear to have changed much in over 10 years.

[ulf]

Both contribute to baseline drift. Although the Noblelight is cheaper and a small convenient package, I tested it in the past and found it to be unsuitable (at that time) due to limited weak output and output stability. Weak output requires longer collection times which increases the probability of source and detector drift.

 

I always ran calibration immediately before starting data collection and then checked after data collection for drift, then reran the data collection if necessary, depending on the importance of accurate results.

 

Shane,

Thank you for your comments.

 

I agree that the FiberLight is weak and not very stable. It is still usable for some tasks and much better than nothing at all.

With careful usage and good methods one can reach at least a bit further than I first expected.

 

I would very much like to have something more powerful and stable, but think such sources are out of my reach as this just is a hobby for me, not generating any income.

If I found a suitable surplus light source to a reasonable price I would grab it very fast.

 

Your way of collecting data is matching my method exactly.

With such a light source I would not be able to trust the results otherwise.

I hope to receive more constructive feedback of my methods, especially if they can be improved.

[JCDowdy]

The Log view clearly show noise due to dynamic range limitations.

 

As we are not looking for narrow wavelength peaks or very steep transmission changes the graph can be improved.

The spectral resolution of the detector array is ca 0.4nm per pixel.

The program can do a moving average smoothing of the graph and with 9-pixels soothing the graph looks like this instead:

 

Your method appears to have achieved about as much as possible from your spectrometer. The dynamic range you show in your simi-log plot is essentially all that you could hope for. Smoothing makes for a less noisy baseline but really does not extend the dynamic range nor will a stronger source. However a stronger source would help if you add an integrating sphere.

[ulf]

Your method appears to have achieved about as much as possible from your spectrometer. The dynamic range you show in your simi-log plot is essentially all that you could hope for. Smoothing makes for a less noisy baseline but really does not extend the dynamic range nor will a stronger source. However a stronger source would help if you add an integrating sphere.

 

John,

 

I agree that this is close to what I can obtain with the current settings and light source, and I do not trust the results in the graph between 400nm and 700nm or above 850nm.

However I think I can reach a bit further with a slightly different approach that I will discuss in my post Chapter #4 soon.

 

A stronger source would indeed enable me to get shorter measuring times or use an integrating sphere, but I cannot understad the need for one, when measuring flat filter glass in a collimated light path.

What would I gain with that?

[JCDowdy]

I did not mean to imply you needed a sphere for measuring optically flat filters, although you should for measuring lenses.

[ulf]

That I fully agree with!

 

For proper transmission measurements of lenses you need a sphere.

The official measurement standards have an even stricter demand of the equipment.

 

However, by limiting the scope of measurement I think you can get around with less and without a sphere.

If just searching for a lenses UV-cutoff wavelengths in the limited wavelength range of 300-400nm, I believe it is doable.

 

I have obtained some results for lenses by using a big lens collimator and my weak light source.

They reasonably show results that are similar to other published results of the same lens types.

 

If not expecting any absolute transmission results, I think this is at least a reasonably acceptable method.

[ulf]

#4 Leakage measurement between 400nm and 700nn. Preparation:

 

There is not any valid data between 400nm and 700nm in graph #5 in my Post #4 above.

This is where the filters rejection is strong and the result is just noise.

 

To improve the measurements I need a stronger light source with less sharp peaks in the wavelength range of interest.

This will enable me to average sampling graphs to reduce the statistical noise, before I get problems with baseline drift.

 

Ideally there should be little light outside that range, to lessen the internal stray light in the spectrometer.

 

The visual version of Convoy S2+ happen to fill these requirements reasonably well, except close to 400nm.

 

The spectrum is typical for a Warm-white Power LED:

Graph #7

 

The setup:

The flashlight became the new light-source. I used an adjustable aperture, set to a small opening, to create a reasonably parallel light beam.

 

Convoy S2+ < distance - Aperture = = = Collimator =-= UV-VIS Fibre =-= Spectrometer

 

After some thermal stabilization the light from the Convoy became quite stable.

The light intensity was 15 times stronger, giving a sampling time of 14ms with a spectrometer load level close to 87%.

 

The environment:

The room temperature was quite stable and the spectrometer had been running continuously sampling for hours, to assure that the internal temperature of the spectrometer was not changing.

The light in the room was dimmed quite low and the test setup was shaded even further.

Experimentation showed that the baseline and lamp drift was low enough to allow sampling over several minutes without significant problems.

 

More tweaking can improve the results further, but this worked reasonably well.

 

To be continued...

[ulf]

#5 The Leakage measurement:

 

I opted for an averaging of 4000 samples for calibrations and filter measurements, giving less than 1 minute per measurement session.

The filter, when measured, was placed in the beam between the aperture and collimator.

First I repeated the reference lamp calibration and just before the filter measurement I did a fresh background calibration.

The filter measurement was followed by a measurement of the background and finally a verification of the reference lamp graph.

 

The resulting graphs of background- and filter-measurements are quite noisy.

Background - pale brown, Filter - green:

Graph #8

 

The information behind the noise can be found by doing a moving average "boxcar"-filtering of the data.

This will soften peaks, but I do not expect or search for any such features.

It is quite unlikely to find very quick steps in thee transmission of these ionic filters.

 

Averaged with a 12nm boxcar width:

Graph #9

 

Observe the reasonably good distance between the background- and filter-graphs, within the valid lamp range, 400nm - 700nm.

This makes me believe that the rejection of this filter is at least as good there as the green graph shows.

 

Added the original filter graph to show the full transmission:

Graph 10

 

On the grey background the green graph is valid, on the white background the purple graph is valid within 280nm and 900nm.

 

The End

[JCDowdy]

Yes, one can extend the dynamic range by limiting the bandwidth, and thus the internal stray radiation, of the measurement source.

 

A series of overlapping high OD bandpass filters, with FWHM of ~40-60nm, can be used to segment the input spectrum such that a wider wavelength range composite spectrum can be constructed from the segments. To further this one would also make a series of scans at different integration times within each filter segment and derive the contiguous composite curve from the segments within their respective linear response range.

 

Rather like a spectrometer version of an HDR composite image derived from a series of incrementally exposed photographs.

 

Nicely done.