Jump to content
UltravioletPhotography

Stacking of filters may not give the effects we expect


rfcurry

Recommended Posts

A few months ago I was able to re-establish a serial connection to my Hitachi U-1500 spectrophotometer. What a relief, now I am able to once again capture all of the output of the 1500 on my remote computer, rather than having to record each discrete reading, line-by-line, from the 2-line display.

 

One thing I have noted was that the representation of a filter produced by the well-beloved Schott Excel spreadsheet ( https://www.us.schot...ools/index.html ) did not account for the stacking of component glass elements.

 

I was not aware, previous to Ulf's recent photo of his nifty 3D-printed filter holder, that by physical "stacking" was meant the application of one filter in a ring over another filter in a ring. I had always assumed that stacking meant binding two or more pieces of glass together in the same ring, with, or without, cementing. The Schott output represents "Calculation of cemented glass filter combination (up to 5 types)..."

 

Naturally, the 2+ ring solution introduces up to a 4% loss in transmission for each air/glass transition; hence two rings would decrease transmission by 8%, three rings by 16%, etc., from the transmission projected by the output of the Schott software, which relates only to cemented glass.

 

Another issue is that the light used by the Schott spreadsheet goes only from 380nm to 780nm. As I do not know how the Schott formula uses light, I do not know the effect of having no light below 380nm.

 

I hope someone can help me with a strange aberration - if I run a scan of a filter glass and input the results as a user-defined glass, when I then combine that glass in the Schott speadsheet with another user-defined glass, the graph of the resulting combination is very different than an actual U-1500 scan of the cemented combination. As you can imagine, it is very discouraging to build a physical filter which theoretically has a peak of 75% at 380nm - based upon individual measurements - only to find that the cemented, composite filter only reaches 65% at 380. Any ideas? Has anyone else encountered this?

 

Thank you for any help.

regards,

Reed

Link to comment
Andy Perrin
The Schott program is capable of handling the reflection losses approximately but they do not account for multiple reflections. I do not know why they neglect them, although previously we had concluded that the second order and higher terms were too small to matter. Your comment is making me wonder about that. But first, are you sure you are using the approximate reflection correction that they do provide?
Link to comment
enricosavazzi

I hope someone can help me with a strange aberration - if I run a scan of a filter glass and input the results as a user-defined glass, when I then combine that glass in the Schott speadsheet with another user-defined glass, the graph of the resulting combination is very different than an actual U-1500 scan of the cemented combination. As you can imagine, it is very discouraging to build a physical filter which theoretically has a peak of 75% at 380nm - based upon individual measurements - only to find that the cemented, composite filter only reaches 65% at 380. Any ideas? Has anyone else encountered this?

I suppose you have already taken into account the possibility that the optical adhesive is absorbing the "missing" UV.

 

In general, optical adhesives have a refraction index not too different from that of glass, so if anything, cemented stacks should have a lesser loss to internal reflections than non-cemented ones.

 

The only other thing I could think of is that using optical cement on a dielectric coated surface will change the optical properties of the outermost coating (the interface becomes coating-to-adhesive instead of coating-to-air). This, however, will not affect the optical properties of dielectric layers underneath the outermost one. Right now I cannot think of other phenomena that could result in what you observed.

Link to comment

Andy,

I guess that Schott doesn't include multiple reflections because they clearly stipulate that the program is only for cemented filters. Thus, if you have a UG11 glass (RI=1.56), a S8612 glass (RI=1.54), and cement between (RI=1.56), the loss due to refraction index differences is not great, it is principally occasioned by the air/glass and glass/air transitions. OTOH, stacking the two glass components introduces two more air/glass transitions. That will approximately double the transmission loss through reflection..

 

I do not neglect correct reflect(ion), Andy. I multiply my scan points by the inverse of the reflection factor before inputting as a user-defined filter. This gives me the internal transmission that is the Schott program default. Thanks for the input.

Link to comment

Enrico,

 

Yes, I have scanned the same filter in a ring both before and after cementing. The output is nearly identical in both instances. Also, Norland shows my preferred cement to possess 95% transmission at 350nm and 80% at approx. 315nm. A good thought, though.

 

Thanks.

Link to comment
  • 2 weeks later...

Reed Curry: Another issue is that the light used by the Schott spreadsheet goes only from 380nm to 780nm. As I do not know how the Schott formula uses light, I do not know the effect of having no light below 380nm.

 

Reed, you can define your own light source in Filter > Light Source.

To define a light curve between 300-400 nm, you first need to reset the axis to have a range from 300 - 400 by right clicking on the user-defined light source chart and selecting Format Axis. The natural choices for a user defined light source for UV-pass filters might be either a sunlight curve or perhaps a 365nm UV-LED curve.

 

However the light source has to do with the Chromacity diagram of a given filter glass. It doesn't affect the transmittance curve. Besides which filtered UV radiation cannot be assigned a colour locus anyway. So it seems pointless to create a Light Source for uv filters.

Link to comment

Reed Curry: Another issue is that the light used by the Schott spreadsheet goes only from 380nm to 780nm. As I do not know how the Schott formula uses light, I do not know the effect of having no light below 380nm.

 

However the light source has to do with the Chromacity diagram of a given filter glass. It doesn't affect the transmittance curve. Besides which filtered UV radiation cannot be assigned a colour locus anyway. So it seems pointless to create a Light Source for uv filters.

 

One way to add the effect of a light source with the transmission calculations could be to att a virtual custom filter with intensities for different wavelengths of the light source. The result from the program will be the remaining light of the light source after passing the filter.

I think this method is giving valid results.

 

I have done this and defined a "sun"-filter with the output from the solar light simulator discussed on this forum a time ago.

The result of a virtual stack with the "sun"-filter and some real filter glass seam reasonable.

Link to comment

The 2012 Schott filter program, which is what I have used for most of my graphs has no provision for generating Combination T type graphs (T = Transmittance),

it will only generate Combi Ti graphs which don't incorporate surface reflection data in the calculation (Ti = Internal Transmittance).

The 2017 version of the program has the added ability to generate Diabatic Combi T plots.

However, even though the reflection factor filter data (P) for each filter glass type is present in the program, that P data is only used to calculate and draw the individual T filter plots,

but the Diabatic Combi T plot does not use the P data from each individual filter to calculate its plot, unfortunately.

The only way to include the combined reflection factors into the calculation for the Diabatic Combi T plot is to first calculate it manually (Example: P1 x P2...),

and then enter the sum into the Combi filter data input page, at the bottom, where is says,

"effective reflection factor of the combination", which seems to have a default of about 0.93/4 which is about the average P for a lot of the Schott filters.

The 'effective reflection factor', for two filters stacked, without being glued is P x P = Peff, P being the total reflection of both sides of a single filter, Peff being the total reflection in the stack.

P is listed for each single filter in the 'user defined curves' and/or 'Data Table', as "Reflection factor P".

Multiply those, then enter that number in the ERT field, then go to T Diabatic graph.

 

There will of course be a difference between Dabatic Ti and Diabatic T plots, and that difference is the reflection Peff you enter in the ERT field.

Ratio changes, but graph suppression remains the same between Ti and non-glued T transmission.

Exactly how to calculate the Combi T ERF field number for glued filters, I am not sure, since the glued stack still has two surfaces, but Schott has the 0.93 in that field as a default.

 

As I said, most of my graphs were made using the 2012 program, which I have extensively modified over the years to line up Diabatic and Linear graph sizes and ranges with each other cosmetically,

and which I have laboriously entered many other types and brands of filter data.

I use both the 2012 and 2017 versions for various reasons. I use 2012 often to view suppression down to 1E-10, whereas the 2017 version has no way to show suppression below 1E-05,

and many people like to exceed 1E-05 sometimes.

I would take me quite some time to edit the 2017 version to where I could use it in the same ways, and it would still not show suppression below 1E-05.

 

So be happy. :D Happy that such a program even exists for us to use, and happy for all the Combi Ti plots you have in the filter graph sticky.

 

Thanks to Andi Perrin for help understanding P and Peff calculations. :)

Link to comment

Here is a comparison between Jonathan's spectrometer and the Schott 2017 program, using Diabatic unglued T stack graphs.

The Schott graphs shows calculated stack plots for all but the first two (2mm and 4mm) plots, the first two filters are individual T plots, not combi plots, the rest are un-glued reflection calculated combi T plots.

Refection for each U-340 layer is P = 0.91. For two filter layers stacked, Peff = 0.91 x 0.91 = 0.828, for three filter layers stacked Peff = 0.91 x 0.91 x 0.91 = 0.753.

These agree with each other fairly closely.

 

Johanthan's Spectrometer:

Ocean Optics Spectrometer and DH-2000-BAL light source.

The transmission was measured with the incoming and collection fibers at 90 degrees to the surface, and collimators were used on the incoming light source fiber and collection fiber.

Thank you Jonathan for letting me use your graph for this comparison.

post-87-0-72070200-1561802086.jpg

Above graph and test by:

Dr Jonathan Crowther

Director

JMC Scientific Consulting Ltd

 

 

Here is the same using the Schott 2017 program:

post-87-0-64860400-1561802135.jpg

Link to comment

I have a small question. When calculating for an unglued stack, are you assuming that one filter glass is stacked directly against another? Or are you assuming that there is a filter ring's worth of air between them?

 

BTW, for anyone new to the filter stacking game, please know that you should not directly stack one filter glass against another without optical cement. You will get an interesting phenomenon: Newton's Rings. Although I'm willing to bet that there is someone "out there" who is deliberately inducing rings just to make a crazy abstract photo!! B) :P

 


 

Thanks everyone for helping answer Reed's questions about our favorite program!

 


 

I'm thinking that it would be nice to gather data sets for non-Schott filters and post them somehow for people to use in the Schott program. That probably could be done.

 


Link to comment
Andy Perrin
Re the unglued filters: Enough air gap not to induce Newton rings. The calculation neglects partial reflections between each piece of filter glass.
Link to comment

Thanks, Andy!

 

We all should keep in mind that Schott charts and spectrometric measurements represent a curve which may not be achieved in practice with any particular filter. I'm pretty sure everyone knows this, but as Editor & Lead Nerd, I try to add little reminders for newbies just in case.

Link to comment

You should never mount filter glass against each other unless gluing. Unglued filters need an air gap between them to prevent Newton rings.

Separate individual camera filters screwed together have ample air gap.

 

Here is an example involving 2012 and 2017 filter program.

First, lets look at the two data input screens from 2012 and 2017.

 

This is the 2012 data input screen.

Notice at the bottom there is the field to enter the Peff. This field seems to have no use in the 2012, it changes nothing, there is no T Comi provided,

so the Peff in 2012 is essentially useless as far as I know.

post-87-0-59305200-1561827627.jpg

 

This is the 2017 data input screen.

You can enter the calculated Peff at the bottom, and it will change the Diabatic Combi T Plot, if you don't manually do this then the T plot will not be accurate.

This can be done for both glued and air gaped stacks, but I am not sure what the formula is for glued stacks.

post-87-0-68081800-1561827651.jpg

 

Here is the user defined filter curves page, see how I have deliberately changed the U-340 'P' to an erroneous number.

The U-340 P should be 0.91, I have changed it to 0.60 to show how it changes things in the Diabatic Combi T plot.

post-87-0-06017400-1561827677.jpg

 

See how the two individual filters have lower peak transmissions than the Diabatic Combi T plot, this is because because of the erroneous U-340 P = 0.60 data I edited above.

The Diabatic Combi T plot line is 'higher' than the two stacked filters. This is of course impossible, and demonstrates how the Combi T plot line doesn't use the P data of the stacked filters,

but instead uses the Peff that you manually calculate and enter in the data entry page.

post-87-0-82253400-1562286454.jpg

 

Here is the Diabatic Combi Ti graph for the same filters and stack which uses no P or Peff calculations.

post-87-0-19843400-1562286414.jpg

 

Footnote: I did find a way to edit the program to make it display 1E-10, but I am far from able to get that working correctly yet.

Work in progress. Slow process, back burner.

Link to comment

More good info about how to use Peff.

 

I also got stuck on defining the axis stuff for below IE-05. I don't know why Schott can't provide a better tutorial. I suppose they don't think that anyone is ever going to use the program for anything outside the Schott specs. :rolleyes:

Link to comment
Andy Perrin
Anything below 1e-5 cannot possibly be a real measurement, and it’s questionable whether a calculation would be relevant either. I think this is a quest that is probably better not to set off on.
Link to comment

Try this, it will get you there if you fiddle with it, but you will need to resize everything and I am not sure it works correctly yet.

post-87-0-69419000-1561830264.jpg

Link to comment

Anyone know what the 'normalized' graph is used for?

They have the normalized graph feature in both the Schott 2012 and 2017 versions.

I never use the normalized graphs, since they adjust all the plots to 100%, and the rest of the transmission proportionally I guess.

The band width remains the same.

Here is a comparison between a Linear Combi Ti plot, and a Normalized Ti plot.

Unfortunately, they only have the Normalized graph in linear version, I wish they had a diabatic normalized Combi graph, that might be a little more useful.

 

Perhaps it is just a way of visualizing the ratio between peaks and lows (OD and suppression)?

I don't know, I have always wondered what they are used for.

What do you think?

 

 

post-87-0-92729300-1561854421.jpg

 

I have related info in a new topic here:

https://www.ultravioletphotography.com/content/index.php?app=core&module=search&do=viewNewContent&search_app=forums

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...