Filter Transmission Charts #2

[Cadmium]

Post #7

Schott IR-Block Comparison: BG38, BG39, BG40, S8612 @ 2.0 mm

 

There are wrong graphs - one for BG and other for UG

 

I don't understand what you are saying here.

I don't see any UG in that graph.

[Cadmium]

The 2017 version has a combined T diabatic graph.

[Cadmium]

There is actually a typo error in the 2017 program data page. The second equation line should read T instead of Ti.

This is only a screen label, and doesn't change the way it works.

 

It should read like this:

Ti COMBI = Ti1 x Ti2 x Ti3 x Ti4 x Ti5

T COMBI = P_eff x ( Ti1 x Ti2 x Ti3 x Ti4 x Ti5 )

 

[Andy Perrin]

I don’t know how they are calculating the “effective reflectivity factor” but they have to be doing something equivalent to the paper linked above. I will say that I don’t like using any equations where you can’t see the details of the calculations.

 

The correct method involves taking account of all multiple reflections. That means you can’t just multiply the Ti values together. Obviously you could come up with an effective factor by doing it the proper way and then dividing by the product of the Ti’s but I would not trust that method very far (unless the new thicknesses were very similar). If you are going to account for reflections at all why mess with effective this-and-thats anyway? The linear algebra is not THAT hard.

[Andrea B.]

I don't see any UG in that graph.

 

That's cause I removed the misplaced chart. B)

[Cadmium]

Andy, You might want to take a look in the program, the equations are buried in there somewhere. Either than or bring up your concerns with the people who wrote the program.

Either way, the T graph shows a different transmission than the Ti graphs, using COMBI or Single filter mode, and should be closer to the real transmission, at least for glued stacks with no air gap.

[Andy Perrin]

But I don’t care about the program, I just care that people not be misled by inadequate approaches. Reflectivity varies slightly as a function of wavelength also, which I’m not sure they are accounting for. I have my own program which I can contribute but as usual it is written in MATLAB. I may try writing a version in some other language if people really want graphs with properly computed reflections incorporated.

 

I mean, I’m sure their version is better than not including reflections at all, but people should be aware of the limitations.

[Andrea B.]

Do we have some reason to believe that Schott transmittance calculations for combinations are inaccurate?

 

Here is the Schott tech reference on Transmittance: https://www.google.c...6PF4RZdVxoMqM1-

Clicking it will download a PDF.

[ulf]

Reflectivity varies slightly as a function of wavelength also, which I’m not sure they are accounting for.

 

I don't think they have the refractive index as function of wavelength in their numerical data. Isn't that needed for a correct reflectivity calculation?

I assume that this reflectivity discussion is all based on normal incidence of the light. For photography that is a rather rare special case. :)

 

Andy, how big would the error in reflection losses approximately be, if not taking in account the multiple reflections?

Without doing the math, my gut feeling is that it will not be a very dramatic difference to the final result for a two-filter stack.

 

From a principle point of view I agree that one should, if possible avoid any simplifications that add errors.

But if those are known to be small it might be OK as long as you are aware of what has been done.

 

The raw data for these calculations must(?) be based on empirically obtained data, and there are tolerances between batches.

If I remember correctly Schott state ±3nm for the transition wavelengths.

[Andy Perrin]

You can actually calculate the single bounce reflection coefficient from the refractive indices at any angle and wavelength because Schott gives the refractive index at two (sometimes three) wavelengths or they give the Abbe number which can also be used to find the dispersion. Definitely these are small effects for the most part, perhaps changing things by (my guess) 0.5%. But the reflection losses altogether are small too (~9%). So if you care about such small things to begin with, why not get it right?

 

And yes the experimental uncertainties are also significant but why add systematic error on top of that?

 

Short version: if you are going to do rough calculations just use the Ti and call it a day. If you are going to do detailed reflection calculations then solve the darn problem properly and include everything.

 

All of this reflection stuff amounts to less than 1 stop.

[Cadmium]

Do we have some reason to believe that Schott transmittance calculations for combinations are inaccurate?

 

Andrea, no we don't have any reason to believe that, other than Andy's druthers about calculating T graphs using MatLab.

It would not be only for COMBI, because the Schott program has been doing the T graph calculations for single filters for years.

I have never heard any such doubt about the Schott program calculations, and the single filter T calculations have been in the program at least since the 2012 version.

 

I trust that Schott and the authors of their filter program would not have integrated those features and calculations into the program without some solid understanding of the math involved.

I am going to go on trusting the features and improvements of this program, after all it is endorsed by Schott, who I feel is the best filter glass company on the planet.

Schott's data sheets show more detailed data about their filters than ANY other company provides on their data sheets.

This Schott filter calculator is the only such program that I know of provided by ANY filter company.

If others wish to dissect the program, then do it in some other topic, because this is only conjecture.

I am sorry if people can't figure out how to do T graphs with MatLab, but I use the Schott program, not MatLab.

Research your internal druthers about the math used in the Schott program, and then start a topic about MatLab.

I suggest you take these questions up with the authors of the Schott program, their names and emails are available, maybe they will explain to you how to calculate it using MatLab.

 

Here is a screen shot of one of the data pages,

You see the reflection factor P, the reference thickness, the actual thickness, and the λ[nm] by λ[nm] transmission data, all entered as Ti data.

The Ti graphs don't use P as part of the calculation, but the T graphs do.

Other than that, I can't speak about the internal math of this program, but it is there to be found if you want to click around.

But this topic is not about all of that if you ask me.

[Andy Perrin]

Cadmium I don’t know why you are taking this personally but please don’t.

 

I was not referring to single filter calculations — the Schott program does those correctly. I have previously verified that. Whole discussion above was about stacks. And how much approximation is acceptable in a calculation.

[Cadmium]

I am not taking it 'personally'. I am just telling you to use facts.

We note your doubt, but please verify your 'facts' in some other topic before you suggest there is a problem with the Schott filter calculations.

This topic is about something else.

[ulf]

You can actually calculate the single bounce reflection coefficient from the refractive indices at any angle and wavelength because Schott gives the refractive index at two (sometimes three) wavelengths or they give the Abbe number which can also be used to find the dispersion. Definitely these are small effects for the most part, perhaps changing things by (my guess) 0.5%. But the reflection losses altogether are small too (~9%). So if you care about such small things to begin with, why not get it right?

 

And yes the experimental uncertainties are also significant but why add systematic error on top of that?

 

Short version: if you are going to do rough calculations just use the Ti and call it a day. If you are going to do detailed reflection calculations then solve the darn problem properly and include everything.

 

All of this reflection stuff amounts to less than 1 stop.

 

I haven't looked that close to the Schott data sheets.

The ones from Hoya is even less informative.

 

I fully agree that a good attempt to do things correctly is the way to go, but the deeper you dig the more complex it gets.

For non-normal incidents there will be polarisation-effects too, making it a bit too complex for a simple Excel sheet.

 

Not all have access and skill enough to use Matlab.

As I am one of them, lacking that, I instead prefer the slightly better approximation with a single bounce reflection loss, over the Ti, well knowing that it is not the full truth.

For practical use the difference in exposure time is not that important, even with just the Ti.

[Andrea B.]

Andy: ......single filter calculations — the Schott program does those correctly. I have previously verified that.

 

Andy, it's good to know that you have verified the calculations. And thank you for mentioning that. :)

 

---

 

Cadmium, at ease please. :) Nothing going on here. We are just discussing an interesting topic. My question was sincere and not meant as a challenge. I would not do that on my own forum.

 

---

 

Now.....moving on..... B) B) B)

 

I am cranking out a new set of charts for Transmittance only. They will be very nice and useful I think. Still only for Schott. I'll tackle Hoya later.

 

I first needed to learn the Schott program. Not that it is very hard to do that. But there are some little adjustments to the axes and so forth which one needs to figure out how to do. And there is no way to reset the program, so I have to keep an unaltered copy of the program to occasionally replace my working copy.

 

Some of the Transmittance charts I have created have to be combined in Photoshop Elements. So I needed to work out a way to do that. Makes this effort a little more time consuming, but it is doable.

 

SAMPLE of TRANSMITTANCE CHARTS

http://www.ultravioletphotography.com/content/index.php/topic/2861-filter-transmission-charts-temporary-trial/page__view__findpost__p__22388

[Andy Perrin]

Ulf, I basically agree with you that none of this makes much difference for photography. I’m going to drop the subject. It is a debate about a loss of perhaps 0.5% and I think that is the very definition of when you have to let things go...

[Andrea B.]

It is important to hear about these various factors even if we cannot account for all of them in the modeling. But I agree that we are "close enough for all practical purposes" with my little Transmittance chart efforts. (You do know that old story about the King, the Prince, and the King's Daughter, don't you?)

 

I was thinking that if we wanted to know what we were really going to be able to record, then we would have to include the lens characteristics and the amount UV-light. We've seen from Jonathan's attempt that it gets rather messy rather quickly. Andy, wouldn't that would make for some interesting programming? B)

[ulf]

I was thinking that if we wanted to know what we were really going to be able to record, then we would have to include the lens characteristics and the amount UV-light. We've seen from Jonathan's attempt that it gets rather messy rather quickly. Andy, wouldn't that would make for some interesting programming? B)

 

I'm on my way playing with a simple form of that, without programming.

 

I saved a slightly boxcar-averaged version of the sun-light model output as a custom filter into the Schott calculator as a quick test.

It worked well and could be combined in a filter stack to give an usable result.

 

The same could be done with measured filter- and lens-transmission responses.

 

The output from my spectrometer do not save data in the integer-wavelength bins needed by the Schott calculator, but I got help with a script that make that data transformation.

 

It will be nice when I have have a custom model of the Baader U, but I do not know how to handle the reflections from the dichroic BaaderU. I guess here the estimate must be more rough, just using the Ti-values.

 

The only part missing for me is the camera sensor response.

[Andy Perrin]

I've thought about this but it's impossible to directly model the lens. There is simply not enough information about exact distances, curvatures, etc. If we had measured lens transmission we could certainly use Jonathan's results to get estimated sensor response (for his sensor, of course). What I would love is a measured reflection spectrum for the front surface of various lenses, because that would allow me to predict hot spots (maybe).

[ulf]

I can provide my lens transmissions data, if someone wants it. Just let me know.

 

It has a somewhat limited quality due to my measurement setup and methods.

The spectra are not absolute transmission measurements, but are at least good indications of possible transmissions.

They are normalised for an average of 100% in a short wavelength range just into the VIS range.

Not done with integrating sphere or high end spectrometer.

 

Still they are agreeing reasonably with other measurements published on the web elsewhere, made with secret not disclosed methods or equipment made by a former member.

I can in detail describe the setup and my way of doing the measurements to help anyone judging their validity.

[Cadmium]

Just as an example, using the Schott 2017 filter calculator, with the all new Combi Diabatic T graph feature, I have overlaid graphs of the Ti and T versions of the same UV only filter stack.

The graphs are not sized or numbered the same in the calculator program, so I have adjusted the scale to make things line up.

Here you see a comparison between the Internal Transmittance (Ti) and the Transmittance of the same stack.

You do the math. ;)

 

This feature is handy for getting a little better idea of what to expect the real life stack transmission will be.

[ulf]

The difference between the stack peaks seams small, close to 4 or 5%

That is what I would have expected from a cemented stack with transmission losses in the two outer surfaces.

 

The T-graph result is just a scaling of T_i by the Peff value you can change in the Data Input page.

 

A better P_eff can be calculated for each filter material if you know the refractive index of the material in the wavelength range you are interested in.

 

See the section about Reflectivity halfway down the page: https://en.wikipedia...efractive_index

The P_eff = R_o² calculated for each filter material.

An air spaced stack could better be plotted with a P_{eff combined} = P_{eff 1} x P_{eff 2}

 

Just as Andy pointed out before this is not completely correct as it is not taking into account the multiple reflections between the filter surfaces, both internally and between the two filters in the stack.

The correct value would be slightly lower due to those losses.

 

When I measure stacks in my spectrometer setup the results align well with this.

At the next measurement session I will see if I can resolve any evidence of the multiple reflection effect or if that is too small for my setup to detect.

[dabateman]

Andrea, Thank you from the initial title I didn't know you only wanted to go with Hoya/Schott glass. But that makes sense as its the most fun to mix and match with. For dichroic filters I am guessing there are now only really three common ones anyway. The Baader Venus U, the Straight edge U and Straight edge U mark 2. Going down the Omega rabit hole is really only for scientific uses and not intended for photography.

 

Cadmium, Thank you!! your reply post is very interesting. I thought magically you could cut a 4mm into a 1mm and 2mm. I didn't know the limits, as I have never done this and I will never do this. Much less time for me to just order it than try to cut even a simple sheet. The main problem I would say with Microscope filters is heavy or deep pitting. That is typically when they are sold. Students will leave lazers or lights focused on them and you will get holes in the glass. So yes I should say to anyone reading this if this thread survives that buying used microscope filters is a big risk!

 

The rest of this thread is also interesting. I do like the Schott Excel sheet. What I have found is it must be opened in Microsoft Excel. Opening in LibreOffice does not work well. The curves and axes do not populate correctly. Hoya does have an extremely limited software called Hoya Cal. Its an exe file that only provides the internal transmission data for a single filter. It can output a .csv file though to open the data. However, I just typed in the Hoya data from the product sheets into Schott Excel sheet user section, as this although tedious maybe more accurate.

 

With all the added Science discussion hear I do hope you keep this thread now Andrea.

[Cadmium]

Then by all means write a better program using better math that uses your better calculation. :)

I am just pointing out the new feature which is better than not having that feature, as all the previous version didn't have,

and on occasion some have complained about combi Ti graphs that I have posted because they don't represent true transmission.

So... what should we do, post Ti graphs and have people complain, or post T graphs and have people say those arn't correct either?

All I am doing is pointing out a new feature that was not there before, not in the version Andrea is using, and a great added improbvement.

If you don't think the math is correct in the Schott program, then make something else.

[dabateman]

One thing that should be mentioned from a photography point of view, is that going from 80% transmission to 70% will not really be noticed. But going from 10% to 20% would. Also if you have a filter with only 10% transmission, you may need a better filter.

However from a uv photography point of view 5% IR will kill the shot.

I am adding this as for a general discussion, these things should be noted. Photography is about stops which we see as doublings in the spectral peaks.

 

Andrea, when you do an update a brief introduction on this could help a general reader. Also important is the width of the curve. 100% transmission with 10nm range is not as good as 40% transmission with 60nm peak width, in regards to exposure time. Although that too would be dependent on where in the uv spectrum the peaks fall due to camera sensitivity. A 280nm to 350nm filter would have very long exposure time, compared to a filter with 385nm to 395nm range. So the wording would need to be clear.