Lens design #2

[Stefano]

This lens is similar to the previous one, but has a longer focal length and is better corrected. Like the previous lens, it is not corrected for chromatic aberration.

 

The focal length is about 55 mm, and the maximum aperture is f/2.8.

 

Optical scheme:

 

Lens at f/2.8, ray angle of 14° (covers APS-C):

The lens actually covers larger sensors, at the expense of vignetting and barrel distorsion. Here's a 30° ray angle:

At close focus distances (here 150 mm), some field curvature emerges:

Lens at various apertures, 14° ray angle, infinity focus:

 

 

There's plenty of space for the aperture, and also there's room for an helicoid at the rear. The most challenging part is the 0.5 mm gap between element 2 and 3.

[lukaszgryglicki]

That looks very interesting, I'm not sure if I read those drawings correctly, but looks quite good even at f=2.8.

 

[colinbm]

The 4 lenses are US$420 + Freight.
What else is needed, 1' inside diameter tube, adjustable aperture & a helicoid.
This seem to be do-able.
The lenses could be spaced with a sleeve of black metal velvet foil.

[ulf]

@StefanoFrom your calculated diagrams it looks like you are using the lenses all the way to the edge. In the real world the lens elements at edges will have to rest on some spacer rings.

They will reduce the usable optical path.

There are suitable building components for lens assemblies at Thorlabs that could be used: https://www.thorlabs.de/newgrouppage9.cfm?objectgroup_id=3307

I think finding a solution with a big iris like in your diagrams will be tricky. Maybe that will reduce the usable optical diameter even further.

The iris types fitting the 1" build systems have a maximum opening of 12mm https://www.thorlabs.de/newgrouppage9.cfm?objectgroup_id=1479

 

Have you checked how sensitive the design is to minute variations in spacing of the elements?

For normal camera lenses that is often very important, at least for some of the lens elements and adjustments if needed might be done in special optical bench setups.

A mass produced lens can have much of the needed precision solved in high tolerance machining of the parts. Here the build mist be adjusted by tuning with retaining rings.

With the build systen the problem of having the optical axis of the elements coincide is mostly solved. 

 

If high accuracy spacing is needed here then assembling the design and get close to the theoretical performance might be a difficult task.

With threaded retaining rings it be doable, but the calculated ideal design will be very tricky to reach.

 

At Thorlabs I have not found any way to create the 0.5mm space, but that might be done with an O-Ring with a slightly bigger material diameter that can be compressed with at threaded retaining ring on the other side of the lens element.

 

58 minutes ago, colinbm said:

The 4 lenses are US$420 + Freight.
What else is needed, 1' inside diameter tube, adjustable aperture & a helicoid.
This seem to be do-able.
The lenses could be spaced with a sleeve of black metal velvet foil.

The added costs of the mechanical components will make the complete build quite more expensive than US$420!

[colinbm]

This lens would need to be able to match up with M42 x1mm helicoids & M42 to camera mounts.
Is there an adaptor for this ThorLabs 1" tube ?

@ulf I like the compressed O-ring idea too.

[colinbm]

ThorLabs has adaptors for M42x1mm to their 1' system.
 

SM1A48 Adapter with External SM1 Threads and Internal M42 x 1.0 Threads   SM1A49 Adapter with External M42 x 1.0 Threads and Internal SM1 Threads

[ulf]

It would be interesting to see how much the chromatic aberration will cause problems in the limited spectral range created by the combination of a filter like Baader U and the fading UV-sensitivity of the image sensor. 

Maybe the image flaws caused by imperfect lens element positioning will be greater.

 

 The fun of trying to build a workable lens might be the big outcome of this, not the final resulting lens. 

[colinbm]

What other UV lens can you get for under $1,000 that covers #5mm full frame ?

I would like to try this, if I can get a good materials list.
There are two 1" apertures, 1-12mm https://www.thorlabs.com/thorproduct.cfm?partnumber=SM1D12

1.4 - 25mm https://www.thorlabs.com/thorproduct.cfm?partnumber=SM1D25

[Stefano]

5 hours ago, lukaszgryglicki said:

That looks very interesting, I'm not sure if I read those drawings correctly, but looks quite good even at f=2.8.

Yes, it is reasonably sharp at f/2.8.

[Stefano]

3 hours ago, ulf said:

@StefanoFrom your calculated diagrams it looks like you are using the lenses all the way to the edge. In the real world the lens elements at edges will have to rest on some spacer rings.

They will reduce the usable optical path.

There are suitable building components for lens assemblies at Thorlabs that could be used: https://www.thorlabs.de/newgrouppage9.cfm?objectgroup_id=3307

I think finding a solution with a big iris like in your diagrams will be tricky. Maybe that will reduce the usable optical diameter even further.

The iris types fitting the 1" build systems have a maximum opening of 12mm https://www.thorlabs.de/newgrouppage9.cfm?objectgroup_id=1479

Yes, although this happens mostly at wide apertures.

Here's the lens at f/5.6:

At f/2.8 there will be some vignetting.

[Stefano]

3 hours ago, ulf said:

Have you checked how sensitive the design is to minute variations in spacing of the elements?

For normal camera lenses that is often very important, at least for some of the lens elements and adjustments if needed might be done in special optical bench setups.

A mass produced lens can have much of the needed precision solved in high tolerance machining of the parts. Here the build mist be adjusted by tuning with retaining rings.

With the build systen the problem of having the optical axis of the elements coincide is mostly solved. 

 

If high accuracy spacing is needed here then assembling the design and get close to the theoretical performance might be a difficult task.

With threaded retaining rings it be doable, but the calculated ideal design will be very tricky to reach.

 

At Thorlabs I have not found any way to create the 0.5mm space, but that might be done with an O-Ring with a slightly bigger material diameter that can be compressed with at threaded retaining ring on the other side of the lens element.

The gap between elelemt 1 and 2 can be shrinked without affecting the performance much, while the 0.5 mm gap should not be wider than 0.6 mm and if reduced to 0 (the corners of the lenses touching together) the performance is slightly better at narrower apertures like f/5.6 but looks to be worse at f/2.8, but maybe I am being tricked by the diagram.

[Stefano]

Yes, the lens needs adapters to be mounted on a M42 helicoid, and I would also use a M42 iris with adapters. As Colin said, Thorlabs sells them.

[Stefano]

Also I'm aware that this is not a cheap lens, I won't build it now, but it is significantly cheaper than other options and has a moderately wide aperture of f/2.8.

[Stefano]

Also another important aspect are AR coatings. Thorlabs doesn't offer coatings that are effettive in the 200-1100 nm range, but only coatings that are effective in UV, VIS, NIR and SWIR (and laser wavelengths). Those coatings reduce reflections from about 4% to less than 1% in the designed range, but increase reflections to >10% or even >20% outside. A lens with UV coatings wouldn't be very useful in IR.

[ulf]

I applaud anyone daring to try to build a lens like this.

I must admit is difficult to tell if I am pessimistic or realistic about the performance to be expected.

I have never tried to design something like this.

 

Instead I have experience with electronic design of complex circuits and I think there are some similarities to this lens design.

 

The theoretical design on paper always work, sometimes only driven by the pride of the designer!! ;-)

In the real world there can turn up surprising flaws.

They can be caught by a correct simple SPICE simulation or prototyping and testing.

 

Also when several components are involved there are tolerances that interact.

For the lens build that is like lens positions and lens shape tolerances. I expect the Thorlab lenses have a good shape, but the rest...?

When simulating the possible electronic outcome of all tolerance variations you can get very surprised and disappointed.

In some cases a few or several components need to be tuned together to get what you want.

 

My thoughts about this lens project are just based on how I have seen the mechanical design of good older VIS lenses.

There, for some lens elements it seams like the position and centring of their optical axis can be super critical. 

Here there are fewer corrections needed/possible as all elements use the same material. That might make the design easier.

 

33 minutes ago, Stefano said:

Yes, it is reasonably sharp at f/2.8.

How do the result of that design diagram translate to real image sharpness?  

 

I rejected the actuator controlled iris with a bigger opening that Colin found in fear of light leakage, but that might not be a big problem if the filter are rear mounted.

 

I am eagerly waiting for someone to try to build this design to see if I can say "What did I tell you! , or not  ;-).

 

It might be a waste of money to get a good lens , but still fun to try or turn out to be at least marginally usable.

 

With a lot of good luck?, it can be a usable lens, at least when stopped down to f/4 or more.

It will naturally be limited to use with narrow bandwidths, but maybe also OK with a typical UV-stack or a Baader U.

[ulf]

20 minutes ago, Stefano said:

Also another important aspect are AR coatings. Thorlabs doesn't offer coatings that are effettive in the 200-1100 nm range, but only coatings that are effective in UV, VIS, NIR and SWIR (and laser wavelengths). Those coatings reduce reflections from about 4% to less than 1% in the designed range, but increase reflections to >10% or even >20% outside. A lens with UV coatings wouldn't be very useful in IR.

I do not think there are anyone offer that type of super coating, if at all technically possible.

 

As there are very many quite well working alternative lenses available for VIS and NIR this lens build IMHO should be aimed at UV only then an UV coating might improve the performance if the rest of the design is good enough to be detected.

 

Remember that the "increased reflections to >10% or even >20% outside" is for each lens surface. Your design have eight lens to air surfaces, making the added blocking much worse.

[colinbm]

Thanks Ulf
As Stefano suggested, I would prefer to use a M42 aperture that I already have, via the available adapters.

[Stefano]

20 minutes ago, ulf said:

How do the result of that design diagram translate to real image sharpness?  

For that I probably need a more advanced software like Zemax. If anyone has access to better software and wants to simulate this lens, it would be useful.

[colinbm]

I already have M42 adapters for 52mm filters.

[Stefano]

I'm also thinking that the doublet formed by lens 2 and 3 behaves like a single thick lens. I don't think such lenses are available, but using two lenses instead of one increases the degrees of freedom and possibly allows to design a better lens. Maybe this lens can be fine-tuned.

[Stefano]

Does having lens surfaces very close together cause problems? (like between elements 2 and 3, if the edges touch). I think things like Newton rings should be avoided (anyway, those lenses have slightly different curvature radii, as can be seen in the first image, so they wouldn't touch in the middle)

[dabateman]

Elements 2 and 3 can be spaced with black electric tape cut to size. Not really an issue. 

@Stefano,

what is the distance from rear element to focus plane at infinity? 

 

Was this simulated at 550nm?

 

Its possible the distances might be slightly different at 365nm.

 

I might be able to make this lens inside a Canon 50mm f1.8 ii lens. It has spacing for three elements infront of the aperture and a holder behind the aperture for an other lens group. I can't remember the maximum aperture opening,  but its around 17-20mm. Maybe 20mm. So would allow for nearly F/2.8 design. 

The advantage is I would get autofocus and the focal length of 55mm isn't that different from 50mm, so IBIS will not be far off, which was a problem with 75mm element design I have.

 

 

[Stefano]

I don't know the wavelength used by the software, I tried to check it before and I have to figure out where to see it. Probably the software used the yellow sodium line, but I will try to optimize the lens for UV, maybe around 350 nm.

 

I am not at home now, so I can't check the distance between the rear element and the focus plane, but judging from the drawing it is maybe 55-60 mm.

 

Also, the first 3 elements could fit nicely in a Thorlabs 1" long tube.

[ulf]

1 hour ago, dabateman said:

Elements 2 and 3 can be spaced with black electric tape cut to size. Not really an issue. 

I think that is a good idea, if at all needed. I think trying to optimise for f/2.8 with all the artefacts at that speed is a stupid idea. 

 

The image quality will improve when stopping down and I guess the goal is to get good image quality.

If I were to design this I would not try to make a very fast lens. Instead something like f/4 would be a more reasonable goal.

 

Most fast older lenses without aspherical elements are quite fuzzy wide open.

If you say that a short distance between element 2 and 3 improve performance at f/5.6 that is the best approach.

[Stefano]

The software seems to use 587.5618 nm as the test wavelength, which is yellow. All tests and measurements below were done at that wavelength as I don't know how to change it. I will update those if I can change the wavelength.

 

For David, the software doesn't give me that distance (as far as I know), so I measured it in Paint from a large screenshot (I know this isn't the best way, but should be pretty accurate, the screenshot was almost 2000 pixels wide). The distance from the surface of element 4 to the sensor is about 50.9 mm, at infinity focus. In UV, this distance will likely be a few mm shorter.

 

For Ulf, I did some tests with the spacing between elements 2 and 3, with the lens set at f/5.6 to see the performance when stopped down a bit.

 

The images below were strongly magnified, as much as the software allowed. When the distance is 0 mm, it means the outer edges of the lenses are touching. The angle in degrees is the angle of the input ray (from the subject). An angle of 14° covers approximately the diagonal of a Canon APS-C sensor.

Lens at f/5.6.

 

0°, 0 mm

 

0°, 0.1 mm

 

0°, 0.2 mm

 

0°, 0.3 mm

 

0°, 0.4 mm

 

0°, 0.5 mm

 

7°, 0 mm

 

7°, 0.1 mm

 

7°, 0.2 mm

 

7°, 0.3 mm

 

7°, 0.4 mm

 

7°, 0.5 mm

 

14°, 0 mm

 

14°, 0.1 mm

 

14°, 0.2 mm

 

14°, 0.3 mm

 

14°, 0.4 mm

 

14°, 0.5 mm

 

The 0° ray fan shows some spherical aberration, and improves at narrower separation.

The 7° ray fan shows some field curvature, and improves at narrower separation.

The 14° ray fan shows some coma (I think), and improves at wider separation.

 

I would say it is better to make elements 2 and 3 touch.

 

To give an idea of the magnification, here's a full view, at 0 mm separation:

 

Also, the free version of the software I'm using (WinLens3D) offers some tools, like a spot diagram. Here's it at f/5.6:

 

The thin air gap between element 2 and 3 seems to be important, as I tried to replace them with a single lens with the same thickness and curvature radii but without the gap, and the quality was slightly worse.