33 replies to this topic

[ThomasM]

I am trying to check the manufacturing tolerances for lenses in OSLO. There is a Tolerance menu where you can set the trolerance values, but I couldn't figure out how to see the effects. The other way around, if I change the values for e.g. decentration to a large value, it has no effect on the spot size and shape or when I evaluate wavetront the Strehl. It seems that I have misunderstood how it works, I have looked in the program reference but have not found a solution.

 

It would be great if someone could give me some advice.

[IslandPink]

I haven't used Oslo myself, but in CodeV and Zemax ( and our own software ) , tolerance effects would normally be evaluated in MTF. What I mean is that the tolerances might not be deployed until you do an MTF tolerance run. You should also be able to do Strehl ratio somehow. Keep looking in the manual and hopefully someone with specific experience will comment. 

[TOMDEY]

It sounds like something's "turned off" in your software or use of it? It's been a long time since I've done lens design... and Doug was my teacher, advisor, and mentor... I even did some derivations and coding for him way back when. Are you using the full professional version? Someone here should know how to get it to work for you.    Tom

[Mike I. Jones]

I've been designing and tolerancing with Zemax and Code V for decades now, and only key designs into OSLO-EDU for this forum.  I haven't really tried tolerancing in it, so post up a file of interest and I'll see what can be done.

[ThomasM]

I've been designing and tolerancing with Zemax and Code V for decades now, and only key designs into OSLO-EDU for this forum.  I haven't really tried tolerancing in it, so post up a file of interest and I'll see what can be done.

 

Thanks, here is a file of a triplet lens, you know it.  I already checked the decentration tolerances with ATMOS and found that 0.007 mm of the surface with the most curvature, the front of the negative lens, leads to a 10% Strehl loss, so it is very critical. For comparison, I checked a BK7 / F2 achromat, also 100 mm f/4, and the tolerance is even tighter 0.003 mm decentration leads to a 10 % Strehl loss in the green line. So basically this is very critical, and I wanted to check the numbers with OSLO.

Attached Files

•  100mm f4 S-FPL55 S-LAL14 S-FPL55(2).len   848bytes   5 downloads

[jimhoward999]

Oslo Tolerancing is similar to zemax.  Here is an abridged version:

 

On the tolerance tab you have

 

update tolerance data 

Monte Carlo

MTF/WvF tolerancing

 

You hit the tolerance data tab and enter your tolerance set for components and surfaces (radii, cts, air spaces, decenters, wedge, tilt etc)

 

You hit the tolerancing tab and make the selections you want on the dialog screen,    Be sure to select "full report".  click okay and a tolerance sensitivity table should appear.

 

The Monte Carlo routine takes the distribution that you specify and applies it to the tolerance set.  I think "end point" is often most realistic, but that is my personal preference. .  The routine returns the change in the error function (merit function).  So you need to set up the Merit function so it represents whatever specification you are trying to simulate with the Monte Carlo.

[ThomasM]

Oslo Tolerancing is similar to zemax.  Here is an abridged version:

 

On the tolerance tab you have

 

update tolerance data 

Monte Carlo

MTF/WvF tolerancing

 

You hit the tolerance data tab and enter your tolerance set for components and surfaces (radii, cts, air spaces, decenters, wedge, tilt etc)

 

You hit the tolerancing tab and make the selections you want on the dialog screen,    Be sure to select "full report".  click okay and a tolerance sensitivity table should appear.

 

The Monte Carlo routine takes the distribution that you specify and applies it to the tolerance set.  I think "end point" is often most realistic, but that is my personal preference. .  The routine returns the change in the error function (merit function).  So you need to set up the Merit function so it represents whatever specification you are trying to simulate with the Monte Carlo.

 

Thank you,  I am using OSLO edu and only the user-defined tolerance analysis is enabled, all others, e.g. Monte Carlo, are disabled. So if I use the user defined- tolerancing / tolerance sensitivity analysis, I already get an error function of 2.1 without any tolerances, all values set to 0,  whatever that means and to get a small change in the error function e.g. from 2.1 to 2.8, a decentration of 0.3mm is necessary. With ATMOS, even 0.007 mm leads to a significant reduction in the Strehl, -10 %. I know I'm comparing apples with oranges here, as I'm not clear what the error function in OSLO means. On the other hand, I think the 0.007 mm or even only 0.003 mm for the achromat is clearly too small, the 0.3 mm of OSLO seems too large to me.

Comments and advice are very welcome.

 

 

 

[IslandPink]

What is the lens you're using here, Thomas ?

I think you're right that the tolerance should be somewhere in between those extremes.

I just need to get a feel for what you're analysing here, first.

[IslandPink]

....small change in the error function e.g. from 2.1 to 2.8, a decentration of 0.3mm is necessary. With ATMOS, even 0.007 mm leads to a significant reduction in the Strehl, -10 %. I know I'm comparing apples with oranges here, as I'm not clear what the error function in OSLO means. 

I assume OSLO is similar to CodeV and Zemax, typically the 'merit function' is constructed from a few things, but mostly the Rms spot size for the specified field points you have. It will take account also of the wavelength set and any field weighting you have from centre to edge. This value does not directly correspond to any physical measure , but is the number you use to track progress during optimisation, or to compare various design variants, if they are set up the same way.

Your number as quoted suggest that the lens has got about 33% worse ( in terms of Rms spot size ) with the tolerance set applied. 

[ThomasM]

What is the lens you're using here, Thomas ?

I think you're right that the tolerance should be somewhere in between those extremes.

.

The lens is a very fast apo, 100 mm f/4 and for comparision an achromat, also 100 mm f/4, see also post #5.

 

 

Attached Files

•  100mm f4 S-FPL55 S-LAL14 S-FPL55(2).len   848bytes   1 downloads

[MKV]

 

I'm not clear what the error function in OSLO means.  

 

The answer may be in OSLO's Manual. 

[IslandPink]

Ok yes, I suspected so.

Well pending what you get from your reading, and from Jim, you have to have a think about how good this ( these ) lenses need to be.

They are for a pair of binoculars, yes ?

What's the maximum viewing magnification ?

Does the base performance of the design exceed your requirement ?

 

It would be handy if you could get MTF losses from the tolerancing, but not essential.

[ThomasM]

The lenses are for a binocular telescope, actually I think 90 mm f/4. Currently I own two 80 mm f/4.2 achromats, during the day CA is already visible at 20x, and I expect good performance with the apo lenses at 1 mm exit pupil, i.e. 90 x for the triplets. I think that the manufacturing and adjustment tolerances should not degrade the Strehl for green light by more than 10%, so I hope for a Strehl a bit more than 70% or a ~ 10% reduction in contrast in the MFT at 100 cylces/mm. In other words, it is not worth all the effort of making the lenses if a large part of their excellent design performance for such fast lenses gets lost.

[jimhoward999]

Thank you,  I am using OSLO edu and only the user-defined tolerance analysis is enabled, all others, e.g. Monte Carlo, are disabled. So if I use the user defined- tolerancing / tolerance sensitivity analysis, I already get an error function of 2.1 without any tolerances, all values set to 0,  whatever that means and to get a small change in the error function e.g. from 2.1 to 2.8, a decentration of 0.3mm is necessary. With ATMOS, even 0.007 mm leads to a significant reduction in the Strehl, -10 %. I know I'm comparing apples with oranges here, as I'm not clear what the error function in OSLO means. On the other hand, I think the 0.007 mm or even only 0.003 mm for the achromat is clearly too small, the 0.3 mm of OSLO seems too large to me.

Comments and advice are very welcome.

The error function is the same thing as the Merit function.  Oslo just uses a different word for it. 
 

User defined tolerancing is fine.  It lets you basically tolerance on anything.   Whatever ever operands you put in the merit function (error function), it will apply the tolerance and compute the change in that merit function and produce a sensitivity table.  

 

So if you want to tolerance on RMS WFE so you can get at Strehl (1-(2pi sigma)^2) you have to make your error function be RMS WFE.   You can do that either by defining a grid of opd operands at a single field point and making that your error function, or by using one of the pre-defined operands of which RMS WFE is one.

 

 The problem however, is that I don't believe you can do much of that in the .edu version.   I dont think you have access to the rms wavefront error operand and I am not sure you can even define a custom merit function using opd. ...but maybe.

[MKV]

Unless you intend to use these binoculars at an observable Airy disc level, what's the point of chasing the Strehl score? 

[jimhoward999]

One thing you could do is just manually evaluate tolerances by perturbing things one by one and observing the change in Strehl.     Your system is just three elements so there aren't all that many tolerances.

 

Just for fun I took the system that you posted and evaluated it first nominally at 0.55um and then with L1 R2 decentered by 0.007mm.  This means that element is wedged.   I found the monochromatic Strehl at best 0.55um focus to go from 0.990 to 0.979.

 

decentering it by 0.3mm made the Strehl go to zero.

Edited by jimhoward999, 23 March 2025 - 06:06 PM.

[IslandPink]

For x90 , it's not a surprise that it's a bit tweaky.

Most  Apo combinations are more sensitive than Achro doublets, plus you are going for f/4, which makes the whole thing very demanding.

Is there no way you can re-think this for something slower, like f/5 or f/6 ?

[ThomasM]

Unless you intend to use these binoculars at an observable Airy disc level, what's the point of chasing the Strehl score? 

Of course, the BT is not primarily intended for planets, but I want to take advantage of the design. Making the lenses means grinding and polishing 15 surfaces, it would be sad just to spoil the performance you can expect from ray tracing by ignoring the presumably very tight tolerances because you don't know them.

[ThomasM]

For x90 , it's not a surprise that it's a bit tweaky.

Most  Apo combinations are more sensitive than Achro doublets, plus you are going for f/4, which makes the whole thing very demanding.

Is there no way you can re-think this for something slower, like f/5 or f/6 ?

 

What puzzles me is that it is the other way around, the standard achromat is even more sensitive than the triplet apo, decentering the third surface by only 0.003 mm reduces the Strehl by 10%. Meanwhile I think there is a problem with the Atmos results, or I am misinterpreting them, only the third surface is super sensitive to decentering.

[ThomasM]

One thing you could do is just manually evaluate tolerances by perturbing things one by one and observing the change in Strehl.     Your system is just three elements so there aren't all that many tolerances.

 

Just for fun I took the system that you posted and evaluated it first nominally at 0.55um and then with L1 R2 decentered by 0.007mm.  This means that element is wedged.   I found the monochromatic Strehl at best 0.55um focus to go from 0.990 to 0.979.

 

decentering it by 0.3mm made the Strehl go to zero.

 Thank you for looking into the system. I have checked with Atmos the effect of decentering of the first surface, and I am quite surprised, even 5 mm (not a typo, 5 mm) has only a small effect, the Strehl drops from 0.94 to 0.87. I also checked all other surfaces, typically 0.02 mm decentration reduces Strehl by 10%, to 0.85, but the third surface, the concave surface with -173.1 mm radius is very critical, as I wrote before, 0.007 mm decentration reduces Strehl to 0.86. Meanwhile I think there is a problem with the Atmos results, or I am misinterpreting them. So could you please also check areas 2-6 for comparison, that would be great.
 

[MKV]

Of course, the BT is not primarily intended for planets, but I want to take advantage of the design. Making the lenses means grinding and polishing 15 surfaces, it would be sad just to spoil the performance you can expect from ray tracing by ignoring the presumably very tight tolerances because you don't know them.

Of course, how about inhomogeneity tolerance of the substrate blanks? For this project you'll need top grade blank$. 

[MKV]

 Thank you for looking into the system. I have checked with Atmos the effect of decentering of the first surface, and I am quite surprised, even 5 mm (not a typo, 5 mm) has only a small effect, the Strehl drops from 0.94 to 0.87. I also checked all other surfaces, typically 0.02 mm decentration reduces Strehl by 10%, to 0.85, but the third surface, the concave surface with -173.1 mm radius is very critical, as I wrote before, 0.007 mm decentration reduces Strehl to 0.86. Meanwhile I think there is a problem with the Atmos results, or I am misinterpreting them. 

Manually decenter the elements in both ATMOS and OSLO and see how they compare.

[Oregon-raybender]

Wow, 0.007mm decenter, that is very tight and hard to hold or even make.

What happens if you change radius by +/- 5%?  It seems that lens is

correcting a aberration that is causing the issue. I would trace each aberration

( SA, CM, etc...) at each surface and see what that lens is doing.

 

Starry Nights

[IslandPink]

5% will be no use on radius, I would start at about ±0.3% and see how that looks. ( perhaps you meant ± 0.5% ) 

[IslandPink]

Here is a starting set of tolerances I would suggest.

Depending on how committed you are to x90, this may be too slack. 

All dimensions in mm

 

Lens thicknesses :  ±0.05

Air gaps : first one ±0.03 , second one ± 0.04

Radii : ±0.3% 

Centering : call L2 the datum, . then allow 0.03 radial on L1 or L3 

Tlit : call L2 the datum, the 2' arc random on L1 and L2

 

Glass tolerances, I would leave these for now, do a melt fit when ( if ) you buy the glass.