At around 20 seconds in, I say that the Cooke has enough degrees of freedom to account for 'all' aberrations. I should have said 'all third-order aberrations', of course. That point gets made explicitly when we discuss the optimized design, but I don't want to give the wrong impression. The Cooke has enouhg degrees of freedom to account for all third-order aberrations simultaneously, and it's the first design that can do so that we've looked at.
So, as I understood it, the number of rings is set to 5 because the optimizer can use higher degree aberrations to compensate for the lower ones, and for this reason, an attempt to control the aberrations manually would not be as efficient as when done by the automatic optimizer. Thank you for the video. It's really appreciated!
It's my pleasure, thank you. The optimizer is reducing the RMS spot size, and the question is, how many rays do we need to trace to get an accurate representation of the RMS spot? Grids of fixed size suffer from sampling issues, and so you find yourself tracing more rays than needed. Gaussian quadrature works by fitting the rays traced to a polynomial of fixed order, so that as long as this order is higher than the highest radial order of the pupil then we can guarantee that we will accurately measure the RMS spot. In general, n rings will compute the RMS wavefront or spot of a system where the highest order aberration is r^(2n-1). We optimize the RMS spot size to be as small as possible, rather tan optimizing the aberrations directly.
Nice, Mark! I've found that, in some cases, it seems necessary to target certain aberrations specifically, especially primary and secondary axial color (and also sometimes the lateral color). And in also some specific instances, astigmatism needed targeting in order to "shepherd" the lens in the right path. Otherwise, yeah I follow you on letting the optimizer balance everything else :)
It's a fair point Ronian, and of course there are always corner cases. I'm trying to hit the 'Basic Shapes' here. I will be covering getting designs into spec with a fixed number of degrees of freedom later, but since you raise it here's the quick version: Although I think you should not target aberrations directly, distortion is a separate aberration to the image quality of the lens, and so can be used. I'll cover that in the double Gauss video coming soon. It is tempting to add chromatic controls directly in the merit function, but I always suggest you don't do this until you have set the weights of the field and wavelength points. In most cases, we just use weights of 1 for all fields and wavelengths. That's great as a starting point, but in most designs you get the best performance on-axis and at the central wavelength, and the further out you go in both dimensions the worse the performance is . In this case, increasing the weight on the out-of-spec wavelengths or fields lets you move performance from in-spec regions to out-of-spec regions, without increasing the complexity of the merit function or its calculation time. Ultimately all you can do is move performance around, you can't just improve it unilaterally without giving more degrees of freedom.
Hi Mark, this is a great video not only in the main substance but also in the nifty tricks.. I just watched this video but I'll watch all of your videos for sure. I also have a specific question re cooke triplet. Is it possible to design a cooke triplet with about 100x demagnification, let's say taking object size of 20 mm and making it 0.02 mm. The design does not have to be necessarily compact. I'm just wondering if there is a principle problem with this much magnification power. As far as I can tell, for cooke triplet to work, it needs to somehow have some symmetry in power of the lenses and this would make it not a good design for high magnifications. Thanks!
It's always a question of what the performance specification is. Cookes are a good choice for mid-field (say 40 degree field). Best bet is to try it...if it outperforms our spec try two lenses instead ion three, if it underperforms try maybe a Tessar (make the rear lens of the Cooke a doublet)
Is there any constraint which tries to restrain the desired fov in the given detector size or at image surface? If any can you suggest the operand. Thanks.
Very nice videos, looking forward for more content. I have a question: When you finalize the design you force the negative lens surfaces to have same radii to make assembly easier. But if I want to keep the different radii, do you have any comments on having different clear semi diameters or chip zone to indicate which side is front?
Yes indeed, there are several tactics you can use. Different bevel sizes is my favorite, and you can also get markings put on the edge of the lens. When we get to managing edges, I'll cover this is more detail.
Mostly this is prepared automatically based on the constraint selected in merit function wizard. But if you want to add more constraint to your design you have to be clear what optimization you require in your design like what is the target value one is looking for and what will be weightage to that constraint while optimising. But yeah it will be better if we can get an example from MARK himself, It will be helpful. Sir please consider it for a new video.
If softwares like zemax can optimise and give rise to best possible lens then why one need to add multiple lenses why can't we work with single lens as the ultimate goal is to have proper focus?
Hi Anil. The point is that a certain shape can only do so much with a given field and wavelength range. If you need more than a single can give, you have to provide more degrees of freedom. These can be extra surfaces, or aspheric coefficients, or maybe diffractive surfaces. Within the Basic Shapes series I'm keeping to spherical optics and showing how any why more complex shapes give better performance
![[TH] 2024 PMSL SEA Finals D2 | Fall | ต้องรักษาฟอร์ม อย่ายอมให้ใครแซง](http://i.ytimg.com/vi/x38QsGn7A3M/mqdefault.jpg)
At around 20 seconds in, I say that the Cooke has enough degrees of freedom to account for 'all' aberrations. I should have said 'all third-order aberrations', of course. That point gets made explicitly when we discuss the optimized design, but I don't want to give the wrong impression. The Cooke has enouhg degrees of freedom to account for all third-order aberrations simultaneously, and it's the first design that can do so that we've looked at.
Thank you very much for the tutorial & I enjoyed every second of your tutorial to refresh my skills
Very informative! Thanks Mark for this insightful video.
Glad it was helpful!
Hi Mark, very nice video again ! It is very clear and easy to understand, yet it taught me a few things. Looking forward to the next ones !
Best
Glad it was helpful!
So, as I understood it, the number of rings is set to 5 because the optimizer can use higher degree aberrations to compensate for the lower ones, and for this reason, an attempt to control the aberrations manually would not be as efficient as when done by the automatic optimizer.
Thank you for the video. It's really appreciated!
It's my pleasure, thank you. The optimizer is reducing the RMS spot size, and the question is, how many rays do we need to trace to get an accurate representation of the RMS spot? Grids of fixed size suffer from sampling issues, and so you find yourself tracing more rays than needed. Gaussian quadrature works by fitting the rays traced to a polynomial of fixed order, so that as long as this order is higher than the highest radial order of the pupil then we can guarantee that we will accurately measure the RMS spot. In general, n rings will compute the RMS wavefront or spot of a system where the highest order aberration is r^(2n-1). We optimize the RMS spot size to be as small as possible, rather tan optimizing the aberrations directly.
Very nice videos
Nice, Mark! I've found that, in some cases, it seems necessary to target certain aberrations specifically, especially primary and secondary axial color (and also sometimes the lateral color). And in also some specific instances, astigmatism needed targeting in order to "shepherd" the lens in the right path. Otherwise, yeah I follow you on letting the optimizer balance everything else :)
It's a fair point Ronian, and of course there are always corner cases. I'm trying to hit the 'Basic Shapes' here. I will be covering getting designs into spec with a fixed number of degrees of freedom later, but since you raise it here's the quick version:
Although I think you should not target aberrations directly, distortion is a separate aberration to the image quality of the lens, and so can be used. I'll cover that in the double Gauss video coming soon. It is tempting to add chromatic controls directly in the merit function, but I always suggest you don't do this until you have set the weights of the field and wavelength points. In most cases, we just use weights of 1 for all fields and wavelengths. That's great as a starting point, but in most designs you get the best performance on-axis and at the central wavelength, and the further out you go in both dimensions the worse the performance is . In this case, increasing the weight on the out-of-spec wavelengths or fields lets you move performance from in-spec regions to out-of-spec regions, without increasing the complexity of the merit function or its calculation time.
Ultimately all you can do is move performance around, you can't just improve it unilaterally without giving more degrees of freedom.
Dear Mark, very nice videos. I watched them all and learned a lot. Don't you have plans to make some videos about tolerance?
Hi Mark, this is a great video not only in the main substance but also in the nifty tricks.. I just watched this video but I'll watch all of your videos for sure.
I also have a specific question re cooke triplet. Is it possible to design a cooke triplet with about 100x demagnification, let's say taking object size of 20 mm and making it 0.02 mm. The design does not have to be necessarily compact. I'm just wondering if there is a principle problem with this much magnification power. As far as I can tell, for cooke triplet to work, it needs to somehow have some symmetry in power of the lenses and this would make it not a good design for high magnifications. Thanks!
It's always a question of what the performance specification is. Cookes are a good choice for mid-field (say 40 degree field). Best bet is to try it...if it outperforms our spec try two lenses instead ion three, if it underperforms try maybe a Tessar (make the rear lens of the Cooke a doublet)
Is there any constraint which tries to restrain the desired fov in the given detector size or at image surface? If any can you suggest the operand. Thanks.
Sure. Use REAY to have the desired image height on the image (or any other) surface
Very nice videos, looking forward for more content.
I have a question: When you finalize the design you force the negative lens surfaces to have same radii to make assembly easier. But if I want to keep the different radii, do you have any comments on having different clear semi diameters or chip zone to indicate which side is front?
Yes indeed, there are several tactics you can use. Different bevel sizes is my favorite, and you can also get markings put on the edge of the lens. When we get to managing edges, I'll cover this is more detail.
You didn't explain how was the merit function prepared
Mostly this is prepared automatically based on the constraint selected in merit function wizard. But if you want to add more constraint to your design you have to be clear what optimization you require in your design like what is the target value one is looking for and what will be weightage to that constraint while optimising. But yeah it will be better if we can get an example from MARK himself, It will be helpful. Sir please consider it for a new video.
If softwares like zemax can optimise and give rise to best possible lens then why one need to add multiple lenses why can't we work with single lens as the ultimate goal is to have proper focus?
Hi Anil. The point is that a certain shape can only do so much with a given field and wavelength range. If you need more than a single can give, you have to provide more degrees of freedom. These can be extra surfaces, or aspheric coefficients, or maybe diffractive surfaces. Within the Basic Shapes series I'm keeping to spherical optics and showing how any why more complex shapes give better performance
@@marknicholson5508 Thanks. It was helpful to understand.