[Webinar] Cell Phone Lens: The Fundamentals Behind the Optical System Design [Q&A]

Q: In beginning, maybe slide 7 or so, you show an optical ray trace model for a "typical" cell phone lens.  It looks like the aperture stop is on a curved surface where this curved surface goes inside of the front lens.  How is the possible and how can this be defined in Zemax.  Why would you want to do this?

A: Thank you for the question. The lens is actually an iPhone lens from the iPhone 5 or so.  I agree that with the number of rays of the pupil, it can look like the stop is curved and inside the first lens. The lens has two dummy surfaces, where the dummy surface applies to the inner rays but not the outermost rays of the pupil (you can see this by looking at the marginal rays for each field). The patent illustrates the stop surface with a +/- thickness to bring the rays forward and then backward. This is usually done to evaluate performance for tolerancing purposes. It gives more control to the bundle of rays other than the marginal rays to evaluate performance.
 

 
If we delete the two +/- surfaces, you can see the centroid rays change, but not the P=1.00 and P=-1.00 rays at the extremities. The performance doesn't change at all for the nominal design. 

You can find more details in the patent document. 
https://patents.google.com/patent/US20170299845A1/en

Also, if you are interested, please join the discussion related to this topic in a separate community post.

I hope this helps! 
 
-Kats

Hello Katsumoto,

Thanks for giving the webinar. It was very interesting!

Are you familiar with designs when each field enter to the system from different stop:

Trying to optimize when using one ciruclar stop for all fields, fail to converge to anything that works.

Moreover these systems are usally in the range of ~2 WFNO and larger than 50 degree FOV.

Is there any tips that you can give?

Thanks,

Nadav

Q: Is there an advantage of using Even aspheres from a raytracing optimization perspective? Does Zemax optimize Even aspheres better and faster than Q-type aspheres?
A: There is a chance that Zemax optimizes Even Aspheres better, if we are simply looking at convergence of the Merit Function. But the convergence of the Merit Function does not necessarily mean that the optical system is reaching a solution. 

-Kats

Q: I would be interested to know how big  a value of K is too big.
A: You can do some math to see where K becomes dominant, but for typical cellphone optics, +/- 5~10 is getting on the large side. +100  or -100 is out of the question.
-Kats

B: The airgaps between the lenses for the smartphone camera design look very small. What tolerances are you relying on for your physical lenses; and how do you rely on the centering of these lenses? Is it simply the mechanical tolerances of the metal work, or do you use some optical lens centering equipment? Thanks. 
B: For a low-res surveillance lens, it could be mechanical only. For the more high-end lenses, there are optical alignments too. There are some products that require every module to be optically tested after manufacture, and some that take a few from the lot and that is satisfactory. 

-Kats

Q: Does any aspherical coefficient of the Q-type representations conflict with radius of curvature or conic constant, as it happens with Even/Odd aspheres? Thank you.
A: No! That is the beauty of the Q-type 

-Kats