The most important bench mark however is how the lens performs compared to the simulations. A perfectly manufactured lens would produce results exactly the same to the simulations, which is what I need to be able to accurately assess the differences in image quality between my initial prototypes and these new ones. However, it's very very clear in this case that the lenses do not perform according to specification.
Therefore, the question remains: If the lenses do not accurately represent the designs and their simulations, what went wrong?
Okay, lets establish one thing first. The lens designs work. The sample images that I have taken, like the one shown above, are taken through my lenses which either feature 5 or 8 plastic elements depending on which one you're using. The important thing to take from these images is that while the image quality may be bad, the photos do show a focused image. This is proof that the multi-element designs work, and more importantly it proves that the errors are most likely manufacturing based.
Lets look at the image a little closer. The first thing that strikes me is that the errors are spread all over. They aren't confined to the edges like the previous prototype 50.1 PMMA. This is yet another sign that the errors are a result of manufacturing and not the design. There is strong haze, easiest to see in contrasty areas, as if it is almost double imaging. In fact, this is the key. The haze seems to be being caused by multiple versions of the image being cast out of place, turning the image fuzzy. After all, you can see it trying to work, it just doesn't seem to be getting there.
Okay, so the problem is a result of a physical defect caused by manufacturing. Lets have a good look at the optics again.
Above are some of the photos of me assembling the lenses. There are two main problems that you can see in these photos. 1. There are a lot of surface scratches. 2. There is a weird distortion effect near the absolute centre of the image. These factors are the cause of the errors.
Cumulative Errors
It may not be particularly easy to see in the photos I have posted online, but when you look very closely at the lens elements under inspection lights you can still see surface scratches and surface imperfections on the lenses. These micro scratches are insignificant and do not negatively effect image quality when you use the individual element on its own. For example, the initial prototype (50.1 PMMA) was a single element with the same level of surface finish and the results from that were impressive. However, the new designs that I am building are multi-element systems, not singlet systems.
Each and every microscratch on the surface of a lens element scatters the light. This means that if we looked at a ray trace, the rays of light that hit the scratches do not travel to where they are supposed to. The reason this isn't a problem in a singlet system is that there are only 2 surfaces that may have micro-scratches meaning the percentage of scattered rays are low. The majority of the rays are unaffected and carry on to form a high quality image.
A multi-element system is different thought. When a scattered ray from the first element in the system enters into the second in the wrong place, it is further refracted to a point which is even more erroneous. Thus, as the scattered rays caused by scratches at the front of the system refract through the remaining elements, the errors are accumulated and increased. These cumulative errors are much more dramatic than anything you would see in a singlet system, and are the cause of the intense haze. The errors you see on the image are literally the rays hitting the image sensor in the wrong place. Not only that, but each surface adds it's own errors. Therefore, with each element you add, 2 more surfaces that could contribute errors enter into the system, meaning the image quality is further and further degraded. Hence the results from the 90.2 lens (with 8 elements, 16 surfaces that could contain errors) are much worse than the 50.2'8 (5 elements, 10 surfaces that could contain errors). It's error multiplication. Also, with the more elements added, you increase the chance of rays hitting multiple scratches on different surfaces on it's journey through the system, really causing mayhem.
If that was difficult to understand let me quickly explain it with a simple analogy of Chinese Whispers. The message in the game is comparable to the image, and it is whispered from person to person. Each person is like a lens element, and adds small errors accidentally. The more people in the chain, the more errors there will be at the end. By the time the message gets to the final person, there are so many errors accumulated that the message barely resembles what was said at the start, and much the same the image quality barley resembles what it really should be.
Where do these micro errors come from?
When I make an element, I cut the profiles using a CNC machine guaranteed to be accurate up to 0.01mm, and then I take them out and polish them on a manual lathe using papers and paste. Finally, some lenses are buffed on the buffing wheel also.
To put it simply, errors are introduced the moment I touch it. Computers don't get things wrong, errors are a result of human intervention; and in this case they are a result of me polishing the elements by hand. I thought the the surface finish would be high enough for use as prototypes, given that I had already made a singlet prototype which produced adequate image quality using exactly the same methods. I was wrong. The chain effect of errors explained above means that any tiny error I make is multiplied and causes a severe negative effect on the resulting IQ. To make lenses of adequate quality, you simply must eliminate the human factor. Ultimately, these errors can only be eliminated if the elements are polished by a machine.
