Brandon Dube
Offline
• •
Upload & Sell: Off
|
uhoh7 wrote:
Anecdotal, though a good source. We need to see the charts. 11-24 consistency impressive.
These are great tests they are doing. A real service.
The real message is: if you are choosing carefully, which lens to buy, then get it from somewhere it can be returned, and test it right away: if you care.
No numbers, scores, or charts yet because the sample size is very small, but I can tell you I do not think the 35/1.4 ASPH is not more consistent than the 35L in absolute terms, based on a sample size of 5 (this is the highest I have measured of any Leica model, with 15 total models tested on some capacity). That is to say, the consistency score at n=5 is worse, but I don't know how much it will necessarily change between 5 and 10+ samples.
More generally about manufacturing, I would like to offer some insight into this blog post...
No low to medium volume optics manufacture really works with assembly lines. This means if they make less than, say, 10,000 lenses per month they are making them in a similar fashion to how Leica is.
A 1um machining tolerance is, well, something I doubt. 1/1000" is a high precision part. This is 25x finer than that and to me seems unattainable for the price of leica lenses. It's also way better than is needed; the centering quality of most optical parts is sensitive to more like 10-100um, and the metals don't need to be that finely machined to not seize on each other. To say that the lenses would not function if they were manufactured to a lower tolerance to me sounds misinformed. Photolithographic lenses commonly aren't aligned to that tolerance, and they are corrected over a full order of magnitude better than any Leica lens ever designed.
Regarding special glass, 900403 is a glass code and is more properly written 900.403. It means the glass has a refractive index of 1.900, and an abbe number of 40.3. The presence of 12 ingredients is not an extraordinary number, nor is the use of lanthanum. Here you can see an Ohara glass chart, I use them because they have the most transparent naming scheme: http://www.oharacorp.com/images/map-catalog-02.gif
S-LAH breaks down into "S," "La," "H." S just means it's from the new series of lead free glasses, and La is lanthanum. You can order S-LAH58 from Ohara right now which has a code of 883.408. It's slightly less refractive but slightly less dispersive. Realistically you could substitute this glass in for anything designed with the 900.403 glass and get the same performance. A 60x glass is very very expensive (about the same as CaF2) but also not unheard of.
In many cases the factory receives the glass as pressings which roughly approximate the final shape of the lens. Other glasses can only be obtained in blocks which must be sliced into sheets, cut into squares, cemented together and ground to from a stack of discs.
This is ABSOLUTELY not done. Optical materials must be extremely homogeneous and free of any bubbles, impurities, or other defects. If a manufacture does their own cutting they receive slab glass. Blanks are cut from the slab with a circular diamond tool and then processed into lenses. If glass is received in sheet form it may be cut, but it will never, ever, ever be cemented to form a thicker piece.
An inventory of 100,000 grinding and/or polishing heads is also silly. A shop may keep 100, even a thousand, but not a hundred thousand. A head is good to produce elements of a single radius of curvature and a small range of diameters. Leica currently makes fewer than 100 lens models and on average they have less than 10 elements, so they would need at most 2,000 heads to produce their current lens lineup. They have not produced 50 times this number of models in recent time.
The use of CNC equipment from manufactures like Optipro, Satisloh, schneider (not the same as the lens company), and others is both good and bad. CNC is actually faster compared to traditional grind and polish techniques and faster per lens, but you can only work one piece at a time. Traditional grinding and polishing you can work maybe 6 pieces per head, so while the process takes longer and is a bit faster in volume and about 50% cheaper.
CNC is, however, better for good surfaces if you have well tuned machines. The downside is that CNC can also produce Mid Spatial Frequency errors, MSF. Here are two images of MSF: http://www.mflens.com/quality/waviness.html
http://opticalengineering.spiedigitallibrary.org/article.aspx?articleid=1892215
Look familiar? MSF is what has caused people to associate "onion-ring bokeh" with aspherical lenses. In reality, it is an interference pattern brought about by MSF. You can CNC spherical optics, but it is more or less "needed" for good aspheres.
Checking after each step is completely normal and not noteworthy in any respect. Grinding an element may take multiple days of continuous processing and that time is expensive. You do not risk waste by not performing a 10-15 second test plate check before moving on. "Proper tolerances" is a nice way to fluff it up. Essentially they are going to check the center thickness and radius of curvature. The surface quality cannot be evaluated easily until the surface becomes specular, i.e. polished. The radius is also stepped down, the roughest grind to remove tons of material might produce a R=205mm part. The medium grind would produce R=201-200.5mm, and then the fine grind would produce R=200.01-200.000mm. This has to do with padding yourself against error.
Leica applies a standard of ±0.0002% for the accuracy of the refractive index. This compares to the international standard of ±0.001% as applied by other lens manufacturers. The accuracy of the Abbe number, the measure for dispersion, is ±0.2% for Leica compared to ±0.8% internationally.
The author is way off on their tolerance of Nd. Standard tolerance is +- 0.0005 for Nd. You can find this in "How
to Save Time (and Money)
When Specifying Optical Glass" from JML Optical if you wish for a reference, though Schott, Hoya, Ohara, Nikon, Zeiss, Optimax, QED Optics, Corning, Jenoptik, and every other manufacture I can think of use the same tolerance. +- 0.0002 is indeed the high tolerance glass and is about 2-3x more expensive than the standard material.
The given value for Vd tolerance is accurate. I believe Nd +- 0.0002 glass also carries Vd +- 0.2% anyway. In any event, these tolerances are not necessary for Leica's correction level. If their lenses were diffraction limited at maximum relative aperture it would matter, but that is not the case.
A quarter wave surface quality is a standard tolerance for the RMS roughness. If they are referring to Peak-to-Valley, a quarter wave is quite good but still not exceptional. A half wave is not the standard tolerance for other manufactures, it is considered the bare minimum for imaging quality optics. Laser quality begins at 1/20 wave and ranges to 1/200th wave. Flat optics are often tenth wave quality simply due to the ease of their manufacture. In any event, quarter wave is not special. It means the surface is no more than +- 150nm from a reference plane but that is actually not exceptional for an optical surface. Angstrom level RMS roughness is exceptional. It should be noted that the wave reference for these is not the average wavelength, but is 632.8nm. This is the wavelength of a HeNe laser, as used in Zygo and other interferometers, which is how these are measured. http://zygo.com/?/met/interferometers/dynafiz/ Other manufactures exist, e.g. 4D, but Zygo is more or less de facto.
Here is a 1/40th wave part I polished myself:
http://i.imgur.com/ueQEoEw.png
This part was just prepared as a spot part for MRF, a technology used by Leica I will talk about more a bit below. The author is incorrect on the centering and thickness tolerances. Both of these are typically held to about 10um. 1um is a typical result from a CNC machine for centering, and thickness to within 5um. Nanometer precision for those two quantities is highly extraneous and it would be random luck to be that accurate.
Regarding aspheres, I could write chapters about their manufacture and will not go into too much detail, but the author is not very accurate. Precision molding is not just impossible with exotic glasses, the glass must have a low Tg, or glass transition temperature where it becomes a very very viscous liquid. Manufactures each offer about 9 low Tg glasses. They all offer almost 140 different glass varieties. Typical aspheres are still polished for small departure (<5um) or ground and polished.
The CNC generation of optics does require better tolerances, that much is accurate. When I mount a part for deterministic polishing I do not accept more than 1um of runout on my probe for the centering, so I am aligning to about a half micron. I am not sure what Leica accepts, but this is a very easy tolerance to keep for a skilled operator.
The author talks about sub-aperture grinding and polishing. As far as I am aware, there are no sub-aperture grinding machines. His description of MRF is terribly incorrect, though.
You can see Optotech machines in the photos earlier in the article. They are a good manufacture. There really are no bad eggs in the CNC optical manufacturing world. The rolling bottles with QED logos lower are MRF fluid. MRF uses a smart fluid that is about 60% iron by weight, 30% water, 8% cerium oxide or other abrasive (diamond and other abrasives are available), and 2% additives like gelatin to stabilize the fluid.
The machines are in the 7 figure price range and if you are not competitive you do not sell any machines and quickly go bankrupt. Optifab ( http://spie.org/x27531.xml ) was not long ago at all and about a billion dollars of machines, lenses, etc, were on the exhibition floor. I attached two photos from the show at the bottom of the page to give you an idea of some of the monster lenses on display.
MRF works by having a fluid with an embedded polishing agent. The fluid is passed over a wheel, underneath which is an electromagnet. The fluid is drawn tightly against the wheel and the polishing agent is pushed to the surface. The iron can be deformed but is fairly stiff. It provides an infinitely contourable tool and is capable of some of the finest surface figuring available. Angstrom level RMS roughness is easily doable by MRF but more importantly, it can generate any shape you wish, unlike ring tool or other methods. Here's a video of a machine running at Jenoptik:
The fluid must be rolled when not in the machine as it will separate if it stops. Once it separates the fluid is "dead" and not suitable in any capacity for use. The machines use a very complex pump system to keep almost the entire volume in motion at any given time. Each of those large containers visible in the photo is worth about $800. You can see they use C10+, which is the most general fluid. D11 and a couple others are also available. I'm sure they use D11 occasionally, but is almost a 24 hour job to change the fluid, so C10+ is the best to have on the machine at any given time. A key of MRF is that it requires no tooling to adjust to a new polishing demand. It can clean up mid spatial frequency error, or polish the most complex aspheres. Wonderful technology. Also almost $10M per machine. I believe Panasonic recently purchased a few machines, because they put out some PR piece about "solving the onion ring bokeh of aspheres" and MRF is the best way to do that. MRF is not going to be used at someone like Canon or Nikon I don't think due to volume of production issues. Zeiss also uses it in Oberkochen on e.g. the cinema lenses.
Five aspheres a day depends a lot on the volume of machines. To generate a 40-um departure asphere (this is departure from the best fit sphere, 40um is a "moderate" asphere) you may require 18 hours of polishing. If they have 5 machines, well... yeah they can only make 5 a day. The failure rate is almost 0 though.
The quote at the bottom though is total bullshit. Jenoptik is widely recognized in optics as someone at the forefront of aspheric and freeform lens fabrication. Zeiss is also a leader in freeform and aspheres for their photolith business. There are other manufactures as well. Aspheres and freeform are done best in the US. At the Institute of Optics we also produce some of the best surfaces both to support the activities of the laser lab, http://www.lle.rochester.edu/, and for our research in freeform optics, an area in which we are currently leading. My PI is the director of CeFo, http://centerfreeformoptics.org/ , which Jenoptik is a member of, among other companies. Freeform is basically superaspheres and much much much more difficult to fabricate and test than aspheres. I am currently in the process of working on a freeform telescope to launch into space, we have $100,000 to build it.
Here are some big, big, big lenses from Optifab: http://imgur.com/a/zu8bO The navitar guys have front elements about 8" across, and the corning advanced optics lens is almost 4x the size of a 400/2.8.
If anyone is in or will be in Rochester, NY and would like to see some of this stuff, feel free to contact me. I can show you some of the stuff in a lab I work in at the university or try to organize a tour with a manufacture for you. JML put out this video often passed around for lens manufacturing: and is right down the road.
Regards,
Brandon
|
.
