Brian Caldwell

The responses here have been very thoughtful, and have given me much to consider.  I may modify my product direction a bit, but on the whole, it seems that my ideas aren't entirely crazy.

Thanks for your comments.  I agree regarding the bokeh shape.  In the case of my test shot the bokeh is actually nine-sided, which is created by the iris shape in the Nikon 135mm f/2 AiS Nikkor that I used as a spherical backer lens.  Upcoming integrated primes will use a very circular iris to yield nice smooth elliptical bokeh.

Thanks for this - its really useful information!  Out of curiosity, do you know anything about the Cineroid viewfinders?

Thanks for your feedback on the squeeze ratio - I was definitely interested in what you would think.

I own Caldwell Photographic, and I design all the Speed Booster optics.

Monitoring is definitely on my mind.  Nothing is easy.

I agree.  Integrated primes (and zooms) are more difficult to develop and manufacture, but its definitely a solution that I ultimately want to provide.

My plan is to do a compact "normal" attachment, a larger wide attachment, and then a series of integrated lenses.  The attachments will definitely focus much closer than the Iscoramas.  Distortion will be very traditional anamorphic style - definitely not the oddball simultaneous combination of barrel and pincushion shown by the Cookes, although probably slightly less than 2x Panavision lenses due to the difference in squeeze ratio.  

I'm working on two attachments, plus a series of integrated primes.  The smaller of the attachments would have 72mm rear threads, and I want to price it so that independent film makers and videographers can afford it.  I'm not far enough along on the other products to have a good feel for pricing yet, but its my intention to compete with the best products out there (e.g., Hawk, Cooke, Panavision, etc.) at a very compelling price.

Breathing will depend on the product.  I am developing two different attachments with variable diopter front focusing, and these will breathe similarly to the SLR magic and Iscorama.  BTW, this isn't zero because the focal length always decreases as you focus closer a - + diopter arrangement.  See this demo, for example:  https://www.youtube.com/watch?v=_tZK5Nm3Gx4   For rear-focus integrated prime lenses the breathing will be much less.  

AFAIK, the Hawk lenses also use variable diopter front-focusing, so the appearance of breathing will depend on the focal length, the total change in magnification, and to a lesser extent the aperture.  You can certainly tell that the Hawk suffered lots of breathing in the comparison I showed above, since the magnification and bokeh are smaller even though it was shot from the same camera location and supposedly has a longer focal length than my prototype.

You may be right about 4:3, although the GH4 is pretty widespread these days, and hopefully will be joined by other modest-cost 4:3 cameras in the near future.  There's also the possibility of shooting with any 16x9 4k camera and cropping the excess, since anamorphic seems to be as much about its look as about its efficient use of sensor space.  For 16x9 you would supposedly want a 1.34x squeeze, but that just doesn't offer much anamorphic appeal.

I've recently started numerous anamorphic lens projects based on the idea that 1.79x squeeze is the ideal ratio when dealing with a 4:3 sensor .  It's often stated that 2x squeeze on 4:3 gives you the DCI standard 2.39:1 scope ratio, but of course this isn't quite true.  If you really want a perfect mapping of 4:3 to 2.39:1 without having to crop the sides, then the correct math  is:  2.39/(4/3) = 1.7925, which I'll round off to 1.79.  So, for the ARRI Alexa, RED Dragon and Panasonic GH4 used in 4:3 mode it seems to me that 1.79x is ideal.  Also, if you consider the Alexa in its Open Gate format (1.55:1) you get 1.55 * 1.79 = 2.77:1, which is almost exactly equal to the classic Ultra Panavision 70 (2.76:1 aspect ratio).

You might be concerned that 1.79x wouldn't give enough anamorphic artifacts, but based on my experience so far it seems that the artifacts are very similar to 2x, and in addition there are significant advantages in size, weight, cost, and image quality.  Some time ago I built a 1.80x prototype that used very traditional rear-group focusing with counter-rotating astigmatizer aberration compensation, and found that it compared very favorably to a similar-spec 2x 140mm Hawk V-Lite:

So, my question is, since I'm about to start spending money like crazy developing this stuff, am I crazy to be going in this direction?

I like your attitude!

No problem!  The original points made in older resurrected threads like this often get pretty obscured.

Earlier in this very old thread a comparison was made between the Canon 85/1.2 + Speed Booster and the Nocticron for m43.  If indeed the Nocticron is the sharper of the two (likely), then my point was that the problem lies in the Canon 85/1.2 and not the Speed Booster.  After all, the latest generation of Speed Boosters is capable of improving on Zeiss Otus sharpness even while increasing the speed to f/1.0 or faster, and I can prove it.

So, the 85/1.2 is a fine lens, but if pure sharpness is your metric then it is outclassed by a Zeiss Otus, and is even further outclassed by a Zeiss Otus + Speed Booster.  Seriously.

Any chance there will be a 4:3 2880x2160 recording mode for anamorphic, or will cropping on post be the solution?

For anamorphic you may want to do a conventional MTF vs frequency chart for a number of selected field points.  For example, when designing anamorphic lenses for Arri Alexa 4:3 I normally use seven field points:  (0,0), (0,8mm), (0,11.88mm), (6mm,0), (8.91mm,0), (6mm, 8mm), and (8.91mm,11.88mm).  Due to the strange asymmetrical aberrations a plot of MTF vs image height can be pretty confusing.  I suppose you could do three separate MTF vs image height charts; one for X, Y, and diagonal planes, respectively

Regarding medium format focal reducers, IMO the main problem is that medium format optics tend to be big and slow, and their optical correction is often ho-hum.  So, for example, if you take a Hasselblad 80/2.8 and add a 0.7x focal reducer you get a 56mm f/2.0 lens with merely average optical correction.  Certainly not something that would set the world on fire, and it would pale in comparison to a Zeiss 55/1.4 Otus, and would probably have a hard time keeping up with a 55/1.8 FE.  In principle you should be able use a focal reducer to match the FOV and DOF the same, assuming you use the same rectangular shape for both formats.  However, the newer 135 designs by Zeiss, Sigma, etc. are so good that you might as well buy one of these instead of going with an adapted medium format solution.  

Hi Rich:

Your numbers would be for the horizontal dimension of the format, not the diagonal.  My numbers indicate the full diagonal dimension.  According to my data, the diagonal (corner-to-corner) of the BMPCC sensor is 14.32mm, and the diagonal of the GH4 Cinema 4k crop is 17.4mm.

I do appreciate your open minded spirit about using optics beyond their design intent.  But just to be clear about what happens in this case - using the 0.58x Speed Booster on the GH4 in cinema 4k mode - I plotted some MTF curves below.  As you can see, the performance is great over the whole BMPCC format, but if you go beyond that the performance falls off a cliff.  I'm  very cautious about all of this because a few users have purchased the BMPCC S.B. with the intention of using it on a larger format, only to be disappointed with the results.  This is why this particular Speed Booster has always been advertised as being BMPCC-specific, and has "BMPCC" engraved on it.

The 0.58x BMPCC Speed Booster is designed to cover an image circle of about 15mm diameter, which is slightly larger than the BMPCC sensor diagonal.  Aberration correction is extremely good within this 15mm diameter, as is the relative illumination.  However, for the GH4 4k mode you need to cover at least 17.4mm.  Assuming you don't damage your camera by trying to mount the BMPCC S.B. on a GH4, you will encounter vignetting and significant image degradation in the corners.  I do not recommend trying this.  By contrast, the new m43 XL Speed Booster will cover the entire m43 format, not just the reduced 4k crop, and you won't have any problems with image quality.  The difference between the BMPCC and XL is only 1/3 stop.

A caution here:  parfocality in a zoom lens is strongly dependent on sample variation.  So, even if a zoom is designed to be parfocal, then a typical as-built lens won't necessarily maintain precise focus through zoom.  High-end cine zooms rely on post-assembly re-machining of at least one of the cams controlling the various zoom group motions.  Needless to say, this is a very expensive process.

Everything I said is true and is backed up with hard evidence.  There's no magic here - just good optics.  Perhaps you might actually read the white paper I gave you a link to?

Eventually, conservative established lens makers in Japan and Germany will wake up and realize the true benefit of the Speed Booster approach for designing high-speed short BFL optics.  With any luck I'll be retiring on royalties when they do

There are numerous examples of Speed Booster/lens combinations that beat native lenses.  For instance, a Voigtlander 90mm/3.5 plus S.B. gives a 60mm f/2.5 that is better than the latest Olympus 60mm f/2.8 macro lens:  http://***URL removed***/forums/post/51895542 .  A Sigma 35/1.4 plus S.B. gives a 25/1.0 that is much better than a Voigtlander 25/0.95:  http://***URL removed***/forums/post/55298481 .  And these examples use the old version of the Speed Booster.  The new ones referred to in my white paper are even better.  Finally, I defy you to find any f/1.0 lens with better image quality than a Zeiss Otus combined with a Speed Booster Ultra.

I would expect to see some degradation due to the filter stack.  However, the 28/1.4 Nikkor is reasonably telecentric AFAIK, so it should have less of the filter-induced astigmatism that plagues other designs.  Probably most of what you are seeing is due to the d800 having vastly cleaner and more detailed images than 35mm film, so its just showing faults in the lens that were always there.

This is a false generalization most likely based on weaknesses inherent to teleconverters, which magnify aberrations.  A well-designed focal reducer, on the other hand, will *shrink* the aberrations as mentioned above by Araucaria.  And with a little know-how you can even do better than that to actually design a focal reducer that compensates some of the aberrations in the master lens.  Here's a recent whitepaper that I wrote proving that a Metabones Speed Booster significantly increases the MTF of various lenses, including the extremely challenging case of a Zeiss Otus:  http://www.metabones.com/assets/a/stories/The Perfect Focal Reducer (Metabones Speed Booster ULTRA for M43) - Whitepaper.pdf

Disclaimer:  I develop the optics used in Metabones Speed Boosters

The "T" version differs from the "non-T" version only in that the "T" has an improved internal blackening to reduce reflections.  The Metabones website shows the original "non-T" version as discontinued: http://www.metabones.com/products/details/MB_SPEF-BMPCC-BM1 , and the "T" version as the current EF-BMPCC model:  http://www.metabones.com/products/details/MB_SPEF-BMCC-BT1 .  Its possible that B&H are selling old stock of the original version, but I think its more likely that they simply made a mistake on their website.  Since you're in Australia, you may get better service simply by ordering direct from Metabones.

Optics is mostly simple, with some occasional complexity thrown in to make it interesting.  In fact, I've almost completely forgotten all the fancy math I learned after high school, since by and large all the math you need to know to be a successful lens designer is geometry, trigonometry, and a bit of algebra.

In your case, assuming that the anamorphic portion is working at infinity (i.e., parallel light in and parallel light out) the aperture of the optical system is determined by the diameter of the iris diaphragm in the taking lens *unless* there is some other limiting aperture in the system.  Imagine that you take a pin and poke a tiny hole in a large piece of aluminum foil.  Next, open the f/1.2 taking lens wide open and place the aluminum foil in front of your Rectilux so that the pinhole is centered on the optical axis.  Clearly, in this case the f/# of your lens system is determined by the diameter of the pinhole and not by the diameter of the taking lens' iris diaphragm.  So we would say that the pinhole is the limiting aperture in your lens system.

In your case the clear aperture of the back of the anamorphic section is 43mm in diameter (assuming it is round and not rectangular).  This means that the collimated on-axis beam of light exiting your anamorphic section cannot exceed 43mm in diameter, and may be less if any of the other optical surfaces in the Rectilux or your anamorphic group are limiting apertures.  The entrance pupil diameter of your 85mm f/1.2 taking lens (most likely 1/3 stop faster than f/1.4, or f/1.2599 in reality) is 85/1.2599 = 67.5mm.  Since 67.5mm is bigger than 43mm you are underfilling the entrance pupil of the taking lens.  As a consequence, you could stop down the taking lens until its entrance pupil is reduced to 43mm, and have no impact on the actual f/# of the system.  In your case, the actual maximum f/# would be f/(85/43) = f/1.98 ~ f/2, and not f/1.2.  Again, this assumes that the limiting aperture of the system is the rear aperture of the anamorph, and not some other surface in the Rectilux or the anamorph.  If either of the latter is true then your true f/# would be slower than f/2.

Is it possible that you are underfilling the entrance pupil of the 85/1.2, and hence not actually shooting at f/1.2?  For example, if you were shooting with an iscorama-36 the limiting aperture would be the 36mm diameter at the rear of the anamorphic section, thus giving you a maximum aperture of f/(85/36) ~ f/2.4

Hi Andrew:

I'm very interested in understanding the differences you are seeing between s35 (with a focal reducer) and 24x36mm.