Monday 1 December 2014

RX vs non-RX lenses






Ever since Bolex decided to utilize a beam-splitting prism in their cameras to make them reflex, there has been confusion and misinformation about the effects of the prism and the nature of the lenses that were designed to work with it. The first H16 Reflex Bolexes hit the market in 1956, and were supplied with a new range of Kern lenses that were labelled "RX" for focal lengths of 50mm and under. These lenses were also available in versions for non-reflex cameras, usually simply labelled "AR" (which actually stood for "anti-reflective" coating, but distinguishes them from the RX labelled ones). To confuse matters a little, the very first lenses Kern made for the reflex Bolex were labelled "DV" (which stood for "direct vision" I believe) as well as having AR engraved on them, but these are basically RX lenses.

Other companies also made RX lenses for reflex Bolexes. Both Schneider and Som Berthiot released primes that were labelled "RX" or "H16 RX", and some early Angenieux zooms were also designated "RX", or sometimes "Special P" (relating to Paillard).

Differences

Back Focus

The first thing to clarify is that the back-focus distance setting of RX and non-RX lenses is the same. In other words, both are designed to form an image 17.52mm behind the mount flange, which is the C-mount standard. You cannot simply "re-collimate" or "adjust the back-focus" of an RX lens to make it non-RX or vice versa.

Some confusion exists because the prism in a reflex Bolex acts to extend the light rays further back, so the "physical" distance between the mount and the film plane in a reflex H16 is actually 20.76mm, but the simple fact is that any C-mount lens fitted to a reflex Bolex will have its back focus distance extended back to this point.

Aberrations

The problem is that the prism causes the outer light rays from a lens to be bent at different angles to the central ones, so they don't converge at the same point, creating blurry images at the film plane. These aberrations are what the RX lenses have been designed to counteract. In the absence of a prism in the light path, an RX lens will exhibit the same sort of aberrations it was designed to negate, causing blurry details and flare around bright highlights and softness in the corners.

The aberrations are more pronounced at wider apertures, because this is when the light rays pass through the edges of the lens elements as well as the centre, exhibiting the most variation in how they are bent. When a lens is stopped down, the aberrations are reduced. 

Official Bolex literature claimed that the focal length of a lens would also affect how pronounced the aberrations were. 50mm was the longest focal length prime that any manufacturer designed as RX, because it was decided that beyond 50mm the effects of the prism were not noticeable enough to worry about. It's actually a little more complicated than that. A crucial causal factor is the position of the exit pupil (the apparent position of the iris as seen from the rear), which along with the rear element size determines the angle of the cone of light departing the rear of the lens. A deeper exit pupil creates a more parallel cone which will produce less aberrations when a refractive material like glass is placed in its path. While a longer focal length will often have a deeper exit pupil than a wide angle lens, this is not something that is always directly related to focal length. As we shall see, the 10mm RX Switar for example exhibits less aberration without a prism in the light path than the 25mm RX Switar does, because the 10mm has a much deeper exit pupil than the 25mm. The longer Kern primes all happen to have deep exit pupils and slower apertures, which is the reason they didn't need RX versions.

Aperture

Another area of confusion is the f stop markings on RX lenses - some people believe that the aperture rings have been marked to compensate for the light loss caused by the reflex prism diverting about a quarter of the light to the viewfinder. This is incorrect. The f stop marks are the same on an RX lens and a non-RX lens, and relate to the lens only. The light lost to the viewfinder needs to be factored in to the stop calculation for any lens fitted to a reflex Bolex. (This should be obvious when we remember that longer focal lengths were not marked as RX, so having exposure compensated marks on some lenses but not others would be a very ridiculous and confusing state of affairs.) 

The question of whether one lens might seem brighter than another at the same f stop relates to the coatings and the internal light loss caused by reflections and scattering. Many cinematography lenses are marked in T (for "transmission") stops to account for this. Typically a prime lens might lose a third of a stop, while a zoom might lose a half (or more for older models). An older version of the same lens may have earlier, inferior coatings - leading to more light being reflected or scattered at each glass to air surface, and a dimmer, less contrasty image at the same f stop. Damaged coatings will do the same.

Measuring the difference

Many people are aware of the difference between RX and non-RX lenses, but few can say exactly how much the image quality may suffer when the wrong type is used on a particular camera. This blog post is an attempt to show in simple terms the different optical characteristics of some lenses that are available as RX and non-RX variants, when used without a 9.5mm thick glass prism in the light path. 


Comparisons


The following images are photographs of a lens test projection. In simple terms, a test slide is placed at the film plane and light is shone through it and then through the lens and onto a screen. We are effectively reversing the normal light path from subject, through the lens and onto the film plane. It allows us to see exactly how a lens recreates the fine pattern details of the test slide, and thus how well it would resolve or distort a subject placed at the screen distance.

The photos were taken with a Nikon D200 camera, and much of the actual resolution of the test image is finer than the Nikon camera/lens can capture, but even so it is easy to see the difference between an RX and a non-RX 25mm Switar projection.

Only a quarter of the whole test pattern projection is shown, since the image deterioration is quite symmetrical (unless a lens is faulty). For the 25mm projections that follow, the crop is the bottom left quarter. The centre of the lens is always the sharpest, seen in these photos at top right.


Kern Switar 25mm AR @ f/1.4




Ken Switar 25mm RX @ f/1.4


At f/1.4 (wide open) the RX lens exhibits considerably more halation and flare than the AR lens.



Kern Switar 25mm AR @ f/2.8




Kern Switar 25mm RX @ f/2.8


Notice that while stopping down improves the corner sharpness of the non-RX lens considerably, the RX lens is still quite soft in the corners. Even at f/5.6 it is still not as sharp in the corners as the AR lens. While the RX lens halation at f/1.4 is mainly caused by spherical aberration which disappears fairly quickly as you stop down, the corner softness seems to be mainly the result of astigmatism, which is not affected as rapidly by aperture changes.  

The following photos are of 10mm Switar projections, cropped to the top left quarter. The test pattern centre is therefore at bottom right. As mentioned previously, there is less difference between these RX and non-RX images because the exit pupil is quite deep inside the lens, so very little astigmatism is introduced by the prism that needs correcting. The maximum aperture is also slightly (a third of a stop) slower than the 25mm.



Switar 10mm AR f/1.6




Switar 10mm RX f/1.6

Notice that the centre of the AR lens is sharper, but the corners are slightly worse. Wide angle lenses of this era often have issues with corner sharpness, due to various aberrations including field curvature, coma and chromatic aberration, I suspect the prism corrections in the RX lens have reduced some of the aberrations causing corner softness in the non-RX lens. The centre of the RX lens shows considerably more halation however. 



Switar 10mm AR f/2.8




Switar 10mm RX f/2.8



At f/2.8 the AR lens has improved markedly in the centre, while the RX lens still has spherical aberration halation. Stopping down has not improved the corner resolution of the AR lens all that much, in that part of the image the RX lens is still superior. It's interesting to note that the spherical aberration at f/2.8 is better in the 25mm RX than in the 10mm RX at the same stop, but the astigmatic corner sharpness of the 25mm RX is much worse.

I don't have a 50mm AR Switar on hand at the moment to show comparisons with the RX version, but I have projected and analysed the 50mm RX. It has an exit pupil depth halfway between the shallow 25mm and the deep 10mm.  Considering the focal length and the fact that it's an f/1.8 lens, one might expect the 50mm RX to exhibit minimal aberration, but in fact it shows about as much halation and actually more corner degradation than the 10mm RX, which reinforces the idea that the exit pupil position is a more critical factor than the focal length.

Conclusions


It can be difficult quantifying the effects of aberrations that vary with both aperture and lens design, but from these projections we can at least get a sense of the degrees of difference between RX and non-RX lenses. They show that RX lenses exhibit halation caused mainly by spherical aberration, and astigmatic corner softness in lenses with shallow exit pupils. The aberrations are most pronounced at full aperture, still visible at f/2.8 and the halation reduces more quickly than the astigmatism.

For reflex Bolex users wanting a guideline on non-RX lens choice, I would say any lens with an exit pupil distance of more than about 50mm from the film plane should be fine stopped down to f/2.8 and beyond, with the effect diminishing as the pupil distance increases. A lens that has the iris apparently positioned directly beneath the rear element (so an exit pupil distance of maybe 20mm from the film plane) will need more stopping down, and may still exhibit soft corners at f/5.6. Naturally this is all dependent on the final viewing parameters, these guidelines assume a pretty critical Circle of Confusion. And of course many old C-mounts have inherent aberrations anyway, so it can all be rather moot.

For digital camera shooters wanting to use C-mount lenses the issue gets even more complicated because of the fact that almost all digital cameras do in fact have an extra block of glass in the light path, situated between the rear of the lens and the sensor. It's the sensor stack, consisting of cover glass, optical low pass filter and IR filter, which can vary from 0.5mm to over 4mm in thickness. Micro 4/3 cameras typically have the thickest sensor stacks at around 4mm. While this is less than half the thickness of the 9.5mm Bolex prism, it does mean that RX lenses are probably not much worse than non-RX lenses on Micro 4/3 cameras, since one type is over-correcting for the sensor stack and the other is not correcting at all. For a good discussion of the effects of sensor stack thickness on image quality (and the importance of exit pupil distances), see Roger Cicala's LensRentals articles at:
 http://www.lensrentals.com/blog/2014/06/sensor-stack-thickness-when-does-it-matter

One last note about RX lenses. For their 8mm reflex models, Bolex commisioned Kern to make 3 primes -5.5mm, 12.5mm and 36mm - and 2 zooms, all designated "H8RX". For reasons best known to Bolex, these lenses have the standard 1" C-mount thread and so are often advertised as C-mounts, but the back-focus is actually 2mm shorter than the C-mount standard. This means that if they are used on C-mount cameras (other than a H8RX) or with standard C-mount adapters they won't focus any further than a few feet away, in fact the 5.5mm won't focus past an inch.  Plus of course they only really cover a tiny 8mm frame, perhaps the 36mm covers more but with severe fall-off. Considering that I have seen H8RX Macro-Switars go for extraordinary sums on ebay, with their short back-focus never mentioned, I think this should be more commonly known. Though I guess it's not a problem if the buyer only wants these lenses to photograph flowers and insects anyway.

Thanks to Dennis Couzin for first questioning the official Bolex literature nearly 40 years ago, and later suggesting the importance of rear exit pupil depth. See:
https://sites.google.com/site/cinetechinfo/