[pathdoc]

Why is it that all the ultra-fast lenses (f/1.2 or wider) seem to hover around the 50-55mm mark? (Let's talk in film terms to keep things easy and forget APS-C for the moment; AFAIK there aren't any APS-C lenses of that speed for Pentax in any case.)

When I think about it, my logic goes like this. The aperture of a 50mm f/1.2 lens is 50/1.2 or 42mm, plus then we have to have additional diameter around that for the aperture blades to retract into and all the associated mechanisms, aperture ring, etc. etc. Surely if we have a 42mm lens, we could either make the f/1.2 lens smaller or - going to extremes - fit an f/1.0 lens inside the same basic diameter? Going further, would it not be possible to make even shorter focal lengths even more extreme in aperture terms? What are the limitations that prevent this from happening? Why haven't we, for example, got 24mm f/0.5 lenses all over the place?

[PJ1]

Why haven't we, for example, got 24mm f/0.5 lenses all over the place?

As I understand it (and putting it very simply) the larger the aperture, the greater the distortion towards the edge of the image because of the angle at which the light strikes the sensor. With wide angle lenses, this effect will be greatly increased. Aspherical correcting lenses can go some way to fixing this, but obviously lenses are going to get bigger and more complicated the more correction is required. Most ultra-fast lenses are not very sharp wide open. There is a point at which you could go to a lot of expense and still not get acceptable quality. Curved sensor technology might bring some changes in that area although I am not sure what the advantages would be unless one wanted (needed?) an ultra-thin depth of field. I think you would need a sensor which adjusts its curve to the focal length - probably not an impossibility but I am not sure I will ever need it.

[Digitalis]

Surely if we have a 42mm lens, we could either make the f/1.2 lens smaller or - going to extremes - fit an f/1.0 lens inside the same basic diameter?

The entrance pupil and aperture at the optical center of a 50mm f/1.0 lens has to be at least 50mm - there is no way of getting around that.

Going further, would it not be possible to make even shorter focal lengths even more extreme in aperture terms?

It is possible, however the imaging performance of lenses like this will be absolutely atrocious.

What are the limitations that prevent this from happening?

Lens mount diameter,Optical design constraints, Flange distances, Glass chemistry, Financial constraints,R&D, complaining geriatric consumers with aching back muscles, and the standard model of particle Physics

Why haven't we, for example, got 24mm f/0.5 lenses all over the place?

Because no one really needs them.

[Na Horuk]

Yeah, there are all sorts of other considerations, mostly having to do with image quality and actual construction. Making large pieces of glass is difficult, and adding a bunch of other pieces of glass to correct the image increases costs. And then you need to make sure the whole thing fits inside the mount diameter and has the correct register distance, that it allows focusing and doesn't get in the way of the DSLR mirror, and so on.
55mm and 50mm were traditionally the most common, simplest lenses available for this kind of sensor size. You can look up posts about the "nifty fifty" and why that focal length is often the cheapest with best IQ and fastest aperture. Just look at a regular 50mm f1.8 lens design and compare it to a 14mm f2.8 design. The 14mm will be much bigger, more complex.

That said, Pentax hasn't really used its APSC advantage to the fullest. Many Pentax DA lenses are actually full frame designs, not newly made designs, fast as possible for APSC. Fuji is making some wonderful, fast lenses that are optimized for its sensor size. But they have a hefty price tag, as well.
Photography is all about compromises - what can be manufactured, at what cost, and is there a market for it? If Pentax makes a $10,000 super fast lens for APSC.. how many will they sell? Even the 560mm is not exactly flying off the shelves, and its a good lens with a relatively reasonable price. Its already too niche. Imagine if they made it f3.5 and tripled its price (and weight)

[pathdoc]

Imagine if they made it f3.5 and tripled its price (and weight)

Yep. I own a Sigma 70-210mm f/4.0-5.6 in K-A mount and a Bushnell 70-220mm f/3.5 (constant maximum) in M42 - the latter's size and heft certainly aren't helped by the macro feature (which takes the form of a built-in, wind-out extension tube thingy) and it being from the all-brass-and-glass era, but the contrast is truly striking. It's the only lens I've got that has a tripod mount.

Just look at a regular 50mm f1.8 lens design and compare it to a 14mm f2.8 design.

I'll have to take your word for it as I've never had the opportunity to handle (for example) the Pentax DA 14mm f/2.8, but I know just how small they can make a 50mm f/1.8 or even an f/1.4 as I have both.

Lens mount diameter,Optical design constraints, Flange distances, Glass chemistry, Financial constraints,R&D, complaining geriatric consumers with aching back muscles, and the standard model of particle Physics

I'm not quite sure how the last one works in, but the rest make sense.

The entrance pupil and aperture at the optical center of a 50mm f/1.0 lens has to be at least 50mm - there is no way of getting around that.

True. But this would logically mean that the entrance pupil and aperture at the optical centre of a 28mm f/1.0 lens only has to be 28mm, yes? And shouldn't that open the door for a generally more compact arrangement than a fifty of the same f number? By all means tell me why I'm wrong - that's why I'm asking this question.

[Aristophanes]

Why haven't we, for example, got 24mm f/0.5 lenses all over the place?

The focal plane on an image from a lens like that would be millimetres thin so as to make nothing look in focus, much less sharp.

Also, a lens like that would have a front element something like 10cm across and it would still have massive distortion unless the design put corrective elements in, reducing the t-stop substantially, so by the time light hit the sensor it would be measured in stops less regardless.

[romay]

Why is it that all the ultra-fast lenses (f/1.2 or wider) seem to hover around the 50-55mm mark? (Let's talk in film terms to keep things easy and forget APS-C for the moment; AFAIK there aren't any APS-C lenses of that speed for Pentax in any case.)

When I think about it, my logic goes like this. The aperture of a 50mm f/1.2 lens is 50/1.2 or 42mm, plus then we have to have additional diameter around that for the aperture blades to retract into and all the associated mechanisms, aperture ring, etc. etc. Surely if we have a 42mm lens, we could either make the f/1.2 lens smaller or - going to extremes - fit an f/1.0 lens inside the same basic diameter? Going further, would it not be possible to make even shorter focal lengths even more extreme in aperture terms? What are the limitations that prevent this from happening? Why haven't we, for example, got 24mm f/0.5 lenses all over the place?

There are multiple reasons.
The most important one is flange distance. For most DSLR cameras, the flange distance is in the order of 50mm. That's no coincidence.
That means you don't need any complicated retrofocus design for lenses around that mark, which makes it much easier to construct fast "normal" lenses for a resaonable price.
That's also one reason why APS-C sucks, - the lens sweet spot is no longer a "normal" focal length.

Fast longer focal lengths are not so complicated to build since shortening the optical path is quite straightforward. However, the longer the lens the bigger the entrance pupil needs to be for a certain aperture. You need more glass which makes everything more expensive. More expensive means smaller market, so you're naturally seeing less of them.
But that's not everything. Larger diameter glass also means thicker lenses. If you look at some super fast designs you'll notice some very thick elements. At some point when their elements become too big, optical designs just break. You could counter that by using materials with higher refractive index, but there are only so many kinds of transparent materials. So unless we use fresnel lenses, cardioptic designs or infinitely dense glass, there are some physical upper limits on lens speed for telephoto lenses with refractive designs.

Side note: If you use cardioptic designs, these limitations do not appear. You can get affordable mirror lenses way faster than f/0.5 which you call "telescopes"

For wide angle lenses, it's more complicated.
Again, the market dictates what is developed. Few people really need fast wide angle lenses as they are mostly used for landscape which you need high DOF for. So, producing a fast wide can backfire if people aren't willing to pay the premium.
The physical limitation is the need for a retrofocus group to get the nominal focal length extended to the flange distance. Technically, that group is just an inverted telephoto lens. However, if you have a sensor larger than the entrance pupil, this becomes super hard to design in a fashion that still yields acceptable image quality.The reason is simply that you need very high refracion to both "extend" a focal point to like 5 times it's focal distance and still be able to reach the sensor corners from an entrance pupil much smaller than the sensor.
However, if the sensor is small in relation to the entrance pupil or you have a shorter flange distance, you can also have fairly fast wide angle lenses. For example, look at the leica lenses. Their register distance is around 20mm (if I'm not mistaken) since they are used on rangefinder cameras and you can have designs that are very fast (f/1.4 ish) around that focal length. I also have c-mount lenses at home that are 17mm/0.95 and 25mm/0.95 - Register distance is 17mm and image circle is like 1", so the designs are super simple (look like a gauss doublet to me). For that flange distance and image circle, you can even get affordable lenses as fast as f/0.7. However due to the crop factor, they are technically speaking no wide angles.

[pathdoc]

Aha, all is beginning to be clear now. Thank you.

[Na Horuk]

One interesting thing to keep in mind is that most lenses can be traced to just a couple optical design lineages, each with its strengths and weaknesses. Some lenses we have are already pushing the limits

[ScooterMaxi Jim]

Also, keep in mind that all digital sensors begin to lose light acceptance properties around f/2.8, and beyond around f/1.7 the fall off becomes significant (to varying degrees depending on the particular sensor). Anything beyond around f/1.4, the loss is severe - so you aren't gaining much at all. Because this has a lot to do with angle of acceptance, the wider angle lenses are going to result in even more light acceptance loss than the 50mm lenses.

[WPRESTO]

An odd thing about fast lenses: If you compare IQ ratings for the same manufacturer, same or nearly the same FL, the faster lens will almost always score better than the slower: 50~55 f1.4 lenses are almost always rated higher the f1.8~2.0; 300mm f2.8 lenses are almost always rated higher than 300mm f4.0, etc. etc. and this seems to apply at equivalent apertures (both lens stopped to say f5.6). Some manufactures will offer three lenses differing by one stop steps, for example a 28mm f1.4, an f2.0, and an "bargain priced" f2.8. Independent lab testing will almost always rate the f1.4 best and the f2.8 poorest in such a set from one manufacturer. There are, of course, exceptions. For example, a 50mm f2.8 macro will usually outscore a 50mm f1.4 on IQ.

[Madaboutpix]

An odd thing about fast lenses: If you compare IQ ratings for the same manufacturer, same or nearly the same FL, the faster lens will almost always score better than the slower: 50~55 f1.4 lenses are almost always rated higher the f1.8~2.0; 300mm f2.8 lenses are almost always rated higher than 300mm f4.0, etc. etc. and this seems to apply at equivalent apertures (both lens stopped to say f5.6). Some manufactures will offer three lenses differing by one stop steps, for example a 28mm f1.4, an f2.0, and an "bargain priced" f2.8. Independent lab testing will almost always rate the f1.4 best and the f2.8 poorest in such a set from one manufacturer. There are, of course, exceptions. For example, a 50mm f2.8 macro will usually outscore a 50mm f1.4 on IQ.

There may be some truth in that looking at conventional laboratory tests. In real-life shooting, however, slower lenses with their typically lower element count will often deliver in areas that go largely unnoticed or are given less weight in laboratory test verdicts than they deserve: low and less complex distortion, superior flare resistance, pleasant microcontrast, better colour differentiation/saturation. This may help to explain why the relatively slow DA Limited primes perform so nicely, or why an inexpensive consumer lens like the mere 12-element DA55-300 holds its own even on a 24MP sensor.

[Adam]

inexpensive consumer lens like the mere 12-element DA55-300 holds its own even on a 24MP sensor.

True, but that doesn't mean that more complex designs like the DA* 60-250mm aren't optically superior I just think the days of bargain lenses equating to bargain image quality are over.