Can lenses get much better??

George Zip wrote in post #18209160
I was thinking about this.

Have we reached a point where future improvements will be so marginal to IQ it will not really matter?? Possibly the improvement might be more in weight, better IS and that sort of thing.

I am referring to Canons lenses like the 35mm, 24-70 2.8 II, 70-200 2.8 II, 16-35 2.8 III, 100-400II.

It just seems hard to make significant improvements to the above.

For all the trolling that goes on about how non innovative canon are with their cameras you can not deny they generally make the world class lenses.

The one lens I would really like to see given an overhaul is the 28-300 L, it would be great if they could make this a bit lighter and improbable some of the inevitable compromises on a zoom like this.

I thought that when I bought my Nikkor 300mm f/2.8 IF ED way back in the 1980's..

Lenses can be made better but at a significant cost. We are seeing lenses getting bigger and heavier rather than smaller at present (Sigma 50 Art anyone?), in preparation for ever denser sensors. Trouble is a 200MP sensor is going to need a very steady tripod excellent technique and ideal conditions to maximise sharpness. While IS may improve a little,it's really pretty much stalled at 4-5 stops because there is only so much amplitude to human physiological tremor and no IS can compensate for movement which meaningfully changes the framing of a shot. I expect the costs of R&D may soon exceed the extra profits from larger, even sharper lenses whose practical advantages are progressively smaller and mainly of appeal to amateur pixel peepers (of whom I'm one). Advances in sensor design and fps will probably be more meaningful for most people. Really clean shots in, indoor dimly lit venues may become possible. As an example the Canon 35L 1.4II is an improvement in every way over the 35L but the latter is good enough for many and is a very old lens. The former is larger and much more expensive-I don't see many amateurs forking out for the latest and greatest at these prices and much of the income generated by CaNikon in the last 10-12 years has been from enthusiastic amateurs and incomes of pros in general has been falling. I think the pace of progress is going to slow with respect to lens design and releases.

snapperz wrote in post #18274246
Trouble is a 200MP sensor is going to need a very steady tripod excellent technique and ideal conditions to maximise sharpness.

Not if the sensor is twice as wide and twice as high
Even on the same sensor format, moving from 50MP to 200MP only requires twice the shutter speed to achieve the same sharpness.
Don't worry, it all be nice and fun.

On the topic of 'sharpness' if we equate that to detail resolution (which is not the real definition of 'sharpness')...
If we assume that

• today's 120 line-pairs/millimeter (e.g. photozone.de tests for Sigma 20mm and 35mm ART on 50 Mpixel camera show 113 l-p/mm from those lenses) are blasé performance,
  
• and we need 185 line-pairs/millimeter on a 24mm tall FF sensor, so we need a 118 Mpixel sensor to capture what that lens could deliver.
  

Wilt, you are making a lot of arguable assumptions, and seem to forget that the best color sensors have a Bayer array slashing per-color resolution.

About shutter speed, we were obviously talking about per-pixel blur, not image-level blur.

CheshireCat wrote in post #18274463
Wilt, you are making a lot of arguable assumptions, and seem to forget that the best color sensors have a Bayer array slashing per-color resolution.

About shutter speed, we were obviously talking about per-pixel blur, not image-level blur.

But given that the single-color sensel is interpreted via its neighboring sensels into the R-G-B factored pixels, that per-color resolution is somewhat better than the single-color pixel spacing. But I will concede that the article upon which I based my prior post, while considering pixel density on a single color wavelength, did take a simplified approach in the oversimplification of ignoring Bayer. OTOH, factoring in Bayer is probably over the heads of 99.9% of those reading POTN!

George Zip wrote in post #18209160
The one lens I would really like to see given an overhaul is the 28-300 L, it would be great if they could make this a bit lighter and improbable some of the inevitable compromises on a zoom like this.

To respond to just this part, I wish we could get a 24-200 IS STM (or maybe USM). Don't worry about the red ring or blue goo or the longest end of the 28-300L; just get me the FF equivalent of the 18-135 IS STM without the L premium (and with the wide end a little wider). If I want weathersealing and higher quality I'll pick the right lens for that; this would be for packing a single, useful lens with a FF body to handle most situations, or to leave on the body so that if I see something I can just pick it up and shoot instead of needing to also put on the right lens, by which time the target may have moved or the moment may be gone.

Wilt wrote in post #18274548
But given that the single-color sensel is interpreted via its neighboring sensels into the R-G-B factored pixels, that per-color resolution is somewhat better than the single-color pixel spacing. But I will concede that the article upon which I based my prior post, while considering pixel density on a single color wavelength, did take a simplified approach in the oversimplification of ignoring Bayer. OTOH, factoring in Bayer is probably over the heads of 99.9% of those reading POTN!

Wilt actually the colour resolution will be quite a bit less than the sensel spacing limit, not higher. The Bayer array effectively gives you RGGB colour quads at half resolution. The thing being that you can effectively get two RGGB sensels and a GRBG sensel between them, by alternating the way you combine the colours in the filter. So at a first approximation you will end up with your colour Nyquist resolution being = 2/3× Unfiltered Nyquist instead of 1/2× Unfiltered Nyquist. You might be able to bring this up a bit more by some really clever interpolation algorithms, but I doubt you will get it over about 3/4× Unfiltered Nyquist.

Alan

BigAl007 wrote in post #18276983
Wilt actually the colour resolution will be quite a bit less than the sensel spacing limit, not higher. The Bayer array effectively gives you RGGB colour quads at half resolution. The thing being that you can effectively get two RGGB sensels and a GRBG sensel between them, by alternating the way you combine the colours in the filter. So at a first approximation you will end up with your colour Nyquist resolution being = 2/3× Unfiltered Nyquist instead of 1/2× Unfiltered Nyquist. You might be able to bring this up a bit more by some really clever interpolation algorithms, but I doubt you will get it over about 3/4× Unfiltered Nyquist.

Alan

Thanx Al. <my brain just went numb>

Wilt wrote in post #18276985
Thanx Al. <my brain just went numb>

What is it they say Brain Fart = Essential bodily function. I have them too at times.

What really gets me is that audio types will want to push the sampling rate up to Nyquist×2, while photographers seem to want to push things so that the lens is delivering more than Nyquist? Are so many photographers really that dumb? My personal view is that given lenses that will deliver useable 130 LP/mm to the sensor, I would want a 35mm sensor with around 200 MP, if I remember correctly from the last time I ran the numbers.

Alan

See I had one too, before I corrected it! It's Nyquist times to oversample.

Interestingly, Alan, just recently I posted about lab tests measuring close to 120 line-pars/millimeter with some recent lenses measured on 50 Mpixel cameras. Then I referenced an article on Luminous Landscape about the f/4 diffraction limits translating to cameras with sensors at 118MPixels, and that would be pretty well aligned lenses to delivering about 180 line-pair/mm as well.

Please post links.

CheshireCat wrote in post #18277222
Please post links.

You gotta pay for access to articles on Luminous Landscape now, unlike in the past. The change came when Michael Reichmann passed away in May 2016

https://luminous-landscape.com …nsors-out-resolve-lenses/

I could not locate the prior post that I referenced, so I will recreate the highlights of its content:

• recent lens tests of some lenses on 50MPixel bodies by photozone.de indicate resolution of 225 line-pairs per millimeter at f/4 (Sigma 24mm f/1.4 ART) has MTF value of 5416 line-pairs/picture height
  
• On Luminous Landscape, Reichmann article chart shows that at f/4 (using the Green light diffraction limits) a lens-diffraction limited sensor would have >118 MPixels
  
• to fully resolve 225 line-pairs per millimeter per Nyquist theory takes 451 pixels/mm, or 175.8 Megapixels!
  
• and to match 118MPixel sensor for resolution (ignoring diffraction effects) would take a lens capable of 185 line-pairs/mm
  

Wilt wrote in post #18277370
I could not locate the prior post that I referenced, so I will recreate the highlights of its content:

• recent lens tests of some lenses on 50MPixel bodies by photozone.de indicate resolution of 225 line-pairs per millimeter at f/4 (Sigma 24mm f/1.4 ART) has MTF value of 5416 line-pairs/picture height
  
• On Luminous Landscape, Reichmann article chart shows that at f/4 (using the Green light diffraction limits) a lens-diffraction limited sensor would have 118 MPixels
  
• to fully resolve 225 line-pairs per millimeter per Nyquist theory takes 451 pixels/mm, or 175.8 Megapixels!
  
• but to match 118MPixel sensor for resolution (ignoring diffraction effects) would take a lens capable of 183 line-pairs/mm
  

Makes the head spin a bit

Nevermind the storage space point (never been cheaper) we'd also need much faster circuitry in the body to keep the framerate up. What would the throughput be for 10fps RAW at 175mpx?