Why can't there be IS/OS in all lenses?

[sagittariansrock]

With current high pixel density sensors, shutter speeds of 1/ 2xFL, 1/ 3xFL, or even faster are often required (before adjusting for the crop factor).

Previously I thought the 1/nFL rule is constant for a specified focal length. Why is the correction for crop factor necessary?

 

[Drizzt321]

With current high pixel density sensors, shutter speeds of 1/ 2xFL, 1/ 3xFL, or even faster are often required (before adjusting for the crop factor).

Previously I thought the 1/nFL rule is constant for a specified focal length. Why is the correction for crop factor necessary?

Sensors smaller than 35mm FF have a narrower Field of View which _simulates_, to an extent, a longer focal length. It's really the narrow field of view which will give bigger differences in your view for the same angular movement than it would on a wide angle view. So the narrower Field of View will have greater visible movement for the same shake than would a wide angle. Thus, you need a faster shutter for crop sensor cameras.

 

[Loswr]

Previously I thought the 1/nFL rule is constant for a specified focal length. Why is the correction for crop factor necessary?

No, you need to apply the 1.6x crop factor (or whatever is appropriate for your sensor size). Using the same focal langth lens with a smaller sensor yields a smaller angle of view, therefore the relative effect of a given amount of angular motion (aka camera shake) is larger due to the smaller sensor. Multiplying the guideline shutter speed by the crop factor compensates for that difference.

 

[mrsfotografie]

(also worth noting that Sigma's new 50/1.4 apparently does have a different design that the typical double-Gauss, although it lacks OS).

I think it has more of a retro-focus design like their 35 mm, hence the increase in length over the old model.

First, the 1/FL 'rule' is a legacy from film days. The resolution of current digital sensors exceeds the practical resolution of film (plus, you can view pixels at 100% on a large monitor, but you can't normally view film grains at 100%). The higher resolution means the effect (blur) of a given amount of angular motion is more easily visible (similar to how your hand shake is much more visible when pointing a laser pointer at the wall compared to a broad beam flashlight). That is part of the reason why 40D owners who upgraded to the 7D initially complained of soft images – their shutter speeds weren't high enough; the same was true for Nikon D700 users who moved to the D800. With current high pixel density sensors, shutter speeds of 1/ 2xFL, 1/ 3xFL, or even faster are often required (before adjusting for the crop factor).

Interesting, to me this sheds new light on why Canon put IS in the 24, 28, 35 primes.

 

[sagittariansrock]

Thanks Drizzzt and Neuro!

 

[Mt Spokane Photography]

Previously I thought the 1/nFL rule is constant for a specified focal length. Why is the correction for crop factor necessary?

No, you need to apply the 1.6x crop factor (or whatever is appropriate for your sensor size). Using the same focal langth lens with a smaller sensor yields a smaller angle of view, therefore the relative effect of a given amount of angular motion (aka camera shake) is larger due to the smaller sensor. Multiplying the guideline shutter speed by the crop factor compensates for that difference.

And, several years ago in a paper that Canon published for the 18mp 7D, they recommended that you use a 1 /2x preferably faster shutter speed due to the small pixels, just a tiny movement prevents you from getting pixel sharp images. As crop cameras continue to get smaller photosites, IS becomes more important. FF photosites are still pretty big, but when they get to 42MP, then they will match the 7D in size, and faster shutter speeds or IS become helpful for normal focal lengths.

 

[Ruined]

(1)If you are using an appropriate shutter speed (reciprocal of focal length on full frame, reciprocal of focal length*1.6 on crop) then IS basically does nothing. So, to give an example, if you are using a 50mm lens, a shutter speed of about 1/50 or faster will result in no camera shake and no benefit from image stabilization.
(2)Again, the slower your shutter speed the more moving objects blur, which IS cannot counteract. And, with my own testing I further noticed that once you start to get real slow shutter speeds handheld everything starts to look mushy even when it is within the range of IS to correct.
(3)When do you need a slower shutter speed? When you don't get enough light.

1. The "reciprocal rule" is a rule of thumb. If after the "best we can do", sometimes a rule of gnat's a## is preferable, if available.

Consider this challenge - mount a 50mm lens, in plenty of light, shoot at 1/50, and see if you can shake the camera as you press the button enough to create a less sharp than desired image.

if you agree that this can be done, then we are talking about the degree of shake applied at the camera. if you think that it will not exceed the arbitrary standard-shake which the "rule" will cover, I imagine you to be free of any of the tremor that, in varying degree, effects many an older photographer, and that you do not hurry up inclines or across distance to change position relative to moving subjects(wildlife), and therefore are not ever shooting winded or tired (or old) enough to be shaky.

Of course you can still make 1/50 blurry by excessive shake. But in those same instances, IS comes out a mess also in my experience, with ghosting abound. using the reciprocal rule and reasonable camera holding (i.e. not purposely trying to shake it around), I have found IS offers no benefit at said appropriate shutter speeds for photography. IS is a huge benefit for video in all cases IMO, though. I took about 100 shots with the 35mm f/2 IS USM and 70-200mm f/2.8 II IS USM. The former supposedly offers the latest and greatest version of IS. But I could not find any improvement at all even zoomed in on the pixel level when shooting w/ the reciprocal rule. There were times I would get a bit of shake on the pixel level without IS, but not all of the shots with IS were shake-free either.

The only time I saw a repeatable, noticable difference is when the shutter speed was significantly slower than the reciprocal rule.

2. If the subject is immobile, the focus is correct,and the photo is still mushy, then the camera shake is patently NOT "within the range of IS to correct". The amount of correction available is finite, and can be inadequate if more is required than available.

Again, through real world experimentation I have found that IS is less effective at very slow shutter speeds in delivering a sharp image with wide angle lenses. Meaning, although the image is sharper than it would be without IS, it is still blurry compared to a higher shutter speed. So, while IS can do a lot, it appears it can't do it all once the shutter speed gets slow enough, even if it is within the number of stops that the IS is supposed to work. Real world example, shooting with a 24mm f/2.8 IS - 1/30 looks reasonably sharp without IS, but drop down to 1/8+slower and even with IS it starts to look like mush on the pixel level handheld in low light. So, in this example, I'd rather use the 1/30 without IS than the 1/8 with IS. I did not experience this with IS and faster shutter speeds. IS seems to work really well at faster shutter speeds when the shutter speed is significantly slower than the reciprocal rule.

3. I have plenty of light. I WANT the moving-object blur (which the IS "cannot correct"), and I want surrounding subject area Pin-sharp (which IS can help provide). The subject is a waterfall. Other situations could be imagined.

The main issue with this scenario is at least Canon's IS does not render moving-object blur as smoothly as when IS is turned off. Even the latest generation IS, such as in the 35mm f/2 IS. While it detects motion such as panning shots where you want the subject sharp and background blurred, the background instead of being a smooth blur often looks a bit less natural, like several ghosted images placed over one another. It just does not appear as natural as when panning without IS. So while the non-moving subject may actually be sharper in a case like this with IS, the background tends to look less natural due to some mechanism of the IS processing. I can post an example of this, but if you've done panning shots with IS I'm sure you know what I mean. I guess in that case it depends if you'd rather have a sharper subject or a less processed looking background.

 

[Badger]

To OP,
I'm guessing the simple answer is, "It's complicated". Was just reading a review of the "cost is no object" Zeiss Otus 55mm f/1.4 for $4000. I'm guessing those guys know a thing or two about lenses and if they thought slapping a $50 or $500 IS module on this lens would make it better, they would have.
I don't pretend to know the technical reasons, but at some point, I have to trust that someone at Canon does.

 

[Ewinter]

Speaking of missing something, despite Sony being big proponents of sensor-shift IS, that feature is missing from their new FF mirrorless a7 and a7R.

I guess there has to be compromises in these bodies, as well. It's the same with most of the NEX I believe- maybe there just isn't the room/budget?

 

[Loswr]

Speaking of missing something, despite Sony being big proponents of sensor-shift IS, that feature is missing from their new FF mirrorless a7 and a7R.

I guess there has to be compromises in these bodies, as well. It's the same with most of the NEX I believe- maybe there just isn't the room/budget?

I suspect they determined that current implementations of sensor shift IS aren't sufficiently effective with the larger FF sensor. I'd chalk that up as a win for lens-based IS (note that optically stabilized lenses were announced alongside the a7/R), at least until they improve the in-body type sufficiently.

 

[sdsr]

I must be missing something - why not put the IS in the camera instead of putting it in each of the lenses?

Sony, Olympus, and several others do just that. It has the advantage of working with any lens, in theory. In practice, in-body stabilization is much less effective with longer lenses (the sensor can only be shifted so far and so fast), doesn't stabilize the optical VF, and doesn't help with AF since the AF sensor isn't stabilized. I suppose Canon and Nikon would also consider not being able to charge more for IS in each lens as a disadvantage.

Speaking of missing something, despite Sony being big proponents of sensor-shift IS, that feature is missing from their new FF mirrorless a7 and a7R.

One reason I'm keeping my Olympus OM-D and hesitant to try a Sony 7/7r is the excellent performance of the OM-D's IBIS, which stabilizes the EVF (if you're lucky enough to figure out the correct setting...) and, because it's mirrorless, doesn't need an AF sensor; it works superbly on the 100-300 Panasonic lens (better, probably, than the IS in that lens, which I keep turned off). This is not only good for "native" m43 lenses, but wonderful for legacy lenses - it's easy to manually focus an IBIS-stabilized 135mm legacy prime on my OM-D, quite an ordeal on my Fuji x-e1, despite the reduced crop factor (there's no point even trying on my FF Canons). Whether IBIS can cope with really big, heavy lenses, I don't know (though while I owned a Pentax K5 it seemed to do a good job with the unstabilized version of the Tamron 70-200 2.8). Apparently it's not in the A7s because the IBIS needed for a FF sensor would require a bigger camera body than Sony wanted for this line of camera.

As for the superiority of the Canon 24-70 2.8 II to the stabilized Tamron 24-70, it rather depends on how you use it, doesn't it? I'm not disputing that in many (most?) situations the Canon is a better lens, at least marginally, on a Canon body. But if you shoot in very low light you may, thanks to the Tamron's stabilization, get better results with the Tamron; if sharpness matters that much to someone, it's worth noting that the Tamron on a Nikon D800e is sharper than the Canon on is any currently available Canon body (according to Roger Cicala, at any rate); and it's not obvious that, say, the images Dustin Abbott has posted here taken on his 6D with the Tamron would look any better if they had been taken with the Canon lens.

 

[dgatwood]

So where is the cutoff really that you need IS? On crop, almost every lens could benefit from it unless they are very wide or very fast. On full frame, I would say you'd probably start to really benefit from IS at 100mm f/2.8 and longer lenses. So, this is why you see IS really being employed in modern full frame fast lens designs at 100mm and over, as 1/100 is probably the lowest you'd want to get a sharp image by avoiding motion blur (in addition to camera shake) - plus it lets a good amount of light in at f/2.8 or wider.

Whatever cutoff you pick, remember that it is arbitrary, and depends on the sensor's pixel size. For pixel peeping purposes, if the sensor's PPI doubles, with everything else being equal, the focal length at which the lack of IS results in motion by an objectionable distance (in pixels) is roughly halved.

But it gets worse. The main reason for IS is low-light photography, and if you have 4x as many pixels per unit area, the amount of light that hits each one drops by a factor of four. Thus, in low light, the whole "everything else being equal" thing falls apart, and the focal length at which you need IS actually drops by up to a factor of 8, assuming I'm doing the math correctly.

In other words, the focal length at which you need IS is approximately proportional to one over the cube of the sensor's linear resolution. Does that sound plausible?

 

[Ruined]

As for the superiority of the Canon 24-70 2.8 II to the stabilized Tamron 24-70, it rather depends on how you use it, doesn't it? I'm not disputing that in many (most?) situations the Canon is a better lens, at least marginally, on a Canon body. But if you shoot in very low light you may, thanks to the Tamron's stabilization, get better results with the Tamron

Personally, I think if you are shooting in "very low light" you need either A) a lens faster than f/2.8, or B) a flash. Otherwise you'd *really* have to crank the ISOs or use a shutter speed that would result in motion blur. My primary thing is events and I would either go the flash route, or use my 50mm f/1.2 - in either case, IS would be a moot point for that type of scenario IMO...

Where I see IS perhaps being most useful is telephoto lenses where the shutter speed needs to be fast to begin with, and still life photos.

 

[sagittariansrock]

As for the superiority of the Canon 24-70 2.8 II to the stabilized Tamron 24-70, it rather depends on how you use it, doesn't it? I'm not disputing that in many (most?) situations the Canon is a better lens, at least marginally, on a Canon body. But if you shoot in very low light you may, thanks to the Tamron's stabilization, get better results with the Tamron

Personally, I think if you are shooting in "very low light" you need either A) a lens faster than f/2.8, or B) a flash. Otherwise you'd *really* have to crank the ISOs or use a shutter speed that would result in motion blur. My primary thing is events and I would either go the flash route, or use my 50mm f/1.2 - in either case, IS would be a moot point for that type of scenario IMO...

Where I see IS perhaps being most useful is telephoto lenses where the shutter speed needs to be fast to begin with, and still life photos.

Indeed, it is amazing how useless the IS of my 17-55 was unless I was shooting still subjects. I would happily go about with 1/20 sec shots only to find subject motion on my large screen. I think the higher resolution of modern cameras (the pixel density of the 7D) has contributed to this.

 

[Ewinter]

I must be missing something - why not put the IS in the camera instead of putting it in each of the lenses?

Sony, Olympus, and several others do just that. It has the advantage of working with any lens, in theory. In practice, in-body stabilization is much less effective with longer lenses (the sensor can only be shifted so far and so fast), doesn't stabilize the optical VF, and doesn't help with AF since the AF sensor isn't stabilized. I suppose Canon and Nikon would also consider not being able to charge more for IS in each lens as a disadvantage.

Speaking of missing something, despite Sony being big proponents of sensor-shift IS, that feature is missing from their new FF mirrorless a7 and a7R.

One reason I'm keeping my Olympus OM-D and hesitant to try a Sony 7/7r is the excellent performance of the OM-D's IBIS, which stabilizes the EVF (if you're lucky enough to figure out the correct setting...) and, because it's mirrorless, doesn't need an AF sensor; it works superbly on the 100-300 Panasonic lens (better, probably, than the IS in that lens, which I keep turned off). This is not only good for "native" m43 lenses, but wonderful for legacy lenses - it's easy to manually focus an IBIS-stabilized 135mm legacy prime on my OM-D, quite an ordeal on my Fuji x-e1, despite the reduced crop factor (there's no point even trying on my FF Canons).

I seem to be able to get focus on my legacy 135mm just fine on the a7, even with the viewfinder at 5x magnification. Getting the shots sharp is another matter shutter wise is a problem, but often I find that people move more than I get hand shake, so that higher ss is not a bad thing. Also, the sensor can be pushed much further than where i'd push any m4/3 to ISO wise.
It's all compromises, I guess