F0.95 ???

[Great Bustard]

It is very easy to test

Just focus the lens to infinity.

Then take an image of a nearby point source, i.e. a small hole in a black cardboard, or even better a pin hole if you have one.

Take images with different apertures and see the size of the circle go up when opening the aperture.

At some aperture, it might stop being larger. Then - that is the limit.

You probably even can see this on the LCD directly.

Here ya go:

http://www.dpreview.com/forums/post/36768003

 

[Roland Karlsson]

http://www.dpreview.com/forums/post/36768003

No.

The lens needs to be focussed at infinity to do this measurement.

Otherwise it is a perfect experiment.

 

[Roland Karlsson]

Making some drawings of the ray traces ....

Yeah .... that the micro lenses limits the aperture do not decrease the blurriness. I was wrong.

But ... the test should really be done with the lens focussed at infinity. That test is much harder to do.

But ... it will show the same thing

 

[Roland Karlsson]

I am now a wiser man
I made an error in my thinking. Now corrected.

The limited aperture acceptance by the micro lenses do not affect the DOF all that much.

The out of focus parts of the image will hit the sensor from many angles, some of them within the acceptance angle of the micro lenses.

So --- a F0.95 will have a shallower DOF than a F1.4, even if the micro lenses only accepts F1.4.

Therefore the limited acceptance angle of the micro lenses will only affect the brightness and not the look.

 

[bobn2]

I am now a wiser man
I made an error in my thinking. Now corrected.

I think your first thoughts were right.

The limited aperture acceptance by the micro lenses do not affect the DOF all that much.

The out of focus parts of the image will hit the sensor from many angles, some of them within the acceptance angle of the micro lenses.

But I think if you work through it, you find that the pixels in the outer sides of the blur circle are only getting light from the edges of the exit pupil, so they just can't see that light.

 

[Roland Karlsson]

I think your first thoughts were right.

But I think if you work through it, you find that the pixels in the outer sides of the blur circle are only getting light from the edges of the exit pupil, so they just can't see that light.

Bob,

Interesting!

You usually are right, as far as I can remember. So ... should I accept your statement ... and claim I was originally right

It is the intuitive answer after all. If you cannot see the light rays, they cannot affect the result. Seems flawless.

So ... I am tempted to take my correction back.

But ... it is late ... and I am tired.

So ... lets sleep on it.

 

[bobn2]

I think your first thoughts were right.

But I think if you work through it, you find that the pixels in the outer sides of the blur circle are only getting light from the edges of the exit pupil, so they just can't see that light.

Bob,

Interesting!

You usually are right, as far as I can remember. So ... should I accept your statement ... and claim I was originally right

It is the intuitive answer after all. If you cannot see the light rays, they cannot affect the result. Seems flawless.

So ... I am tempted to take my correction back.

But ... it is late ... and I am tired.

So ... lets sleep on it.

I don't think it's an easy question, I have swung both ways on it, but in the end, convinced myself that it affected both DOF and brightness. After all, if the light from the edges of the exit pupil isn't being registered, how can it affect DOF?

However, I think that the practical truth is somewhere in between. My analysis assumes that the microlenses are perfect lenses, when if fact they are very poor lenses, so it's quite possible that in extremis they capture light from a ragged, irregular portion of the exit pupil, in which case the perceptual bokeh might not be what you'd expect from the simpler view.

 

[Roland Karlsson]

bobn2 wrote:
I don't think it's an easy question, I have swung both ways on it, but in the end, convinced myself that it affected both DOF and brightness. After all, if the light from the edges of the exit pupil isn't being registered, how can it affect DOF?

However, I think that the practical truth is somewhere in between. My analysis assumes that the microlenses are perfect lenses, when if fact they are very poor lenses, so it's quite possible that in extremis they capture light from a ragged, irregular portion of the exit pupil, in which case the perceptual bokeh might not be what you'd expect from the simpler view.

OK - time to go to bed for me.

One thing is sure though. The T-stop you get when focussing the lens on infinity has very little to do with the bokeh you get when focussing at something nearby.

 

[Horshack]

I've done a light transmission test (t-stop) on my A7r for all the 35mm to 50mm lenses I have on hand. The purpose of the test is to see how much light is reached at the sensor when the lenses are wide-open, which accounts both for lens transmission (t-stop) as well as sensor micro-lens transmission loss (DxO Article). This information can be used to make weight/size/price decisions for those considering a purchase of these lenses and wanting the best effective light transmission (ie, best low-light performance).

The test was performed by shooting a raw image of an X-Rite Color Passport, with each lens at the largest aperture supported and the chart centered in the frame to prevent vignetting from affecting the results. All raws were loaded into CS6 with identical white-balance. The Leica Noctilux f/0.95 was used as the reference lens since it is the fastest. The 2nd darkest gray patch on the color chart (3rd patch on top row from left) was sampled in CS6 to obtain the reference luminance. I then loaded the images from the other lenses and adjusted the ACR exposure compensation for each so that the 3rd patch matched the luminance from the Noctilux reference...the delta from reference to the f-stop shows the relative transmission of each lens.

Side note: By comparing the light transmission of the Canon 50L f/1.2 and f/1.4 lenses mounted on a generic EF adapter (no electrical contacts) vs those same lenses mounted on a Metabones Smart Adapter (w/electric contacts, so f/stop data passed to camera), I determined that the A7r does a "silent" ISO gain of .35 stops for f/1.2 and .33 stops for f/1.4. The values in the chart for EF lenses exclude those gains (ie, show the real EV difference).

How to read the chart:
* The "Expected Exp Comp" column shows the expected EV difference between each lens tested vs the Noctilux based on the nominal aperture difference. For example, f/1.2 -> f/1.0 represents 1/2 stop nominal difference. For ease of viewing I've rounded the Noctilux up to f/1.0
* The "ACR Exposure Comp Necessary" column shows how much positive exposure compensation was required to match the brightness of the Noctilux. This value is expected to be less than or equal to the nominal f/stop difference.
* The "EV Diff" shows the exposure compensation necessary vs the expected exposure comp. A positive value indicates that the amount of ACR exposure comp was less than indicated by the nominal f/stop difference. A negative value indicates that the amount of ACR exposure comp was more than indicated by the nominal f/stop difference.

If you want to see the effective transmission difference between two lenses, compare them in the "ACR Exposure Comp Necessary" column. For example, the Sony FE55 @ f/1.8 is 1.30EV darker than the Noctilux @ f/0.95.

 

[Great Bustard]

I've done a light transmission test (t-stop) on my A7r for all the 35mm to 50mm lenses I have on hand. The purpose of the test is to see how much light is reached at the sensor when the lenses are wide-open, which accounts both for lens transmission (t-stop) as well as sensor micro-lens transmission loss (DxO Article). This information can be used to make weight/size/price decisions for those considering a purchase of these lenses and wanting the best effective light transmission (ie, best low-light performance).

The test was performed by shooting a raw image of an X-Rite Color Passport, with each lens at the largest aperture supported and the chart centered in the frame to prevent vignetting from affecting the results. All raws were loaded into CS6 with identical white-balance. The Leica Noctilux f/0.95 was used as the reference lens since it is the fastest. The 2nd darkest gray patch on the color chart (3rd patch on top row from left) was sampled in CS6 to obtain the reference luminance. I then loaded the images from the other lenses and adjusted the ACR exposure compensation for each so that the 3rd patch matched the luminance from the Noctilux reference...the delta from reference to the f-stop shows the relative transmission of each lens.

Side note: By comparing the light transmission of the Canon 50L f/1.2 and f/1.4 lenses mounted on a generic EF adapter (no electrical contacts) vs those same lenses mounted on a Metabones Smart Adapter (w/electric contacts, so f/stop data passed to camera), I determined that the A7r does a "silent" ISO gain of .35 stops for f/1.2 and .33 stops for f/1.4. The values in the chart for EF lenses exclude those gains (ie, show the real EV difference).

How to read the chart:
* The "Expected Exp Comp" column shows the expected EV difference between each lens tested vs the Noctilux based on the nominal aperture difference. For example, f/1.2 -> f/1.0 represents 1/2 stop nominal difference. For ease of viewing I've rounded the Noctilux up to f/1.0
* The "ACR Exposure Comp Necessary" column shows how much positive exposure compensation was required to match the brightness of the Noctilux. This value is expected to be less than or equal to the nominal f/stop difference.
* The "EV Diff" shows the exposure compensation necessary vs the expected exposure comp. A positive value indicates that the amount of ACR exposure comp was less than indicated by the nominal f/stop difference. A negative value indicates that the amount of ACR exposure comp was more than indicated by the nominal f/stop difference.

If you want to see the effective transmission difference between two lenses, compare them in the "ACR Exposure Comp Necessary" column. For example, the Sony FE55 @ f/1.8 is 1.30EV darker than the Noctilux @ f/0.95.

Nicely done! Your efforts here are much appreciated!

 

[RussellInCincinnati]

Horshack, what a great post. It looks like F/0.95 Leica lens really does hustle more photons to the pixel wells than do all the other lenses you tested (a nice set by the way) that feature smaller maximum nominal F-stop ratings like F/1.2 etc.

So it is not true on the Sony A7r, that ultra fast lenses like the Leica F/0.95 give you no dimmer light capabilities.

 

[Allan Olesen]

One thing is sure though. The T-stop you get when focussing the lens on infinity has very little to do with the bokeh you get when focussing at something nearby.

Why?

Do you expect the incident angle of the light rays on the sensor to change?

Or do you think that it has an influence whether the light rays meet each other in a point in front of the sensor instead of behind the sensor?

Anyway, there is one aspect which haven't been covered much yet:

The corners of the sensor will behave differently than the centre of the sensor.

In the centre, the light rays are symmetrical around an axis which is perpendicular to the sensor.

At the corners, they are symmetrical around an axis which is angled. This does not affect the image if the microlens is perfect and can catch light from any angle. But if the microlens cuts off light rays outside a given angle, you will get an asymmetrical OOF disc. And you actually often see exactly that in digital photos taken with large aperture lenses.

 

[Roland Karlsson]

One thing is sure though. The T-stop you get when focussing the lens on infinity has very little to do with the bokeh you get when focussing at something nearby.

Why?

Do you expect the incident angle of the light rays on the sensor to change?

Of course they do. You have moved the lens further from the sensor when focussing on something nearby. Then the F0.95 lens no longer has an efficient aperture of F0.95.

The corners of the sensor will behave differently than the centre of the sensor.

In the centre, the light rays are symmetrical around an axis which is perpendicular to the sensor.

At the corners, they are symmetrical around an axis which is angled. This does not affect the image if the microlens is perfect and can catch light from any angle. But if the microlens cuts off light rays outside a given angle, you will get an asymmetrical OOF disc. And you actually often see exactly that in digital photos taken with large aperture lenses.

That is correct. All discussions so far assumes the the light hit the sensor per pendular to the surface.

There are two ways of fixing this.

1. You can have a tele centric lens.
2. You can have a skewed micro lens array, like in the Leica.

--
/Roland
X3F tools:
http://www.proxel.se/x3f.html
https://github.com/rolkar/x3f

 

[Allan Olesen]

One thing is sure though. The T-stop you get when focussing the lens on infinity has very little to do with the bokeh you get when focussing at something nearby.

Why?

Do you expect the incident angle of the light rays on the sensor to change?

Of course they do. You have moved the lens further from the sensor when focussing on something nearby. Then the F0.95 lens no longer has an efficient aperture of F0.95.

If you are taking two photos with the same OOF object in the same distance from the front element, then the incident angle of light emitted from that object should not change when you change the focus distance.

You seem to be confusing this with what is happening for an object in focus.

 

[Roland Karlsson]

One thing is sure though. The T-stop you get when focussing the lens on infinity has very little to do with the bokeh you get when focussing at something nearby.

Why?

Do you expect the incident angle of the light rays on the sensor to change?

Of course they do. You have moved the lens further from the sensor when focussing on something nearby. Then the F0.95 lens no longer has an efficient aperture of F0.95.

If you are taking two photos with the same OOF object in the same distance from the front element, then the incident angle of light emitted from that object should not change when you change the focus distance.

You seem to be confusing this with what is happening for an object in focus.

I thought we were talking about how much light leaving the lens that could be catcher by the sensor.

And that increases if we move the lens further away from the sensor.

--
/Roland
X3F tools:

Proxel X3F Tools

GitHub - rolkar/x3f: Tools for manipulating X3F files from Sigma cameras

Tools for manipulating X3F files from Sigma cameras - rolkar/x3f

github.com

 

[Allan Olesen]

One thing is sure though. The T-stop you get when focussing the lens on infinity has very little to do with the bokeh you get when focussing at something nearby.

Why?

Do you expect the incident angle of the light rays on the sensor to change?

Of course they do. You have moved the lens further from the sensor when focussing on something nearby. Then the F0.95 lens no longer has an efficient aperture of F0.95.

If you are taking two photos with the same OOF object in the same distance from the front element, then the incident angle of light emitted from that object should not change when you change the focus distance.

You seem to be confusing this with what is happening for an object in focus.

I thought we were talking about how much light leaving the lens that could be catcher by the sensor.

And that increases if we move the lens further away from the sensor.

You still seem to be thinking about objects which are in focus. But we are talking OOF objects here.

The sensor will receive the same amount of light from an OOF object in the same distance from the lens, no matter how the lens was focused, and the light will arrive at the same incident angles. This is at least true for a simple lens with only one element.

 

[Roland Karlsson]

You still seem to be thinking about objects which are in focus. But we are talking OOF objects here.

No - I am not thinking about objects at all. I am only thinking about what the sensor can see. If you move the lens further from the sensor, then the aperture will look smaller. And if you move it far enough, all the light from the light cone from the aperture will be able to hit the sensor. You could say that the F0.95 stop might actually be e.g. F2.0.

The sensor will receive the same amount of light from an OOF object in the same distance from the lens, no matter how the lens was focused, and the light will arrive at the same incident angles. This is at least true for a simple lens with only one element.

Yes, and I fail to see why that is relevant.

 

[Allan Olesen]

You still seem to be thinking about objects which are in focus. But we are talking OOF objects here.

No - I am not thinking about objects at all. I am only thinking about what the sensor can see.

Yes, the OOF disc on the sensor. That is what the sensor sees if you have an OOF point.

 

[bobn2]

You still seem to be thinking about objects which are in focus. But we are talking OOF objects here.

No - I am not thinking about objects at all. I am only thinking about what the sensor can see. If you move the lens further from the sensor, then the aperture will look smaller.

The angle of the light cone focussed to a point in the focal plane depends only on the f-number of the lens, not the distance of the exit pupil.

 

[Anders W]

The Voigtländer 25 mm Nokton for u43 is F0.95. That is cool, sort of. A great potential for shallow DOF, or ... ?

I thought, correct me if I am wrong, that the photo sensor, with micro lenses, Bayer filters, etc, etc, cannot see more than F1.5 or so. Everything else is wasted.

Have I got it right?

In that case, nothing will happen when opening the lens from F1.4 to F1.0.

Have a look here for a test of the light loss of the CV Nokton 25/0.95 on MFT. The results suggest that the loss wide open is only 0.4 EV.

http://www.dpreview.com/forums/thread/3209678