Leica D Summilux 25 mm f/1.4 review

FM Forums \| Alternative Gear & Lenses \| Join Upload & Sell	Search Used
1 2 end

Archive 2009 · Leica D Summilux 25 mm f/1.4 review

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #1 · Leica D Summilux 25 mm f/1.4 review

theSuede wrote:
Could you link that thread, Jonas? I'm certain you made a good comparison out of it :-) Thank you. My "bla bla bla" was primarily meant for people that have gone blind into religious beliefs centered on their own system without actually comparing. I'm certain that the 25F/1.4 is an impressive lens, on the 4/3-format. I haven't tried it, but I'm hugely impressed with the Oly 50F/2 and the 150F/2.

Maws - we're not talking about the same thing - and you're taking the word "picture" totally out of its context. My "boundaries" for the equivalence test with "same picture" are:
*same total picture resolution (how can it be "the same picture" otherwise?)
*same shutterspeed - which together with:
*same DoF - remember that the pictures are "same resolution", so this is easy
*same "FoV" - gives the same amount of light onto both sensors if the two criteria above are satisfied.

I'm hardly taking the word picture out of context. I'm normalizing based on exposure at a given ISO, which is the generally accepted way of doing so. And you're simply ignoring identical exposure with that setup. By normalizing total illumination you've created two images that may look identical under certain circumstances but can greatly differ under others depending on exactly where on the sensor's noise curve each image is placed. And you're getting that normalization effectively by accident (since what gives you the identical DoF is the fact that the image magnification and physical aperture are the same given the format shift for the former and the fact that a 25mm at f1.4 and a 50mm at f2.8 both have a ~17.9mm aperture for the latter). Actually even DoF doesn't exactly scale like that due to acceptable CoC differences between formats.

If I give you a compact with 8.4�m pp, it will have a resolution of ~800x600 - and that picture will hardly be comparable with the E-30 or the D700, seen as a total picture's worth. But at F/2.8 and a certain shutterspeed, it will indeed have the same noise per pixel as the D700, only 25 times less pixels... For me that's 25 times less picture (or a 5 times smaller picture - when measured linearly).

But if you think 800x600 is equivalent to a D700's 12MP - please go ahead and reduce all of your picture libraries to that resolution - it will save you MASSIVE amounts of space.

No, what it will give you is 5 times less linear resolution, not 5 times less picture (since we're assuming identical FoV's here). Never said it was going to deliver the same resolution. But the difference between 4/3rds and FX is far less than the difference between that frankly ridiculous suggestion. And what resolution you need is controlled by the desired output, which will be a major contributing factor in camera choice.

The point I'm trying to make and you're busy ignoring is that the difference in noise performance scales slower than the increase in sensor area. This is because it's NOT directly linked to total illumination. Nothing with regards to photography is directly linked to total illumination, it's a secondary factor.

Jun 06, 2009 at 04:00 PM

theSuede
Offline
Upload & Sell: Off

p.2 #2 · Leica D Summilux 25 mm f/1.4 review

YES!
- the noise scales slower than the area increase - due to electronic noise and losses! But it does scale - with a "reasonably" linear effect!
The "losses" compared to a perfectly linear scaling is not even close to the effect that most 4/3-users would want it to. Of the pictures above, which is D700@ISO1600 and which is 4/3@ISO400? They're developed EAXCTLY the same - except for the colour correction I applied as mentioned. I've also increased colour saturation to be able to see "blotchiness" better (at the loss of blowing channels at places), and contrast to slightly amplify noise. Same (very low effect) sharpening. Equal for both pictures.

I can't really say that one is "more noisy" than the other. I can see other differences, due to the "personalities" of the equipment used, but no difference in noise.

Jun 06, 2009 at 04:16 PM

theSuede
Offline
Upload & Sell: Off

p.2 #3 · Leica D Summilux 25 mm f/1.4 review

And btw - I was a bit quick on the keyboard earlier - you were right to criticize my choice of words. Maybe a better phrasing would be:
"25 times less picture information", and "5 times smaller picture dimensions"

Jun 06, 2009 at 04:20 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #4 · Leica D Summilux 25 mm f/1.4 review

theSuede wrote:
Smooth tonality comes from low photon shot noise per picture element (pixel) combined with high resolution. Low shot noise = lots of light. Light amount is equal to the space angle of the view times the area of the aperture. Same aperture, larger space angle of view = more light. More light and same resolution = less noise.

The high resolution would be a total waste, if the light amount over the frame was the same as on a 36x24 sensor. It would all be noise - the equivalent of using a VERY very high ISO.

Then please explain why a 22MP Digital back with a 36x48 sensor is visibly noisier than a 24MP 24x36 sensor at ISO 400-800. This is despite the twofold increase in total illumination. Of course we're talking about different sensor technology, and sensor technology, along with sensor site size are the determining factors in noise performance.

Smooth tonality comes from the output being at a lower magnification relative to the recording media or from the increase in resolution of a given media size.

In both cases total illumination is a secondary factor. It can certainly be used as a shorthand to predict performance given a known set of variables but it is not actually the source of the performance.

Jun 06, 2009 at 04:22 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #5 · Leica D Summilux 25 mm f/1.4 review

theSuede wrote:
YES!
- the noise scales slower than the area increase - due to electronic noise and losses! But it does scale - with a "reasonably" linear effect!
The "losses" compared to a perfectly linear scaling is not even close to the effect that most 4/3-users would want it to. Of the pictures above, which is D700@ISO1600 and which is 4/3@ISO400? They're developed EAXCTLY the same - except for the colour correction I applied as mentioned. I've also increased colour saturation to be able to see "blotchiness" better (at the loss of blowing channels at places), and contrast to slightly amplify noise. Same (very low effect) sharpening. Equal for both pictures.

I can't really say that one is "more noisy" than the other. I can see other differences, due to the "personalities" of the equipment used, but no difference in noise.

I'd have to say the first image is likely the 4/3rds, based on the lower dynamic range and lower resolution (this is assuming that you haven't mucked too much with the tone curves and that the softer first image is not due to other factors).

Of course, using an E-520 means that you're also using a sensor that's nearly 2 generations behind the D700 sensor and lower resolution to boot (although the latter is not as large a factor given the small resolution difference, nominally 2MP but larger in actual resolution as the E-520 is fairly low resolution for a 10MP camera while the D700 is fairly high resolution for a 12MP camera, mostly a result of differing AA filters).

Newer, lower-noise sensors would perform better in this comparison, providing a cleaner image at ISO 400. You'd also see much the same comparison if you compared say a 1Ds or a DCS14 at ISO 400 to the D700 at ISO1600 despite all 3 cameras being FX.

Edited on Jun 06, 2009 at 04:38 PM · View previous versions

Jun 06, 2009 at 04:32 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #6 · Leica D Summilux 25 mm f/1.4 review

theSuede wrote:
And btw - I was a bit quick on the keyboard earlier - you were right to criticize my choice of words. Maybe a better phrasing would be:
"25 times less picture information", and "5 times smaller picture dimensions"

Agreed.

I'm very much NOT arguing that there aren't differences between sensor sizes, but rather that Total Illumination is not a controlling factor in them but rather an ancillary factor that can be useful in predicting performance of a given sensor technology across different sizes or roughly predicting differences between format sizes using a given generation of image capture technology.

Jun 06, 2009 at 04:35 PM

theSuede
Offline
Upload & Sell: Off

p.2 #7 · Leica D Summilux 25 mm f/1.4 review

The Kodak DCS pro from 2003' is also noisier at ISO800 than the modern compact LX-3, even though it's a "full 36x24mm" camera.

AFAIK the D700 and the 520 in the example above could be said to be "same generation, same level of sensor refinement". But it is also true as you say that you can optimize for a certain range of relative exposure by balancing dark-current, read noise and amplification noise in the construction/development phase. This had limited impact though, and should be considered a minor variation if you're comparing sensor of different sizes but same level of technological refinement.

Now please - which is which of the pictures? I solemnly swear that I have NOT tampered with the pictures (individually) in ANY way other than the colour correction - the rest i saved as an action and just hit "replay" for the other picture.

Jun 06, 2009 at 04:38 PM

theSuede
Offline
Upload & Sell: Off

p.2 #8 · Leica D Summilux 25 mm f/1.4 review

Of course if you mean "there are other choices to be made during the development process that impact the sensor (or indeed "picture") performance more than small differences in sensor area" I'm totally with you.
But the difference from 4/3 up to FX isn't small, it's roughly four times the area.

Four times the area is the same difference as between a standard compact cameras' 8x6mm sensors and the 17x13mm 4/3 sensors.

Jun 06, 2009 at 04:46 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #9 · Leica D Summilux 25 mm f/1.4 review

theSuede wrote:
The Kodak DCS pro from 2003' is also noisier at ISO800 than the modern compact LX-3, even though it's a "full 36x24mm" camera.

AFAIK the D700 and the 520 in the example above could be said to be "same generation, same level of sensor refinement". But it is also true as you say that you can optimize for a certain range of relative exposure by balancing dark-current, read noise and amplification noise in the construction/development phase. This had limited impact though, and should be considered a minor variation if you're comparing sensor of different sizes but same level of technological refinement.

Now please - which is which of the pictures? I solemnly swear that I have NOT tampered with the pictures (individually) in ANY way other than the colour correction - the rest i saved as an action and just hit "replay" for the other picture.

IMHO the first is the E-520, based on other factors (apparent dynamic range, less resolution), as I noted above.

the E-520 is an older sensor design than the Nikon D700 as well. Note the D700 sensor is inherited from the D3, a late 07 launch and one of the first of the current generation designs. It really was the breakthrough design for current generation high ISO performance. The E-520 inherits its sensor from the E-510/E-410, of early 07 but also NMOS rather than CMOS (the latter does perform better overall) and was about a generation behind in refinement compared to other models of similar vintage and level. That's an effective 2 generation difference in sensor design (crop a D2x down to 4/3rds and you'll see similar noise performance to the E-520)

Jun 06, 2009 at 04:48 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #10 · Leica D Summilux 25 mm f/1.4 review

theSuede wrote:
Of course if you mean "there are other choices to be made during the development process that impact the sensor (or indeed "picture") performance more than small differences in sensor area" I'm totally with you.
But the difference from 4/3 up to FX isn't small, it's roughly four times the area.

Four times the area is the same difference as between a standard compact cameras' 8x6mm sensors and the 17x13mm 4/3 sensors.

Sensor size difference is 3.48x between 4/3rds (18x13.5mm) and FX. It's 5x even between an 8x6mm sensor and 4/3rds. But frankly I'm just nitpicking there, it's not a big issue. The real deal-breaker between compact performance and even 4/3rds is sensor tech. Most P&S's are still using CCD designs due to the ability to use electronic shutters. They also typically use always-on designs which increases noise.

But my point is that the other choices matter more than moderate differences in sensor area although a large difference will tell notably up to a certain point (noise performance peaks at FX and trends downwards with sensors larger than FX).

If you want a really good example of how much the other factors matter crop a T1i shot down to 4/3rds and compare it to that E-520. Or even a 50D shot. Then compare to say an A350 shot also cropped down to 4/3rds. I'd bet solid money that T1i shot at 400 and cropped down to 4/3rds will look better than your ISO 1600 D700 shot with all else normalized. The A350 may well look worse than the E-520 which is roughly equivalent. note example cameras chosen based on pixel density (E-520 is 4.1MP/cm2, T1i is 4.5MP/cm2, the Sony A350 is 3.8MP/cm2).

Jun 06, 2009 at 05:01 PM

Makten
Offline
Upload & Sell: Off

p.2 #11 · Leica D Summilux 25 mm f/1.4 review

mawz wrote:
Then please explain why a 22MP Digital back with a 36x48 sensor is visibly noisier than a 24MP 24x36 sensor at ISO 400-800. This is despite the twofold increase in total illumination. Of course we're talking about different sensor technology, and sensor technology, along with sensor site size are the determining factors in noise performance.

Sensor technology is the key, yes. The MF cameras do have lower native ISO, and therefore the circuitry is optimized for lower amplification of the signal. The photosites may have larger full well capacity due to their size, which means that if you want to use the full DR at native ISO (which is low), the s/n ratio will decrease more rapidly when using higher ISO speeds.

If we use "the same" sensor technology, the larger sensor will always win at a certain exposure per area, since the total area is larger. I say "the same" because no one would just scale up or down a certain sensor to fit a different format.

I have many times wondered why the MF camera manufacturers don't make sensors that would kick 3D butt. But probably the MF users are more interested in good DR at low ISO than low noise at high ISO speeds.

Jun 06, 2009 at 05:16 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #12 · Leica D Summilux 25 mm f/1.4 review

Makten wrote:

Sensor technology is the key, yes. The MF cameras do have lower native ISO, and therefore the circuitry is optimized for lower amplification of the signal. The photosites may have larger full well capacity due to their size, which means that if you want to use the full DR at native ISO (which is low), the s/n ratio will decrease more rapidly when using higher ISO speeds.

If we use "the same" sensor technology, the larger sensor will always win at a certain exposure per area, since the total area is larger. I say "the same" because no one would just scale up or down a certain sensor to fit a different format.

The 5D sensor is essentially a scaled-up 1DmII sensor, the Kodak sensors used in the Leica's are descended straight from scaled-down MF sensors. Yes, such things have actually made it to market although it is certainly rare. And yes, a larger sensor will always win for performance if all other aspects are equal. The extra resolution effectively reduces apparent noise. But the increase isn't linear, you need enough extra resolution for the difference to matter. Going say 10-12MP isn't going to do much. Really you need around 50% more pixels to see an appreciable difference.

I have many times wondered why the MF camera manufacturers don't make sensors that would kick 3D butt. But probably the MF users are more interested in good DR at low ISO than low noise at high ISO speeds.

I've long wondered why Nikon or Sony hasn't done a 36x48mm version of their flagship FF sensors for the MF market. Especially Nikon, A 24MP 36x48 version of the D3 sensor would be a world-changer in the MF market.

Jun 06, 2009 at 05:27 PM

Makten
Offline
Upload & Sell: Off

p.2 #13 · Leica D Summilux 25 mm f/1.4 review

mawz wrote:
The 5D sensor is essentially a scaled-up 1DmII sensor, the Kodak sensors used in the Leica's are descended straight from scaled-down MF sensors. Yes, such things have actually made it to market although it is certainly rare. And yes, a larger sensor will always win for performance if all other aspects are equal. The extra resolution effectively reduces apparent noise. But the increase isn't linear, you need enough extra resolution for the difference to matter. Going say 10-12MP isn't going to do much. Really you need around 50% more pixels to see an appreciable difference.

By "scale up" I actually meant the whole sensor. Not cutting a larger piece out of the same wafer.

I've long wondered why Nikon or Sony hasn't done a 36x48mm version of their flagship FF sensors for the MF market. Especially Nikon, A 24MP 36x48 version of the D3 sensor would be a world-changer in the MF market.

Probably because ISO 200 isn't ideal for what most MF users do. I'd like it though!

Jun 06, 2009 at 05:54 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.2 #14 · Leica D Summilux 25 mm f/1.4 review

Makten wrote:

By "scale up" I actually meant the whole sensor. Not cutting a larger piece out of the same wafer.

If you mean taking the same pixel count and just making everything bigger, that requires an entirely new sensor design. Sensor scaling is actually done by using a larger mask on the wafer (each sensor is imaged separately on the wafer) and keeping the sensor site size the same. changing the sensor site size (and the microlens size) is where much of the engineering in the sensor exists.

Probably because ISO 200 isn't ideal for what most MF users do. I'd like it though!

On the flip side, you'd gain back the backbone of the MF market before digital, the wedding/event shooters who all went to DSLR's and would be quite happy with an ISO 200 base if it meant they got clean ISO 1600 from their Hassy/Mamiya. And there's a LOT more of them than the current lot of MF shooters.

Jun 06, 2009 at 06:03 PM

1 2 end
FM Forums \| Alternative Gear & Lenses \| Join Upload & Sell 1 2 end