M9 Noise - another thread!

[dougi]

Interestingly, it seems someone has done objective comparisons already between noise in M8 and M9:

http://www.h2hreviews.com/article/H...s-Leica-M8-2-vs-Olympus-E-P1/Image-Noise.html

Never heard of this mob before, and unsure of the test methodology, but noise looks very similar for both. Presumably this is for RAW, and doesn't consider any improved smoothing in JPEG.

Of course, M9 will likely look better re noise in print as there is less enlargement so the noise will appear finer for a given print size.

 

[user237428934]

Strange. When you look at the chart you notice a measuring point at ISO 800 and ISO 2000 for the M8 but that's technically not possible. The test seems not very reliable.

 

[jaapv]

No- there are very impressive graphs - but many of them flatly are at odds with those of reputable reviews - I would take this whole thing with a pinch of salt.

 

[ulrikft]

It seems like they have done a pr pixel comparison, which is by definition faulty when comparing noise levels of diffeirng sensors.

 

[Ben Z]

0.75x 18=13.5 vs 10.1 megapixels for the M8 means there's more pixels crammed in per square millimeter on an M9, which generally translates to higher noise levels and therefore greater need for in-camera NR. (And Leica says there's actually less density toward the edges, meaning even higher concentration in the area equivalent to the M8's crop. Add the repositioned thicker IR filtration and in-camera cyan correction, and ostensibly the M9 has to cook the raw capture more than the M8 does. So if the noise and image quality are just equal, it's already a feat well done. And if they are improved, kudos should go to Leica's design partners for a job well orchestrated.

Tests are entertaining, but the proof is in the pudding, or in this case, the final prints. Right now I don't think there's enough experience out there post-processing M9 files to draw an accurate conclusion as to its maximum potential.

 

[ferider]

0.75x 18=13.5 vs 10.1 megapixels for the M8 means there's more pixels crammed in per square millimeter on an M9 .....

Pixel density is equivalent. Got to do

crop_factor * crop_factor * nr_pixel

for scaling.

 

[mani]

It seems like they have done a pr pixel comparison, which is by definition faulty when comparing noise levels of diffeirng sensors.

You seem to have only one comment to make - across at least two forums!

Hilarious.

 

[Ben Z]

Pixel density is equivalent. Got to do

crop_factor * crop_factor * nr_pixel

for scaling.

If I were to cut the M9's full-frame sensor down to the M8's size, I would have a sensor 3/4 the area, with 3/4 the # of pixels in it. Please explain me why that isn't so, unless a) the size of the pixels themselves is not the same M8-M9, which has not been stated by Leica, or b)the pixel concentration is unequal throughout the sensor, which according to Leica is the case, but the stronger concentration is in the center, making the density higher within the cropped area equivalent to the M8's sensor.

 

[ferider]

Here you go. You are mixing properties proportional to area (Mpixel) vs. width/height (crop factor).

 

[ferider]

Per Reichmann's interview of Stephan Daniel, the M8 and M9 sensors in the center are identical, but outside of center, microlenses are shifted and the pixel architecture was changed. The 1-2 stops additional dynamic range were achieved with different/other components on the new Jena Optics PCB (different A/C converter, among others).

Also, due to pre-DNG vignetting correction, in theory, the M9 must have different dynamic range for center and for corners, for wide angles at least. How much, time and users will tell.

Roland.

 

[Ben Z]

Here you go. You are mixing properties proportional to area (Mpixel) vs. width/height (crop factor).

Got it. Thanks. According to your math the M9 still has greater pixel density, but not by much. Although Leica does say the M9's pixels are less dense in the outer region (i.e. the part beyond the M8 cropped area) so that would indicate a higher density within the cropped area. Again, not as much as I suspected. Now it makes sense to me why my 1.6-crop Canon 20D images are almost indistinguishable from the 5D, which had always puzzled me. Another myth of the purported superiority of full-frame busted. Thanks again.

 

[Pickett Wilson]

Ben, the 5D's FF sensor produces far superior results than the 20D. Whatever the theory, I can tell you as someone who has used every 1.6 crop Canon since the D30 and well as two 5D's for several years, the images from the 5D are a lot better.

 

[ulrikft]

You seem to have only one comment to make - across at least two forums!

Hilarious.

If more than one forum has people that make the same mistake, i correct it in both places, and I do have quite a few other comments to make, so begone troll.

 

[ulrikft]

To make the point clear: pixel density has yet not shown any relation to noise, it is an assumption people make beacause they look at per-pixel crops, and hence, gimp the higher megapixel sensor. Compare the 5d MkII and the d700, you'll soon see that good high iso capability is not for low resolving sensors only.

And if people don't understand why i choose to compare the leica M9 to the best high iso sensors out there.. well, show me the high iso expert digital rangefinder, and I'll make you a personal test, for your eyes only, ok?

 

[mani]

If more than one forum has people that make the same mistake, i correct it in both places, and I do have quite a few other comments to make, so begone troll.

"Begone troll!" 😀
Dude - until you own at least one rangefinder camera, then at least be slightly polite to those that do, and know what they're talking about (and have a few more posts under their belts on this and other fora, for that matter).

 

[ulrikft]

"Begone troll!" 😀
Dude - until you own at least one rangefinder camera, then at least be slightly polite to those that do, and know what they're talking about (and have a few more posts under their belts on this and other fora, for that matter).

I have two rangefinder cameras that i use quite often, thank you. And I have quite a few posts on this and other foras. Any other comments?

 

[Rob-F]

Pixel density is equivalent. Got to do

crop_factor * crop_factor * nr_pixel

for scaling.

No, Ben Z. got it right. First, the M9 has 1.75 times as many pixels, but the sensor is only 1.33 times larger. Thus it is intuitively obvious that the pixels are spaced closer together. How much more closely spaced are they? Why, simply 1.75/1.33. There are 1.32 pixels in each interval in which there are one pixel in the M8. So they can't be equivalent.

 

[Rob-F]

Well, let's take another look. The M8 sensor, I take it, is 18 x 27mm. That's 486 square mm. And the M9 sensor is 24 x 36mm. That's 864 square mm. (I'm using my 1956 Post 10" slide rule, so cut me some slack) So on an area basis, the M9 sensor is 1.75 times the area of the M8 sensor. So when I look at it that way, the ratio of the sensor areas, matches the ratio of the megapixels. So it looks like the pixel density must be the same.

So now I think ferider has it right. Apologies are in order.