Read Sean's ISO results and quite disappointed.

[padraigm]

Just read Sean's ISO write-up, and well it's disappointing that this is all Leica could come up with. It really does seem to me to be a case of in camera smoothing at higher ISO and at a loss of some detail. I am so disappointed and I think I am now out of the market for an M9. It's just not good enough in this day an age. I so wanted to have a digital RF with good high ISO. Matching the M8 to the M9 could probably be achieved with noise ninja or difine...For sure the full frame is nice but the price!!!! I might just get another M8.. Just my opinion of course.

[Bo_Lorentzen]

We talking about different cameras.? I was just looking through the 1250 iso shots from Samy's and those faster.. comparing to shadows in these and M8 images. its at least 1 stop better. full frame. Lovely camera.

.

[adan]

I haven't read Sean's report, but I suspect if he's seeing some smoothing or processing at ISO 1600 and above, he's likely right. I sense a slight CaNikoness to the 1600-2500 range (in my so-far limited 4 gigs of shooting). And that's too bad, unless Leica produces a firmware upgrade to control noise smoothing by the users.

OTOH, the big sensor means I can shoot my 21 f/2.8 instead of the C/V 15mm f/4.5 - so I can skip ISO 1600 and use ISO 400 in the same light (or spring for the 21 Summilux and shoot ISO 160).

Also remember that at the same print size (with more, smaller pixels from the M9) the noise speckles will be 1.33x smaller.

[sean_reid]

Yes, about one stop better in prints. I think people had hopes of Canon/Nikon level high ISO performance but even a gain of one stop works well for me.

Cheers,

Sean

[sean_reid]

Originally Posted by adan

I haven't read Sean's report, but I suspect if he's seeing some smoothing or processing at ISO 1600 and above, he's likely right. I sense a slight CaNikoness to the 1600-2500 range (in my so-far limited 4 gigs of shooting). And that's too bad, unless Leica produces a firmware upgrade to control noise smoothing by the users.

OTOH, the big sensor means I can shoot my 21 f/2.8 instead of the C/V 15mm f/4.5 - so I can skip ISO 1600 and use ISO 400 in the same light (or spring for the 21 Summilux and shoot ISO 160).

Also remember that at the same print size (with more, smaller pixels from the M9) the noise speckles will be 1.33x smaller.

There's careful discussion about how this relates to print size and yes that is important.

Cheers,

Sean

[adan]

Hi, Sean. So are we going to have to relive the whole Digilux 2 thing? (Shoot 2 stops under to dodge the noise suppression, and then push it ourselves?). I thought Leica had learned from that.

[sean_reid]

No, I don't think that is needed at all.

Basically, for a given print size and a given level of noise, the M9 can go a stop faster in ISO than the M8. A stop gain is good even if it doesn't make the M9 a world champ at high ISO.

As for the high ISO smoothing - its not heavy handed at least.

Cheers,

Sean

[sean_reid]

Originally Posted by padraigm

Matching the M8 to the M9 could probably be achieved with noise ninja or difine...

No it couldn't match them because one has 18 MP. Check out that last part of the ISO section again.

Cheers,

Sean

[jl4069]

Andy,

thanks for the reply in the other thread.

You wrote just now:
"OTOH, the big sensor means I can shoot my 21 f/2.8 instead of the C/V 15mm f/4.5 - so I can skip ISO 1600 and use ISO 400 in the same light (or spring for the 21 Summilux and shoot ISO 160)."

Maybe there is more to IQ improvement you mentioned due to wide lenses? Maybe also the sensor? JL

[dalippe]

Originally Posted by adan

Also remember that at the same print size (with more, smaller pixels from the M9) the noise speckles will be 1.33x smaller.

Although the M9 will have a noise advantage in prints, I don't think that is quite the reason. Noisy areas of the photograph will be shrunken, but there will be more of them. I think the right technical explanation is what Sean said: down-sampling, which is just averaging, increases signal to noise. When you average N pixels, noise should go down by sqrt(N) while signal remains constant.

On the other hand, I believe that Leica's choice isn't as good for noise as simply having fewer pixels to begin with on the full size sensor. My understanding is that the relevant noise is read noise and doesn't grow with pixel size (dark current, on the other hand, is proportional to pixel area so larger pixels do nothing to help with signal-to-noise when dark current is the limiting factor).

To take a simple example, if you start with twice as many pixels and then down-sample by making every output pixel the average of two input pixels, you increase signal to noise by a factor of only sqrt(2) because signal stays the same and noise goes down by sqrt(2). But if you start with half as many pixels, giving each pixel twice the area, then you increase signal to noise by a full factor of 2 since signal doubles and noise is constant. Of course actual down-sampling algorithms aren't this simple, but I don't think that changes the basic insight.

I'm disappointed with Leica's choice. I may not be representative, but I don't need more than 10 megapixels. The largest prints I've ever made are 20x30 and those are rare. But I often shoot under dim, tungsten light and love the freedom my D3 gives me to skip the flash and still choose my DOF. When I can get away with it, I prefer to shoot my M8 and was hoping the M9 would more substantially increase the fraction of the time I can get away with shooting the Leica instead of the Nikon.

[sean_reid]

Originally Posted by dalippe

Although the M9 will have a noise advantage in prints, I don't think that is quite the reason. Noisy areas of the photograph will be shrunken, but there will be more of them. I think the right technical explanation is what Sean said: down-sampling, which is just averaging, increases signal to noise. When you average N pixels, noise should go down by sqrt(N) while signal remains constant.

On the other hand, I believe that Leica's choice isn't as good for noise as simply having fewer pixels to begin with on the full size sensor. My understanding is that the relevant noise is read noise and doesn't grow with pixel size (dark current, on the other hand, is proportional to pixel area so larger pixels do nothing to help with signal-to-noise when dark current is the limiting factor).

To take a simple example, if you start with twice as many pixels and then down-sample by making every output pixel the average of two input pixels, you increase signal to noise by a factor of sqrt(2) because signal stays the same and noise goes down by sqrt(2). But if you start with half as many pixels, giving each pixel twice the area, then you increase signal to noise by a factor of 2 since signal doubles and noise is constant. Of course actual down-sampling algorithms aren't this simple, but I don't think that changes the basic insight.

I'm disappointed with Leica's choice. I may not be representative, but I don't need more than 10 megapixels. The largest prints I've ever made are 20x30 and those are rare. But I often shoot under dim, tungsten light and love the freedom my D3 gives me to skip the flash and still choose my DOF. When I can get away with it, I prefer to shoot my M8 and was hoping the M9 would more substantially increase the fraction of the time I can get away with shooting the Leica instead of the Nikon.

I'm sure a lot of us would have loved truly clean ISO 2500 but you're still gaining a stop which is worthwhile certainly.

We agree on the down-sampling thing and its obvious in the test pics when they're matched in pixel dimensions.

Cheers,

Sean

[charlesphoto99]

People sure are funny. Leica essentially do the impossible (make a FF digital M) and give an extra stop to a stop and half iso but still not good enough. But then again it doesn't fit in my pocket, doesn't focus to 6", doesn't zoom from 18-300mm, and so on.

Get real people! Learn how to take a photograph and the noise won't be a problem. Go shoot some high speed color neg or TMAX 3200 under low light and then come back and talk about noise. Or better yet lug around a D3 everywhere (and I'm a D3 owner).

There really is no single camera that will "do it all." And thank heavens for that.

[dalippe]

Originally Posted by charlesphoto99

People sure are funny. Leica essentially do the impossible (make a FF digital M) and give an extra stop to a stop and half iso but still not good enough. But then again it doesn't fit in my pocket, doesn't focus to 6", doesn't zoom from 18-300mm, and so on.

Get real people! Learn how to take a photograph and the noise won't be a problem. Go shoot some high speed color neg or TMAX 3200 under low light and then come back and talk about noise. Or better yet lug around a D3 everywhere (and I'm a D3 owner).

There really is no single camera that will "do it all." And thank heavens for that.

That really isn't very fair. Given the fact that Leica solved the FF problem, there was a simple design decision they could have made to increase the high ISO abilities of the camera so that we wouldn't have had to lug our D3's around as often. Had they made that decision, then they would have been giving up something else (the ability to make large, detailed prints). What' s wrong, on a discussion forum, with users discussing which trade-off works better for them?

[dalippe]

Originally Posted by sean_reid

I'm sure a lot of us would have loved truly clean ISO 2500 but you're still gaining a stop which is worthwhile certainly.

We agree on the down-sampling thing and its obvious in the test pics when they're matched in pixel dimensions.

Cheers,

Sean

Yes, I saw that in your review update. Been waiting for it all day

[sean_reid]

Originally Posted by dalippe

Had they made that decision, then they would have been giving up something else (the ability to make large, detailed prints). What' s wrong, on a discussion forum, with users discussing which trade-off works better for them?

That definitely is the other side of the coin. Consider my friend Tod who will now do work with the M9 that he used to do with a 6 x 9 Fuji. Big prints - serious work for galleries and museums. 18 MP at lower ISO levels is just what he needed.

And of course with down-sampling one gets a much cleaner 10 MP file. So maybe it was good choice in the end...

Cheers,

Sean

[sean_reid]

Originally Posted by dalippe

Yes, I saw that in your review update. Been waiting for it all day

What a coincidence, I've been waiting all day to get that off my to do list.

Cheers,

Sean

[robertwright]

Sean, in that picture that immediately follows the noise section, under Wedding, of the woman on the phone next to the wall, what is going on with the wall to her left-is there a ridge on the wall going horizontally? or does it actually extend into the black on the left like I think it does?

could be weird compression artifact? or I won't say it...

[sean_reid]

Originally Posted by robertwright

Sean, in that picture that immediately follows the noise section, under Wedding, of the woman on the phone next to the wall, what is going on with the wall to her left-is there a ridge on the wall going horizontally? or does it actually extend into the black on the left like I think it does?

could be weird compression artifact? or I won't say it...

It seems like banding to me and that's how I describe it above the picture. Its one of just two high ISO pics I've found like that. Both were consecutive and Leica is clued into this (but we can't replicate it yet). Thousands of other pics at same ISO look fine in that respect.

I included the picture because I like it and it also nicely connects the noise section to the wedding section. It was published out of context on TOP (Mike Johnston just liked the pic) and now there's a little discussion going there about "discovering" the banding. That's the problem when people freely comment on a review they've never read. Kind of funny that it is a banding *illustration* in the review.

Cheers,

Sean

[charlesphoto99]

Originally Posted by dalippe

That really isn't very fair. Given the fact that Leica solved the FF problem, there was a simple design decision they could have made to increase the high ISO abilities of the camera so that we wouldn't have had to lug our D3's around as often. Had they made that decision, then they would have been giving up something else (the ability to make large, detailed prints). What' s wrong, on a discussion forum, with users discussing which trade-off works better for them?

Okay, then maybe Leica should offer the M9 with either a CCD sensor or a CMOS? No problem discussing, but it really seems a damned if you do, damned if you don't situation for Leica. As far as I can figure the only sensor that works for the severe angle of light in an M RF is a Kodak designed CCD. A CCD sensor just will not allow for super high iso's. So why all the moaning then? It is what it is and either you spring for it (I most certainly am) or you don't. The 1600 iso I tested on the M9 is great. Maybe not quite as good as the D3 but good enough for real world situations.

I'm not saying the camera is perfect - the AWB is still over the place and where's the goddamn pc socket? And they didn't have to put sapphire glass on the lcd - how about some good ol' tempered glass like Nikon instead of plastic? Yes, Leica do seemingly live in their own stratosphere at times, but damn if they aren't wonderful tools to make photographs with.

[robertwright]

Originally Posted by sean_reid

It seems like banding to me and that's how I describe it above the picture. Its one of just two high ISO pics I've found like that. Both were consecutive and Leica is clued into this (but we can't replicate it yet). Thousands of other pics at same ISO look fine in that respect.

I included the picture because I like it and it also nicely connects the noise section to the wedding section. It was published out of context on TOP (Mike Johnston just liked the pic) and now there's a little discussion going there about "discovering" the banding. That's the problem when people freely comment on a review they've never read. Kind of funny that it is a banding *illustration* in the review.

Cheers,

Sean

goodness I read it twice and missed it both times.

"discover' is a funny word. I could show you pics where I 'discovered' banding on my 5DII but who cares.

If I were to make one comment it is that unless you know for a fact that Leica is doing in camera smoothing on ISO 1250 and 2500, why write that you think that they are?

It is conjecture.

We are awash in opinion on the net (like mine!) but we really need core reporting. Basic factual reporting. Unfortunately the kind that requires phone calls and people who know saying so. Stuff like that. I love your reviews but why guess at something? Unless Leica told you in the junket, we are doing some hocus pocus at high iso. The pixel pitch may be the same but everything else about the sensor is different, the microlenses, the filter thickness. So they will not image the same way. Also the signal processing is different.

I'm sure there is a reason for what you are seeing, just we don't really know what that is. If Leica says 'secret sauce' to guard their RD fine, but at least it is from the horses you know what

I also would like a really thorough explanation of the sources and causes of image defects like noise, banding, hot pixels etc. Just for the geek-factor of it. What do engineers have to content with in terms of things like RF interference, heat, signal path length, quantization whatevers, I think it would be enlightening.

done now.