R10 and Full-Frame Sensor

[imagist]

Advertisement (gone after registration)

I feel a bit leery about the extremely high ISO performance which is coming from CMOS sensors. I hope that the R10 achieves more CCD real estate and maintains bit depth; and I don't want to give up extreme detail rendering to achieve high ISO ratings. Right now my DMR and M8 give me images with unique character; somewhere between 26x36 CMOS and and medium format CCD's - I'd like this unique territory to remain and improve in Leica land.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[ptomsu]

Sorry Peter but that is truly hilarious! Did you really have a DMR?

 

Yes I did and I know what I am talking about. Do you own a DMR? Do you own a M8? And can you thus tell the difference?

 

The M8 files are not only as good as the DMR files, they look even better in my eyes.

 

And it all depends to a high degree on the post processing SW you use as you might be aware. I can even easily trim NEF files out of the D2X to look the same as files out of a DMR, the only thing it needs more work. But having said that, the DMR look is not always best, so there are situation I really liked it, but also other I preferred more the D2X.

 

But what is really superior to both are the files out of the M8.

[andybarton]

If you think that M8 images look better than DMR ones, Peter, that's fine. It's how you perceive your work that counts to you.

 

As you can see from Conrad's post above (and he owns a DMR, btw), there are others who disagree. No problem with that.

[marknorton]

Bit depth is a bit like mega-pixels, the more you've got, the better it's got to be, right?

 

The Kodak specification for the DMR sensor specifies 67 dB dynamic range which corresponds to about 11 bits, so measuring the signal to 16 bit accuracy makes no sense - you're measuring noise. We do not know what analog-to-digital converter was used in the DMR but measuring to 16 bit accuracy over the required number of samples in the time available in the electrically noisy environment of a digital camera is a tough thing to do and I don't think the DMR does it for one minute.

 

Even if you use an A-D converter which can theoretically measure to 16 bit, whether you actually achieve that depends critically on circuit layout and screening. I used to develop digital audio stuff where we were measuring to 16 bit accuracy 48000 times a second and that was tough enough, nevermind the 10M times a second in the DMR. At that sort of conversion rate, what you think of an electrical ground reference is all over the place.

 

What the DMR does do is keep the measured data as 16 bit binary numbers rather than the lossy compression used on the M8 which was done in the name of speed and storage (less data to write to the card) but you pay the price in terms of frame rate. Never used a DMR but my understanding is that it trundles along at about 1 frame a second, a real slug compared to the 9 fps 14 bit I get on my D3.

 

I expect an R10 will do the same, measure the data to whatever bit depth makes sense for the signal to noise available - but it's unlikely to be true 16 bit - and keep the data as 16 bit instead of the nasty compression used in the M8.

[EH21]

Peter,

I've seen your posts on this subject on a couple different boards and feel compelled to post a response. I'm another one who thinks the DMR files are the equal of the M8 or even better under certain circumstance. The M8 files might be better than the DMR at ISO 800 and higher, but not at ISO 100. Some of the M lenses may be a tiny bit better than their R counter parts so perhaps that's what you've noticed? Also you can shoot the M camera at slower shutter speeds than the R so if you were more accustomed to shooting a rangefinder you might not have shot the DMR at a fast enough shutter speed. Of course how one handles the files in post can also make a big difference, so if you've used one RAW processor for the DMR and a different one for the M8 you could have easily had different experiences - especially if you used ACR which didn't do the DMR files any justice. If you truly shot both cameras and have some files saved from each, I suggest you go back and pull some of those DMR files and run them through flexcolor, RAW developer or C1. The DMR is a very high quality piece of gear but requires some experience of the user to get the most of it.

Eric

[masjah]

Mark

 

I'm an unreconstructed film person here, so please can you put me right! When the sensor spec says 67dB dynamic range, that, as I understand it, is the (logarithmic) light intensity ratio between black (noise level) and white (saturation). But at this stage (i.e. on the sensor) isn't this an analogue figure, with a continuous range of intensities at each pixel. Doesn't a higher bit depth then enable you to digitise this continuous range with more, smaller, steps between black and white, that is, reduce what I think is called "quantisation error"?

 

I'd be delighted to be put right where I've got this wrong - I'm asking this naive question out of genuine puzzlement!

[ptomsu]

Advertisement (gone after registration)

Bit depth is a bit like mega-pixels, the more you've got, the better it's got to be, right?

 

The Kodak specification for the DMR sensor specifies 67 dB dynamic range which corresponds to about 11 bits, so measuring the signal to 16 bit accuracy makes no sense - you're measuring noise. We do not know what analog-to-digital converter was used in the DMR but measuring to 16 bit accuracy over the required number of samples in the time available in the electrically noisy environment of a digital camera is a tough thing to do and I don't think the DMR does it for one minute.

 

Even if you use an A-D converter which can theoretically measure to 16 bit, whether you actually achieve that depends critically on circuit layout and screening. I used to develop digital audio stuff where we were measuring to 16 bit accuracy 48000 times a second and that was tough enough, nevermind the 10M times a second in the DMR. At that sort of conversion rate, what you think of an electrical ground reference is all over the place.

 

What the DMR does do is keep the measured data as 16 bit binary numbers rather than the lossy compression used on the M8 which was done in the name of speed and storage (less data to write to the card) but you pay the price in terms of frame rate. Never used a DMR but my understanding is that it trundles along at about 1 frame a second, a real slug compared to the 9 fps 14 bit I get on my D3.

 

I expect an R10 will do the same, measure the data to whatever bit depth makes sense for the signal to noise available - but it's unlikely to be true 16 bit - and keep the data as 16 bit instead of the nasty compression used in the M8.

 

Mark,

 

I cannot agree more! And I only can reiterate, I am getting much better results from the M8, also at ISO 100.

[Jamie Roberts]

{snipped}What the DMR does do is keep the measured data as 16 bit binary numbers rather than the lossy compression used on the M8 which was done in the name of speed and storage (less data to write to the card) but you pay the price in terms of frame rate. Never used a DMR but my understanding is that it trundles along at about 1 frame a second, a real slug compared to the 9 fps 14 bit I get on my D3.

{snipped}.

 

Mark--I don't think the buffer is that bad. It's not 1fps; more like 2.5, which

is a lot faster. I'm happy to be wrong; can't actually go and check right now (which I'd usually do before I post).

 

But I did want to say that 9fps is simply wayyyy overkill for me. I had a 1d2 at 8fps, and sold it; it's replacement (the 1ds2) was "only" 5 fps. Other than certain sporting situations, I just don't get the veritable machine-gun abiliity of some cameras these days.

[mgcd]

Mark,

 

I cannot agree more! And I only can reiterate, I am getting much better results from the M8, also at ISO 100.

 

 

Utter rubbish. As you now know, I own a DMR and I have the M8 on long term loan. Mark you have no experience with the DMR as you stated yourself, and it trundles along at 2-2.5 fps by the way... As for image quality, the best of the DMR trumps the M8 at every turn.

 

Cheers,

[ptomsu]

Peter,

I've seen your posts on this subject on a couple different boards and feel compelled to post a response. I'm another one who thinks the DMR files are the equal of the M8 or even better under certain circumstance. The M8 files might be better than the DMR at ISO 800 and higher, but not at ISO 100. Some of the M lenses may be a tiny bit better than their R counter parts so perhaps that's what you've noticed? Also you can shoot the M camera at slower shutter speeds than the R so if you were more accustomed to shooting a rangefinder you might not have shot the DMR at a fast enough shutter speed. Of course how one handles the files in post can also make a big difference, so if you've used one RAW processor for the DMR and a different one for the M8 you could have easily had different experiences - especially if you used ACR which didn't do the DMR files any justice. If you truly shot both cameras and have some files saved from each, I suggest you go back and pull some of those DMR files and run them through flexcolor, RAW developer or C1. The DMR is a very high quality piece of gear but requires some experience of the user to get the most of it.

Eric

 

Eric,

 

sure that there are always subjective ways to see results. But as I had both (DMR sold) and M8 now, I think I can in a pretty objective way say what I see (have seen).

 

I only use (used) LR 1.3, C1 Pro and Aperture. I have used C1 Pro for both DMR and M8 and stopped using it, because I finally did not get much difference in RAW compared to LR - and Aperture today (talking about Aperture 2, which IMHO is even much better than C1 4 Pro (which I also tried). Please note I am only talking about the RAW engines of these SW tools, nothing else. I must admit I never used Flexcolor, maybe that this is the last final bit in difference for the DMR.

 

But why argue about this, what I say I are my impressions and these impressions also lead to my decisions finally and this is to sell the DMR and keep the M8 and go for a D3.

 

But please also note that these decisions did not only depend on the IQ of the systems, but on many other facts, one very important the future I see for a specific system - and I do not se any future for the R system and for the DMR other than you will be able to maybe reuse the R lenses on a future R10 in an inferior manual mode only. Which I am not willing to do. But also here - many users have no issue with it and this is where the R glass and also a DMR makes sense.

 

For me it did not. And as I wrote in the answer to Mark, I also agree with him that the D3 is currently one of the best tools in terms of IQ and color accuracy (cll it bit depth) on the market, not only for high ISO, but at low ISO as well. We only can hope the R10 will come close to that in IQ.

[ptomsu]

Utter rubbish. As you now know, I own a DMR and I have the M8 on long term loan. Mark you have no experience with the DMR as you stated yourself, and it trundles along at 2-2.5 fps by the way... As with image quality, the best of the DMR trumps the M8 at every turn.

 

Cheers,

 

Well, if you found this out, just be happy with your choice, what sense does it make to try convince others, who have built up a different opinion???

 

And BTW, I have played extensively with the D3 (will receive mine unfortunately in some 4 weeks) but what I have seen this beast outperforms the DMR easily in IQ, not only at high ISO. The main reason I have the M8 and use it is because of compactness, some really unique and great primes and my general love of the M system. But otherwise there would be no reason as well if you compare to the IQ results of a D3.

[andybarton]

Eric,

 

sure that there are always subjective ways to see results. But as I had both (DMR sold) and M8 now, I think I can in a pretty objective way say what I see (have seen).

 

I only use (used) LR 1.3, C1 Pro and Aperture. I have used C1 Pro for both DMR and M8 and stopped using it, because I finally did not get much difference in RAW compared to LR - and Aperture today (talking about Aperture 2, which IMHO is even much better than C1 4 Pro (which I also tried). Please note I am only talking about the RAW engines of these SW tools, nothing else. I must admit I never used Flexcolor, maybe that this is the last final bit in difference for the DMR.

 

Aperture does not work with DMR files

[ptomsu]

Aperture does not work with DMR files

 

It does if you convert them to DNG in PSCS3 for example. Then you can import them in Aperture. Of course not the recommended way. But since I do not use the DMR any longer, I do not have any issues with that.

 

Besides that I must say I am overwhelmed by the Aperture 2 RAW engine, and I thought that I would not want o change away from LR any more. But now I am seriously thinking to do that

[mgcd]

It does if you convert them to DNG in PSCS3 for example. Then you can import them in Aperture. Of course not the recommended way. But since I do not use the DMR any longer, I do not have any issues with that.

 

Besides that I must say I am overwhelmed by the Aperture 2 RAW engine, and I thought that I would not want o change away from LR any more. But now I am seriously thinking to do that

 

DMR files are not supported in Aperture. Period.

[andybarton]

It does if you convert them to DNG in PSCS3 for example. Then you can import them in Aperture. Of course not the recommended way. But since I do not use the DMR any longer, I do not have any issues with that.

 

That doesn't work. The only way to get Aperture to work with them is use Sandy's software to downgrade them to M8 depth files.

[marknorton]

Mark

 

I'm an unreconstructed film person here, so please can you put me right! When the sensor spec says 67dB dynamic range, that, as I understand it, is the (logarithmic) light intensity ratio between black (noise level) and white (saturation). But at this stage (i.e. on the sensor) isn't this an analogue figure, with a continuous range of intensities at each pixel. Doesn't a higher bit depth then enable you to digitise this continuous range with more, smaller, steps between black and white, that is, reduce what I think is called "quantisation error"?

 

I'd be delighted to be put right where I've got this wrong - I'm asking this naive question out of genuine puzzlement!

 

John, you can of course measure the signal you get out of the sensor to whatever accuracy you like and that measurement will have a quantisation error - the difference between the true signal and the measured value but once the accuracy of your measurement is such that the noise level is around 1/2 LSB, there's no point in going any further - you are just measuring noise.

 

In digital audio, you actually add a small amount of noise with a particular probability density function which leads to rounding of the quantisation error and is subjectively is less harsh to listen to.

 

If the sensor has a dynamic range of 67dB, a 12 bit A-D converter will do the job. Those 12 bit samples can then be retained as 16 bit values rather than scrunching them down to 8 bit values.

[marknorton]

Mark--I don't think the buffer is that bad. It's not 1fps; more like 2.5, which is a lot faster. I'm happy to be wrong; can't actually go and check right now (which I'd usually do before I post)

 

As I said, I have no experience of the DMR but the Kodak specification of the sensor used in it says the readout time is 1247mS which suggests a frame rate of less than one a second. If the DMR as implemented does better, it may be the actual sensor used in the DMR improves on the published Kodak specification.

 

The readout time of the M8 is around 250mS, partly achieved using 2 output channels and higher clock rates; the Nikon D3 sensor has 12 output channels.

[ptomsu]

DMR files are not supported in Aperture. Period.

 

You can convert the DMR DNGs in Photoshop to DNG and then Aperture 2 can open these.

 

Apple - Aperture - Technical Specifications - Raw Support

 

"Aperture works with most DNG files generated from cameras that support this format and DNG files generated by the Adobe DNG Converter with the “Convert to Linear Image” option turned off."

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[andybarton]

You can convert the DMR DNGs in Photoshop to DNG and then Aperture 2 can open these.

 

Apple - Aperture - Technical Specifications - Raw Support

 

"Aperture works with most DNG files generated from cameras that support this format and DNG files generated by the Adobe DNG Converter with the “Convert to Linear Image” option turned off."

 

Note the fourth word... Have you tried this yourself? I have... I doesn't work. We went through all this when Aperture 2 first came out.

 

http://www.l-camera-forum.com/leica-forum/digital-forum/45757-aperture-2-out.html

 

I do not knw where you generate your wisdom from???? Is is the same source where you know that DMR files look better than M8 files ??? :D

 

Yes. Personal experience. I generate this wisdom by trying it for myself and engaging the help of Eoin and Sandy. Only by using Sandy's DMR2M8 converter can Aperture be fooled into reading DMR files.

[mgcd]

You can convert the DMR DNGs in Photoshop to DNG and then Aperture 2 can open these.

 

Apple - Aperture - Technical Specifications - Raw Support

 

"Aperture works with most DNG files generated from cameras that support this format and DNG files generated by the Adobe DNG Converter with the “Convert to Linear Image” option turned off."

 

Peter this is my problem with many of your posts, did you try this? Had you tried, surely you would have taken note of a notice of unsupported format... In my case, as I do this for a living, I tested the latest Aperture version with the DMR. Results: unsupported. Period.