Leica news: LFI #1/2007

[papimuzo]

The journal LFI nbr. 1/2007 has 13 pages related to the glitches of the M8, well described with examples, as well as the proposed solutions. This shows to me the openess of this company.
Have a good time all, particularly those who own a M8

[scott kirkpatrick]

Can it be viewed online?

scott

[papimuzo]

Only an introduction at LFI International
Regards.

[carstenw]

LFI is independent, I believe, although they receive a lot of support from Leica.

[scott kirkpatrick]

Originally Posted by papimuzo

Only an introduction at LFI International
Regards.

That's my impression, as well. But it promises to address the question of how the "16 bit" data is transformed into the 8 bit DNG files as well as discuss the engineering choices that led up to the IR sensitivity and perhaps how the bleeding occurs.

I have now tried on both Firefox and IE to subscribe, but my online shopping experience with LFI is, to say the least, disappointing. Once I wound up with four subscriptions in my cart and other times the cart had emptied itself while I typed in my address and payment details. I'd like to see the whole article, so i will keep trying. If there is a trick to this, I will report it here.

scott

[sdai]

Check your cookie settings, Scott ... I'm also interested in what they have to say about the 8-bit vs. 16-bit ... good for them, at least, here's a great company who is listening.

[carstenw]

Yeah, their website is a bit difficult. I just bought the magazine and read most of the article (in German), and they do discuss fully, with examples, both magenta colours and banding. The banding was a hardware bug and will be fixed, as we knew. The reason for the IR problem is as follows, again, known to us here:

There are two kinds of IR filters, whether for lenses or sensors: coloured filters and the hot-mirror style filter. The latter is highly dependent on the angle of incidence, and hence had to be rejected for the sensor, not only because of colour shifts with lenses with rear elements at different distances and which project light rays onto the sensor at different angles. This could be solved in firmware, after all. A nasty side-effect is that all lenses would have to be coded, which was quite a severe restriction given the large number of lenses which were not planned to be coded, or which come from other manufacturers.

So the decision fell on a coloured filter, by necessity. The problem was now the thickness of the filter. With a thin filter, only a certain amount of infrared could be filtered out. With a thicker filter, more could be filtered out, but then there were internal reflections, (as can be seen on the R-D1,) which was deemed unacceptable. Some tests were made and the M8 was determined to be more sensitive to infrared than usual cameras, but somehow the very bad cases did not appear early, and it was deemed acceptable, given the possibility of adding a filter on the lens later. Apparently most testers who saw the problem either assumed that the white balance algorithm still needed tuning, or that Leica would simply fix it.

Once the potential severity of the problem was discovered, only after initial shipments, Leica went to work under a lot of pressure, and the two-free-filter deal was the only way to "fix" it, as previous analysis had been accurate. I suppose we simply need to accept it as the price for working with such a camera, just like many medium format photographers do. Perhaps increased research will find ways of more accurately cutting more infrared light with a thin filter, and then a better solution can be implemented.

By the way, the article names Jamie Roberts for his work on profiles! Congratulations, Jamie!

The next point: 8-bit vs. 16-bit. When they heard of this, they managed to get ahold of an early test M8 without this encoding (with resultant slow write-out and 20MB files), and compared image quality of the two M8s. They did a series of photographs of high-key subjects, since the bright areas "suffer" most from this encoding. They were unable to see any difference with the eye, by analysis, or in prints.

Anyway, the article goes on and is clearly very carefully written. The company is independent of Leica, I believe (IDC Corporate Publishing GmbH), but there is clearly a very close relationship. They did not want to overreact, but especially not to underplay the problem. I would recommend that anyone interested buy the magazine and read the article. The magazine is great, by the way. After reading a few issues (even before becoming interested in the M8), I went online and tracked down back issues almost completely into the mid-90s and many issues even back to the 80s and a few in the 70s. I hope eventually to have a relatively complete collection back to some cut-off date I haven't decided yet. The old ones are too expensive and rare, and are in the hands of collectors.

[marknorton]

Thanks Carsten for the synopsis, I agree LFI is a good magazine and it sounds like they have taken the right approach to documenting the problems and giving some background. It's good too not to have to peer through the layers of Leica PR-speak.

I see they are reviewing the WATE in the same issue, what are they saying about it?

[sdai]

Originally Posted by carstenw

The next point: 8-bit vs. 16-bit. When they heard of this, they managed to get ahold of an early test M8 without this encoding (with resultant slow write-out and 20MB files), and compared image quality of the two M8s. They did a series of photographs of high-key subjects, since the bright areas "suffer" most from this encoding. They were unable to see any difference with the eye, by analysis, or in prints.

Comparing to the magenta cast problem, this seems to be a non-issue ... did they mention anything about Jenoptik? My gut feel is that whole electronics part of the camera was done by that company ... not just its firmware.

Thanks a lot for the preview, Carsten.

[carstenw]

I can't comment on the WATE or Jenoptik yet, although I expect that the lead time for the latter issue means that it won't be in there until the next release. I have only read the M8 article, as I am still at work

[sdai]

It's simply hard to understand that Leica let one company (or just themselves) design the DSP but hire another company to program it.

Thanks again, Carsten.

[Peter D Lux 2]

Thanks Marcel for the link. And thanks to Carsten for giving a preview of the M8 article. I just bought my first LFI a couple of days ago, and it seems to me that the magazine is a very interesting read. I'll subscribe by old fashioned paper mail..

Carsten, do you think that when you get your camera back, you will ever find the time to read all the old issues?

Cheers, Peter

[marknorton]

Originally Posted by sdai

It's simply hard to understand that Leica let one company (or just themselves) design the DSP but hire another company to program it.

Leica were starting a long way back and even now, it's not clear how much of the critical knowledge of electronics and DSP they have built in house. Sub-contracting may be a short term convenience, but if Leica are serious about this digital photography thing, they need to develop their own unique store of IP rather than warmed-over IP from someone else which, by definition, is Old News.

The quality of Leica's lenses is based on expertise achieved over decades with optical design and fabrication. They need to do the same with DSP.

[loretdem]

Originally Posted by carstenw

So the decision fell on a coloured filter, by necessity. The problem was now the thickness of the filter. With a thin filter, only a certain amount of infrared could be filtered out. With a thicker filter, more could be filtered out, but then there were internal reflections, (as can be seen on the R-D1,) which was deemed unacceptable.

Although I agree with Leica regarding the M8's IR sensitivity and their filter solution, I must mention that although the R-D1 has its own problems, unacceptable internal reflections is not one of them. Please enlighten me if I'm mistaken.

-Carlos

[chrism]

Originally Posted by marknorton

... I agree LFI is a good magazine and it sounds like they have taken the right approach to documenting the problems and giving some background. It's good too not to have to peer through the layers of Leica PR-speak.

But the English version is rather odd to read - it has been translated out of German, but not into English, if you know what I mean!

Chris

[elansprint72]

Originally Posted by chrism

But the English version is rather odd to read - it has been translated out of German, but not into English, if you know what I mean!

Chris

Chris,
I know exactly what you mean; Porsche used to drive me mad by sending a superb quality magazine which appeared to have been translated by a machine. I offered to do them a gratis translation into "real English" if they could give me a week turn-around time; they replied, rather stiffly, that they had no need of my services because they had an excellent technical translator called Graf Berndt von Somethingothic! It's a culture thang!

Cheers,

Pete.

[newyorkone]

Thanks Carsten...

The R-D1 reflection issue is real and someone had posted example photos a while ago either here or on RFF. Could easily be seen for points of light at certain angles.

I think the most severe magenta examples that we saw were bad samples coupled with bad firmware and bad color profiling in Capture 1. That's why it wasn't a red flag during testing. The problem was there but much more subtle. I seriously doubt that Leica would have shipped the camera if they had seen purple/fuchsia faces and severe magenta cast across the entire frame like we have seen from the worst examples. It defies logic and it wasn't some kind of conspiracy. That's nonsense.

The few photos that I have seen from the new batch of M8's look a lot better straight out of the camera without IR filters and with some color profile tweaking should get things very close to where we need to be for general shooting and people can use IR filters for more critical work.

[marknorton]

German translated into English is generally bad news and comes across as all formal and ambiguous. Carsten is an exception - don't know if he is English or German but his English is completely fluent.

[jrc]

Originally Posted by marknorton

Leica were starting a long way back and even now, it's not clear how much of the critical knowledge of electronics and DSP they have built in house. Sub-contracting may be a short term convenience, but if Leica are serious about this digital photography thing, they need to develop their own unique store of IP rather than warmed-over IP from someone else which, by definition, is Old News.

The quality of Leica's lenses is based on expertise achieved over decades with optical design and fabrication. They need to do the same with DSP.

But can't you just buy the expertise? Get the number 4 guy at Nikon and number 4 guy at Canon and write an iron-clad contract that pays them x no matter what happens? Or something like that? I have no idea how this stuff works, being self-employed forever, but I do know that high-end computer talent seems to metastasize out of Silicon Valley...it doesn't work this way with cameras?

JC

[marknorton]

Companies like IBM and Canon (Numbers 1 and 2 in terms of patents granted) exist to use and licence their IP. Actually making computers/cameras/??? is a messy business, far rather licence your inventions and let someone else take care of delivering working product to the Great Unwashed.

Leica cannot survive using someone else's IP, they need to develop their own. They have it in spades in lens design, optics and so on and their latest business numbers suggest good profits are to be made from licencing that IP. In terms of making cameras, they need to increase the proportion of in-house IP to increase their profits. Subcontracting out to third parties is a short term expedient but it will not build a value base for future expansion.