Bug in M10 6-bit coding

[UliWer]

Advertisement (gone after registration)

vor 5 Stunden schrieb verbivore:

No, not true. With a coded lens, the camera would recognize that lens. With an R-adapter, the camera would recognize that it was an R-adapter (unlike what it does now). With a macro-adapter, the camera would recognize that was a macro-adapter (unlike what it does now). With an uncoded lens, the camera would recognize that it was an uncoded lens. That is, the camera would register the lens correctly, instead of incorrectly.

Of course you are right.

The purpose of lens detection is to show what is attached to the camera - even if it is just an adapter. 

Very important for lens usage and nice to look at in the EXIF. 

 

[Advertisement]

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

[lct]

Just checked on my M240. Same exif data as the CL's except that in auto lens mode, the default focal length is not 90mm (CL) but 0mm (M240). I prefer the CL from this standpoint and i suspect the M10 works the same but i have no experience with the latter.

Manual lens profile (35/2 asph):
Camera Model Name            : LEICA M (Typ 240)
Focal Length                 : 35.0 mm
Lens Info                    : 35mm f/2-16
Lens Model                   : Summicron-M 1:2/35 ASPH

Auto lens profile:
Camera Model Name            : LEICA M (Typ 240)
Focal Length                 : 0.0 mm
Lens Info                    : 0mm f/0
Lens Model                   : Macro-Adapter-M

[jaapv]

14 hours ago, verbivore said:

I

The entire 6-bit coding system is a charming but anachronistic kluge. (Even providing a 7th bit would have doubled the number of codable lenses from an inadequate 64 to a reasonable 128.) If that is Leica's concession to the 21st century, they should at least implement it in a way that's useful to their loyal and patient customers. 

Sorry, it is seven bits on M series cameras, and the seventh bit even has three values. That is the frameline selector. It was implemented that way to accommodate the MATE. That will result in 190 lenses.

The reason Leica tries to limit the use to 64 are the SL and CL, which appeared long after the system was designed. I hope that you are not suggesting that the M mount should be redesigned necessitating a recoding of all existing lenses.

[pico]

15 hours ago, verbivore said:

the image is altered in the camera according to the lens code (vignetting and distortion), unless the user remembers to undo that correction in post-processing.

Is there an option in PP to show a version without the in-camera applied correction?

 

[pico]

4 hours ago, jaapv said:

Sorry, it is seven bits on M series cameras, and the seventh bit even has three values.

Oh, as a programmer what I would have given to make a binary represent three values. Or is Leica using qubits?

[jaapv]

45 minutes ago, pico said:

Is there an option in PP to show a version without the in-camera applied correction?

 

AFAIK those corrections are applied before the DNG is written, unlike the Q where they are contained in a separate file. 

[lct]

Advertisement (gone after registration)

1 hour ago, pico said:

Is there an option in PP to show a version without the in-camera applied correction?

Some softwares allow to disable lens profiles if needed. I can only quote Iridient Developer i'm using from time to time. Shows some effect on distortion when using ultra wides (SE 21/3.4 asph for inst.) or lenses with little optical corrections (7art 35/2). Differences are not huge though, in my experience at least.

[farnz]

2 hours ago, pico said:

Oh, as a programmer what I would have given to make a binary represent three values.

Oh but you always did, Pico: 1, 0, and indistinct. 😊  

As a long distance optical fibre systems design engineer I deal with this all the time owing to natural attenuation of light in the fibre presenting three outcomes at the far end: 'pulse detected', no 'pulse detected', and 'ooo there might be a pulse but equally there might not I can't tell for sure so I'll guess.  Drat, too late'.

Pete.

[UliWer]

vor 7 Stunden schrieb lct:

Just checked on my M240. Same exif data as the CL's except that in auto lens mode, the default focal length is not 90mm (CL) but 0mm (M240). I prefer the CL from this standpoint and i suspect the M10 works the same but i have no experience with the latter.

The M10 with lens detection in auto mode shows the last lens you have chosen manually when you attach the adapter. .

The M10 differs from former M-models as in "auto" mode it "overrides" any other detection, if the camera recognizes a certain 6-bit code of an M-lens. If there is no code the last manually chosen lens is detected. This was often described as useful in former threads here, and I agree, if you just use one uncoded lens. You choose the lens detection of your single uncoded lens and use "auto" mode. Your coded lenses will activate their proper codes; if you switch to your uncoded lens, the manual detection is also right. Of course with more than one uncoded lenses this does not work, since a binary logic cannot offer you a third or fourth option besides 0. last manually chosen code or 1. lens detection for a 6-bit code.

As the adapters for Macro or R-lenses have a certain 6-bit code, you expect that this code triggers something. Though it doesn't and the camera reacts as if the adapters were uncoded - so the M10 shows your last manually chosen lens.

To avoid the "frustration" that coded adapters have no effect, the camera would need a new function, which I'd call "prefered lens". When the camera  detects the adapter's code it switches to the code you entered as "prefered lens" e.g. the Makro-Elmar-M or any specific R-lens. Of course this would not help if somebody uses more than one "prefered" lens, though we might have some new interesting threads here.

 

 

[UliWer]

vor 3 Stunden schrieb pico:

Is there an option in PP to show a version without the in-camera applied correction?

 

You just have to make two photos: one with and one without lens detection with the same lens. The menu allows you to switch lens detection off. I'd like to show two examples, though we have a strong thunderstorm at the moment and I prefer not to go out until it is over. 

[UliWer]

In spite of thunderstorms there is a "laboratory" way to show the effects of lens detection. Two photos with the M10 and the Super-Elmar 1:3,4/21mm Asph. with an opaque object immediatley in front of the lens hood and an electric lamp. The results show a gross exaggeration of the results, in reality the differences are much more subtle. Though with the M9 the result for lens detection off would be horrible.

With lens-detection:

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

Lens detection off:

 

 

 

 

[verbivore]

UliWer and others: Yes, this is the exact test one should do. Next step: Bring the "With Lens Detection" image into Lightroom or Adobe Camera Raw; go to the Lens Corrections tab (indicated by the little drawing of a lens cross-section); go to the Profile tab; un-check Enable Profile Corrections. If the image is now identical to the one with "Lens Detection Off," that means the camera does not apply corrections to the image in the DNG file, but just adds metadata to the file, which post-processing applies to "undo" what it thinks are the lens's distortions. If the two images are still different, it means that the camera has applied the corrections to the pixels in the image itself -- which would mean that the bug I documented would be truly damaging, since it would distort the original raw files when the wrong lens was coded, with no way of recovering the true image.

Last time I tried this, I recall that the former situation (corrections applied through metadata, reversible in post-processing) was the case. I'd be curious what you find. 

I'm not sure that your "preferred lens" option is needed for the R-adapter or for uncoded lens -- the camera should simply remember the last R lens or M lens that was set by the user, and apply it. Yes, you'd need something like this feature to distinguish the manually set M lens used with the Macro-Adapter from the uncoded one.

To the various bit-counters: if Leica were to make full use of every combination of painted pits and frameline indicators, it would be a 7.585-bit coding system: 6 bits for the six pits, each black or white, plus 1.585 (base-2 log of 3) bits for the choice among three framelines (assuming all combinations are equally likely). 

Edited August 27, 2019 by verbivore
eliminate an ambiguity

[UliWer]

The profile "correction" in Lightroom only addresses distortion, which the in-camera lens correction for the M does not change (other Leica cameras apply quite a lot of distortion correction in-camera). You may change vignetting with oher tools in Lightroom, but you cannot effectively correct colour shift, which the in camera correction does. You can see this in my example with lens detection switched off which shows some cyan drift on the right side (much exaggerated by using an opaque object in front of the lens). With an M9 you would also see strong red shift on the left ("Italian flag" syndrom). 

vor 4 Stunden schrieb verbivore:

I'm not sure that your "preferred lens" option is needed for the R-adapter or for uncoded lens -- the camera should simply remember the last R lens or M lens that was set by the user, and apply it. Yes, you'd need something like this feature to distinguish the manually set M lens used with the Macro-Adapter from the uncoded one.

The camera remembers the last lens set by the user and applies the coding when you use the R-to-M-adapter. You criticized this behaviour in your first posting in this thread.

My suggestion for "preferred" lens was only meant for adapter usage: coding of the adapter+preferred setting.

[lct]

11 hours ago, UliWer said:

The M10 with lens detection in auto mode shows the last lens you have chosen manually when you attach the adapter. .

The M10 differs from former M-models as in "auto" mode it "overrides" any other detection, if the camera recognizes a certain 6-bit code of an M-lens. If there is no code the last manually chosen lens is detected. This was often described as useful in former threads here, and I agree, if you just use one uncoded lens. You choose the lens detection of your single uncoded lens and use "auto" mode. Your coded lenses will activate their proper codes; if you switch to your uncoded lens, the manual detection is also right. Of course with more than one uncoded lenses this does not work, since a binary logic cannot offer you a third or fourth option besides 0. last manually chosen code or 1. lens detection for a 6-bit code.

As the adapters for Macro or R-lenses have a certain 6-bit code, you expect that this code triggers something. Though it doesn't and the camera reacts as if the adapters were uncoded - so the M10 shows your last manually chosen lens.

To avoid the "frustration" that coded adapters have no effect, the camera would need a new function, which I'd call "prefered lens". When the camera  detects the adapter's code it switches to the code you entered as "prefered lens" e.g. the Makro-Elmar-M or any specific R-lens. Of course this would not help if somebody uses more than one "prefered" lens, though we might have some new interesting threads here.

Interesting thank you but i can't seem to see a difference between coded and uncoded lenses as far as adapters are concerned. In both cases, the camera can only read the code of the adapter. When i choose manually a lens profile, say 35/2 for instance, i set the camera in manual lens mode for that so it is not in auto lens mode anymore as a matter of fact. Next time i restart the camera, it will stay in manual lens mode and keep that same 35/2 lens profile consequently. If i want the camera to revert in auto lens mode, i just have to order it that's all. AFAIC i simply choose an user profile on my CL to do that. All my user profiles including the auto lens mode, the simple fact to choose one of those user profiles suffices to set the camera in auto lens mode and to remove automatically the 35/2 lens profile this way. Takes a couple of seconds and i don't even have to chimp for that. Cannot you do the same with the M10? Just curious.

[mdemeyer]

As someone with only one regularly-used R lens (an 80mm Summilux), I would love to see the coded R adapter behave as suggested here. Seems like a trivial firmware change to associate the last selected R lens with the adapter. 

Come on, Leica.  Fix (or enhance, if you prefer) it!