Where is my thread ?

[t024484]

Advertisement (gone after registration)

I cannot find my thread. was it removed ?

Anyhow, I investigated the 15 tables in the firmware, each 2048 bytes long, and my feeling

was right.

These tables are used for JPEG only.

When all set to 00, the RAW picture stays completely intact, but the thumbnail, as well as the full JPEG are now completely black.

This means to my feeling, that for JPEG's only 11 bits from the 14 bits are used for compression, since the tables are 2^11 in size.

 

Hans

[fefe]

Well i was trying to read it and it poofed and said to contact the administrator...

[Shootist]

Yeah it go removed. Probably because of copywiter stuff.

Thanks for all the work and I'm PM'ing you.

[LUF Admin]

Hi,

 

sorry, I had to hide Hans' thread. It's not deleted or removed.

 

Hans' work touches some legal aspects of intellectual property of Leica.

I first want to ask Leica about it before the software gets distributed.

 

I already told Hans about it.

 

Best regards from Bonn

Andreas

[diogenis]

I was expecting this to come earlier. Anyway..

Let's just hope again that something good will turn out from Hans's work. i will miss his thread though

[dpattinson]

I think Leica would do well to consider open-sourcing their firmware, or at least parts of it. It seems likely that they would benefit more from the innovation that might result, than by retaining any IP in-house.

[SJP]

Advertisement (gone after registration)

I think Leica would do well to consider open-sourcing their firmware, or at least parts of it. It seems likely that they would benefit more from the innovation that might result, than by retaining any IP in-house.

On the whole I would agree with that but there are some issues to consider:

 

1 One needs to worry about warranty. If a M8 has non-standard 'experimental' firmware installed & for some reason this caused damage to the hardware (e.g. does firmware control shutter/rewind actuation?) or a crash of the M8 operating system that needs a factory reset then no-one should expect Leica to pick up the bill for that under warranty.

Note: How can Leica know whether this was the case? Firmware upgrades can be reversed after such an event.

 

2 If a M8 has non standard firmware installed then it would not be supported by Leica anymore. So if such a camera was resold then Leica should expect/demand it to be resold only with standard factory firmware installed.

 

3 There should be a message in the LCD on startup indicating the fact that the firmware is non-standard, or not supported etc.

 

Of course many Leica owners are modifying the hardware all the time (thumbs-up, non-Leica lenses, different covers, novel baseplate designs) & firmware modding is a natural extension of that inherent to a digital camera system. I do this frequently with Windows & Palm OS so why not on my (mine not Leica's) camera. It is an interesting question in general for the digital age.

 

Anyway the cat is out of the bag & experience shows that cats resist being stuffed back into a bag with great vigor. So Leica should not consider that route IMHO.

[dpattinson]

It's important to remember that open-source does not necessarily mean free-for-all. Linux is a very good example of a piece of software that has benefited greatly from contributions, but is used in mission critical systems all over the world.

 

The key is to ensure that verifying any contribution does not break existing functionality is made as fast and cheap as possible. Usually this involves automated regression testing, which in the case of firmware would require a software simulation of the camera I expect. I presume Leica already have such a simulator, and if it were to be released along with the source, would allow contributors to test their work before submitting it.

 

A contribution model, where users with the interest and capability to make changes has greatly benefited many software products. Especially if Leica regard the firmware as non-core IP which is solely there to support their hardware, it makes sense to open it up and allow more innovation.

[mjh]

I think Leica would do well to consider open-sourcing their firmware, or at least parts of it.

It’s not going to happen. No camera vendor would do that.

 

But the modification we are arguing about here is a rather modest one. Hans didn’t modify any part of the code. He didn’t even have to know anything about the code, about how the firmware is talking to the hardware – all the really difficult stuff. Thankfully, some aspects of the M8’s image processing are controlled by look-up tables, so all Hans needed to do was to find those tables, do a little detective work to determine their internal structure and their use, and then proceed to change these tables according to his fancy. He was just changing tables of data which, as firmware hacks go, is quite straight forward.

 

Now I am obviously not a lawyer, but I don’t think Leica could prevent anyone from changing their firmware and uploading it to the camera – not effectively anyway. Anyone attempting this would do so at their own risk and would certainly void their warranties. One would have to face a dead M8 at worst, but not a prison sentence. Very reassuring, I know, but in reality, the chances of damaging the M8 this way are rather slim. Provided that all you change is data, that is.

 

So the legal issue is merely this: If you offer some version of Leica’s firmware for download, be it an original version or a hacked one, you are violating Leica’s copyright. So you should refrain from doing so, and you should also refrain from discussing your intention of doing so in a public forum, since the owners of this forum could be held responsible as well.

 

Now merely describing how, exactly, anyone could modify the firmware freely available from Leica’s website might be innocuous, but then again, I am not a lawyer. Moreover, I am sometimes wrong.

[diogenis]

I think Leica would do well to consider open-sourcing their firmware, or at least parts of it. It seems likely that they would benefit more from the innovation that might result, than by retaining any IP in-house.

 

No they shouldn't ever allow this to happen. Not just because they have Intellectual property to protect, but because this is a product with 50 years of legacy behind it.

Also, with open source systems, you lose your focus.

Leica should be the maestro here, and us its happy customers who will always ask them to fix things and all will be the recommendation to their work.

[t024484]

A small update from my side.

Jaap had been so kind, as we all know, to provide me with a nice landscape picture with a green wagon that started the whole compression thing for me.

 

Since my feeling was that a human face is probably the most sensitive and thus the least forgiving in picture degradation, I wanted to test this also.

I found someone who was so nice to allow me to experiment on a number of fantastic pictures of faces of beautiful ladies made with a Leica DMR.

He a well known professional, his name is Eric Hiss.

 

The much I tried, and I was really trying quite hard, I could not find any degradation between the original and the compressed/decompressed pictures with the SQRT,

Already the statistical information showed me that almost all deviations were within the band of +/- 1.8% that I used before as a reference.

 

So after all, as a surprise to me, Jaaps picture turns out to be more critical than a razor sharp picture of a human face. Is it because of the green, where our eyes have their highest sensitivity ?

I don't know, but all the experiments have learned me quite a bit.

I still believe that a log compression is better, but only marginal, and only under certain conditions,

 

Hans

[marknorton]

I'm really surprised Leica didn't wrap the firmware load in a checksum protected wrapper to prevent unauthorised update. If the loader can't validate the file checksum, no go.

 

You might do an exclusive-OR of all the 32 bit words in the file and scatter the resulting 32 bits throughout the file which only the hard coded loader knows how to find.

[gravastar]

I'm really surprised Leica didn't wrap the firmware load in a checksum protected wrapper to prevent unauthorised update. If the loader can't validate the file checksum, no go.

 

You might do an exclusive-OR of all the 32 bit words in the file and scatter the resulting 32 bits throughout the file which only the hard coded loader knows how to find.

 

I was more than surprised by that as well and was going to comment on it. How does the camera check the validity/non corruption of the firmware - scary.

 

 

Bob.

[ammitsboel]

I've tried the log compression and I think it looks different to the original.

If the difference is for the better only time will tell.

 

The test was not purely scientific, but still the shadows seemed somewhat clearer visible through the noise floor. The digital "noise" problems(not ISO noise) I didn't like about the M8 might have something to do with the original compression, but I really have to do some more testing to be sure.

 

Thanks to Hans for displaying the test results in the earlier posting.

[mjh]

The test was not purely scientific, but still the shadows seemed somewhat clearer visible through the noise floor.

Compared to the sqrt compression, a logarithmic compression favors the shadows – at the expense of slightly less differentiation in the highlights.

[ammitsboel]

Compared to the sqrt compression, a logarithmic compression favors the shadows – at the expense of slightly less differentiation in the highlights.

Thanks, I was looking for a better explanation to the differences.

[andybarton]

 

So after all, as a surprise to me, Jaaps picture turns out to be more critical than a razor sharp picture of a human face. Is it because of the green, where our eyes have their highest sensitivity ?

 

Could it not just be that DMR files are different from M8 ones, in that they are not compressed in-camera like M8 ones?

[t024484]

Could it not just be that DMR files are different from M8 ones, in that they are not compressed in-camera like M8 ones?

 

Hi Andy

 

I just investigated the effect of compressing and expanding, as an isolated subject.

The M8 is not a DMR, so apart from compressing the data, there are many things that we don't know about what and how data is being processed, and to what happens after compression..

What strikes me, although it may be suggestion, is that the M8 with log compression produces a sharper and less noisy pictures, especially in darker areas. Other people trying log compression more or less came to the same careful conclusion.

This effect however cannot be reproduced with DMR pictures.

So all I can say, the experiments with the DMR were very instructive, and have no further use in better understanding the M8.

I include a crop of one of Eric Hiss pictures, to give you an impression how detailed these pictures were. Could this have been a M8 picture also ?

If not, it is again the story of apples and pears.

 

Hans

 

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!

[andybarton]

I don't know whether that could be an M8 photograph

 

To be honest technically breaking down DMR files is a moot exercise these days

[Advertisement]

Hallo,

Look here: Leica M

[t024484]

Thanks, I was looking for a better explanation to the differences.

 

Here is a graphical presentation of the two different compression schemes, illustrating MJH's statement.

Horizontally is the uncompressed digitized sensor value being proportional to the luminance , and vertically the value after compression.

The slope of the graph tells you the sensitivity to a change in luminance.

From 5000 to 16000 (bright to very bright) , the sqrt has a higher resolution, below 5000 the Log is equal or better.

As from a luminance of 50 (very dark) and upwards, the log has an almost steady slope.

 

Hans

 

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!