M8 technical exploration, Luminous Landscape

[john_brewton]

Interesting article on M8 has just been posted on Luminous Landscape. Just

passing this along.

[paul hart]

Conclusion: 'Rangefinder cameras are a different thing'

 

Hmmm, very profound. I would have read the rest of the article, but the white on black hurts

my eyes and might impair my ability to take photographs with my different thing.

[cpj]

You missed the concluding sentence. Not only are rangefinder cameras a "different thing,"

but "the rangefinder user is in fact a different photographer."

 

Vive la difference!

[karl_graf]

And what if the rangefinder user also uses other types of cameras in equal amounts?

[bobatkins]

Multiple Personality Disorder?

 

Well Rounded Photographer?

 

Take your pick...

[g-man1]

quote: "Leica M8 has adopted several compromises. They must to be known, understood and accepted."

 

How about Leica forget about compatability, reinvent itself, join up with a smart Japanese company, and start over again and invent a camera with a new sensor, new interchangeable, autofocussing, image stabilised lenses -- oh, wait....

[karl_graf]

"Leica M8 has adopted several compromises. They must to be known, understood and accepted."

 

There is always the G2. A few shops still have them in stock. Oh, wait. The M8 is the only digital rangefinder M on the market, unless there is an RD2 in the works.

[anthony_brookes5]

Gary. Of course that is the complete opposite of the truth. Leica users accept no compromises. Only a Leica will do. Every camera is a compromise so the statement is almost valueless. My Ikoflex is a compromise as is my M6. Silly comment from Luminous Landscape.

[stan_belyaev]

I was amaized with this article. What a nice way to deal with a problematic camera. Why should one overcome/correct the problems? Let's accept them, moreover let's make a controvertial camera a new "gold standard" and urge people to enjoy "benifits" (limitations) of it. (like black and white "extended spectrum" -visible and infrared photography). It looks like the other manufacturers are far behind in this area and it will take some time for them to catch up. Does it mean that all digital cameras manufactured to date are less capable and need to be upgraded? )

 

I can not understand why the authors keep referring to Put's atricles, which are methodologically incorrect.

How can one compare 75mm lenses with full frmame and 1.33X sensors at the same distance to the object? Can not they realize that in this setting Leica magnifies the object by 33%?

Why would the resolution of the Leica's sensor be 73.5lp/mm?! It is only possible if one assumes that 1lp=2 pixels.

It is generaly accepted that 1lp=3pixels. That bring resolution of the sensor to of less that 50/lp/mm. Consequently it will bring the resolution of the lens/sensor system to of about 50lp/mm. Reala ($2.5 a roll) resolves much better.

 

So a litte bit of distortion, a litle bit of technical inaccuracies and an expensine, unique (in some ways), and controvertial product becomes a stellar performer.WOW!

Contax ND was controvertial too. You know what happened to it....

[astral]

"Leica M8 has adopted several compromises. They must to be known, understood and accepted."

 

The same is true about Holga/Lomo 'photography' - so that puts them on a n equal footing with the M8? Don't ya just love glib statements .... Internet: ultimate fountain of wisdom!

[Brad_]

<I>Leica users accept no compromises.</I><P>

 

Please, not while I'm drinking milk. It's not pretty...

[thomas_sullivan]

oh, now we not only ignore reality, we re-invent it to suit our own misgivings. I guess this comment will get deleted also. If Canon or Nikon user's did this, you guys would be all over them. ....whatever,

[jerry_kirkwood]

Leica users <del>accept</del> admit no compromises.

[blakley]

<i>Why would the resolution of the Leica's sensor be 73.5lp/mm?! It is only possible if one assumes that 1lp=2 pixels. It is generaly accepted that 1lp=3pixels.</i>

<p>

Stan, let's be fair to Erwin. He didn't say that the Leica's sensor's <i>resolution</i> is 73.5mm, he said it's <i>Nyquist limit</i> (maximum sampling frequency in pixels/mm divided by two) is 73.5 lp/mm.

<p>

As far as "generally accepted" is concerned, authorities other than Erwin, including <a href="http://216.239.59.104/search?q=cache:a4DThLnHq38J:www.edmundoptics.com/techSupport/DisplayArticle.cfm%3Farticleid%3D290+nyquist+limit+digital+sensor&hl=en&gl=us&ct=clnk&cd=10&client=firefox-a">Edmund Optics</a> and our own <a href="http://www.bobatkins.com/photography/digital/size_matters.html">Bob Atkins</a>, define the sensor's maximum resolution as its Nyquist limit. I was unable to find your proposed definition (sampling frequency divided by 3) in use <i>anywhere</i> so whatever agreement you're referring to certainly isn't the generally accepted usage.

[Brad_]

<I>...define the sensor's maximum resolution as its Nyquist limit.</I><P>

 

No such definition was put forward in the Edmunds material. What was said, was:<I> The

limiting resolution of the CCD camera is determined by its Nyquist limit.</i> There's a huge

difference. <P>

 

Let me know if you'd like an explanation why, and why using a

value of 1/3 the sampling frequnecy is more realistic.

[rj]

A whole lot of technical mumbo jumbo to tell me that the camera isn't all that bad. Big woop, show me some pics from it that aren't a "test" please.

[blakley]

Brad, I understand why you and others want to take 3 lines rather than 2 as the base unit of resolution (for example, when you're talking about the human visual system's ability to distinguish details in prints) but the fact remains that this is NOT the way the industry measures the resolution of sensors. Neither Edmund, nor sensor manufacturers, nor camera manufacturers measure resolution based on groups of 3 lines. Erwin's use of terminology is perfectly conventional in this respect, and to accuse him of technical inaccuracy for using perfectly ordinary terms from optical engineering is just silly.

[Brad_]

Bob, it has **nothing** to do with the human visual system.

 

It is driven by fact-of-life alias considerations in discreet sampled data systems - only

applicable to digital capture. Again, read carefully what Edmunds actually said - without a

technical background in signal processing, it's easy to miss the distinction.

[Brad_]

<I>Big woop, show me some pics from it that aren't a "test" please.</I><P>

 

No argument there. So far, haven't seen any visually interesting photos.

[stan_belyaev]

Edmund is wrong. Pls, read carefully!

 

"Nyquist Limit

The limiting resolution of the CCD camera is determined by its Nyquist limit. This is defined as being one half of the sampling frequency (i.e. #pixels/mm). For example, the Sony XC-ST30 is a monochrome high resolution CCD camera and its sensor's active area has 768 horizontal pixels in 4.8mm. This represents a horizontal sampling frequency of 160 pixels/mm. The Nyquist limit in this example would be defined as 80 lp/mm. One should note that there are image processing methods, such as sub-pixel sampling, which enable the user to statistically extrapolate higher resolution than the Nyquist limit"

 

So sampling frequency is 160pix/mm. Nyquist limit should be half of 160. It is 80pix/mm. Traditionally, resolution is expressed in LP/MM, but not in L/MM. Lets assume that 1lp=2pix (it should be 3), then 80pix/mm is eqial to 40lp/mm. What 80lp/mm are they talking about?!

The actual resolution of the system should be lower than its Nyquist limit.

 

The same mistake was made in the article.

[blakley]

*sigh*. Neither Edmund nor Erwin is wrong, and I'm not the one who's not reading carefully.

<p>

<a href="http://www2.olympus.dk/corporate/images/Micro_Series_Part_3.pdf">Here</a> is a great article describing the relationship between the Nyquist limit and the actual resolution one can get out of a device under various circumstances. It talks about the difference in resolution measures you get when you decide that you prefer to measure resolution by allocating 3 pixels per line pair rather than 2 pixels per line pair, and why you might prefer 3 pixels if you're worried about losing resolution because of a lack of alignment between the subject and the rows and columns of the image sensor.

<p>

Sensor manufacturers use the 2 pixels per line pair definition when they talk about resolution. Do they do this because it produces higher resolution numbers, which sound more impressive to customers? Probably. But who cares? You guys want Erwin to use terminology differently than the manufacturers do. He's not going to do that. If you want that, run your own tests and use the terminology you prefer.

But don't criticize him, or Edmund, for errors neither of them has made.

[blakley]

Re: <i>Bob, it has **nothing** to do with the human visual system.</i>

<p>

Actual experts don't agree with you - for examples, you might look

<a href="http://research.opt.indiana.edu/Library/GlennFryLecture/GlennFryLecture.html">here</a>,

<a href="http://www.spie.org/web/oer/october/oct97/eye.html">here</a>, and

<a href="http://webvision.med.utah.edu/KallSpatial.html">here</a>.

[blakley]

And to answer your specific question, Stan:

<p>

<i>Traditionally, resolution is expressed in LP/MM, but not in L/MM. Lets assume that 1lp=2pix (it should be 3), then 80pix/mm is eqial to 40lp/mm. What 80lp/mm are they talking about?!</i>

<p>

They're talking about "pairs of adjacent lines, one light and one dark", which is a perfectly normal and common definition of "line pair". When your image and your sensor are properly aligned and the lines are each one pixel wide, you can get 80 of these line pairs in a millimeter on a sensor with a horizontal sampling frequency of 160 pixels per millimeter, exactly as the Edmund article states.

<p>

The fact that you prefer to define line pairs as "pairs of dark lines separated by a light line" does not make Edmund's (and Erwin's, and the sensor manufacturers') use of the other definition "a mistake".

[stan_belyaev]

1. http://en.wikipedia.org/wiki/Nyquist_frequency

160pix/mm - 80pix piars/mm - 80lp/mm. Samnpling frequency is 160. Nyquist frequency is 50% of sampling frequency - 80pix/mm = 40lp/mm.

 

2. http://www.clarkvision.com/imagedetail/sampling1.html

There is a key phrase: "The Nyquist theorem ONLY applies to data and sampling that are in phase. To be "in phase," it means that sampling occurs only at the peaks and valleys of the image data, and a pixel would not fall on a boundary. Clearly this condition is not met in image data of real scenes in general".

That is why 3pix/lp is much realistic number that 2pix/lp.

[blakley]

Stan, repeating yourself doesn't make Erwin wrong, and it doesn't make anyone disagree with him, either. Erwin's terminology is consistent with every reference returned by <a href="http://www.google.com/search?q=%22line+pairs+per+millimeter%22&start=0&ie=utf-8&oe=utf-8&client=firefox-a&rls=org.mozilla:en-US:official">this Google search</a> (type "line pairs per millimeter" into the Google search bar to get the same result yourself).

<p>

<a href="http://www.normankoren.com/Tutorials/MTF2.html#Nyquist">Here</a> is one of the best explanations. It agrees with Erwin and disagrees with your formula - just like all the others.