Poor scans - May explain why some switch to Digital

[mauro_franic]

<p>Scott, that is a very nice crop (was it provia with Coolscan 5000?).</p>

[mauro_franic]

<p>Scott, I agree with your comments. </p>

 

[matthew_newton]

<p>Wow, talk about a lot of posts.<br>

I get the point, Medium format film, at least 6x7 is better then digital SLRs...at least 35mm frame and smaller. I would bet still that something like the new 60mp medium format backs are better then even the best 6x7 scans at least rivaling if not beating drum scans with anything other then maybe some slow B&W films.<br>

If you have the money or a friend with one I'd love to see the test. Beyond that though, when it comes down to it the latest generation of full frame digital SLRs are going to meet or beat the best of the best 35mm film scans under 99% of conditions.</p>

[robert lee]

<blockquote>

<p>I just looked up Aztek starter kits: $639 + exorbitant shipping = close to $700. ... Before I even begin to think about ordering from Aztek to try wet mounting, I have to know: will it give near drum scanner quality, as Ellis Vener says?</p>

</blockquote>

<p>First the caveats: try this because you're comfortable tinkering, and are willing to accept the potential for equipment or media damage. No whining if something goes wrong.</p>

<p>Alright, instead of Kami, give mineral oil a try. See if you can get hold of the food grade stuff; "baby oil" from the drug store is perfumed. The fragrance probably doesn't change optical properties one way or another, but it's also an unnecessary component for the present purpose.</p>

<p>I'm hacking together a MF/LF film scanner off and on. The light source is intensely bright, so I was concerned with the evaporation rate of Kami. I dug around Google and it turns out that mineral oil is a potential substitute (and a component of Kami actually.) </p>

<p>The one test scan I did showed improvements on par with a more traditional kit. However, the process itself is time consuming and more messy than I think I can regularly tolerate. Let me add though this is probably more an indictment of wet mounting in general than anything else.</p>

<p>By the way, I was also uncomfortable with Kami from a health and safety perspective. I downloaded its material safety data sheet (MSDS.) Kami is about 90% naptha (camper stove fuel) and the rest n-hexane and mineral oil. It may not leave messy stuff on the film, but that means that the evaporates goes into the room air. I dunno... I don't think huffing camp fuel is such a great idea.</p>

 

[Landrum Kelly]

<p>Thanks, Robert. Mineral oil is a very helpful suggestion, since it is surely sufficiently chemically inert not to hurt the emulsions--and its boiling point is high enough that it presents no particular fire hazard.</p>

<p><strong>I do have one other question for anyone who has done wet mounts:</strong> is it possible to do it with the glass holders from Nikon? I already have those. Or does it require some kind of special holder such as those from Aztek? I notice that the $600+ starter kit from Aztec includes the holder, and <em>the holder alone sells from Aztek for about $500.</em><br>

<em> </em><br>

In other words, it sounds as if it might be possible to rig up a wet mount system with the existing Nikon holders so that one would not need to buy the Aztek holder. Then again, since I have never seen the Aztek holder and have never tried wet mounting, I really do not know.</p>

<p>Again, thanks.</p>

<p>--Lannie</p>

[steve m smith]

<blockquote>

<p>I'm hacking together a MF/LF film scanner off and on. The light source is intensely bright, so I was concerned with the evaporation rate of Kami. I dug around Google and it turns out that mineral oil is a potential substitute (and a component of Kami actually.)<br>

The one test scan I did showed improvements on par with a more traditional kit. However, the process itself is time consuming and more messy than I think I can regularly tolerate. Let me add though this is probably more an indictment of wet mounting in general than anything else.</p>

</blockquote>

<p>I have not had cause to try it yet but if I wanted to dedicate a flatbed scanner to film use I was thinking that instead of wet mounting, I could replace the glass with a piece of Perspex (Plexiglass) with the image area cut out. As the film holder suspends the film just above the glass then it should still be in the right place (assuming correct thickness of Perspex) and there are two less surfaces to collect dust and reflect light and no potential for Newton's rings.<br>

Has anyone tried this or something similar?</p>

[rubo]

<p>Rishi - just curious about something you posted earlyer.<br>

What did you mean by saying "12.6MP measured-resolution" of the 5D MkII?<br>

I thought it has 21MP???</p>

[mauro_franic]

<p>Rubo,</p>

<p>Digital cameras resolve less information than the nominal pixels they produce. Depending on the demosaicing algorythims it is around 2/3rds.</p>

<p>My canon 40D (outputs 10MP) resolves 6MP worth of information.</p>

[mauro_franic]

<p>TMX 35mm for e.g. (my own test) resolve 6 times the detail of my 40D. This is 36MP iof information.</p>

<p>Obviously, you are limited by your scanner on how much of this you can extract. Also grain comes together with detail the deeper you scan the film (although it can be removed with software).</p>

[mauro_franic]

<p>TMX 6x7 (developed by me) captures about 20-30 times the detail that I can capture with my 10MP DSLR. </p>

[mauro_franic]

<p>Here is a comparison of my 10MP DSLR vs Velvia 67</p><div></div>

[mauro_franic]

<p>Velvia (just with a Coolscan, an Imacon would give you significantly better results):</p><div></div>

[swilson]

<blockquote>

<p ><a href="http://www.photo.net/photodb/user?user_id=4089386">Rubo Aristakesyan</a> <a href="http://www.photo.net/member-status-icons"></a>, Mar 20, 2009; 04:11 p.m.</p>

 

<p>Rishi - just curious about something you posted earlyer.<br />What did you mean by saying "12.6MP measured-resolution" of the 5D MkII?<br />I thought it has 21MP???</p>

 

</blockquote>

<p>Most digital cameras, including the 5DMkII, have a byer pattern for color. Whereas there is no exact number that can be given there is some loss of resolution from this pattern.</p>

<p>Opions vary as to how much loss there is, some will try to tell you that you only get 25% of the cameras pixels as the real resolution, som will try to tell you that you get 100%, neither of these number is correct.</p>

<p>I use about 75% of the sensors resolution. So what does this mean? It means I can take a 8 MP image from my camera and shink it down to 6MP and see almost no loss of detail. BTW images from scanners, for other reasons, have a even larger number that they can be shurnk buy before loosing detail. </p>

<p>I have not tested the 5D MkII but I would have guessed the number to be between 15 and 16 MP. Some of this depends on the Anti-alias filter, where I don't know for this camera.</p>

[rubo]

<p>So my 12MP XSi resolves around 7-8MP of information?<br>

OK, i'll buy that.<br>

But why does 21.1MP 5D MkII resolve only 12.6? Shouldn't it be more like 14-15MP?</p>

<p>Just asking questions, don't shoot me, PLEASE :)</p>

[swilson]

<blockquote>

<p ><a href="http://www.photo.net/photodb/user?user_id=3995956">Mauro Franic</a> <a href="http://www.photo.net/member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/3rolls.gif" alt="" /></a>, Mar 20, 2009; 04:38 p.m.</p>

 

<p>Rubo,<br>

Digital cameras resolve less information than the nominal pixels they produce. Depending on the demosaicing algorythims it is around 2/3rds.<br>

My canon 40D (outputs 10MP) resolves 6MP worth of information.</p>

 

</blockquote>

<p>Ok, we might be getting to splitting hairs here but in test I have don't I figure my 350D produces about 6MP starting out at 8MP. I would have thought that the 40D would have been a bit over 6MP. Some of this depends on the lens used, I mostly use prime lenses when I am going for good resolution, is it posible that some of your loss is from using a zoom lens, even a L zoom lens is not close to as good as a L prime.</p>

[mauro_franic]

<p>The 5DII resolves around mark 4 in the chart above.</p>

<p>Velvia resolves around 14 (11 Captured by the Coolscan).</p>

<p>Velvia 6x7 captures (high contrast) over 100 MP of information, the 5DII in comparison between 12MP and 14MP.</p>

[mauro_franic]

<p>Rubo, depending the demosaicing algorythm you choose, you can increase detail at the expense of color or artifacting. In general between 60% and 75% is a normal range.</p>

[swilson]

<blockquote>

<p ><a href="http://www.photo.net/photodb/user?user_id=4089386">Rubo Aristakesyan</a> <a href="http://www.photo.net/member-status-icons"></a>, Mar 20, 2009; 04:53 p.m.</p>

 

<p>So my 12MP XSi resolves around 7-8MP of information?<br />OK, i'll buy that.<br />But why does 21.1MP 5D MkII resolve only 12.6? Shouldn't it be more like 14-15MP?<br>

Just asking questions, don't shoot me, PLEASE :)</p>

 

</blockquote>

<p>I would have thought that it would have been closer to 14-15 as well.</p>

<p>Testing of this kind of thing is not an exact science, anyone who says it is does not understand all the complications.</p>

<p>Getting into this can also stir up the byer pattern/foveon debates. There is no question that a foveon sensor get more detail per pixel (real pixel not what they call a pixel) then a byer camera, but it is not even close to the 3X number Sigma would have you believe.</p>

<p> </p>

[rishij]

<p>Rubo -- Sure, it's 21MP, but no way in heck does it actually resolve 21,000,000 real pixels worth of information. This is why places like Dpreview.com take these digital cameras and shoot resolution test charts with them. From a resolution test chart shot using the 5D Mark II, they calculate the actual measured resolution the camera is capable of. In my eyeball calculation, I measured 12.6MP for the 5D Mark II, which comes close to what Dpreview also measured (though they only report their result in resolved 'lines per picture height'... I just took those numbers & translated them into megapixel values by multiplying the horizontal and vertical LPPH values).</p>

<p>Remember, the camera has 21,000,000 pixels, but only 1/2 of them are green pixels, 1/4 red, and 1/4 blue.<br>

-Rishi</p>

[swilson]

<blockquote>

<p ><a href="http://www.photo.net/photodb/user?user_id=3995956">Mauro Franic</a> <a href="http://www.photo.net/member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/3rolls.gif" alt="" /></a>, Mar 20, 2009; 04:57 p.m.</p>

 

<p>The 5DII resolves around mark 4 in the chart above.<br>

Velvia resolves around 14 (11 Captured by the Coolscan).<br>

Velvia 6x7 captures (high contrast) over 100 MP of information, the 5DII in comparison between 12MP and 14MP.</p>

 

 

</blockquote>

<p>The tricky part here is translating between a test chart and effective pixels. <br>

I would be happy deeply to get into the math of sampling systems that are sampling data close to Nyquist if you really wnat to, but for now I will give the short version</p>

<p>

<p>Most is not all digital cameras have AA filters that allow some spatial frequencies past Nyquist to get to the sensor, this is a good thing most of the time. If they did not then contrast at frequencies lower then Nyquist would be much less then they are with the weaker AA filter. But the test chart has very high contrast data past Nyquist as so there is a folding back of the higher frequencies, this make some rather odd things show up, like in the test chart. It can also make high contrast small text seem to change in wird ways.<br>

<br />If it were a simple monochrome imaging system then the fact that the system is showing aliasing would be proof that all the pixels are being use, a 12 MP camera would have 12MP of real data. But the bayer pattern makes this determination much harder, as aliasing gets pretty complicated.</p>

</p>

 

[swilson]

<blockquote>

 

<p ><a href="http://www.photo.net/photodb/user?user_id=2381463">Rishi Sanyal</a> <a href="http://www.photo.net/member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></a>, Mar 20, 2009; 05:03 p.m.</p>

 

<p>Rubo -- Sure, it's 21MP, but no way in heck does it actually resolve 21,000,000 real pixels worth of information. This is why places like Dpreview.com take these digital cameras and shoot resolution test charts with them. From a resolution test chart shot using the 5D Mark II, they calculate the actual measured resolution the camera is capable of. In my eyeball calculation, I measured 12.6MP for the 5D Mark II, which comes close to what Dpreview also measured (though they only report their result in resolved 'lines per picture height'... I just took those numbers & translated them into megapixel values by multiplying the horizontal and vertical LPPH values).<br>

Remember, the camera has 21,000,000 pixels, but only 1/2 of them are green pixels, 1/4 red, and 1/4 blue.<br />-Rishi</p>

 

 

</blockquote>

 

<p>This is a very bad way to determine effective pixels, try the following experiment, I have.<br>

Take a test chart in pdf format and change it to a high-res bitmap, the pixels should be about perfect here, reduce the number of pixels to say 10MP, these should be as good as you can get pixels, now use your technique to read the number of real pixels and you will end up with a number below 10.</p>

 

[mauro_franic]

<p>That's right. I often wonder why canon doesn't issue monochrome versions of their DSLRs (without the color filters). It could be a nice quality boost for B&W shooters.</p>

[rishij]

<p>Mauro -- I don't understand how you upscaled Scott Wilson's dSLR image of the biker to match Lex's film scan if you don't know what the original size/resolution is...</p>

<p>For that matter, Scott, we've all been sharing what format/resolution we've been posting. Please tell us the actual resolution of your images; in particular, the biker one. As well as the house one you posted next to my upscaled film shot.<br>

-Rishi</p>

[rishij]

<p>Here's a crop from the 5D Mark II resolution test chart shot from dpreview.com:<br>

<a href="http://a.img-dpreview.com/reviews/CanonEOS5DMarkII/Samples/Compareto/Res/5D2-Res-002.jpg">Click here</a><br>

<em>Taken from dpreview.com</em></p>

<p>Zooming in I rate the vanishing resolution point around 29... multiply by 100, you get 2900 lines per picture height (LPPH). For all practical purposes, horizontal & vertical resolution were on par with each other, so I then multiply this number by the sensor aspect ratio: 36/24 (roughly), to get 4350.</p>

<p>4350 x 2900 = 12.6MP</p>

<p>Scott, this is what dpreview uses to determine resolution. I'm not sure what you were trying to get at re: the PDF --> BMP, etc. What were you trying to prove there?</p>

<p>Re: AA filters & Nyquist, I get what you're saying. But, regardless, what's wrong with measuring the point at which I can no longer distinguish the 9 lines (pretty much getting to 0% contrast) if I use the same technique to measure the resolution of my film scans?</p>

<p>-Rishi</p>

[rubo]

<p>OK.<br>

Now at least i know what you were talking about.<br>

Every day you learn some thing new :)</p>