Questions regarding medium format photography and pixels

[kevinbriggs]

<p>On Mamiya's <a href="http://www.mamiya.com/photography-why-medium-format.html">webpage</a>, they state the following: "Similar to film formats, larger sensors offer higher resolution, greater dynamic range, more color depth and bigger pixels as compared to smaller sensors."</p>

<p>Question(s): do ALL medium format digital cameras possess bigger pixels for their sensors...?</p>

<p>Secondly, are all medium format pixels the same size...? Or do different manufacturers produce different pixel sizes for their medium format cameras...?</p>

<p>Let's say I'm comparing the Mamiya DM33 with the Hasselblad H4D40 (yes, let's pretend I have a few tens of thousands of dollars to spare): does one literally need to inquire with the manufacturers as to exactly how large their individual/respective pixel sizes are and then compare the two...?</p>

<p>Lastly, and to switch to the subject of FILM and medium format (the following question is nonetheless related to a similar "optics" question):</p>

<p>If you were shooting with medium format film as opposed to digital, and you just happened to be using the Nikon Coolscan 9000 (for instance), do the optics associated with the scanners -- either the Nikon or drum scanners or whatever -- also have different "pixel sizes" as it were...?</p>

<p>Meaning, if you are shooting with medium format Fuji Velvia 50 or 100, for instance, and are trying to soak up all of the color and saturation and tonality... and on and on and on... with regard to the actual film, must you then choose a scanner that similarly incorporates the largest "pixel sizes" (again, as it were) for scanning such landscape masterpieces...?</p>

<p>Thanks in advance for any and all replies!</p>

 

[swilson]

<p>You can work out the pixel size for the Mamiya from there<br>

<a href="http://www.mamiya.com/dm33-dm40-and-dm56.html#Specifications">http://www.mamiya.com/dm33-dm40-and-dm56.html#Specifications</a><br>

The DM33 has pixels that are around 7.1 micron for example<br>

For Masselblad you can get the data here<br>

<a href="http://www.hasselblad.com/downloads/datasheets/h-system.aspx">http://www.hasselblad.com/downloads/datasheets/h-system.aspx</a><br>

The Nikon Coolscan 9000 scans with pixels that are around 6.35 microns in size.<br>

Keep in mind that the sensor/film area is different from system to system.</p>

<p>Of course film pixels are digital sensor pixels are not the same, and there can be endless arguments over which is better. The best thing is download samples from each system.</p>

[ondebanks]

<p>To some extent, touting large pixels as an advantage is putting "spin" on something the manufacturers couldn't really do much about. The trend in medium format digital is actually towards smaller pixels - it's easier for CCD manufacturers to do this than to make larger area sensors. MFDBs started with 12 micron pixels (e.g. Lightphase C645); spent a lot of time at 9 micron pixels (there are still a few like this); and the more expensive ones now have 7.2, 6.8 or 6.0 micron pixels.</p>

<p>You can find all this by looking at various sources online - principally the brochures, user manuals & tech spec pages for the MFDBs.</p>

<p>Re. your second question - film scanners do not need large pixels because they do not need to handle the sort of wide dynamic range that occurs in natural light. The film medium has already compressed or clipped the dynamic range, to something much more manageable. Negative film compresses it, while slide film like Velvia clips it.</p>

[kevinbriggs]

<blockquote>

<p>You can work out the pixel size for the Mamiya from there
http://www.mamiya.com/dm33-dm40-and-dm56.html#Specifications
The DM33 has pixels that are around 7.1 micron for example
For Masselblad you can get the data here
http://www.hasselblad.com/downloads/datasheets/h-system.aspx</p>

 

</blockquote>

<p>Hi Scott,</p>

<p>Thanks for the links. Unfortunately, I'm not seeing pixel size listed anywhere on any of these documents and/or webpages.</p>

[swilson]

<p>You can figure it out from the sensor size and the number of pixels.</p>

[kevinbriggs]

<blockquote>

<p>You can figure it out from the sensor size and the number of pixels.</p>

 

</blockquote>

<p>What's the formula to follow...?</p>

<p>Thanks!</p>

[kevinbriggs]

<p>If I'm performing the calculations correct, the pixel sizes associated with the medium format sensors aren't really all that larger than what I'm seeing on the Canon 1Ds Mark III (which I'm currently using), which is around .0062.</p>

<p>For instance, when rounded to the nearest ten-thousandth, I get a pixel size of .0060 for the Hasselblad H4D-40, .0089 for the Mamiya DM22, .0071 for the Mamiya DM28, .0060 for the Leica S2, etc.</p>

<p>So I must admit I'm not totally understanding Mamiya's claims about the benefits of larger pixel sizes in medium format cameras....?</p>

<p>... maybe I'm not calculating these correctly.....</p>

<p>Or, when Mamiya is referring to "smaller sensors" are they meaning point-and-click "Coolpix" and "Powershot" cameras...?</p>

[kevinbriggs]

<blockquote>

<p>....and the more expensive ones now have 7.2, 6.8 or 6.0 micron pixels.</p>

 

</blockquote>

<p>Thanks Ray for this confirmation.</p>

[ondebanks]

<blockquote>

<p>Or, when Mamiya is referring to "smaller sensors" are they meaning point-and-click "Coolpix" and "Powershot" cameras...?</p>

</blockquote>

<p>Not really - they're referring rather to 35mm FFDSLRs and APS DSLRs. MFDB makers have to carve out their market by selling themselves as bigger and better than 35mm full frame DSLRs, in particular. Which they are; but not by as much as they were in the film days.</p>

<blockquote>

<p>So I must admit I'm not totally understanding Mamiya's claims about the benefits of larger pixel sizes in medium format cameras....?</p>

</blockquote>

<p>You're right, sort of. You won't necessarily get <em>every</em> one of the benefits on the list ("higher resolution, greater dynamic range, more color depth and bigger pixels"), because some of them trade off in opposition to each other - greater dynamic range vs. higher resolution, for example. But say you hold the pixel size constant - then you have more pixels on a MFDB than a FFDSLR. Or you hold the pixel count constant - then you have larger pixels and more dynamic range on the MFDB than the FFDSLR.</p>

[gregory_king1]

<p>As Ray said, 9 micron pixels is a kind of "baseline". It's what you get at 6mp for APS, 11mp for FF, and 22mp for the 36x48mm MF sensor. <br>

<br />Many of the higher resolution MF sensors are higher resolution on SMALLER sensors (like 44x33mm, for example), which obviously results in quickly shrinking pixels.</p>

<p>An interesting related chart...too bad he doesn't show negative film resolution.</p>

<p><img src="http://www.clarkvision.com/articles/film.vs.digital.1/film.vs.digital.6x4.5.gif" alt="" width="613" height="505" /></p>

[leighb]

<blockquote>

<p>What's the formula to follow...?</p>

</blockquote>

<p>Pixel size = sensor width / number of pixels</p>

<p>For example, given 49mm width and 7,200 pixels, pixel size = 49 / 7212 = 49/7200 = 6.8 µm.</p>

<p>- Leigh</p>

<p> </p>

[Ed_Ingold]

<p>The difference between images from a top of the line, small format sensor (e.g., a Nikon D3) and a medium format digital back (e.g., Hasselblad CFV-16) is more than pixel size. AFIK, all MFDs are CCDs, but the design is different than for small cameras. For one, they are so-called "full frame" sensors, which have no internal gating (i.e., electronic shutter), are more efficient in gathering light, hold more electrons (i.e., collect more photons), and have a much lower noise level, but take a lot longer to read. Components are hand-selected for optimum performance and sensors are individually tuned and optimized. There is no anti-aliasing (low pass) filter, which alone gives you about 50% more resolution per pixel. Most MFDs have cooling, either fan or thermoelectric, which further reduces noise. The result is 12-13 stops of useable dynamic range, compared to 7-9 for a first-rate DSLR.</p>

<p>The difference in image quality is immediately obvious on screen, but especially in print. MFD is much better at rendering color, shadow detail and textures than it's smaller brethren. Even a casual observer can see it in an 8x10. I have attached a picture I took of a motorcycle, all black and chrome. Notice the texture of the leather seat, even at web resolution. The rendering of textures like black leather, white fabric and skin is probably the most compelling reason to use MFD. In landscapes, there is as much shadow detail as you care to print - more than you can capture even with negative film. I have done HDR merging with medium format, which I can only describe as "redundant."</p>

<p>There are far more situations for which I would choose my D3 over MFD. But for fussy work, it's hard to beat.</p><div></div>

[Ed_Ingold]

<p><em>Meaning, if you are shooting with medium format Fuji Velvia 50 or 100, for instance, and are trying to soak up all of the color and saturation and tonality... and on and on and on... with regard to the actual film, must you then choose a scanner that similarly incorporates the largest "pixel sizes"</em></p>

<p>I'm OK with the epithets "color" and "saturation", but I think "tonality" is misplaced when referring to Velvia. If you look at the characteristic curve published by Fujifilm for Velvia, you see that the density on film ranges from 0.2 to about 3.4. The maximum, or DMax, is well within the capacity of most film scanners, including the Nikon LS-9000 and Imacon (DMax = 4.2). Don't look for shadow detail in those deep blacks - there isn't any.</p>

<p>The range of exposures corresponding to this density range is -0.5 to about -2.0 log(lux-seconds), the equivalent of about 5 stops, which represents the dynamic range of the subject which can be captured. The touted dynamic range of Velvia is consequently an illusion due the high contrast in the resulting slide. The contrast is boosted so that every stop in the subject becomes nearly two stops change on film. "Tonality", or the ability to render subtle shades and gradients, goes out the window, along with shadows and highlights.</p>

[dave_redmann]

<p><em>An interesting related chart...too bad he doesn't show negative film resolution.</em></p>

<p>Unless I'm totally misreading the chart, it is at best misleading, and at worst utter nonsense. Where to start?</p>

<p>Okay, well, resolution for Velvia. Look at Fuji's own data sheet (<a href="http://www.fujifilmusa.com/shared/bin/AF3-0221E2Velvia50PIB.pdf">http://www.fujifilmusa.com/shared/bin/AF3-0221E2Velvia50PIB.pdf</a>, the last page, under #21) and you will see it shows (presumably for typical conditions) that MTF response is down to 50% at about 45 lp/mm. Given that a typical 645 frame is about 56 x 41.5 mm, that's about 19 MP. (Yes, I see elsewhere that Fuji claims considerably higher resolving power with high-contrast subjects, but first, it doesn't specify any MTF response, which probably means there is just barely some detectable difference, not necessarily a useful one; and second, real subjects don't usually present that much contrast.)</p>

<p>Now TMX--TMX will comfortably outresolve Velvia, but this chart suggests much lower resolution.</p>

<p>Also, while I might agree that 645 Tech Pan can resolve 50 MP, to do so you need a lens with very stong MTF response at around 75 lp/mm. To put that in perspective, <em>Popular Photography</em> found that the Hasselblad HC 80mm lens for the H1/2/3/4 did not produce 75 lp/mm anywhere in the frame at any aperture; their tests found 66 lp/mm in the center at f/11 and falling off from there, and the corners never achieved more than 47 lp/mm. (Interestingly, this would suggest that, without a better lens, a medium format digital back with a 49 x 36.7 mm sensor will produce very little extra resolution once it exceeds about 40 MP.)</p>

<p>Any supposed pixel count for film necessarily involves simplifying assumptions and semi-arbitrary cut-offs, because film simply does not have pixels and does not behave like it does. Within a certain range, you can dispute what assumptions and cut-offs are appropriate, and reasonable and knowledgeable people can disagree. But there is a fair amount of nonsense out there, and you have to be careful what you take from what you read.</p>

 

[Ed_Ingold]

<p><strong>Useable dynamic range </strong>is that over patches in a step wedge can be distinguished from adjacent patches - the entire step wedge in a single exposure. It is not, as DXO asserts, saturation level minus the noise level - this exaggerates the performance of DSLRs, which employ extensive noise reduction measures, and diminishes performance of MFD's, which do not. DPReview and Luminous-Landscape use step wedges in their evaluations, perhaps others as well.</p>

<p>Use of a step wedge in a single exposure avoids the misleading results involving complex subjects (e.g., tin figures and crayon boxes), which are then examined for any visible detail, or separate exposures in which each frame is adjusted to optimize exposure and contrast. While these home-brew tests might demonstrate latitude (a rather nebulous term), they bear little relationship to dynamic range.</p>

[Ed_Ingold]

<p>(double post)</p>