Pentax and Phase one

[benjamin_kim2]

<p>I used Phase one with P45+ today and it was fantastic because of colors, detail, and more. I know that Phase one is so expansive that I can not even afford it for my career. I'm quite curious about the image quality and color for Pentax 645Z. I know that Phase one has 16 bit color but does anyone have any thoughts about Pentax 645Z's color or color tone and image quality?</p>

[howard_m]

<p>http://blog.mingthein.com/other-reviews/</p>

[ondebanks]

<blockquote>

<p>I know that Phase one has 16 bit color but does anyone have any thoughts about Pentax 645Z's color or color tone and image quality?</p>

</blockquote>

<p>The 16 bit thing is a red herring; it's really just over-engineering; but it makes good marketing copy. The Kodak KAF-39000 CCD in the P45+ can capture 60,000 electrons per pixel, close to 2^16 (= 65,5356), which seems to suggest that a 16 bit output is appropriate.</p>

<p>But when you consider that the readout noise is 16 electrons (2^4) per pixel, leaving a dynamic range of 2^16 / 2^4 = 2^12 by the standard engineering definition, then a 12 bit output is sufficient, 14 bit would be oversampling the noise (useful only in certain situations), and 16 bit is completely overkill.</p>

<p>The bigger issues here are:</p>

<p>1) Colour response: Kodak's colour filter array in the P45+ vs Sony's colour filter array in the 645Z. I've seen nothing wrong with the Sony, but the Kodaks are legendary. I use an old Kodak-chipped back myself.</p>

<p>2) Tonality: medium format CCDs like in the P45+ have a "high signal, high noise" characteristic. This gives clean highlights, and film-like grainy darker tones. Like film, they also suck at being pushed to higher ISOs. The Sony sensor is CMOS, which has a "moderate/high signal, low noise" characteristic. This is much better suited to being pushed to high ISO, and gives a better dynamic range at low ISO.</p>

<p>But even when these two types of sensors capture a scene at the same signal to noise, it looks different. For example, exposing the CCD at ISO 100, and the CMOS at ISO 400 with 1/4 the exposure time and identical lens settings, might yield similar signal to noise. But high signal in the presence of high noise looks more organic and crunchy than low signal in the presence of low noise, which looks a bit flat and manipulated (looks de-noised even though it wasn't). Signal to noise may be a numerical ratio, but it has both quantity and quality aspects and I frequently point out the latter.</p>

<p>3) Size: the P45+ is a larger sensor (49 x 36mm, vs. 44 x 33mm). It's not a huge gap but something to be aware of if you are considering lens crop factors etc. A given focal length wideangle will retain more of its wideness on the larger sensor. The P45+ also has fewer, larger pixels (6.8 microns vs. 5.3 microns) so the demands on lens MTF are a little easier.</p>

[Ed_Ingold]

<p>Where does "60,000 electrons per pixel" come from?</p>

<p>16 bit word depth has nothing to do with signal to noise ratio. It is the number of steps between black and white.</p>

[rodeo_joe1]

<p>You can compare the Pentax 645Z with other DSLR cameras here: https://www.imaging-resource.com/IMCOMP/COMPS01.HTM</p>

<p>The Sony A7 RII comes pretty close. 4 grand (in UK £) difference in quality? I don't think so. You could get some pretty nice glassware for the Sony with that amount of money. And you can bet that both cameras will be obsolete in a few years time and worth a fraction of their current cost. Professional high-end digital gear like that is best hired as required and not owned outright IMHO. Unless you're sure of making the outlay back within a few months.</p>

<blockquote>

<p>"16 bit word depth has nothing to do with signal to noise ratio. It is the number of steps between black and white."</p>

</blockquote>

<p>Actually it defines the number of levels between "black" and the saturation point of each RGB channel, and in the linear analogue space before a digital gamma curve is applied, any A/D converter bit depth under the noise level is pretty much wasted. If only more attention was paid to the analogue signal, then a log amplifier could be used to apply a gamma curve <em>before </em> the A/D conversion took place. Then we might be cooking with gas. No camera maker uses such a system AFAIK.</p>

[Ed_Ingold]

<p>I could have added "per channel." Each pixel only registers black, white or in between. Color is added in the translation. The sensor is inherently linear (or nearly so) and almost free of reciprocity failure. Whether you add a log curve when reading or afterwards is of little matter.</p>

<p>S log n gamma curves are popular in video cameras for improved dynamic range. The results are not very attractive without further processing (in video, called "grading"). S log 2 gamma is actually an option for the Sony A7Rii (and others), primarily for video but can be applied to stills too.</p>

[rodeo_joe1]

<blockquote>

<p>"Color is added in the translation."</p>

</blockquote>

<p>Colour is <em>subtracted</em> in the RGB Bayer filtering before light even gets to the photosites.</p>

<blockquote>

<p>"Whether you add a log curve when reading or afterwards is of little matter."</p>

</blockquote>

<p>Do you design camera electronics by any chance Edward? Because that's the sort of thinking that results in a restricted dynamic range and useable bit depth.</p>

<p>Once an analogue signal is digitised you can only throw information away to transform it - i.e to adjust the tone curve. Applying a desired tone curve (gamma function) to the analogue signal before digitisation better preserves the full A/D converter depth. Can you not see that?</p>

<p>No practical colour space uses a linear tone curve. The result would look flat or require a display brightness, contrast and CT equal to the original scene to display it. Since the sun isn't available in portable form yet, a gamma curve is absolutely required. There is no camera on the market that doesn't apply a gamma curve to the digitised image data before displaying or saving them to a directly-viewable file. How else are you going to show a 10 or more stop subject brightness range in an 8 bit JPEG? Or print a similar SBR on a piece of paper with a Dmax of maybe 1.2D (-4 stops)?</p>

[ondebanks]

<blockquote>

 

<p>Where does "60,000 electrons per pixel" come from?</p>

 

</blockquote>

 

<p>From <a href="http://www.kodak.com/ek/uploadedFiles/Content/Small_Business/Images_Sensor_Solutions/Datasheets%28pdfs%29/KAF-39000LongSpec.pdf">the CCD's datasheet</a>: Saturation Signal spec. is 60 k e-</p>

 

<blockquote>

 

<p>16 bit word depth has nothing to do with signal to noise ratio. It is the number of steps between black and white.</p>

 

</blockquote>

 

<p>But choosing the <em>minimum </em>number of steps between black and white signal does depend on the ratio of signal to noise - the maximum signal to the minimum noise; which is the saturation signal to the readout noise; which is the dynamic range of the sensor. Find the power of 2 nearest to (but larger than) this ratio, and that power is the minimum number of bits required to digitize any signal while correctly sampling the full dynamic range.</p>

<p>Higher bit depth than that (oversampling) doesn't do anything further for signal discrimination, but does resolve the noise levels, which can be helpful in combined/stacked images or combined/binned pixels.</p>

<p>All the cameras I am aware of use even numbers of bits in their ADCs (the most common being 10, 12, 14 or 16 bit). ADCs with odd numbers of bits do exist, but don't seem to be used in the photographic industry. So in the example I gave above for the P45+, I referred to 12 bits being the appropriate bit depth, 14 bits oversampling to a generous degree, and 16 bits being overkill. </p>