Ektar 100 versus Alpha 900, Velvia 100, Portra 160VC and TMax 100

<p>Sorry, too many abbreviations being thrown around. What's N.I. and NE?</p>

<p>NI = Neat Image</p>

<p>NE = NI misspelled</p>

<p>Also, Mauro, I'm having some trouble understanding this LPPH vs. lpmm business.<br>

<br /> Marco claims that 2000 LPPH where the resolution test chart takes up the full frame translates to 40 lpmm. Since a frame has 24mm height, that's 2000 lines per 24mm, or ~83 lines per mm. Which he then translates to 40 line <em>pairs</em> per mm.</p>

<p>So then that essentially means that when they talk of LPPH, they're actually talking about 2000 successive lines with no blank spaces in between? Kind of like saying 2000 rows of pixels per picture height. OK, I guess that makes sense. Let me know if I'm wrong.</p>

<p>In which case, your calculation now makes sense to me. Because if you're seeing 2700 LPPH (well, that's what I see anyway for your Ektar example), then that's only 1350 line pairs per picture height, which is 56.25 lpmm. If you then assume that 2 pixels are needed to represent each line pair, that gives you 112.5 pixels per mm. Or 2700 pixels per 24mm and 4050 pixels per 36mm. Multiply those together and you get 11 megapixels. Which is what you're assigning the Sony & what I'm assigning the Ektar scan since I think they're on par with one another.</p>

<p>So now it'll be interesting to see what the Imacon tells us about the Ektar film.</p>

<p>But now I'm really curious... If at 2700 LPPH for your Ektar scan (my subjective resolution limit estimation), it's resolving 56.25 lpmm, and under your estimation of 3700 LPPH it's resolving 77 lpmm (3700/2/24), then how the heck are you saying that you've seen 160 lpmm on Velvia?</p>

<p>That seems like insanity, given that I already think your 3700 LPPH is an overestimation, no offense :)</p>

<p>Confused,<br>

Rishi</p>

<p>"So then that essentially means that when they talk of LPPH, they're actually talking about 2000 successive lines with no blank spaces in between? Kind of like saying 2000 rows of pixels per picture height. OK, I guess that makes sense. Let me know if I'm wrong."</p>

<p>You are correct. Just like true-pixels with actual information (not DSLR fake-pikes).</p>

<p>3700 lines per picture height = 154 lines per mm. (not line pairs).</p>

<p>Velvia has 160-170 lines per mm. (not line pairs).</p>

<p> </p>

<p>Haha, man I am thick in the head today.</p>

<p>So, out of curiosity, what's the correct abbreviation for 'line pairs per millimeter' and 'lines per millimeter'?</p>

<p>A great piece of work. Well done and thank you Marco. It is amazing to see the differences between film and digital, but even more striking is how all the pictures look good in their own way. That's why so many people shoot both formats I guess.<br /><br />Like some of the other readers, I'd love to see a test of similar quality using MF.</p>

<p>Marco -- how come you haven't commented on Mauro's findings? It really does seem to me like maybe your Ektar example suffered from focusing issues? Especially in light of the fact that the Velvia seemed to have resolved all the lines... I doubt Ektar is that *that* much worse as your results show...</p>

<p>Still, isn't it interesting to see how one person's interpretation of the results leads to a resolution calculation of 11 megapixels (mine, since I put the Ektar/LS-9000 combo at 2670 LPPH), whereas someone else's interpretation of the <em>same</em> results leads to a resolution calculation of 20 megapixels (Mauro's, since he puts the Ektar/LS-9000 combo at 3700 LPPH). No wonder we can't reach a consensus!</p>

<p>Also, Mauro, I do think you need to concede that if you think the Sony A900's 'real resolution', according to <em>your own interpretation</em> of the data over at dpreview.com, is only 11 megapixels, then that's also the resolution of your Ektar/LS-9000 combo. Or, conversely, if you think that your Ektar/LS-9000 combo resolution is 20 megapixels or more, then the Sony really is close to its rates resolution of >20 megapixels. Because when putting those line chart at 800% next to one another, the Ektar certainly does not exceed the Sony in resolving ability for the same LPPH. Wouldn't you agree?</p>

<p>Cheers,<br>

Rishi</p>

<p>"So, out of curiosity, what's the correct abbreviation for 'line pairs per millimeter' and 'lines per millimeter'?"<br>

I don't know. It is probably best to clarify on every post...</p>

<p>"Or, conversely, if you think that your Ektar/LS-9000 combo resolution is 20 megapixels or more, then the Sony really is close to its rates resolution of >20 megapixels."<br>

<br /> Not true. DSLRs are advertised knowingly that only 2/3rd of the pixels have actulal luminosity/color data. The only true statement from DSLRs is when they state for e.g. 24megapixels (true), but it becomes a lie when they add resolution to it (although they ussually avoid it to prevent lawsuits). Still confuses the buyers ingeniously.<br>

<br /> On the other hand, a Coolscan 9000 provides very close to true 20mp in 35mm and 90+mp in 6x7. The rest is out for the film's resolving power.</p>

<p>If it weren't because of the above, it wouldn't make sense to test the resolution of a 24 megapixels camera, would it? That is why DPreview, etc test every camera to provide buyers with a value for resolution.</p>

<p> </p>

<p>e.g: the A900 ha a 24 megapixel output and has aproximate 12 megapixels of RESOLUTION.<br>

You can double that with a 35mm Coolscan, if you have the film providing the data.</p>

<p>Also the 12 mp resolution of the A900 is a soft number because DPreview's test was not optimal. It could be higher... say 16 megapixels..?<br>

A correct way to evaluate resolution is by placing the chart so it extiguishes resolution in the lower numbers. Say 1-5 not 10 to 20. The two main reasons for this are:<br>

<br /> 1) the jump in resolution at the lower numbers of the chart is much higher so it makes it possible to evaluate resolution better. For e.g. If a camera resolves step 2 of the chart, it comfortably resolves step 1 (half the linear resolution), and will clearly show it doesn't resolve step 3 (150% of the linear resolution).<br>

When you are trying to evaluate resolution between the mark 20 of the chart and the mark 21, the jump in resolutions are just marginal. It becomes very subjective.<br>

2) the second reason is that the cross patterns at the corners of the charts are useless otherwise since the camera will outresolve them.</p>

<p>Interestingly enough, Popular Photography back in 2004 rated the LS-9000 at 3414 ppi (67.2 line pairs per millimeter), which would give you 15.6 megapixels scans at most.</p>

<p>Which places your >20 MP resolution of the Ektar/LS-9000 combo a bit high -- but now I'm curious to see what the Imacon will resolve.</p>

<p>Mauro, do you, or anyone else, have a link to a LS-9000 review that places its true optical resolution near 4000 ppi? I remember reading this somewhere, but can't find the reference.</p>

<p>Rishi</p>

<p>The Coolscan is not true 4000. My particular scanner resolves 3900 in one direction (the one I use for film testing) and 3650 in the other direction. These are what I use to calculate true resolution.<br>

35mm=> 20.2MP<br>

and 6x7=> 91.7MP<br>

This is a link to an extensive review with similar results:<br /> http://www.filmscanner.info/en/NikonSuperCoolscan9000ED.html<br>

Also remember Rishi that each pixel of a Dslr captures only one color per pixel whereas the coolscan captures every color on every pixel. Massive difference.</p>

<p>True Mauro... wait a minute. So when Canon advertises 21.1 megapixels... are they saying that the effective resolution is 21.1 megapixels, or that there are exactly 21,100,000 pixels on the CMOS sensor, or that are actually many more pixels but that after interpolation & Bayer pattern processing and whatnot, you end up with an effective 21.1 megapixels?</p>

<p>Thanks,<br>

Rishi</p>

<p>Les, thank you for the links.</p>

<p>Rishi, I hope the question is sarcastic... obviously you have only 21 million pixel but the effective resolution is much lower. This has always been the case - not a new development.</p>

 

<p>Les, can you explain the 3650v 4100h resolution to me. That's better than mine grrrr!</p>

<p><em>Daniel, sadly you are quoting flawed or ill itendened marketing oriented examples.</em><br>

Mauro, not one URL I pasted has anything to do with marketing. They are all real comparison images performed by people just like you and I.<br>

<em>If you have a true interest I'd advice you to test yourdself.</em><br>

Been there, done that, have the prints to show for it. The thing is though, I test real scenes printed to real paper. I'm not all that interested in high contrast test chart resolution because I don't shoot high contrast test charts.<br>

If all you're looking for is a number on a chart, then I'm sure you can soup a B&W emulsion to beat the A900. If you're interested in overall image quality, then just look at Marco's samples again. The A900 looks like a larger format compared to the 35mm, with a target that favors the film.<br>

Not to throw a wrench into a healthy discussion about line pairs, just pointing out a flaw with the idea that line pair numbers tell the whole story about image quality.</p>

<p>Daniel, yes it is all about the print. <br /> A couple questions regarding how you have formed your experience:<br>

1) To form your experience with digital; What digital camera have you shot more often and what printer and size you print?<br>

2) To contrast your digital experience; What film format, camera and films you use for printing on the the same printer and size that led you to your conclusions?<br>

3) What scanner do you have?</p>

<p>If possible, after you tell us about your equipment, please be kind and post a "real life" example comparison of the same subject you shot with both your digital and your film equipment.</p>

<p>Two things come to mind:<br>

First, the Imacon is good, but it isn't a great scanner and far from being the best drum scanner using fluid mounting; Second, scanning negatives requires LOTS of interpretation and judgment.</p>

<p>This means that in the first case, the smoothness as well as the sharpness of the scans from film are compromised and in the second case, the color, saturation and contrast of all films and negatives in particular are all subjective.</p>

<p>You've done a wonderful <em>looking </em> report, but it means very little. To confirm your results, I'd like to see high quality, totally RAW, 16 bit drum scans of each film where there's no interpolation whatsoever. The conclusions might be the same, but at least the test would be more accurate and fair.</p>

<p>Mauro,</p>

<p>I received the film(s) -- thanks!</p>

<p>Unfortunately, I have some extremely disappointing news. I went all the way over to our scanning facility yesterday, with high hopes, only to find that they're closed until January :(</p>

<p>So I'm going to have to hold your frames of film hostage for a bit longer than originally planned -- I hope that's ok?</p>

<p>In the meantime, though, I've been doing some LS-4000 scans at home, as well as dropping them under the microscope. You're right, I flipped when I placed both the TMX and the Velvia under the 'scope. Not so much the Ektar, which paled in comparison to the previous two.</p>

<p>Anyway, I'm putting together some LS-4000 scans as well as my own calculations from under the microscope. But before I post back with that info, let me make sure I'm making my calculations properly.</p>

<p>Note the line on the right edge of the chart and the two crosshair points -- is that the 'chart height' I use to calculate the Line Pairs Per Height? I'm using the measurement tool in Photoshop to get an exact measurement. If so, I'm getting 8.6 chart heights per full height of the frame. If I measure from literally the top of the chart to the bottom I get 7.2 chart heights per full height of the frame, but I'm pretty sure that's the wrong way to measure it right?<br /> <br /> Please confirm. Thanks!<br>

Rishi</p>

<p>Thanks for the links to those scanner tests, Les. Those were exactly what I was looking for.</p>

<p>I'm going to have to say that I feel the resolution for the LS-9000 is overstated, whereas the resolution estimation for the LS-5000 seems spot on.</p>

<p>Take a look here:<br>

<img src="http://staff.washington.edu/rjsanyal/Photography/LS-9000_vs_LS-5000.jpg" alt="" width="292" height="250" /></p>

<p>For each scan, the upper right red circle indicates what I'll call the '6.2 horizontal' (vertical lines indicate horizontal resolution) element, and then lower left red circle indicates what I'll call the '6.3 vertical' (horizontal lines indicate vertical resolution) element. For the LS-5000, it's clear that 6.2 is resolved horizontally, and 6.3 is resolved vertically, which is indicative of 4095 dpi. 6.3 is resolved vertically, indicative of 3650 dpi. Which gives ~20MP worth of information.</p>

<p>For the LS-9000, though, while I agree that it can resolve '6.3 vertical', giving it 4095 dpi in the vertical direction, I feel that '6.2 horizontal' is an overestimation. I can hardly make out those vertical lines. And this is at 300% magnification.<br>

<br /> So does the LS-9000 really perform worse just because of the optical diffuser? In my own scans, the LS-5000 appears sharper than the LS-9000 (but to me, offers too much offensive pepper grain with certain emulsions).</p>

<p>Mauro, I'll try and scan your resolution test chart frames with and without the Scanhancer optical diffuser, to see if I see any differences in resolving power.</p>

<p>Also, does 'vertical resolution' refer to the resolving capability of the actual row of pixels on the CCD, whereas 'horizontal resolution' refers to that resolution achievable by the stepping motor? Think about it for a while, orient yourself and the film and the scanner, etc., then get back to me :)</p>

<p>Cheers,<br /> Rishi</p>

<p>Also, doesn't Nyquist sampling theory hold that in order to truly discern 4095ppi of data (that is, 4095 lines per inch, we'd need to sample at twice that or, namely, 8190ppi?</p>

<p>Let's think in line pairs:</p>

<p>Say the black and white lines don't perfectly match up with the pixels. Then instead of one pixel getting the black line and one pixel getting the white line, one pixel will get, say, 75% of the black line while the next pixel gets 25% of the black line and 75% of the white line. This'll give 75%:25% for the two adjacent pixels (instead of the ideal 100%:0%). In other words, loss of contrast.</p>

<p>I guess that's why the lines in the image above are not black/white at 6.3 and 6.2... but rather 'darker gray/lighter gray'... if we sampled higher (8000ppi), perhaps the contrast would be better?</p>

<p>Am I understanding this correctly?</p>

<p>And if so, doesn't that mean that 8000ppi really will extract more USABLE (not dye clouds, but actual resolution test chart lines) information out of film, given that my own (and Mauro's own) assessment of true optical resolution of film under a light microscope indicates ~26MP for TMX and 21MP for Velvia (and 12.5MP for Ektar 100)... sorry I haven't posted my observations yet, but I will in my next post...</p>

<p>Rishi</p>

<p>Let me state that last question a little more simply:</p>

<p>Mauro and I have clearly seen 20 megapixels worth of information on resolution test chart shots on Velvia under a light microscope (no scanner involved!). In other words, 3665 lines per picture height or 3880 lines per inch. Which is ~1940 line pairs per inch, which would need 3880 pixels per inch to resolve. But then according to sampling theory, wouldn't that mean that we'd actually need to sample at 7760ppi to really get all that data?</p>

<p>That is, an Imacon :)</p>

<p>-Rishi</p>

<p>This thread is really getting down to the discussion of the ultimate resolution (contrast 1000:1). More dpi (1.5 to 2X) for a scanner is needed to resolve the same amount of dpi (line pairs). Otherwise, fuzzy line pairs will appear.</p>

<p>Richard </p>