I’ve read countless threads on various forums with people debating about the quality of one or the other, but have yet to see actual photos posted of the same images digitized multiple ways. Has anyone done any such testing with 35mm negs and would you be willing to post results, including 100% crops?
There are several YouTube channels with such comparisons. Here’s one I saw earlier. https://www.youtube.com/watch?v=to1rAZEELcs
I think it’s probably because to do so with any degree of objectivity and present the results clearly would be very time consuming, and you’d have to have all the requisite gear to hand as well. If you are comparing colour negative inversions using different software by different methods that brings in yet another variable and one that is especially prone to subjectivity since there are so many ways to arrive at a finished result whether it is in NLP, Silverfast, Flextight, Noritsu, Frontier etc. etc. Youtube videos are fine but inevitably they seem to be short on detail and long on entertainment.
Thanks to Vlad’s Test Target (and to a certain extent the more rigorous but less widely used USAF 1951 etched target) you can at least get a good idea of the relative resolutions possible from various cameras, lenses and scanners but yes, you do have to surf the forums a bit. Richard Karash is obviously a great resource for camera scanning and for macro lenses there is also coinimaging and closeuphotography though there are quite a few potentially eligible lenses that don’t appear on either so it’s back to the forums for those.
The lens testing is definitely very useful! Thanks.
I’d be happy to see a side-by-side of an Imacon or even Plustek scan vs the same neg scanned carefully with a camera. It would be worth about 10,000 pages of internet chatter :). I’ve heard people claim they’ve bested drum scans with a d800, but have yet to see the evidence. So far I found one Youtube video comparing a Primefilm scanner with dSLR, and the results were pretty close, but that’s just viewing on Youtube.
I watched that video earlier this week. Interesting stuff, but I wish there was an Imacon or Plustek thrown in the mix.
I see that I’d already watched that one also as I’d bookmarked it. I like his straightforward approach but thought it pretty strange that he down-sized both files to 3000 x 2000 pixels (6MP from around 35MP) and re-saved them, then compared these against each other at 600% or more. Difficult to come to any serious conclusions really though I did think that the Canon R5 camera scan with the 100mm Macro looked cleaner to my eyes in terms of detail. For the colour and the ‘look’ then a lot would be down to the sliders and presets used in both NLP & Silverfast.
(Only just realised that the ES-2 thread is yours as well)
I felt the same way. There’s a dirty quality to the scanner output, and I get the same queasy feeling from the Plustek.
This is a little more interesting: Comparing Imacon X5, Creo iQ3 and Sony A7R IV scans | Photrio.com Photography Forums
I think in this the Sony 35mm scan looks good, but the Imacon shows more film grain (unless that is noise?). More grain means it’s resolving more. This almost steers me away from the camera method.
That is indeed an interesting discussion as it brings in all the main strands of topics that surround the differences between high end scanners and camera ‘scanning’, Bayer arrays, pixel-shift, RGB CCD line scanners etc. He (Steven Lee) particularly makes the point that he’s not overly concerned with ultimate resolution but he takes it upon himself to send out the negatives so that they can be scanned on the Eversmart IQ3 and so we can all compare them at 1:1 for just that.
I think that it’s important to note that in the previous thread that prompted this one he states that “the X5 scans all have aggressive sharpening applied, which the lab claims they couldn’t disable. Subjectively, I think it makes grain look unnatural, while I know others may prefer it”. I don’t know what the lab means there, it doesn’t make sense, of course it is possible to control the level of sharpening (USM) from -120 to +120. I imagine that we’re looking at the tiff rather than the 3F but that is way more than ‘0’ to my mind. Interesting also that he has reduced the sharpening for the Sony from the default 40% to 10%.
There’s quite a lot of discussion of whether you can get pure colour from a Bayer array camera and it’s hard to come to any conclusion from this. The Creo guy (Michael) mentions the reds and indeed they do seem to pop on his scans but otherwise unfortunately the Creo scans look somewhat out of sorts.
The X5 is certainly capable of resolving the grain/dye cleanly but to me it’s way too prominent. Personally I’m pretty impressed with the Sony on the basis of what we’ve seen here.
This is a complex subject that needs careful analysis using professional test materials and rigorous procedure. For testing scanners I’ve used LaserSoft Imaging’s USAF 1951 target (as the Flextight/Imacon needs flexible media) and for my Sony a7r4 I used a Thorlabs R2L2S1N1. This latter target is etched in special glass using an exacting proprietary procedure and costs over $200.
Please understand with scanners that the stated optical resolution (which depends on the technical spec of the sensor) usually exceeds the measured performance which includes that of the lens and remainder of the optical system. Likewise with a camera one depends on not only the sensor specs, but also the optical resolution of the lens and any degradation that enters the digitizing set-up. Therefore the only reliable indicator of true resolution from either approach is a measurement of the end result of properly scanning a professionally made target that returns lp/mm, which one can then convert to PPI.
As well, with a camera set-up that permits greater than 1:1 magnification, effective resolution can be considerably increased by capturing the media at greater than 1:1 magnification and stitching the resulting images to produce the whole image. Lightroom makes this easy to do.
The highest resolution I’ve seen from a desktop scanner comes from the Flextight/Imacon (theoretical optical resolution 8000 PPI), and it measured about 5800 PPI from the USAF1951. The highest resolution I achieved with the Sony a7r4/ (61 MP camera) with Schneider Apo-digitar flat-field macro lens, using the Thorlabs target was in the range of 7315 to 8100 PPI using nearly 2:1 magnification, the range depending on how rigorously one views what constitutes a true line pair separation.
It is possible that some of the observed difference of results between the Imacon and the camera tests can be attributed to quality differences of the USAF versus Thorlabs targets. I can’t test for this because a Thorlabs target cannot be used in a Flextight/Imacon scanner. The only other scanner that came close to the Flextight/Imacon in my tests is the Minolta Dimage Scan Elite 5400 at about 5000 PPI. The other desktop scanners (Nikon, Epson, Plustek) I was able to test were well below.
The upshot of all this is that there are no easy answers to the question the O/P put. It all depends on equipment and testing conditions; but the observations I was able to make suggest that it is possible to obtain superior resolution from a higher-end camera/special-purpose lens combination than from desktop scanners - at least the range of them I know of.
I would add that the colour quality and the options and technologies for achieving high colour quality using a combination of NLP and LR Classic can outperform what is achievable confining the whole workflow to any scanner software alone, especially considering the targeted editing features now available in current LR releases.
I agree with all that you say but you might be short-changing the X5, filmscanner.info measured the X5 at 6900ppi in the portrait orientation and 6150 ppi in landscape with a USAF 1951 target, I suppose they must have used the mounted slide holder. Also, I can thoroughly recommend that the O/P reads your article.
Hi Harry, thanks for your comments. I too was somewhat surprised by the results we got from the Flextight test. The device we used was the 848, not the X5; while they are supposed to be essentially the same machine, if my memory serves me correctly, the X5 is more recent than the 848 so perhaps it was improved a bit. Anyhow, the difference of reported results doesn’t surprise me very much, especially as this test is not totally objective - there is a subjective element interpreting the extent of separation of lines and deciding what to accept as “separated”.
Sorry, I hadn’t realised that you had measured it yourself, you may well be right and there is subjectivity involved as you say. I often refer to filmscanner.info because they seem to be generally realistic and more than happy to debunk the often rather spurious claims of manufacturers but I’m sure you know that site well.
Yes I do know that site - Patrick Wagner is normally quite thorough and on-target - one of the better resources on scanning matters.
Harry and Mark,
Thank you both so much for the thoughtful responses. I think based on all of this information, I’m going to give camera scanning a try. The only question now is specifically what equipment to buy. It looks like people have good results with Sony and Nikon both, and some well-regarded macro lenses are not too expensive. The camera is the tricky part. I’m looking at DXO scores and all that stuff, but sometimes the rendering is more intangible than the numbers reveal.
Any thoughts on Nikon vs Sony? The d800e or 810 are roughly the same as the A7R II, but the latter is 42mp vs 36 on the Nikons. Lenses would be either the Sigma 70mm Art or 55mm AIS Micro Nikkor (both from tier one of Richard’s list). Either of these options is in my comfort zone in terms of price. I’m slightly leaning towards Nikon due to the potentially better prices where I live, but the Sony is fine too. I would just have to learn some new kanji and expressions, since Sony kind of softly “region locks” their cameras by menu language (no English on Japanese models and vice versa). Not a deal breaker, just slows me down a little initially, as my Japanese reading skills are still intermediate.
For now I’m going to use the ES-2 holder. If I find any flatness issues, I can revisit. That lets me get a feel for this without getting bogged down in copy stand alignment, etc. If things go well, I may order that Valoi down the road.
Light source will initially be either flash or a high CRI video light. If I get the Valoi later, that can replace this as well.
I think that the original A7R had the same sensor as the Nikon D800 & D800E, certainly the Nikon sensor was made by Sony (no low pass filter on the ‘E’). The 42MP A7R2 came along 18 months later and so possibly, apart from the slightly higher MP sensor, there may have been some operational wrinklles sorted out as well. If you are going for the ES-2 or similar then the bulk of the Nikon won’t matter but on a stand I would prefer mirrorless - lighter, angled screen and focusing aids.