1Ds MKII vs scanned 4x5

[BernardLanguillier]

Hi there,

Very interesting results, thanks for posting.

I should be in a position to do very similar tests soon between correctly scanned Ebony 4*5 images and my Kodak SLR/n.

I am not sure about A2 based on actual results, but for sure A3+.

Regards,
Bernard

[Jack Flesher]

Fair enough Jack.  I have found I can go beyond A3 up to A2 or sometimes bigger if I take the same care with my Kodak 14nx shots as I would wth my large format shots.  

IMO we are getting to the stage of splitting hairs and there is no right or wrong here.  As a practical matter, if we print at 300PPI or 240PPI makes little difference as the prints get bigger, so it becomes an academic call...  

Plus I think the 14n/c has an advantage in the detail department over the Canon due to the AA filter.  A sharper initial file will uprez better, so you can possibly eek the next paper size in many situations.  

Regardless, as Samir alludes, we are getting to the point where the biggest image limitations in 35mm-sized digital sensor technology is lens design, not the sensors themselves.  For this same reason, MF digital will probably always outperform 35mm digital with bigger sensor sites and more sensor real-estate...

In fact, in a similar vein, it may also be true that drum-scanned 4x5 will always outperform MF digital.  But as you point out, convenience in workflow becomes an issue -- as does the per-image costs.  

Regardless, as Michael pointed out in his article, how much better does it need to be if our capture technology is already capable of output sizes that are larger than we regularly print?

Cheers,
Jack

[Jack Flesher]

Have you looked at the product?

Hi Bob:

I mentioned Q-image specifically in the article -- and yes, I have looked at it.

It does a decent job on some images and moreover has several different interpolation algoritms available to choose from.  For example, "Pyramid" does a great job with intricately detailed subject matter, though it falls down with smooth tonal regions.  I found their "Vector" to do a better all-around job with most prints, giving good detail and reasonably smooth tones.  But both methods were pretty slow by comparison to my method, and in the end did not produce meaningful differences in the final print -- at least IMO.

~~~

As for what resolution to print at, this is a topic that warrants another complete article...  And like interpolation, there is little agreement.  

Without writing an entire article, the jist is that printers print at some native print resolution, and any image you send to them at a different resolution gets interpolated to the printers native resolution by the print driver before being printed.  For Canon this is 600PPI and for Epson this is 720 or 1440 depending on the printer.  So instead of sending a 300PPI image to the printer, there is a camp that believes you should use your imaging program to interpolate to the printer's native resolution because then the print driver will not have to interpolate; the assumption is the imaging program can do a better job of interpolation than the print driver, and the user has more control over the final file.  (A corollary belief is that you should at least send even multiples of those resolutions to the printer; 300PPI to Canon and 360PPI to Epson for optimal results.)

I have done the comparisons for myself and suggest you do the same.  I printed multiple 16x24 prints and compared them directly.  The bottom line was that under the loupe, the 720PPI print may have had a tad better detail.  HOWEVER it did not have as smooth of tonal gradations.  Mind you, these differences were small, and the detail difference was NOT visible to the naked eye -- I needed the loupe to see it.  HOWEVER#2, the smoothness issue was visible to me at critical viewing distances with my naked eye -- they were subtle, but nonetheless there.  

So for me, I decided to stick with the conventional method and I size to 300PPI as standard and send it to my Epson 9600.  For a really big print, I may even go to 240 or possibly 200 to help gain some additional size.

Try it yourself with your printer and see what if any differences you detect.  I do believe that perhaps this was a worthwhile effort with older model printers (and still may be if you own such), as the print drivers were probably not as sophisticated as they are today -- but like everthing else, drivers (and printers) keep getting better too  

Cheers,
Jack

[didger]

Both produce outstanding image quality, but one is more 'outstanding' than the other in qualitative terms if not usability and convenience.

This is surely nearly incontestible.  I've wondered, however, based on what we know from 1dsMKII and from MF backs, how do you think an MF DSLR with about the pixel density of 1dsMKII would do?  That would be something on the order of 35 Mpixels or so (depending on exact pixel density and sensor dimensions).  Do you think that  this would rival 4x5 transparency scans for the most demanding large print applications?  I realize you can get still bigger files from the scans, but realistically speaking (taking into account grain) do you think that MF can potentially rival 4x5 sharpness with no caveats?

Quentin, in view of the fact that you've mentioned that the noise related to no AA filter on the Kodak cameras is significant and in view of the fact that I've even got some 1ds images that are completely unsalvagable because of color artifact garbage, do you think that the extra sharpness with no AA filter is really a good tradeoff?  How much hassle has this been for you?

[Lin Evans]

One other factor which needs to be considered is the actual subject matter in terms of both detail and even more importantly field of view.

Let me give you an example. Digital, by its very nature is resolution limited for enlargement potential, while film is grain limited. As we can see in the drum scan of the 4x5 grain intrusion is becoming an issue at this enlargement. Because of the low noise and relatively clean appearance, digital images may be enlarged to the point where resolution deficit begins to manifest as "marker pixels" (pixels resulting from insufficient sampling sites to properly define detail). In the case of the name plate, it's obvious that there were too few pixels with the 1DS Mark II to properly define the detail. If you look carefully at the crops of the word "GROSS" you will see that the well defined font is apparent on the 4x5 but the end curves on the "S" are missing and more or less a "blob" on the 1DS Mark II enlargement.

So there is no argument at all that the 4x5 has a decided resolution advantage - as you say, "no surprises". On the other hand, had the subject been closer and occupying a good deal less "real estate," such as a photo of a piece of computer equipment which filled the frame, enlargements from the 1DS Mark II can be made considerably larger than with the 4x5 while maintaining superior image quality.

We've discovered this by shooting computer mass storage equipment for trade shows (mid-range tape libraries) where we make full life-sized prints for trade shows from our eleven megapixel 1DS which at 96" print dimension (on the long axis) far outperform what we can do with our 4x5. The 4x5 simply has too much grain at this enlargement size to render a nice, clean image. On the other hand the 1DS with only eleven megapixels resolution produces incredibly clean and detailed prints. Why? Simply because eleven megapixels are "sufficient" to capture all relevant detail in the relatively small geography frame.

My point is that the more geography which is contained in the frame, the more resolution is "required" to render detailed enlargements. Which approach (film or digital) is best depends on the individual circumstances.

For highly detailed landscapes printed at 40x60 the 4x5 wins hands down for single frame capture. For small frame geography where resolution isn't challenged, it's another story entirely.

Best regards,

Lin

[Graham Welland]

Just for fun & giggles I took a couple of shots using my Mamiya 645AFD/DCS645M with 55mm f/2.8 and Wista 45SP with 150 Schneider Symmar f/5.6. The view camera shot is nothing special and I didn't set up any movements for maximum DOF so it's a sort of fair comparison shot.

Mamiya 645/DCS 645M/55mm f/2.8 at f/16:
1k x 1k Basic Shot - not too big.
2k x 2k - 50% of full size, capture sharpened & colour matched only.
100% resolution crop
another 100% resolution crop


Wista 45SP/Symmar 150 f/5.6 at f/32:
1k x 1k basic shot.
2k x 2k - 20% of full size, capture sharpened & colour matched.
100% resolution crop of scan - capture sharpened only.
Another 100% detail.

I couldn't upload the full res versions as the DCS image is 16MB and the scan is 520MB 11k x 9k resolution!

The film is Velvia scanned on an Epson 2450 using SilverFast Ai Studio. 2400 dpi. Obviously an Imacon or drum scan would have better shadow detail and DMax. I can't vouch for the 1DsII quality but the Kodak is no slouch in this department. Film still wins in my book - but it's not by a lot.

Enjoy!

[David Mantripp]

FYI I always print at  360 or 720 PPi (dpi) on my Epson 1270, and would not print at 300, the difference between 300 & 360 is clear...

Really ? On my Epson 2100 I'm pressed to see any real difference above 240 dpi.  Colorbyte Software (makers of Image Print) claim that anything above 200 dpi is more than their RIP needs.  And I've certainly got the impression from various experts that anything over 300 is a waste of ink.

[Jack Flesher]

Since it's pretty obvious that 4x5 film can beat a 1ds2 file, how would 6x7 medium format film compare?  I'm curious, since I don't have the tools to do any testing.  I realize Michael has done his "1ds is better than medium format film" article, but I thought that was only for prints up to 13x19, and then with the 1ds2, I believe he said it's equal to MF up to 16x20 max.  Is this correct?

T-1000

When I first got my 1Ds, I compared it to drum scans of my 645 and 6x7 MF images.  What I found there was that the 1Ds was marginally better than the scanned 645 (Contax 645 with Zeiss lenses) and was marginally lagging behind scanned 6x7 (Mamiya RZ and M7 with Mamiya lenses).  The 1Ds fell right inbetween the two, and they were pretty close to begin with.

Enter the 1DsII.  IMO it is enough better than the 1Ds that I can safely say it trumps drum-scanned 6x7.

Cheers,
Jack

[Jack Flesher]

Any direct comparison pictures yet?  The only ones posted so far were by Akiss, but shot at f22 and there was virtually no sharpness difference.  f8 would certainly show a bigger difference, but 2 pictures are worth a thousand opinions.

Yes, but not mine since I had to sell my 1Ds before I could buy my 1DsII -- and I posted the link where you can download DNG versions of them the last time this question was asked:


http://www.outbackphoto.com/reviews....II.html

[didger]

and I posted the link

Yes, and I looked at that pretty carefully too.  What stood out was that the difference between 8 Mpixels and 11 was quite large and the difference between 11 and 16.7 was rather small.  The most important differences noted in that article were at high ISO.  I still haven't seen anything as simple and clear as Akiss' two photographs showing that with a Canon f2.5 compact macro 50mm lens used at f22 there was very little difference for the detailed hard edged material he shot at ISO 100.  The only objection to the test is that it was shot at f22.  However, without seeing more test pairs, I would take this Akiss test and the fact that 8 to 11 is a big difference and 11 to 16.7 is a small one to mean that if you only shoot at ISO 100 then the money and 50% extra processing and storage overhead is nowhere near worth it for me (even if I had the money anyway).  

The article and the Akiss test quite obviously vindicate the supspicion I had right from day 1, months before the camera was released, that we would not see resolution improvement anywhere near reflecting the 50% pixel increase because lens resolution becomes too limiting, even with world class lenses.  A Canon 50mm compact macro is not exactly garbage even at f22.  I would only expect truly world class lenses shot at optimal aperture to give even a relatively minor real improvement in resolution, based on the article and the Akiss test.  I'm open to be shown differently.  It's not just about arguing or intellectualizing for fun.  If I'm shown a big enough difference in a real life type clear test, I could end up saving my pennies.  As far as what I've been show so far... no way.

[didger]

The article from that link you posted mentioned about the much bigger gap between 8 and 11 than 11 and 16.  I don't have a printer, so I can't try that.  In any case, it's at this point only worth just so much hassle, since I don't have the money to buy a 1ds2 any time soon anyway.  I'm inclined to skip 1ds2 and wait for 30+ MF DSLR and then evaluate as carefully as I can whether to buy a by then cheap used 22 MF or a 30+ if the difference is worth the money.  One thing is for sure, there's no easy answers.  You have to do a lot of research and take everything with a grain of salt.

[Jack Flesher]

The article from that link you posted mentioned about the much bigger gap between 8 and 11 than 11 and 16.

Not exactly -- the article reported that one photographer who downloaded all of them and printed them got that result.  And that photographer did not secify how he processed each image...

It took me about 3 MONTHS of playing with my 1Ds files before I got REALLY exceptional results.  The curve with the 1DII was much shorter -- weeks -- but I had several raw converters to experiment with right out of the gate.

With the 1DsII I am currently limited to DPP (totally sucks) and Bibble, which is a great interface, but I have NOTHING viable to compare its output to.  Hence my reluctance to make any concrete claims.

BUT!  In the end, the 1Ds remains a VERY fine -- no, excellent -- imaging tool.  And I expect that even when I get the 1DsII performing at its peak, the differences will not be significant until I get to 16x24 and larger prints.

[didger]

You're all thinking too small.  Remember the guy that converted a big truck into a camera and made almost billboard size prints straight from the camera?  Scan and process that, you wimps.

[Jack Flesher]

Here is a quick and dirty comparison of the two...

First the digital specs: I shot the digital image with my 1DsII and 35mm f1.4L lens at f8, ISO 100. Exposure was 1/320th @f8. I used the 35/1.4 because it is an excellent lens, and I have a comparable focal length, and equally excellent lens for my 4x5. I shot it at its sweet-spot aperture of f8.

Film specs: I shot the film image a few hours later with my Phillips 4x5 (a high quality, rigid, 4x5 field camera), a Rodenstock 135 APO lens (equivalent FOV to 35mm lens on full-frame), and Fuji "Astia" film (ISO 100). A thin cloud had moved in and softened the harsh light, hence my comment that this is not an ideal comparison. However I think the results are pretty clear anyway. This exposure was 1/15th @f22. F22 is the sweet-spot for this LF lens. (While the exposure times are vastly different, they also represent very "real-world" exposure criteria for the two differing systems.) Obviously both cameras were mounted on a tripod for the exposures. I then scanned the 4x5 transparency on my Imacon PrecisionII at it's highest true optical resolution for 4x5, 1800PPI.

IMPORTANT NOTES: I did process both images lightly and apply what I feel was optimal sharpening for each image, but I made no attempt to color match the two images.  Also,
to give you the same size images to view, I had to downsize the 4x5 image by 1/2 in each dimension to match the final size of the "actual pixel" crop out of the 1DsII's native file. I used CS's Bicubic Sharper in one step to do this. So in effect, the scanned 4x5 image is 4 times as large as the 1DsII native image, and has even more detail than what shows here...  

Conclusion: Scanned 4x5 remains King of detail by a pretty significant margin...

Here is the album: http://jack.cameraphile.org/gallery....album06

Cheers,
Jack

[Jack Flesher]

Conclusion: Scanned 4x5 remains King of detail by a pretty significant margin...

To be expected, but perhaps more important is at what print size would the difference become apparent?  I supect you could go quite large before the 4x5 advantage was material.  I would expect the 1Ds II or Kodak SLR/n or c to hold their own up to around A2.

Quentin

I actually plan to do this comparison on Friday.  

I expect however Bernard is correct and A3 is about it -- that happens to be roughly where the native 1DsII file will print at 300PPI...

By contrast, the native 4x5 scan will print almost to full A1 at 300PPI...

Any way you cut it, there is simply a lot more detail in the scan.

[BobMcCarthy]

Wide lenses is where I'm at and is the reason I've remained bi-lingual owning both Canon and Nikon systems.

I do like the look of my wide nikkors over my "L"'s. With film, I like what I see. It doesn't seem as clear cut with digital as more "stuff" comes into play namely unequal sensor tech.

Canons CMOS has given them a major jump ahead on the industry. Nikon may catch up in sensor tech, especially now that they have thrown their lot  in with the "big guy" (Sony) with their high end sensors.

But, I'm fairly convinced that the lens is rapidly becoming the bottleneck. 645 real estate (or larger) just makes the work for the lens easier. I saw somewhere (maybe here) where it would be possible to make a 100mpxl chip using current P&S technology in ff 35mm. But to what end! On the otherhand, being a tech type, I can forsee the day where coke bottle lens can be made to look pretty good with smart software. Good enough is the question!

I suspect we'll learn alot when the D2x hits the market. I sense it's make it or break it time for Nikon. I for one hope they show us something special. My opinion is we NEED two strong competitors in the marketplace to keep each other on their toes and to keep money in our bank accounts.

It looks like everyone else is becoming marginalized.

[mattlamb]

FYI I always print at  360 or 720 PPi (dpi) on my Epson 1270, and would not print at 300, the difference between 300 & 360 is clear...

between 360 and 720 up real cose you can see a difference but at 2ft its gone unless your an owl.


I look forward to a 1Ds and 1Ds mark II print of after rezing up the 1Ds file to match the mark II.

Matt

[mikebinok]

Plus I think the 14n/c has an advantage in the detail department over the Canon due to the AA filter.  A sharper initial file will uprez better, so you can possibly eek the next paper size in many situations.

Do I understand you to mean that the Kodak even outperforms the 1Ds Mark 2 in terms of detail?  I realize that it will be lopsidedly the other way in high-ISO performance, autofocus speed, etc.....

[Quentin]

Quentin, in view of the fact that you've mentioned that the noise related to no AA filter on the Kodak cameras is significant and in view of the fact that I've even got some 1ds images that are completely unsalvagable because of color artifact garbage, do you think that the extra sharpness with no AA filter is really a good tradeoff?  How much hassle has this been for you?

D,

It is an issue.  More work is needed, and I prefer using ACR 2.3 because its colour noise slider removers 90% of the problem in 90% of cases.  Occasionally I get a problem case I can't so easily salvage.

This is one area where an AA filter makes a difference.  Kodaks decision to leave one off the 14n.SLR/n /c series was as much to do with saving costs as with sharpness, and previous Kodak dslrs had a user-removeable AA filter.

The loss of sharpness is a trade-off.  Its one I'm happy with but others are not.  

Quentin

[Graham Welland]

Don't worry, I'm not ditching the digital equipment any time soon ......

....

What I actually thought when I saw your comparisons was that the  Mamiya / digi back solution looked more than "good enough" for most purposes.  I also find scanning LF at 2,000 dpi on my drum scanner pulls out more than enough detail: beyond that and you increase file sizes to silly levels with relatively minor gains (unless you want to cover the side if a house).

Quentin

Quentin,

Realistically, I agree with you. The digital shot results are more than 'adequate'  and if I hadn't taken two shots of the almost identical scene I'd probably never want for anything more in a large print from either, at least at A2/A3 type of size. It was fun to pixel peep though.

4x5 at 2400 is pretty unmanageable even on my dual G5 with 4.5GB ram. The 4x5 ends up as a 4GB 16bit PS file which is ok once it's loaded but takes a while to load/save. It's definitely overkill.