Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: comparison is unfair to 4x5 Velvia  (Read 30294 times)
jrkeat
Newbie
*
Offline Offline

Posts: 8


« on: May 17, 2006, 11:53:55 PM »
ReplyReply

I though that this article did not fairly compare 4x5 Velvia to the Betterlight Super6Kand the P45.  From the crops, the Betterlight looks clearly superior.  The Betterlight shows greater contrast in the fine details.

The problem is that the chromes were only scanned at ~31 lines/mm (~1600 dpi).  Velvia, when combined with good lenses, is capable of reproducing high contrast details in the 50 lines/mm range.  (See, for example, http://www.hevanet.com/cperez/testing.html, for lenses.  Velvia has listed resolutions of 80 and 160 lpmm depending on the contrast.)  I believe that had the 4x5 Velvia been scanned at higher resolution, finer details would have been visible.  Obviously the contrast of those details would be markedly less than in the original--you can see the contrast dropping off at 31lines/mm--but they would be there.  My experience is that if I scan at 1600 dpi, I miss quite a bit of the detail available on my best chromes.

The dollar bill at the size pictured isn't the best vehicle for looking at features smaller than 31lines/mm.  The smallest features are about 20 lines/mm.

I don't know if the effect would be big enough to make Velvia "better" in many cases or not--I'd like to see the comparison.  I suspect that Velvia would show a discernable resolution advantage, but the Betterlight would almost always look better because of its lower noise levels.  (Of course you need a completely stationary object to use the Betterlight).  Against the P45, I would guess the resolution advantage of the film would be readily apparent.  

On a separate issue, I suspect the drum scan for the Velvia was not done to produce the optimal resolution.  The aperture should be about the size of a pixel.  However, the aperture is round and pixels are square, so adjacent pixels generally overlap a little.  If you're looking for the best possible resolution for a comparison like this, the scan should be done at a higher resolution and then down-sampled.
« Last Edit: May 28, 2006, 10:00:20 AM by jrkeat » Logged
Dmitry Reznitsky
Newbie
*
Offline Offline

Posts: 8



WWW
« Reply #1 on: May 18, 2006, 04:17:40 AM »
ReplyReply

Quote
...

And one more thing:
  I think, that USM should be done not just 'same for all', but more individual to each frame.

  Anywa,  velvia.
Logged

WBR, Dmitry Reznitsky
http://www.reznitsky.info/
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4782



« Reply #2 on: May 18, 2006, 07:26:13 AM »
ReplyReply

You can please some of the people all of the time, and all of the people some of the time, but you can't please all of the people all of the time.

With apologies to Bob Dillan.
Logged
pfigen
Sr. Member
****
Offline Offline

Posts: 450


WWW
« Reply #3 on: May 18, 2006, 05:25:05 PM »
ReplyReply

The Tango drum scanner used has a minimum aperture of 10 microns, so any resolution over 2540 optical is interpolated in one direction. Given that you can see improvement in resolved detail in drum scanned Velvia images up to about 6000 ppi, you really need to scan at that resolution and then rez the digital files to match file size for visual comparison. This is how I've done every film/digital comparison of my own - drum scanning at the highest resolution - 8000 ppi with a 3 micron aperture, just to be safe. It's not about pleasing anyone. It's just another example of flawed testing technique, rightly pointed out by the OP.
Logged
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4782



« Reply #4 on: May 18, 2006, 05:53:54 PM »
ReplyReply

Since the scanning was done by Charles Kramer and Bill Atkinson I'll ask them to respond.

Michael
Logged
Stephen Best
Guest
« Reply #5 on: May 18, 2006, 06:21:34 PM »
ReplyReply

Quote
I though that this article did not fairly compare 4x5 Velvia to the Betterlight Super6Kand the P45.  From the crops, the Betterlight looks clearly superior.  The Betterlight shows greater contrast in the fine details.
[a href=\"index.php?act=findpost&pid=65869\"][{POST_SNAPBACK}][/a]

Put it down to personal bias ... but if you look instead at the top right hand corner of the bill (or left in this orientation, namely the larger "1" and surrounds) to my eyes the 4x5 appears to resolve about the same level of detail as the 6K scan and far exceeds that of the P45 (e.g. look at the small dashes between the leaves). The Canons are awfully soft and no amount of sharpening will bring them up to the resolving power of 4x5/6K.

The sampled areas shown (bottom right of the bill) highlight an area for which the 4x5 is out of focus, whether caused by the subject not being square on, swing or the film not being held flat. How else to explain the inconsistency? Most 4x5 owners know that ground glass placement has to be user calibrated (with shims etc) to match the typical film holder (and film thickness) used ... critical here but less of an issue at usual working apertures. But nobody claimed this was a scientific test.

Other than this focus anomaly, the results of these tests are largely what I'd expect. Not that resolution alone is going to dictate anyone's choice of camera/format ... or is it? :-)
« Last Edit: May 18, 2006, 07:14:03 PM by Stephen Best » Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 8575



WWW
« Reply #6 on: May 18, 2006, 07:05:36 PM »
ReplyReply

Quote
Since the scanning was done by Charles Kramer and Bill Atkinson I'll ask them to respond.

Michael
[a href=\"index.php?act=findpost&pid=65951\"][{POST_SNAPBACK}][/a]

I don't personally know Charles but I do know Bill. The man knows how to drive that Tango. Good enough for me!
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Stephen Best
Guest
« Reply #7 on: May 18, 2006, 07:19:54 PM »
ReplyReply

Quote
I don't personally know Charles but I do know Bill. The man knows how to drive that Tango. Good enough for me!
[a href=\"index.php?act=findpost&pid=65962\"][{POST_SNAPBACK}][/a]

The scan looks fine to me.
Logged
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4782



« Reply #8 on: May 19, 2006, 09:05:07 AM »
ReplyReply

I've discussed the above with Charlie and his reply in part is as follows...

"The Tango 4x5 scan was done at 2300 dpi, resulting in a 16 bit file of around 550 MB. (Not 1600 dpi---where did they get that?)   The 645 Velvia was scanned at 4500 dpi, resulting in a file size of around 421 MB at 16 bits."

Charlie also reports that the scan done on the Aztec was at 8000 ppi, which if the full film area would have been scanned would have produced a 6.5Gb file. (Consequently, only a portion was scanned). This portion, according to Charlie, actually provided poorer detail than that from Bill's Tango at 2300.

Michael
Logged
Bobtrips
Sr. Member
****
Offline Offline

Posts: 679


« Reply #9 on: May 19, 2006, 10:14:08 AM »
ReplyReply

"The problem is that the chromes were only scanned at ~31 lines/mm (~1600 dpi)."

"The Tango drum scanner used has a minimum aperture of 10 microns, so any resolution over 2540 optical is interpolated in one direction. Given that you can see improvement in resolved detail in drum scanned Velvia images up to about 6000 ppi, you really need to scan at that resolution and then rez the digital files to match file size for visual comparison. This is how I've done every film/digital comparison of my own - drum scanning at the highest resolution - 8000 ppi with a 3 micron aperture, just to be safe. It's not about pleasing anyone. It's just another example of flawed testing technique, rightly pointed out by the OP."


Have we just witnessed the birth of yet another LL myth?  

I expect we'll see lots of rants about how Michael claims to that the P45 outperforms 8x10 film and how he biased his claims by scanning the film with a Barbi-scanner.   ;o)


(I've just dealt with a post on another forum that stated there had been a "2x" error with the 645 film sample.)
Logged
Stephen Best
Guest
« Reply #10 on: May 19, 2006, 09:43:43 PM »
ReplyReply

[attachment=586:attachment]

The above is a selection from the 4x5 scan. Look at the outer black line border of the note and you'll see that it's fairly sharp on the left hand side but gets progressively more blurry as it travels along the top ending up quite blurry at the top right corner where the note curls backward. Most of the selection is simply out of focus. In fact, because of the rapid fall-off of depth of field, I'd say the plane of absolute focus is probably closer to the front of the Macbeth chart below. Nevertheless ...

[attachment=590:attachment]

For this image I took a selection from the 4x5, BetterLight 6K and P45 images respectively and upsampled them by 200% with Nearest Neighbor. You can draw your own conclusions, but I'd rank them in descending order as shown with the first two pretty close. I also tried sharpening the P45 with a number of settings but couldn't get it to match the other two.
« Last Edit: May 19, 2006, 09:50:32 PM by Stephen Best » Logged
free1000
Sr. Member
****
Offline Offline

Posts: 400


WWW
« Reply #11 on: May 20, 2006, 07:27:04 AM »
ReplyReply

The conclusion I drew from the test was that 5x4 still offers an advantage over the top end digital backs... this thread confirms it.

But I have bought an Aptus 75. Does the news that its not as good as 5x4 worry me? Not unduly. For my purposes the A75 will be as close to 5x4 as I need. And I know from  experience that a beautiful 5x4 can look worse on the printed page than a 1DsII image if the scanner operator (often out of my control) is a typical repro guy.  I know because I have articles where the 5x4s look muddy and soft next to a 1DsII image which I controlled from soup to nuts.

The beauty of shooting a digital back on a view camera is that, if the shot needs it, I can just slip in a quickload and back up on 5x4.

Moreover... out in the field, with a strong wind buffeting my 5x4, who is to say that wind vibration will not soon remove the advantage of film over the Aptus attached to a more rigid platform?

I don't believe the manufacturer hype about P45's and A75's being 5x4 quality... but  even so, they are pretty fantastic.
Logged

@foliobook
Foliobook professional photography folio for iPad
www.foliobook.mobi
Stephen Best
Guest
« Reply #12 on: May 20, 2006, 07:53:55 PM »
ReplyReply

Quote
The conclusion I drew from the test was that 5x4 still offers an advantage over the top end digital backs... this thread confirms it.
[a href=\"index.php?act=findpost&pid=66106\"][{POST_SNAPBACK}][/a]

Maybe. A more interesting comparison would be between a 4x5 (properly focussed!), BetterLight 8K and a multishot digital back. This could narrow things somewhat.
Logged
jrkeat
Newbie
*
Offline Offline

Posts: 8


« Reply #13 on: May 28, 2006, 08:39:23 AM »
ReplyReply

I did some scans that show that Velvia is capable of showing details WELL beyond the 1600 dpi scans done for this article.  I had some shots that I used for calibrating the focus on my camera that I did with Velvia 50.  (Velvia 100 is widely held to have resolution similar to Velvia 50).

The shots consist of alternating rows of numbers and + signs.  The original image was actually not very high contrast, in spite of being black and white, because the lens was wide open.

First I scanned them at 4800 dpi with an aperture of 6 microns.  I then down-sampled the scan to 1600 dpi and compared the two.  The writing is very readable at 4800, but completely unreadable at 1600 dpi.  (Also, as I noted in my post above, scanning at a higher resolution and down-sampling produces a better image than scanning directly at 1600 dpi, as they did in the article.  So I'm actually being more fair to the 1600 dpi scan than they were.)

At 4800 dpi:
[attachment=632:attachment]

At 1600 dpi:
[attachment=633:attachment]

Conclusion:
Velvia resoves well beyond 1600 dpi or about 8000 pixels horizontally.  The BetterLight 6K back is limited to 8000 pixels.  Therefore Velvia is capable of significantly higher resolution than the BetterLight 6K.
« Last Edit: May 28, 2006, 08:46:22 AM by jrkeat » Logged
Jack Flesher
Sr. Member
****
Offline Offline

Posts: 2595



WWW
« Reply #14 on: May 28, 2006, 09:29:15 AM »
ReplyReply

Quote
Maybe. A more interesting comparison would be between a 4x5 (properly focussed!), BetterLight 8K and a multishot digital back. This could narrow things somewhat.
[a href=\"index.php?act=findpost&pid=66161\"][{POST_SNAPBACK}][/a]

A few points FTR:

1) The BL 8K back is not a practical field capture device due to its slower aperture speeds.

1a) The BL super 6K can scan up to 9000 (interpolated from 6000) by 12000 real, for a partially interpolated 108MP.  I can tell you from personal experience, the gains from the high-res scan setting are real. Bill certainly knows this too and could have shown a scan done so, but I suspect he chose not to due to the controversy the interpolation would have generated.

2) Proper focus is an elusive beast in LF photography.  One, DoF is limited due to the longer focal-length lenses required and two, it is almost impossible to be assured the film was perfectly flat when the exposure was made, but...

3) I have seen a LOT of 4x5 scan files -- and with number 2 in mind, I had the opportunity to see the transparency, be present during the scanning process and inspected the final scan from the test image used here.  I can assure you: the chrome was properly focussed and properly exposed, and the scanning was done with more care than any lab would have done. And most importantly, the resulting scan file was as close to perfect as anyone could hope for from 4x5 film.
Logged

michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4782



« Reply #15 on: May 28, 2006, 10:22:59 AM »
ReplyReply

Wait a second. Your premise is that the Tango drum scans were done at 1600 ppi.

Not so. They were done at 3200. Don't know where you got the 1600 figure from.

Also, they were done again on an Aztek drum scanner at 8000 ppi (only a partial scan because the file would have been about 1.5GB, and according to Charlie it was a quibble as to whether one was better than the other in terms of resolving power. Apparently a synthetic target showed higher resolution, but a Velvia transparency didn't.

This mirrors what has been accepted practice by most scanning professionals for the past number of years. There isn't much more image information on a piece of colour film beyond somewhere between 3000 and 4000 PPI. That's why most high end scanners top out at 4000. Beyond that one is just enlarging grain.

Your experience and opinion may of course differ, but I'm comfortable trusting my own eyes, the experience of respected photographers and technologists like Charlie Cramer and Bill Atkinson, and others whose options I trust.

Michael
« Last Edit: May 28, 2006, 10:23:37 AM by michael » Logged
jrkeat
Newbie
*
Offline Offline

Posts: 8


« Reply #16 on: May 28, 2006, 11:23:59 AM »
ReplyReply

Quote
Wait a second. Your premise is that the Tango drum scans were done at 1600 ppi.

Not so. They were done at 3200. Don't know where you got the 1600 figure from.

Michael
[a href=\"index.php?act=findpost&pid=66785\"][{POST_SNAPBACK}][/a]

Here's where the 1600 dpi figure came from: in the article you indicated that you "set the crop tool to 8000x6000".  While I wasn't sure what you meant, I assumed that you had adjusted them to a resolution of all the images 8000x6000 for comparison.  On a 5x4 pice of film, that's 1600 dpi.  This was backed up by your crops of the doll (about 2200 pixels wide) which seemed to contitute just over a quarter of the width of the image, giving a horizontal resolution of 8000 pixels.

Is that not true?  Do the pixels on crops showing the doll and the top of the bill for the Velvia 4x5 represent less than 1/8000 of the total width of the piece of film?  Did you scan at 3200 and then down-sample to about half that resolution for the doll crops?  (If so, I retract my whining about the scans having been done at 1600 dpi, rather than scanned at a higher resolution and down-sampled.)  If so, it would be useful to see a crop with the original scanned pixels.  For the BetterLight dollar bill, does each pixel correspond to one pixel of a 8000x6000 image?  

Anyway, in spite of any confusion, I believe my point still stands: Velvia clearly can resolve details finer than 1600 pixels per inch (ppi, also called dpi), which is corresponds to a resolution of about 8000 x 6400 if your piece of film is 4 inches by 5 inches (minus a little for the edges of the film holder).  If you use a Betterlight 6K at its native resolution of 8000x6000, it cannot possibly resolve finer detail.

Personally, I agree that there's no point in scanning 5x4 Velvia at resolutions higher than 3200 dpi, but that corresponds to a resolution of 16000x12000.
« Last Edit: May 28, 2006, 11:24:56 AM by jrkeat » Logged
Jack Flesher
Sr. Member
****
Offline Offline

Posts: 2595



WWW
« Reply #17 on: May 28, 2006, 12:24:18 PM »
ReplyReply

Quote
snip

If you use a Betterlight 6K at its native resolution of 8000x6000, it cannot possibly resolve finer detail.

It sure as heck can if the sensor's pixel pitch is finer than the lens being used can resolve.   And in the case of most LF lenses, the BL is already there.
« Last Edit: May 28, 2006, 12:25:36 PM by Jack Flesher » Logged

Stephen Best
Guest
« Reply #18 on: May 28, 2006, 06:10:22 PM »
ReplyReply

Quote
3) I have seen a LOT of 4x5 scan files -- and with number 2 in mind, I had the opportunity to see the transparency, be present during the scanning process and inspected the final scan from the test image used here.  I can assure you: the chrome was properly focussed and properly exposed, and the scanning was done with more care than any lab would have done. And most importantly, the resulting scan file was as close to perfect as anyone could hope for from 4x5 film.
[a href=\"index.php?act=findpost&pid=66782\"][{POST_SNAPBACK}][/a]

I do a lot of 4x5 scans (of my own work and that of others) and generally at slightly higher resolution. I have no quibble with the scan in question, but the 4x5 tranny is plainly out of focus in parts ... whether you can see it or not. The reason for it being so is irrelevant in the context of this comparison. Megapixel peep-fests like this don't mirror real-world usage where you're more concerned with practical issues like depth of field than absolute resolving power for flat-field subjects with the lens near wide open.
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7725



WWW
« Reply #19 on: May 31, 2006, 06:10:17 AM »
ReplyReply

Quote
Anyway, in spite of any confusion, I believe my point still stands: Velvia clearly can resolve details finer than 1600 pixels per inch (ppi, also called dpi), which is corresponds to a resolution of about 8000 x 6400 if your piece of film is 4 inches by 5 inches (minus a little for the edges of the film holder).  If you use a Betterlight 6K at its native resolution of 8000x6000, it cannot possibly resolve finer detail.

Personally, I agree that there's no point in scanning 5x4 Velvia at resolutions higher than 3200 dpi, but that corresponds to a resolution of 16000x12000.
[a href=\"index.php?act=findpost&pid=66788\"][{POST_SNAPBACK}][/a]

Not that it matters much, but his point, as I understand it, is that all the digital files should have been up-resed to match the highest resolution of scanned film.

Cheers,
Bernard
Logged

A few images online here!
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad