How does Better Light calculate MP count?

[sanking]

There was some discussion in another thread today about Better Light scanning backs. Never having used one of these I looked at the product line and came away confused about how total MP count is determined.

For example, here is the description of the Model 6000E-HS.

"Model 6000E-HS — The economical solution for pro photographers, museums and art repro companies who are easing into digital capture. Native maximum resolution at 100% = 6000 x 8000 pixels. Maximum file size at 100% = 137 MB in 24-bit RGB (274 MB 48-bit). Price: $9,495 with 2-year warranty."

Now, if you download the .pdf for product information the Model 6000E-HS is said to give a file size of 144 MP. The Model 6K-HS, which is also listed as 6000X8000 pixels, is said to give a file size of 216 MP.

Question, how do you get 144 MP and 216 MP from a scan of 6000X8000 pixels? I would have calculated this as 48 MP for both backs. Are the higher figures interpolated resolution?

Sandy King

[Jim collum]

There was some discussion in another thread today about Better Light scanning backs. Never having used one of these I looked at the product line and came away confused about how total MP count is determined.

For example, here is the description of the Model 6000E-HS.

"Model 6000E-HS — The economical solution for pro photographers, museums and art repro companies who are easing into digital capture. Native maximum resolution at 100% = 6000 x 8000 pixels. Maximum file size at 100% = 137 MB in 24-bit RGB (274 MB 48-bit). Price: $9,495 with 2-year warranty."

Now, if you download the .pdf for product information the Model 6000E-HS is said to give a file size of 144 MP. The Model 6K-HS, which is also listed as 6000X8000 pixels, is said to give a file size of 216 MP.

Question, how do you get 144 MP and 216 MP from a scan of 6000X8000 pixels? I would have calculated this as 48 MP for both backs. Are the higher figures interpolated resolution?

Sandy King

i'm not entirely sure.. but i suspect they're compensating for the fact that each pixel is non-interpolated (there's no Bayer mosaic going on). You get more real detail at the pixel level compared to the rest of the digital cameras

[erie patsellis]

Sandy, from what I remember (always a scary concept) there is a JEDEC or other trade group standard for calculating MP ratings. On a scan back, they are (claimed) to be equivalent to 3x the raw resolution. Apparently this figure compensates for Bayer pattern "irregularities" in photosites vs. actual resolution.

[JeffKohn]

They're pulling the same marketing BS as Sigma did with their Foveon sensors, and tripling the effective MP count since each photosite location samples R, G, and B. But you're not going to get 3x the resolution from one of these backs as you would from a Bayer sensor with the same number of photosite locations. The real world gain is in the 25-50% range.

[sanking]

They're pulling the same marketing BS as Sigma did with their Foveon sensors, and tripling the effective MP count since each photosite location samples R, G, and B. But you're not going to get 3x the resolution from one of these backs as you would from a Bayer sensor with the same number of photosite locations. The real world gain is in the 25-50% range.

That makes sense.

In the BL literature it is noted that at a print size of 20" X 26.7" the resolution would be 300 spi. Since the sensor size is 72mm X 96mm (2.8" X 3.78") a print that size represents about a 7X magnification, which suggests effective resolution of around 2000 ppi, which would only be about 40 lp/mm. That sounds low to me.

Sandy King

[Ben Syverson]

40 lp/mm. That sounds low to me.
Seriously... A flatbed scan of a negative could get near that at a substantially lower cost, and you could use an instantaneous exposure! Step up to a drum scan and there's no comparison. Especially at 8x10.

[Sevo]

"Model 6000E-HS — The economical solution for pro photographers, museums and art repro companies who are easing into digital capture. Native maximum resolution at 100% = 6000 x 8000 pixels. Maximum file size at 100% = 137 MB in 24-bit RGB (274 MB 48-bit). Price: $9,495 with 2-year warranty."

Now, if you download the .pdf for product information the Model 6000E-HS is said to give a file size of 144 MP. The Model 6K-HS, which is also listed as 6000X8000 pixels, is said to give a file size of 216 MP.

Question, how do you get 144 MP and 216 MP from a scan of 6000X8000 pixels? I would have calculated this as 48 MP for both backs. Are the higher figures interpolated resolution?


You seem to mix up the MB (Megabytes) in their specs with MP (Megapixels) - a 16 bit sensor line will deliver six times as many bytes as pixels (two bytes for each of three colours) of file size.

Sevo

[sanking]

You seem to mix up the MB (Megabytes) in their specs with MP (Megapixels) - a 16 bit sensor line will deliver six times as many bytes as pixels (two bytes for each of three colours) of file size.

Sevo

The figures I cite are direct from Better Light literature. The 6000E-HS is said to give 6000X8000 pixels for a rating of 144 MP, with file size of 137MB in 8-bit, 274MB in 16-bit.

And this is said to give a file size of 20" X 26.7" at 300 ppi. That indicates resolution of about 2000 ppi, which is confirmed by the linear pixel count divided by size of the sensor (8000 ppi / 96mm (3.78"), which gives resolution of 2116 ppi, or just over 41 lp/mm.

Sandy King

[Sevo]

The figures I cite are direct from Better Light literature. The 6000E-HS is said to give 6000X8000 pixels for a rating of 144 MP, with file size of 137MB in 8-bit, 274MB in 16-bit.


How old is that thing? If it is ancient, that (only superficially impressive) figure might be brutal fact. 6000x8000 are 48MP - but with multi-pass with offset scan line, many backs can do a multiple of their native scan line resolution if you have the patience. A 144MP colour image of 274MB uncompressed size would be true 80's 16-bit with 5 bit/channel - back in the late eighties that used to be state of the art (though I don't believe that they already had 6k scan lines back then). On the other hand, 274 is close enough to the 288MB file size of 48-bit 48MP images - if it is new, they are lying or their marketing has made a mess out of their figures...

[erie patsellis]

Megapixels aren't everything, even though it's far from state of the art, my Phase One Studio kit with it's lowly 9mp images still look amazing printed double truck.

[sanking]

Megapixels aren't everything, even though it's far from state of the art, my Phase One Studio kit with it's lowly 9mp images still look amazing printed double truck.

What does "printed double truck" mean?

Sandy King

[jkuska]

Many people naively make this mistake without being familiar with Better Light systems and thoroughly thinking it through. The equation is not merely 8000x10660, it's 3x8000x10660 in normal mode and 3x8000x15990 in enhanced mode. You are not simply taking one picture every time you make a scan, you are taking 3 pictures. There are not just 8000 sensors scanning the scene, there are 24,000 sensors, three rows of 8000 sensors each, one row for each primary color, RGB. This is why Better Light provide superior results over instant capture cameras which only record 1/3 of the color data for each point in the scene and then have to interpolate every image up 300% before it can be displayed or printed. Better Light captures the true RGB value for every point in the scene with no interpolation (& no sharpening, no compression, no anti-moire or anti-aliasing algorithms either - just pure, clean accurate, high information density captures).

Since Better Light captures 3 pictures simultaneously (1 each for red, green & blue – RGB), each measuring 8000x10600 in normal mode then this becomes 3x8000x10600=254.4 megapixels per capture in NORMAL mode.

And then in enhanced mode this becomes 3x8000x15990=383.76 megapixels per capture.

Now in enhanced mode to maintain the correct proportion in the scene Better Light has to interpolate those 8000 sensors (pixels) up to 12000, which becomes a 150% interpolation in the final image, only half the interpolation as used in all instant capture and only in the one axis of the image. Thus even the ENHANCED images are still far superior to any instant capture image and superior to any resizing up that could be done in Photoshop. Only the fussiest of archivist, scholars or academicians object to this modest degree of interpolation and most folks agree it is undetectable.

I hope the proceeding is clear but in any case it is all restated here on Better Lights website as the first topic:
http://www.betterlight.com/faq_sales.html

[jkuska]

What does "printed double truck" mean?

Sandy King

2 pages up, a double spread, not just 8½x11 but 11x17

[Gordon Moat]

That makes sense.

In the BL literature it is noted that at a print size of 20" X 26.7" the resolution would be 300 spi. Since the sensor size is 72mm X 96mm (2.8" X 3.78") a print that size represents about a 7X magnification, which suggests effective resolution of around 2000 ppi, which would only be about 40 lp/mm. That sounds low to me.

Sandy King

It might help to know their pixel cell site dimension, and then that might equate better to the optical resolution (not file sizes). Some of the Better Light literature reminds me of Foveon for some reason. Also, somewhat related, here is an article I ran across recently the discusses pixel sizes and optical resolutions:

http://www.vision-systems.com/display_article/356315/19/none/none/Feat/Matching-Lenses-and-Sensors

I do think the lack of Bayer patterning could improve that, though light transmittance is still affected by colour filtration over each pixel cell site. The other factor would be the dead zone between pixel cell sites, though determining that on the Better Light scan backs is difficult.

Ciao!

Gordon Moat Photography (http://www.gordonmoat.com)

[jkuska]

There was some discussion in another thread today about Better Light scanning backs. Never having used one of these I looked at the product line and came away confused about how total MP count is determined...

...Question, how do you get 144 MP and 216 MP from a scan of 6000X8000 pixels? I would have calculated this as 48 MP for both backs. Are the higher figures interpolated resolution?

Sandy King

Many people naively make this mistake without being familiar with Better Light systems and thoroughly thinking it through. The equation is not merely 8000x10660, it's 3x8000x10660 in normal mode and 3x8000x15990 in enhanced mode for the Super 8. You are not simply taking one picture every time you make a scan, you are taking 3 pictures. There are not just 8000 sensors scanning the scene, there are 24,000 sensors, three rows of 8000 sensors each, one row for each primary color, RGB. This is why BL provides superior results over instant capture cameras which only record 1/3 of the color data for each point in the scene and then have to interpolate every image up 300% before it can be displayed or printed - not to mention all the other algorithms at work reprocessing and stepping on all the data in so many ways. BL captures the true RGB value for every point in the scene with no interpolation (and no compression, no sharpening, no anti-moire, no anti-aliasing algorithms either - just pure, clean, accurate, high information density files).

Since BL captures 3 pictures simultaneously (1 each for red, green & blue – RGB), each measuring 8000x10600 in normal mode then this becomes 3x8000x10600=254.4 megapixels per capture in NORMAL mode. Then in enhanced mode this becomes 3x8000x15990=383.76 megapixels per capture.

Now in enhanced mode to maintain the correct proportion in the scene BL does have to interpolate those 8000 sensors (pixels) up to 12000, which becomes a 150% interpolation in the final image, only half the interpolation as used in all instant capture and only in the one axis of the image. Thus even ENHANCED images are still far superior to any instant capture image and superior to any resizing up that could be done in Photoshop. Only the fussiest of archivist, scholars or academicians don't care for any enhancement but nobody ever really detects this relatively low level.

For the Super 6 this becomes 3x6000x8000=144mp in normal mode or 3x6000x12000=216mp in enhanced mode.

I hope the proceeding is clear but in any case it is restated here on the BL website as the first topic:
http://www.betterlight.com/faq_sales.html

[jkuska]

Seriously... A flatbed scan of a negative could get near that at a substantially lower cost, and you could use an instantaneous exposure! Step up to a drum scan and there's no comparison. Especially at 8x10.

BL backs are effectively flatbed scanners that have been redesigned to fit into a form factor similar to the Polaroid 545 back so that they may be used in basically any 4x5 view camera ;-)

[jkuska]

It might help to know their pixel cell site dimension, and then that might equate better to the optical resolution (not file sizes). Some of the Better Light literature reminds me of Foveon for some reason. Also, somewhat related, here is an article I ran across recently the discusses pixel sizes and optical resolutions:

http://www.vision-systems.com/display_article/356315/19/none/none/Feat/Matching-Lenses-and-Sensors

I do think the lack of Bayer patterning could improve that, though light transmittance is still affected by colour filtration over each pixel cell site. The other factor would be the dead zone between pixel cell sites, though determining that on the Better Light scan backs is difficult.

Ciao!

Gordon Moat Photography (http://www.gordonmoat.com)

72mm/6000=a 12 micon sensor size for the 6K and 72mm/8000=a 9 micron sensor size for the 8K systems. At a 1:1 repro ratio (life size on the imaging plane) with the Super 6K this equals 2116 ppi or 1058 line pairs per inch or about 42 line pairs per mm and with the Super 8K this equals 2822 ppi or 1411 line pairs per inch or about 56 line pairs per mm. It should be noted that LF lenses use a LOT of glass and project a very large image circle to cover all possible camera movements and consequently large format have only diminishing returns to be gained through redesign. Smaller format lenses mostly only cover a fixed and small image area using much less glass so could realize greater gains through redesign. If BL made, for instance, a Super 14K, there is no LF lens that could deliver that level of detail at the imaging plane so while you would have more pixels you would NOT have more detail.

[jkuska]

They're pulling the same marketing BS as Sigma did with their Foveon sensors, and tripling the effective MP count since each photosite location samples R, G, and B. But you're not going to get 3x the resolution from one of these backs as you would from a Bayer sensor with the same number of photosite locations. The real world gain is in the 25-50% range.

It's is NOT BS, it is the industry standard and it is REAL.

[jkuska]

What does "printed double truck" mean?

Sandy King

PS... Hi Sandy... it's Jerry the carbon printer and former owner of the long defunct Limited Edition Photographics

[Ben Syverson]

BL backs are effectively flatbed scanners that have been redesigned to fit into a form factor similar to the Polaroid 545 back so that they may be used in basically any 4x5 view camera ;-)
Right, but if they're only getting 40 lp/mm, then there's no image quality advantage to using them, and many many workflow disadvantages.

You know what else can be used in basically any 4x5 view camera? Sheet film. :)

[jkuska]

How old is that thing? If it is ancient, that (only superficially impressive) figure might be brutal fact. 6000x8000 are 48MP - but with multi-pass with offset scan line, many backs can do a multiple of their native scan line resolution if you have the patience. A 144MP colour image of 274MB uncompressed size would be true 80's 16-bit with 5 bit/channel - back in the late eighties that used to be state of the art (though I don't believe that they already had 6k scan lines back then). On the other hand, 274 is close enough to the 288MB file size of 48-bit 48MP images - if it is new, they are lying or their marketing has made a mess out of their figures...

Super 6K has 18000 sensors NOT 6000, and the Super 8K has 24000 sensors NOT 8000 so the equation becomes 3x18000x8000 and 3x24000x10600 with NO interpolation, NO compression, NO anti-moire, NO anti aliasing and NO sharpening algorithms stepping all over all the data all the time trying to GUESS what it should be; just pure, clean, accurate, high information density files.

[Gordon Moat]

72mm/6000=a 12 micon sensor size for the 6K and 72mm/8000=a 9 micron sensor size for the 8K systems. ....

Unfortunately not that simple. All sensors have dead area between pixels. If it was possible to know the sensor manufacturer, and the model number of the sensor, then that information would be accessible. My guess is that their best system uses an 8µm pixel cell site size.

[jkuska]

Right, but if they're only getting 40 lp/mm, then there's no image quality advantage to using them, and many many workflow disadvantages.

You know what else can be used in basically any 4x5 view camera? Sheet film. :)

Sheet film, while having a Dmax of 4.0 implying a 12 stop range, only has about a 5 or 6 stop range of printable detail, BL has an 11+ stop dynamic range with printable detail basically through out. And when you are focusing which one of the many multiple layers of emulsion in the film are you focusing on and is the film ever held flat in the film holder? The tri-linear array is perfectly flat flying across a flat plane.
Can you do this with film?
http://www.betterlight.com/zoomify/zoom_bigMoney.html
Or this?
http://www.artisan-digital-services.co.uk/digcapture-7.html
Or this?
http://www2.iath.virginia.edu/panorama/watson_uva.html
This last shot by Tom Watson was shot a 2/3 rds rez with a 6K, zoom in on the central quad and look at all of the folks sitting about and the blades of grass in the lawn.

[jkuska]

Unfortunately not that simple. All sensors have dead area between pixels. If it was possible to know the sensor manufacturer, and the model number of the sensor, then that information would be accessible. My guess is that their best system uses an 8µm pixel cell site size.

The Super 8 sensor: http://www.kodak.com/global/plugins/acrobat/en/business/ISS/productsummary/Linear/KLI-8023ProductSummary.pdf

The Super 6 sensor: http://www.kodak.com/global/plugins/acrobat/en/business/ISS/productsummary/Linear/KLI-6013ProductSummary.pdf

[sanking]

PS... Hi Sandy... it's Jerry the carbon printer and former owner of the long defunct Limited Edition Photographics

Jerry,

Are you still printing carbon on glass?

Sandy

[jkuska]

Jerry,

Are you still printing carbon on glass?

Sandy

Hi Sandy, I seldom ever printed carbon on glass, my preferred support was Melinex, the same base used by Ilford for Cibachrome. I used the single transfer method, it was challenging enough beating my head against the wall making 12 20x24's a week this way.

I don't know if you ever bothered but I authored the better part of this article on Wiki:
http://en.wikipedia.org/wiki/Carbon_print
starting with the third sentence down "An Overview and History of Carbon (Pigment) Printing" and including the "Chronological History of Carbon (Pigment) Printing" Table at the end of this brief article and the External Links listed at the bottom.

[jkuska]

Super 6K has 18000 sensors NOT 6000, and the Super 8K has 24000 sensors NOT 8000 so the equation becomes 3x18000x8000 and 3x24000x10600 with NO interpolation, NO compression, NO anti-moire, NO anti aliasing and NO sharpening algorithms stepping all over all the data all the time trying to GUESS what it should be; just pure, clean, accurate, high information density files.

I screwed up on this one, duh! I'm dyslexic and have a distracted personality, the equations could be either 3x6000x8000 and 3x8000x10660 OR 18000x8000 and 24000x10660... typical of me, still astonishing figures in a single capture.

[sanking]

Hi Sandy, I seldom ever printed carbon on glass, my preferred support was Melinex, the same base used by Ilford for Cibachrome. I used the single transfer method, it was challenging enough beating my head against the wall making 12 20x24's a week this way.

I don't know if you ever bothered but I authored the better part of this article on Wiki:
http://en.wikipedia.org/wiki/Carbon_print
starting with the third sentence down "An Overview and History of Carbon (Pigment) Printing" and including the "Chronological History of Carbon (Pigment) Printing" Table at the end of this brief article and the External Links listed at the bottom.

Hi Jerry,

Thanks for the information. I apologize for the misunderstanding but I had confused you with another Jerry who prints primarily on opal glass.

Regarding the BL image files you posted, do you have any technical details as to how many files were assembled to make those images I don't personally work in color so don't have any idea if this work could be done with LF color film.

However, I am fairly certain that the quality could be matched or exceeded shooting B&W separations and merging them in Photoshop. The shooting would take less time, the Photoshop work more, but it would be quite feasible, except for the shot with the people. I know this is old procedure and BL has replaced it, but with Photoshop shooting B&W separations is much more practical than it was twenty years ago.

Sandy King

[jkuska]

That makes sense.

In the BL literature it is noted that at a print size of 20" X 26.7" the resolution would be 300 spi. Since the sensor size is 72mm X 96mm (2.8" X 3.78") a print that size represents about a 7X magnification, which suggests effective resolution of around 2000 ppi, which would only be about 40 lp/mm. That sounds low to me.

Sandy King

hmmm... a capture measuring 6000x8000 pixels can be sized and resized at any resolution you want with affecting or reinterpolating any of the original data, it's just become a matter of scale. So 6000x8000 sized at 300ppi yields a print that's 20x26.7 inches then sized at 200ppi yields a print that's 30x40 inches. You choose the resolution and the output size changes accordingly so at 150ppi the print would be 40x53.4 inches and at 100ppi - 60x80 inches, but the original pixel size of the capture would always remain 6000x8000. 200ppi always produces a remarkably reasonable print off of any stochastically screened micro dot (2880x5760 at the print head) sprayed ink jet printer.

[jkuska]

Hi Jerry,

Thanks for the information. I apologize for the misunderstanding but I had confused you with another Jerry who prints primarily on opal glass.

Regarding the BL image files you posted, do you have any technical details as to how many files were assembled to make those images I don't personally work in color so don't have any idea if this work could be done with LF color film.

However, I am fairly certain that the quality could be matched or exceeded shooting B&W separations and merging them in Photoshop. The shooting would take less time, the Photoshop work more, but it would be quite feasible, except for the shot with the people. I know this is old procedure and BL has replaced it, but with Photoshop shooting B&W separations is much more practical than it was twenty years ago.

Sandy King

Well the Big Money shot is a single Super 8 capture.

The tapestry was 6ft high and 24ft long and was shot on location in 6 tiles along it's length with a Super 8 with the camera and lights traveling in a custom portable track. The tiles were later stitched together for the results you see.

The panorama on the Virgina campus was one continuous pano scan (you can pick up a few motion artifacts here and there) and the PanoAdaptor head can scan from 10° up to 400° (allowing a 40° overlap for finding a convenient stitching point) while recording a MAXIMUM of around 65000 lines of data. Now Tom shot this on Friday, April 14th 2006 at 9:12 a.m. using a 150mm lens with a Super 6 set at 62% rez for a 3750 x 49086 pixel capture with a Line Time of 1/240 and an ISO setting of 980 so that the total Scan Time was 1/240 x 49086 or 204.5 seconds OR 3 minutes and 24.5 seconds. The height to width ratio ended up at about 1:13. So 3x3750x49086=552.2175 megapixels, cool huh?

[jkuska]

Hi Jerry,

Thanks for the information. I apologize for the misunderstanding but I had confused you with another Jerry who prints primarily on opal glass.

Regarding the BL image files you posted, do you have any technical details as to how many files were assembled to make those images I don't personally work in color so don't have any idea if this work could be done with LF color film.

However, I am fairly certain that the quality could be matched or exceeded shooting B&W separations and merging them in Photoshop. The shooting would take less time, the Photoshop work more, but it would be quite feasible, except for the shot with the people. I know this is old procedure and BL has replaced it, but with Photoshop shooting B&W separations is much more practical than it was twenty years ago.

Sandy King

Sandy, we've emailed a little, I also used to manage Ataraxia... how's that for a name for a photo lab? Anthraxia? Anorexia? Atawhatia?

It's been pointed out that I should disclose that I work for Better Light but we don't misrepresent our product, when you are the best you don't need to... just ask NASA (they use a BL back to record the state of the heat tiles before every launch as a QC check before reentry, chew on that), the Smithsonian (they own 2 Super 6's and a Super 8), the FBI and Army Crime lab for forensics... blah, blah. We just don't shoot sporting events, weddings, beverage pour shots or models flouncing down runways very well but some folk do use us for portraiture though I would seldom recommend this unless you were familiar the old time directorial skills that most photographers were familiar with

[Jeremy Moore]

some folk do use us for portraiture though I would seldom recommend this unless you were familiar the old time directorial skills that most photographers were familiar with

I've started doing this again: SO MUCH FUN!

[D. Bryant]

Or this?
http://www2.iath.virginia.edu/panorama/watson_uva.html
This last shot by Tom Watson was shot a 2/3 rds rez with a 6K, zoom in on the central quad and look at all of the folks sitting about and the blades of grass in the lawn.

I must admit that example is pretty incredible, but there is one woman walking that looks very strange.

Don Bryant

[rugenius]

If you can live with the limitations of the scanning back, Betterlight surely has a top notch product.

[Findingmyway4ever]

Good to have someone that can actually explain things about the best digital system for the context on the market today. I would love to see prints from this system, even at the 8X10 size.

[Jim collum]

Good to have someone that can actually explain things about the best digital system for the context on the market today. I would love to see prints from this system, even at the 8X10 size.

If you're ever in the Santa Cruz, Ca area.. send me mail.. I'll show you some (I've had one since 2001.. and it is a top notch product.. as well as having a customer support/sevice model that i've yet to see another company equal )

[JeffKohn]

It's is NOT BS, it is the industry standard and it is REAL.
No, sorry, it's BS. I don't care what standard you claim to follow, marketing a device as capturing 144 megapixels, when the image files only contain 48 million pixels, is misleading. I understand the desire to differentiate a 48mp full-RGB capture from a 48mp Bayer-filtered capture, because there is a difference. But it would have been better to come up with some different term or measurement to communicate that difference, because the term 'pixel' already had a pretty specific meaning long before digital photography took off, and the way Betterlight and Sigma are using the term doesn't fit that definition.

Maybe they should have used the term 'megadots', since the analogy to the R, G, and B 'dots' in an LCD would have been a lot more honest. For instance, the 3" LCD's on many current DSLR's have 960K dots, where each dot emits red, green or blue light and the true RGB resolution is 320K.

It's true that a 48mp RGB capture will have more resolution and detail than a 48mp Bayer-filtered capture. But it's not a 300% increase. You can't possibly say that with a straight face if you know anything about digital capture. For that claim to be true, the missing values in the Bayer-filtered image would have be filled in with random guesses. But that's not how Bayer demosaicing works. It actually uses quite sophisticated interpolation routines, which are getting better with every new release of the top-end raw converters. Such a claim also ignores the fact that the Bayer filter sacrifices color detail in favor of luminosity detail, which is far more important to our perception of detail and resolution. So while Betterlight and Foveon are better than Bayer at a given mp count, they're not 3 times better, not even close. It's more like half that.

None of this is to say that the Betterlight backs can't produce impressive files. They can, assuming you can work with the long exposure times. But the megapixel ratings are bogus.

[jkuska]

No, sorry, it's BS. I don't care what standard you claim to follow, marketing a device as capturing 144 megapixels, when the image files only contain 48 million pixels, is misleading. I understand the desire to differentiate a 48mp full-RGB capture from a 48mp Bayer-filtered capture, because there is a difference. But it would have been better to come up with some different term or measurement to communicate that difference, because the term 'pixel' already had a pretty specific meaning long before digital photography took off, and the way Betterlight and Sigma are using the term doesn't fit that definition.

Maybe they should have used the term 'megadots', since the analogy to the R, G, and B 'dots' in an LCD would have been a lot more honest. For instance, the 3" LCD's on many current DSLR's have 960K dots, where each dot emits red, green or blue light and the true RGB resolution is 320K.

It's true that a 48mp RGB capture will have more resolution and detail than a 48mp Bayer-filtered capture. But it's not a 300% increase. You can't possibly say that with a straight face if you know anything about digital capture. For that claim to be true, the missing values in the Bayer-filtered image would have be filled in with random guesses. But that's not how Bayer demosaicing works. It actually uses quite sophisticated interpolation routines, which are getting better with every new release of the top-end raw converters. Such a claim also ignores the fact that the Bayer filter sacrifices color detail in favor of luminosity detail, which is far more important to our perception of detail and resolution. So while Betterlight and Foveon are better than Bayer at a given mp count, they're not 3 times better, not even close. It's more like half that.

None of this is to say that the Betterlight backs can't produce impressive files. They can, assuming you can work with the long exposure times. But the megapixel ratings are bogus.

The Japan Camera Industry Association (JCIA) established guidelines for determining the number of “effective pixels” in a digital camera that most digital camera vendors, including Better Light, use to ascertain the number of pixels in their devices. The basic JCIA guideline states that the number of “effective pixels” is:

“The number of pixels on the image sensor which receive input light through the optical lens, and which are effectively reflected in the final output data of the still image.”

This simple statement establishes that a “pixel” must be physically present on the image sensor, it must be light-sensitive, and the information it provides must be relevant to (if not directly present in) the final image data. This last qualification allows instant-capture digital cameras to include so-called “ring” pixels surrounding the area that actually becomes part of the final image, since the “ring” pixels are used to determine some characteristics of the final image. The JCIA guideline also considers devices with multiple image sensors, as well as scanning camera designs:

“For DSCs [digital still cameras] that use an image sensor ( including a linear sensor) that capture an image by spatial or optical movement, the number of effective pixels of the image sensor and the number of sampling positions should be noted. If the total number of sampled pixels (ring pixels can be included) is also noted, the description should clearly state that the noted number is the total number of effective pixels, to avoid consumer misunderstanding.”

Just as instant-capture digital camera vendors can count all three colors of pixels physically present on their image sensors, Better Light can count all three colors of pixels physically present on the trilinear image sensors we use in our scanning backs. There are no “ring” pixels on our trilinear sensors, although there are additional pixels on the sensor that are not counted as “effective pixels”. To arrive at the “total number of effective pixels”, we multiply the number of “effective pixels” on the image sensor by the number of sampling positions.

Since a BL back takes 3 distinct pictures each (RGB) using 48mp then the final image is composed of 144mp, it's pretty simple and straight forward with no strange accounting at all. If each of the channels was converted to a distinct grey scale image each of those 3 B&W images would certainly be 48mp each and each would be distinct in apperance. If you recombined them the resulting image would be composed of 144mp, no magic, no tricks, no deception and no strange accounting.

[Gordon Moat]

Interesting selective usage of definitions. I don't see anyone challenging the Better Light claims in court, though all this is misleading. Maximum file size is vastly more useful, and certainly not 144MP. Better Light are not 3 times greater resolving than a PhaseOne or Leaf back of near 48MP, but the marketing implies that claim.

The Better Light scan backs are good devices, but I fail to see this need to be selective about the description. I think that does more harm than good. Compare to Foveon, who make similar marketing choices, and yet still do not enjoy much market success. I would have hoped the digital capture market was mature enough to move past this oddity, but it appears some still cling to the ways of the (recent) past.

[jkuska]

I must admit that example is pretty incredible, but there is one woman walking that looks very strange.

Don Bryant
That strange look is a motion artifact created by the scan, like the gentleman walking up to the podium. If you think those are strange check these out:
http://people.rit.edu/andpph/photofile-c/periport-1.jpg
http://people.rit.edu/andpph/travel-exhibit/ad-700.jpg
http://people.rit.edu/andpph/photofile-c/helene-model-4093a.jpg
http://people.rit.edu/andpph/photofile-b/slitscan-1.jpg
http://people.rit.edu/andpph/photofile-misc/twist-12a.jpg
But to demonstrate what sort of real portraiture can be done with a BL scanning back:
http://www.ronfinleystudio.com/BLshots/BLshots11.html
http://www.ronfinleystudio.com/BLshots/BLshots01.html
http://www.ronfinleystudio.com/BLshots/BLshots08.html
Ron Finley likes to use a BL scanning back for portraiture because his clients typically want giant prints and only a BL back can provide the information densities and dynamic ranges necessary to go as large as they want without falling apart.

And a couple more pro BL shooters:
http://www.richpix.com/portfolio/05_panoramic/panoramic_01.html
http://www.bohonus.com/pages/panoscan_vrs.php

[bob carnie]

On my monitor the Motorcycle shot on Ron Finleys site has tons of banding, looks unprintable to me.

That strange look is a motion artifact created by the scan, like the gentleman walking up to the podium. If you think those are strange check these out:
http://people.rit.edu/andpph/photofile-c/periport-1.jpg
http://people.rit.edu/andpph/travel-exhibit/ad-700.jpg
http://people.rit.edu/andpph/photofile-c/helene-model-4093a.jpg
http://people.rit.edu/andpph/photofile-b/slitscan-1.jpg
http://people.rit.edu/andpph/photofile-misc/twist-12a.jpg
But to demonstrate what sort of real portraiture can be done with a BL scanning back:
http://www.ronfinleystudio.com/BLshots/BLshots11.html
Ron Finley likes to use a BL scanning back for portraiture because his clients typically want giant prints to hang over their fireplaces in their very big homes and only a BL back can provide the information densities and dynamic ranges necessary to go as large as they want without falling apart.

[jkuska]

Interesting selective usage of definitions. I don't see anyone challenging the Better Light claims in court, though all this is misleading. Maximum file size is vastly more useful, and certainly not 144MP. Better Light are not 3 times greater resolving than a PhaseOne or Leaf back of near 48MP, but the marketing implies that claim.

The Better Light scan backs are good devices, but I fail to see this need to be selective about the description. I think that does more harm than good. Compare to Foveon, who make similar marketing choices, and yet still do not enjoy much market success. I would have hoped the digital capture market was mature enough to move past this oddity, but it appears some still cling to the ways of the (recent) past.

File size might be a more interesting comparison, the BL Super 6 in normal mode produces a 274 MB file and in enhanced mode a 618 MB file and the Super 8 in those two modes a 488 MB and 1.1 GB file... and in panorama mode those file sizes could be 5 to 8 times as large when recording up to 65,000 lines of data in a single scan.

[jkuska]

No, sorry, it's BS. I don't care what standard you claim to follow, marketing a device as capturing 144 megapixels, when the image files only contain 48 million pixels, is misleading. I understand the desire to differentiate a 48mp full-RGB capture from a 48mp Bayer-filtered capture, because there is a difference. But it would have been better to come up with some different term or measurement to communicate that difference, because the term 'pixel' already had a pretty specific meaning long before digital photography took off, and the way Betterlight and Sigma are using the term doesn't fit that definition.

Maybe they should have used the term 'megadots', since the analogy to the R, G, and B 'dots' in an LCD would have been a lot more honest. For instance, the 3" LCD's on many current DSLR's have 960K dots, where each dot emits red, green or blue light and the true RGB resolution is 320K.

It's true that a 48mp RGB capture will have more resolution and detail than a 48mp Bayer-filtered capture. But it's not a 300% increase. You can't possibly say that with a straight face if you know anything about digital capture. For that claim to be true, the missing values in the Bayer-filtered image would have be filled in with random guesses. But that's not how Bayer demosaicing works. It actually uses quite sophisticated interpolation routines, which are getting better with every new release of the top-end raw converters. Such a claim also ignores the fact that the Bayer filter sacrifices color detail in favor of luminosity detail, which is far more important to our perception of detail and resolution. So while Betterlight and Foveon are better than Bayer at a given mp count, they're not 3 times better, not even close. It's more like half that.

None of this is to say that the Betterlight backs can't produce impressive files. They can, assuming you can work with the long exposure times. But the megapixel ratings are bogus.

If I take one "RED" picture by itself (which I can do by simply turning off the other two channels) it will contains 48 mp. Then if I take two more pictures (a BLUE and a GREEN) which contain 48 mp each then the sum (combined pixel density) of all three pictures is 144 mp and NOT 48 mp. I believe it is generally accepted throughout the industry that "a sensor" equals "a pixel." So I can go on location a capture a single exposure using the red channel alone to darken the sky then come back and make black and white print of this single channel file and since the file was recorded using 48 million sensors that print would be the result of a 48 mp capture. If I did the same thing with the other two channels so I had three separate B&W prints hanging on the wall there would be a total of 144mp worth of data hanging on that wall in those three prints. Then when these three images were combined to produce a color print none of those sensors or pixels are thrown away or discarded or restructured so the sum would be 144mp. “The eye altering, alters all” -WB

[sanking]

If I take one "RED" picture by itself (which I can do by simply turning off the other two channels) it will contains 48 mp. Then if I take two more pictures (a BLUE and a GREEN) which contain 48 mp each then the sum (combined pixel density) of all three pictures is 144 mp and NOT 48 mp. I believe it is generally accepted throughout the industry that "a sensor" equals "a pixel." So I can go on location a capture a single exposure using the red channel alone to darken the sky then come back and make black and white print of this single channel file and since the file was recorded using 48 million sensors that print would be the result of a 48 mp capture. If I did the same thing with the other two channels so I had three separate B&W prints hanging on the wall there would be a total of 144mp worth of data hanging on that wall in those three prints. Then when these three images were combined to produce a color print none of those sensors or pixels are thrown away or discarded or restructured so the sum would be 144mp. “The eye altering, alters all” -WB

So the sensor in the BL scanning back is similar to the Sigma Fovenon. Are Fovenon type sensors produced by just one source, or several?

Sandy King

[JeffKohn]

If I take one "RED" picture by itself (which I can do by simply turning off the other two channels) it will contains 48 mp. Then if I take two more pictures (a BLUE and a GREEN) which contain 48 mp each then the sum (combined pixel density) of all three pictures is 144 mp and NOT 48 mp. I believe it is generally accepted throughout the industry that "a sensor" equals "a pixel." So I can go on location a capture a single exposure using the red channel alone to darken the sky then come back and make black and white print of this single channel file and since the file was recorded using 48 million sensors that print would be the result of a 48 mp capture. If I did the same thing with the other two channels so I had three separate B&W prints hanging on the wall there would be a total of 144mp worth of data hanging on that wall in those three prints. Then when these three images were combined to produce a color print none of those sensors or pixels are thrown away or discarded or restructured so the sum would be 144mp. “The eye altering, alters all” -WB
None of this changes the fact that your "144mp images" do not have 3 times the resolution of a 48mp bayer sensor. So you can hide behind standards and use all the contrived justifications that you want, it doesn't change the reality that your specs are misleading. I find it especially ironic that the standard you quote from talks about avoiding consumer misunderstanding. All you have to do is look at the post that started this thread to see that there is confusion on this matter.

[Gordon Moat]

Imagine if Creo (Kodak) or Dainippon Screen claimed three times the capability of their scanners. If they used the same marketing language that Better Light and Foveon use, they could rightly claim that the moving tri-linear CCD is capturing three times the normal data. The Better Light is not functionally that different than a scanner. It would not surprise me if the sensor was made by Kodak, though knowing which CCD would definitely clear up the true capability.

To the Better Light people: there are no people arguing that the Better Light is not a very capable scan back that can produce great results. The issue is that the usable amount of information contained in a file is the true measure of your device; so why is it so vastly difficult for you to state that, without resorting to marketing speak that is NOT clear?

[Sevo]

If I take one "RED" picture by itself (which I can do by simply turning off the other two channels) it will contains 48 mp. Then if I take two more pictures (a BLUE and a GREEN) which contain 48 mp each then the sum (combined pixel density) of all three pictures is 144 mp and NOT 48 mp. I believe it is generally accepted throughout the industry that "a sensor" equals "a pixel."

All you have is a 48mp colour sensor. The advantage of being trilinear is that your chroma resolution is (almost - depending on the aliasing filter) as good as the luminance resolution, while Bayer sensors have only half to 1/4 of their luminance resolution for each colour. You might advertise that thing as 48 (48-48-48) MP whereas manufacturers of a 48MP Bayer sensor would in all truth have to admit that their sensor is 48 (24-12-12) MP. But 144MP your sensor is not, unless used for black and white photography of purely black and white subjects using a subpixel to pixel conversion...

[George Stewart]

From my understanding, there is only a small difference (almost unnoticed) between between a Bayer sensor without an antialiasing filter and that of a monochromatic sensor. Here is a link:http://www.luminous-landscape.com/reviews/cameras/achromatic.shtml

The same should apply for a scanning back. In other words, it is the antialiasing filter that typically degrades a sensor's resolution, and not the fact that it uses a color-Bayer design.

Here is another interesting link: http://www.luminous-landscape.com/reviews/cameras/better-light.shtml

In my biased opinion, the BL system is about a 10-20% improvement, overall, over what can be had in a similar resolving DMF system.

[jkuska]

So the sensor in the BL scanning back is similar to the Sigma Fovenon. Are Fovenon type sensors produced by just one source, or several?

Sandy King

No BL is not similar to the Foveon and Foveon is produced by one source. The technology that Foveon uses is based on the fact that the different wave lengths of light (RGB) penetrate silicon to different depths, so from the same single sensor they take the B data from the top, the G data from the center and the R data from the bottom (boy, if Foveon would ever make a full frame sensor they would kill):
http://en.wikipedia.org/wiki/Foveon_X3_sensor

The Kodak tri-linear array has 3 rows of 6000 (or 8000) sensors each about 8 pixels apart so while while scanning there are 3 pictures being taken, one each for R, G & B so that the true RGB data is recorded for every point in the scene with no interpolation, again that's 144 million sensors recording data during the scan. Each of these pictures is 48mp in itself.

With a Bayer Pattern sensor the true RGB data is only known for the area of a 2x2 grid composed of 2 green sensors and 1 each blue and red. To properly display or print this Bayer pattern data the 2 missing data values for each point in the scene has to be interpolated through algorithms (guessed at, speculated on), that 300% interpolation out the gate in every picture. The algorithms are very good but try photographing a diamond with a Bayer patterned sensor then with a BL back, the false color artifacts created by processing the file through these algorithms become immediately apparent and deny the true clarity of diamonds.

So for both BL and Foveon one point on the imaging plane contains ALL the true RGB values for every point in the scene while it takes 4 points on the imaging plane to interpolate the possible RGB values for those same 4 points in the scene. Sort of sounds like both Foveon and BL are 4 times as accurate as Bayer patterned sensors.

But there are obviously both nay-sayer and yea-sayers who in spite of published industry standards have their biases and opinions. The proof is always in the pudding, just try out shooting, out enlarging and out printing a BL file (the notable exceptions would be from a Cruse or Anagram system which equal BL but are 4 to 10 times more expense) every thing else falls apart. The entire instant capture market target the aforementioned "double truck" - an 11x17" spread, all that's needed for publication.

[jkuska]

Imagine if Creo (Kodak) or Dainippon Screen claimed three times the capability of their scanners. If they used the same marketing language that Better Light and Foveon use, they could rightly claim that the moving tri-linear CCD is capturing three times the normal data. The Better Light is not functionally that different than a scanner. It would not surprise me if the sensor was made by Kodak, though knowing which CCD would definitely clear up the true capability.

To the Better Light people: there are no people arguing that the Better Light is not a very capable scan back that can produce great results. The issue is that the usable amount of information contained in a file is the true measure of your device; so why is it so vastly difficult for you to state that, without resorting to marketing speak that is NOT clear?

Interesting... I posted links to the Kodak sensors BL uses (which clearly state the 12 and 9 micron pixel size), and I posted file sizes as well, which clearly exemplify a great deal more data coming from some where, if not from the great increased pixel density then where? Remember there are NO algorithms at work on a BL file; no anti-aliasing, no anti-moire, no sharpening, no compression, no interpolation (in normal mode), just pure, clean, accurate and unadulterated data. Think how much your data is being stepped on and pushed around with ALL of those algorithms at work in most instant capture systems. Enough so that many academicians, scholars, archivists, technicians and photographers prefer BL files when it comes to recordings images of maps, fragments of the Dead Sea Scrolls, the heat tiles on the underside of the Space Shuttle, the Book of Kells, great works of art, stress fracture analysis in jet engine components, uranium rod cross sections out of nuclear reactors, diamonds and yes, even landscape and commercial photography - now that's "marketing speak."

Talking is one thing... taking pictures is another... just get out there and shoot with whatever kind of camera you have and follow your bliss

[jkuska]

All you have is a 48mp colour sensor. The advantage of being trilinear is that your chroma resolution is (almost - depending on the aliasing filter) as good as the luminance resolution, while Bayer sensors have only half to 1/4 of their luminance resolution for each colour. You might advertise that thing as 48 (48-48-48) MP whereas manufacturers of a 48MP Bayer sensor would in all truth have to admit that their sensor is 48 (24-12-12) MP. But 144MP your sensor is not, unless used for black and white photography of purely black and white subjects using a subpixel to pixel conversion...

hmm... this could point to an interesting way of looking at it. Each point on the imaging plane in a Bayer sensor has a 16 (14) bit depth (or less) original record associated with it (prior to post capture processing) while each point on the imaging plane of a scanning system has a 48 (42) bit depth original record associated with it with no post capture processing. Whether you want to attribute the results of that to pixels, sensors, file size or what ever, it clearly points to data density rich and accurate file clearly superior to instant capture by a wide margin.

[jkuska]

Imagine if Creo (Kodak) or Dainippon Screen claimed three times the capability of their scanners. If they used the same marketing language that Better Light and Foveon use, they could rightly claim that the moving tri-linear CCD is capturing three times the normal data. The Better Light is not functionally that different than a scanner. It would not surprise me if the sensor was made by Kodak, though knowing which CCD would definitely clear up the true capability.

To the Better Light people: there are no people arguing that the Better Light is not a very capable scan back that can produce great results. The issue is that the usable amount of information contained in a file is the true measure of your device; so why is it so vastly difficult for you to state that, without resorting to marketing speak that is NOT clear?

I've think I've got it... see a previous post: they do claim 3x the capability: they claim a "captured" 48 bit depth per point point in the scene for color whereas ALL Bayer patterned (instant capture) sensors can only claim a "captured" 16 bit depth OR LESS per point in the scene PRIOR to post capture processing.

[Gordon Moat]

Interesting... I posted links to the Kodak sensors BL uses (which clearly state the 12 and 9 micron pixel size), and I posted file sizes as well, which clearly exemplify a great deal more data coming from some where, if not from the great increased pixel density then where?

Sorry I missed the links. Those Kodak sensors individual pixel cell site sizes are your limit. Compare to PhaseOne MFDBs that use 6µm up to 8µm pixel cell site sizes, and no aliasing filter. You can see in the specification links you provided that the 9µm sensor you use is more sensitive than the 12µm sensor you use. PhaseOne also use Kodak CCDs. The sweet spot in sensitivity of CCDs is in the 6µm to 8µm realm; going much smaller tends to increase noise issues, while going larger reduces sensitivity. There is also a factor of the strength of the colour filter over each pixel cell site, in that each colour (Red, Green, Blue) blocks a certain amount of light, and reduced the fill factor of each well. So your claim of "more data" is down to an argument of Bayer interpolation being vastly degrading.




Remember there are NO algorithms at work on a BL file; no anti-aliasing, no anti-moire, no sharpening, no compression, no interpolation (in normal mode), just pure, clean, accurate and unadulterated data. Think how much your data is being stepped on and pushed around with ALL of those algorithms at work in most instant capture systems.

Several years ago I worked on a development camera for a couple manufacturers. The test box camera had three full frame (24mm by 36mm) imaging sensors, each with a single colour over the entire chip (red on one, green over another, and blue on the third). The test camera was used to evaluate interpolation algorithms and colour accuracy on the final product camera, which used a single chip and Bayer pattern. I was subcontracted on this as the colour evaluator and camera tester, and spent most of my time working with a programmer.

So I don't know where you are getting your information that Bayer interpolation, and your implied message that somehow it is corrupting data so greatly. Perhaps you would care to provide a White Paper on this subject?



Enough so that many academicians, scholars, archivists, technicians and photographers prefer BL files when it comes to recordings images of maps, fragments of the Dead Sea Scrolls, the heat tiles on the underside of the Space Shuttle, the Book of Kells, great works of art, stress fracture analysis in jet engine components, uranium rod cross sections out of nuclear reactors, diamonds and yes, even landscape and commercial photography - now that's "marketing speak."

Yes, that is marketing speak. You also have Anagramm as a competitor in the more scientific market. They use a similar oddity of claiming huge MP count, though to their credit the final file sizes are larger than what Better Light offer.

http://www.anagramm.de/neu/produkte/scanback/david/index_david.html





Talking is one thing... taking pictures is another... just get out there and shoot with whatever kind of camera you have and follow your bliss

You should not take my discussion comments as being that I am against the Better Light, nor that I think it is a bad choice in imaging. Quite simply my problem with Better Light is the same as stated in the posting that started this thread: the claim of MP sizes does not match the file sizes. That makes those claims useless for someone considering whether or not to purchase a Better Light.

The majority of my images that go to corporate and advertising uses will get into final file form based upon an expectation of at least 300 dpi printing. So I can take that 300 figure, and divide it into the maximum file size dimensions of any digital capture device, and figure out what print size will be the maximum without needing upsizing. That is useful information. So again, one more time, why do your MP claims not match the file size reality?

[JeffKohn]

Remember there are NO algorithms at work on a BL file; no anti-aliasing, no anti-moire, no sharpening, no compression, no interpolation (in normal mode), just pure, clean, accurate and unadulterated data. Think how much your data is being stepped on and pushed around with ALL of those algorithms at work in most instant capture systems.You're vastly overstating the image degradation in Bayer-filtered images.

The quality of BetterLight captures is not at question. I don't think anybody is disputing that the image quality of BL scanners is very good. The issue is that your specs are confusing, plain and simple.

[sanking]

The Kodak tri-linear array has 3 rows of 6000 (or 8000) sensors each about 8 pixels apart so while while scanning there are 3 pictures being taken, one each for R, G & B so that the true RGB data is recorded for every point in the scene with no interpolation, again that's 144 million sensors recording data during the scan. Each of these pictures is 48mp in itself.



The Kodak tri-linear 8000 array would be the same one in my Eversmart Pro scanner. When I scan a 4X5 sheet of color negative or transparency film at 2000 ppi the resulting image size is 80mp (8,000 X 10,000 pixels). The file size would be 229mb (8 bit color), 457mb (16 bit color).

Regardless of the final file size in mb, resolution remains 80mp.

I don't understand why count in mp should be any different for the same tri-linear array when used in a Better Light back?

Sandy King

[jkuska]

Sorry I missed the links. Those Kodak sensors individual pixel cell site sizes are your limit. Compare to PhaseOne MFDBs that use 6µm up to 8µm pixel cell site sizes, and no aliasing filter. You can see in the specification links you provided that the 9µm sensor you use is more sensitive than the 12µm sensor you use. PhaseOne also use Kodak CCDs. The sweet spot in sensitivity of CCDs is in the 6µm to 8µm realm; going much smaller tends to increase noise issues, while going larger reduces sensitivity. There is also a factor of the strength of the colour filter over each pixel cell site, in that each colour (Red, Green, Blue) blocks a certain amount of light, and reduced the fill factor of each well. So your claim of "more data" is down to an argument of Bayer interpolation being vastly degrading.

Yes, that is marketing speak. You also have Anagramm as a competitor in the more scientific market. They use a similar oddity of claiming huge MP count, though to their credit the final file sizes are larger than what Better Light offer.

http://www.anagramm.de/neu/produkte/scanback/david/index_david.html

You should not take my discussion comments as being that I am against the Better Light, nor that I think it is a bad choice in imaging. Quite simply my problem with Better Light is the same as stated in the posting that started this thread: the claim of MP sizes does not match the file sizes. That makes those claims useless for someone considering whether or not to purchase a Better Light.

The majority of my images that go to corporate and advertising uses will get into final file form based upon an expectation of at least 300 dpi printing. So I can take that 300 figure, and divide it into the maximum file size dimensions of any digital capture device, and figure out what print size will be the maximum without needing upsizing. That is useful information. So again, one more time, why do your MP claims not match the file size reality?

It's the total sum of all of the post capture processing that "alters" and reinterprets the data, admittedly in often subtle ways, obviously the process has to be fairly good or it wouldn't be consumable... but "consumable" targets the double truck standard and smaller, but for folks and industries who are particular it's a significant enough difference. It's my understanding (I don't retain details only generalities) that larger sensors have a greater "well depth" - that is they can collect more photons and have a great electron volt capacity thus increasing their dynamic range, but I would have to re-research this. For large format the smaller sensors are useless because of the resolution limits of LF lenses, like Kodaks 14K sensor.

I recollect the problem with Anagram (besides cost and dedicated application) is that they offer the 14K sensor which cost a lot and delivers more pixels but does not deliver any more detail because no LF lens that any of us can afford can resolve 14K. After having an optical firm try to design a lens that could deliver more detail to the 10K sensor they came up with a lens that could only deliver a few more line pairs per mm but only at f8 for $5K per lens AND if you stopped up or down ½ a stop you lost ALL of the designed in gains.

And while not taking any of this personally I have simply been trying to explain and exemplify how BL, and apparently others like Cruse and Anagram, determine the pixel count. I understand you don't seem to agree but I did cut and paste the industry reference BL used to make this determination and as I noted they seem to allow for 1 sensor = 1 pixel.

I did mention file size: 274/618 MB for the Super 6 and 488/1100 MB for the Super 8 and I think that does support the pixel count, it certainly support the bit depth. But somewhere in this thread we are simply going to have to agree to disagree. BL uses 3 input sensors, which the JCIA guidelines consider to be 3 pixels, to create 1 output pixel so that a 274 MB file can produce a 20x26.7 inch print at 300ppi BUT with a stochastically screened, micro dot ink jet printer 200ppi produces excellent results so that print size becomes 30x40. Press may be another issue but since the limits of human resolution are approximately 300ppi at 10 inches or so for 20/10 vision a lot is this is mostly moot when considering publication. And for "giclee" prints, stochastic screening and the micro dot size at the print head (typically 1440x2880 or even 2880x5760) pushes everything way below the limits of human vision and that's not even considering the changes in the near point of focus as we age (from 3" as a child to 18" as a young adult to many feet as a senior citizen) or the aesthetically proper viewing distance for a print, typically stated to be 1 to 2 times the diameter of the piece, all of which allows more latitude and flexibility to the printer regarding size.

Note that as a carbon printer my color separations were made at 512lpi for a 20x24 print... I'm talking old school 2 decades ago.

[jkuska]

The Kodak tri-linear 8000 array would be the same one in my Eversmart Pro scanner. When I scan a 4X5 sheet of color negative or transparency film at 2000 ppi the resulting image size is 80mp (8,000 X 10,000 pixels). The file size would be 229mb (8 bit color), 457mb (16 bit color).

Regardless of the final file size in mb, resolution remains 80mp.

I don't understand why count in mp should be any different for the same tri-linear array when used in a Better Light back?

Sandy King

We are back to considering whether a single point in the scene and at the imaging plane is the "definition" of a pixel or whether a single sensor at the imaging plane is a pixel and the differences between capture, display and output pixels. I'll post the standard BL uses it again, I can only present the qualification, if you don't agree with the JCIA guidelines I can't help that but it is a published standard and it does allow for other unusual circumstances. It's not an subjective judgment call or justification, it's a standard and a method:

“The number of pixels on the image sensor which receive input light through the optical lens, and which are effectively reflected in the final output data of the still image.”

This simple statement establishes that a “pixel” must be physically present on the image sensor, it must be light-sensitive, and the information it provides must be relevant to (if not directly present in) the final image data. This last qualification allows instant-capture digital cameras to include so-called “ring” pixels surrounding the area that actually becomes part of the final image, since the “ring” pixels are used to determine some characteristics of the final image. The JCIA guideline also considers devices with multiple image sensors, as well as scanning camera designs:

“For DSCs [digital still cameras] that use an image sensor ( including a linear sensor) that capture an image by spatial or optical movement, the number of effective pixels of the image sensor and the number of sampling positions should be noted. If the total number of sampled pixels (ring pixels can be included) is also noted, the description should clearly state that the noted number is the total number of effective pixels, to avoid consumer misunderstanding.”

Just as instant-capture digital camera vendors can count all three colors of pixels physically present on their image sensors, Better Light can count all three colors of pixels physically present on the trilinear image sensors we use in our scanning backs. There are no “ring” pixels on our trilinear sensors, although there are additional pixels on the sensor that are not counted as “effective pixels”. To arrive at the “total number of effective pixels”, we multiply the number of “effective pixels” on the image sensor by the number of sampling positions.

Isn't this fun? Hey! I use a Canon 850 IS for tourist stuff.

[Gordon Moat]

I did mention file size: 274/618 MB for the Super 6 and 488/1100 MB for the Super 8 and I think that does support the pixel count. But somewhere in this thread we are simply going to have to agree to disagree.

This from your website: "Native maximum resolution at 100% = 8000 x 10600 pixels."

Seems to me that is 84.8 MP

Using a 300 dpi printing method would give a 26 2/3" by 35 1/3" print.

Then for the another: "Native maximum resolution at 100% = 6000 x 8000 pixels."

Seems to me that is 48 MP

Using a 300 dpi printing method would give a 20" by 26 2/3" print.

So where did all those other MP you claim go? How come I cannot use those extra MP you claim in a final print?

[jkuska]

This from your website: "Native maximum resolution at 100% = 8000 x 10600 pixels."

Seems to me that is 84.8 MP

Using a 300 dpi printing method would give a 26 2/3" by 35 1/3" print.

Then for the another: "Native maximum resolution at 100% = 6000 x 8000 pixels."

Seems to me that is 48 MP

Using a 300 dpi printing method would give a 20" by 26 2/3" print.

So where did all those other MP you claim go? How come I cannot use those extra MP you claim in a final print?

That's output/print resolution, not input/capture resolution; apples and oranges
You'll also note the reference to the 48 bit file, that bit depth is generated from the 3 sensors used to capture that point in the scene, not from interpolation and, referring you back to the JCIA guidelines I've posted twice now, 1 sensor = 1 pixel
BL uses those extra pixels to capture the true RGB value in the scene so the data does not have to be interpolated or further processed in any of the ways already mentioned so you are using those extra pixels in the print. I'll I can do is refer you back to those guidelines.

"Native maximum resolution at 100% = 8000 x 10600 pixels."
"Using a 300 dpi printing method would give a 26 2/3" by 35 1/3" print." - that's correct and that's what BL publishes (http://www.betterlight.com/downloads/Product-Info/Spec-Chart.pdf).

But I never print at 300, usually no more than 200 and sometimes 185. I think it's somewhere around 17x22 (certainly 20x24) that prints from small format cameras (Canon mk III) start to fall apart or so I noticed at FotoKino last year. Some BL files can go up to 60x80, though 40x60 and 30x40 are more typical, with no sweat and I have a 3.5x17 foot pano of the Himalayas that's stunning, you can walk up to it an put your nose in it (as though you should), no problem. You, or anyone else, are always welcome to come by the studio, we've got a ton of prints hanging all over the place, just let us know when to expect you... I'm out of here till Monday... it's been real guys, have a good weekend and just remember... Pluto used to be a Planet!

[sanking]

We are back to considering whether a single point in the scene and at the imaging plane is the "definition" of a pixel or whether a single sensor at the imaging plane is a pixel and the differences between capture, display and output pixels. I'll post the standard BL uses it again, I can only present the qualification, if you don't agree with the JCIA guidelines I can't help that but it is a published standard and it does allow for other unusual circumstances. It's not an subjective judgment call or justification, it's a standard and a method:



I appreciate the fact that you have taken the time to explain the reason for the difference between "native resolution" and the other mp figures that are in the BL literature.

Ultimately it makes no difference whether I agree with the JCIA guidelines or not since I am not in a position to impact in any way those guidelines. However, I am fairly certain that the only true indicator of "real" resolution is the MP figure given for native resolution. One can count the R, G and B sensor point as three pixels, but in order to make an image the three have to be collapsed to a single resolution point, which defines native resolution.

My original question was not motivated by a desire to question the capability of Better Light scanning backs, only to understand the reason for the different MP figures. I believe that is now clear.

Sandy King

[Gordon Moat]

......

My original question was not motivated by a desire to question the capability of Better Light scanning backs, only to understand the reason for the different MP figures. I believe that is now clear.

Sandy King

I agree, and that was my motivation for joining this discussion.

[jkuska]

I agree, and that was my motivation for joining this discussion.

Thanks guys, it was fun!

In closing I would fall back to the position that ALL instant capture systems are interpolated UP 300% from the get go and BL files are not interpolated at all in the normal mode and only up 150% in the enhanced mode and then only in one axis.

Then there is also the gross MB difference in files sizes.

Both of these points are STRONG indicators of why you can expect vastly superior print quality at display sizes and larger AND why BL files stand up under the most demanding scrutiny for scholastic & academic studies, forensic investigations or scientific research. The intricate details and color nuances found in fine art, manuscripts, tapestries, artifacts, specimens, on-site locations and many commercial products deserve the highest resolution, the widest color gamut and the longest dynamic range possible.

[JeffKohn]

Then there is also the gross MB difference in files sizes.
Umm, just to correct. If you're implying that the lossless compression that some manufacturers use in their RAW formats in any way degrades image quality that is not the case. What matters is the pixel count and bit-depth, not the size of the file in megabytes. Storing an uncompressed file is just wasteful when there are lossless compression options that cut file size by 50% or more.

[AFSmithphoto]

I just want to make sure that I understand what Betterlight is claiming. The official stance is that by taking three separate color passes at the same scene it has three times the resolution?

That doesn't really increase resolution as the term is traditionally used. (Resolvable detail.) It increases the COLOR resolution, but that’s more appropriately measured in bit-depth rather than megapixels.

For those familiar with sound, megapixels, as MOST of us mean it, is analogous to sample rate, and COLOR resolution is analogous to bit rate. Each is important to image/sound quality but they are important in very different ways.

The long and the short of it is that the number of pixels in the final file is 1/3 what Betterlight marketing might have you believe, but the color of that pixel is 100% determined by the scene. (No interpolation.)

Would anyone at Betterlight dispute that if I took three lock down shots with my Canon T1i, (15mp) each with a different color filter, and joined them in Photoshop I could then claim to own a 45mp camera?

[theBDT]

Would anyone at Betterlight dispute that if I took three lock down shots with my Canon T1i, (15mp) each with a different color filter, and joined them in Photoshop I could then claim to own a 45mp camera?

But your Canon T1i already has a color filtering grid (of sorts) over ITS sensor. You'd be taking three 5mp images effectively (or maybe 5 red 10 green and 5 blue). I think the whole issue of the color filtering grid inherent on most dSLR/digital capture systems today is precisely Better Light's claim to superiority.

[Jeremy Moore]

Would anyone at Betterlight dispute that if I took three lock down shots with my Canon T1i, (15mp) each with a different color filter, and joined them in Photoshop I could then claim to own a 45mp camera?

Yes, because your Canon T1i does not have a sensor that produces pure RGB data from a R sensor, a G sensor, and a B sensor.

Regardless of whether they are using the term resolution as it "is traditionally used", Betterlight is basing their numbers on industry standards.

[srbphoto]

Well this has been some fun technical masterbation!

Jkuska - Doesn't Stephen Johnson uses your backs?

The important thing is how do the images come out. Stephen Johnson's prints are huge with amazing detail. They have been for years. If anyone is in Pacifica, CA (just south of San Francisco) you can see a whole gallery of images done purely digital LF (and I believe using BL backs).

Are there limitations/challenges, yes, but you will get an image as sharp and large as you want. Plus hiking is easier because you won't have all that heavy money in your wallet!

[Jim collum]

the problem started with the traditional digital camera makers inflating their resolution numbers. if they had started out saying their 12 Mp camera really is an 8Mp camera.. then Foveon and Betterlight would just have to say 48Mp.

color bit depth isn't what's better. The Betterlight back captures/resolves detail that would be lost (interpolated) from a Bayer back.

regardless of the marketing dancing that all of the camera makers do in this arena... The Betterlight back still puts out a higher quality image than anything else out there (commercial)


I just want to make sure that I understand what Betterlight is claiming. The official stance is that by taking three separate color passes at the same scene it has three times the resolution?

That doesn't really increase resolution as the term is traditionally used. (Resolvable detail.) It increases the COLOR resolution, but that’s more appropriately measured in bit-depth rather than megapixels.

For those familiar with sound, megapixels, as MOST of us mean it, is analogous to sample rate, and COLOR resolution is analogous to bit rate. Each is important to image/sound quality but they are important in very different ways.

The long and the short of it is that the number of pixels in the final file is 1/3 what Betterlight marketing might have you believe, but the color of that pixel is 100% determined by the scene. (No interpolation.)

Would anyone at Betterlight dispute that if I took three lock down shots with my Canon T1i, (15mp) each with a different color filter, and joined them in Photoshop I could then claim to own a 45mp camera?

[AFSmithphoto]

color bit depth isn't what's better. The Betterlight back captures/resolves detail that would be lost (interpolated) from a Bayer back.

(commercial)

That's not really true. Any camera will capture detail at each pixel. The only part that it can't be sure of is the color/luminance. The "red" pixel isn't an on/off proposition. It sees how bright the light in the red is, and it doesn't need purely red light to be activated, it just won't read as much light from a blue source as it would a red one.

Unless the source is TRULY pure blue, which just plain doesn't occur in nature. Things that we think are blue are really predominantly blue, with a wider range of wavelengths in smaller doses’ than the blue wavelengths’.

So the red pixel sees SOMETHING. It’s a weak signal so it could be a dark red object, or a bright blue object. It has now way to know on its own, but the detail is there. The good news is that the pixel next door is blue, and it registering a very strong signal, so the camera assumes (or interpolates if you like) that the red pixel is seeing the same blue object, and unless the edge of the object is exactly between the two pixels, the camera is correct.

Now, this is WAY oversimplified, and I don't doubt that the betterlight method produces more reliable detail than any other solution, (I don't think anyone has) but to claim a 300% increase is simply ludicrous.

I guess my T1i analogy isn't exactly correct because the sensor is pre-color-filtered.

However, megapixels are an incredibly misleading way of describing increased color information.

Despite what some regulatory comities may say, a Mega-pixel is, simply put a million pixels.

I guess, if you INSIST on describing the advantages of a Betterlight through megapixels, the most accurate thing to say is that it is a 144 megapixel camera that produces a 48 megapixel file.

[Jeremy Moore]

The Betterlight back still puts out a higher quality image than anything else out there (commercial)

Jim, have you surfed Anagramm (http://www.anagramm.com/produkte/scanback/david/index_david.html)'s site in a while? :D

[8x10 user]

I wonder what betterlight is going to do now that Kodak discontinued the trilinear sensor.

[JeffKohn]

The Betterlight back still puts out a higher quality image than anything else out there (commercial)But does it really? I have to wonder how the comparison would come out if you compared a P65+ with Rodenstock HR-W lenses against the 48mp BetterLight with a 4x5 lenses. I'm not saying I know what the outcome would be, just that I think it would be an interesting comparison. Of course the BetterLight seems a bargain compared to the PhaseOne if you take price into consideration.

[jkuska]

Jkuska - Doesn't Stephen Johnson uses your backs?

Yes Stephen Johnson was one of the earliest LF (BL) digital scan back implementers and you are right, his prints are huge and beautiful.

There seems to be some cross over here confusing pixels and resolution. Most of this discussion has been about BL MP count, not resolution, and that MP count is determined by the originally published JCIA INDUSTRY STANDARD which ANSI & ISO adapted in it's entirety.

As to file sizes, one would have to know what size a Canon or Nikon file was PRIOR to ANY post capture processing and compression for comparison but it quickly seems that a 22 MP capture might at best produce a 44 MB file from which 2/3rds of all the color data was missing. (Assuming each sensor produce 16 bits of data or 2 bytes). BL standard, most popular system (the 6000 series) operating in normal mode with NO post capture processing produces a 274 MB file with ALL of the color data present. And BL is a SINGLE pass system, not a three-pass. Certainly the DATA DENSITY prior to post capture processing could be considered to be at least 6+ times greater, for what ever reasons and by what ever means. (274/44=6.23)

Compression is not a concern for BL as it would not speed up our capture times nor tax available hard drive space over the course of a days shoot, even in pano mode. It is never an issue or so far off in the realm of the long tail of diminishing returns it's a moot point.

[JeffKohn]

Your last post contains quite a bit of nonsense, but this thread has grown tiresome so I'll skip the detailed response (and will also be unsubsribing from this thread). Suffice to say that when you start talking about file sizes in megabytes they way you are, it tells me that either you don't know what you're talking about, or you are intentionally trying to cloud the issue.

[jkuska]

I wonder what betterlight is going to do now that Kodak discontinued the trilinear sensor.

Kodak discontinued production of their 6K, 10K & 14K tri-linear sensors but not the 8K sensor (yet). Kodak also still produces a 4K & 2K sensor as well but BL doesn't use these. Kodak alerted BL well in advance and allowed BL to place a final order prior to ending production to permit BL to lay in a stock of these sensors.

These arrays serve the flatbed scanner market, the large format view camera photography market was/is incidental when compared to the scanner market and of no significant consequence to Kodak in the grand scheme of things. Sony discontinued all production of their tri-linear arrays some time ago. No one seems to know what this means... do they believe the flatbed scanner market is saturated? Has there been a paradigm shift in scanner technology? If you know please advise.

It's not clear how Cruse and Anagram are taking this either. There certainly still seems to be a "limited" market for the extreme files these scanning systems produce: the archivist, scholars, academicians and scientist at museums, libraries and in forensics certainly appreciate them as well as many commercial pros, the fine art repro/giclee folks and many landscape photographers. I believe these niche markets could easily be served for another 5 to 10 years and instant capture certainly seems to be fast approaching a limit (determined by the shrinking size of the sensor and the sensors "well depth" as they pack more and more of them into a given area).

[AFSmithphoto]

As to file sizes, one would have to know what size a Canon or Nikon file was PRIOR to ANY post capture processing and compression for comparison but it quickly seems that a 22 MP capture might at best produce a 44 MB file from which 2/3rds of all the color data was missing. (Assuming each sensor produce 16 bits of data or 2 bytes). BL standard, most popular system (the 6000 series) operating in normal mode with NO post capture processing produces a 274 MB file with ALL of the color data present. And BL is a SINGLE pass system, not a three-pass. Certainly the DATA DENSITY prior to post capture processing could be considered to be at least 6+ times greater, for what ever reasons and by what ever means. (274/44=6.23)



I'm not sure if this is meant as a reply to the idea that the betterlight is a 144mp camera that makes a 48mp file or not.

If it is, your claims to file size aren't really relevant. A megapixel is a million pixels, and I don't see how you could dispute that the betterlight system produces a 144mp file, considering it only contians 48 million pixels, no matter how much hard drive space (MBs) its gonna chew up.

You could say that the 22mp file generated by a DSLR is an INTERPOLATED 22mp file, but that is no secret, and it sure as hell doesn't mean the betterlight file is 144mp.

Beyond that . . . additional MB is not really a benifit. As was mentioned before, lossless compression reduces MBs but in no way degrades the image. It just makes the file easier on your computer, so less MBs should be seen as a benifit. Data Density should never be taken as measure of quality.

[theBDT]

As someone whose head is swimming from trying to follow all the different arguments, assertions, and statements in this thread, let me see if I have these basics down correctly:

* Better Light scanners produce a higher quality image, megapixel vs. megapixel, compared to most instant-capture digital imaging devices, as the instant capture devices are interpolated, however—

* Because the nature of Better Light scanners imaging technology, there is no real "apples to apples" comparison of megapixels with, say, dSLRs, and—

* The Better Light marketers claim 140+ megapixels "equivalent" images, though many reasonable people on this board would assert that they are really superior-quality 48 megapixel images, with more faithful reproduction of color, especially in fine details

* Many Better Light scanners have extended sampling modes, which create larger file sizes, and involve a slight interpolation scheme which may or may not be more faithful than dSLR interpolation

* Judging a Better Light scanner image based on sheer megapixels is kind of ridiculous, and judging it based upon file size (megabytes) is completely ridiculous

Is all this basically right?

[jkuska]

As someone whose head is swimming from trying to follow all the different arguments, assertions, and statements in this thread, let me see if I have these basics down correctly:

* Better Light scanners produce a higher quality image, megapixel vs. megapixel, compared to most instant-capture digital imaging devices, as the instant capture devices are interpolated, however—

* Because the nature of Better Light scanners imaging technology, there is no real "apples to apples" comparison of megapixels with, say, dSLRs, and—

* The Better Light marketers claim 140+ megapixels "equivalent" images, though many reasonable people on this board would assert that they are really superior-quality 48 megapixel images, with more faithful reproduction of color, especially in fine details

* Many Better Light scanners have extended sampling modes, which create larger file sizes, and involve a slight interpolation scheme which may or may not be more faithful than dSLR interpolation

* Judging a Better Light scanner image based on sheer megapixels is kind of ridiculous, and judging it based upon file size (megabytes) is completely ridiculous

Is all this basically right?

Your first point: Yes, whether you wish to consider MP, MB or the number of sensors, or the final print, BL produces a superior image, this issue has been addressed and settled a number of times over and over again during the last decade. BL is the standard that PhaseOne and Hasselbad is chasing and trying to emulate by sensor jogging in their MS modes (4 or 16 shot) which, by the way, limits them to "still" shots in a similar way as BL is limited (no beverage pour shots in MS mode). They would like some of the museum and library business BL has captured for the prestige of it.

Your second point: Sort of... there is a published industry standard for determining INPUT pixel count which allows that is 1 sensor = 1 pixel AND which allows for other considerations as well (i.e. "ring pixels" and such), regardless of the ongoing fuss about all this (lodge your complaints with the appropriate governing authorities).

Your third point: It's not a "claim" - it's an industry standard (see above - there is nothing whimsical or opinionated or happenstance about it). In BL's 6000 series scan backs the Kodak tri-linear array has 3 row of 6000 sensors each (1 row each for Red, Green and Blue) to capture the entire scene in one pass, that's 18000 total sensors on the chip. A scan samples 8000 lines across the imaging plane capturing the true RGB values for every point in the scene. So 3 colors x 6000 sensors x 8000 sampled lines equals 144,000,000 sensor sites which include the true RGB value for every point in the scene without any further need for interpolation or Bayer pattern filtration or anti-moire or anti-aliasing processing AND BL does not sharpen or compress the file in any way but leaves those choices completely up to you (as fussy as some folks can be about "raw" processing every shot it is very interesting that they are not more fussy about these other choices as well). It is this sensing of the true color value at every point in the scene that contributes to the vastly superior print sizes or subtle degrees of scrutiny that are possible from BL files, particularly in CRITICAL applications... like evaluating the condition heat tiles on the underside of the Space Shuttle prior to re-entry against a reference file taken with a BL back prior to lift off. There is a reason they don't use any old Nikon or Canon or Hasselblad for this purpose.

Your forth point: Yes, even BL shooters sometimes want MORE so instead of sampling 8000 lines the system can sample up to 12000 REAL lines of real data but in the other axis BL cannot add another 3000 sensors to maintain the 3:4 aspect ratio of the composition so BL interpolates a third "tween" color value for every 2 sensors along the CCD creating a final image that now measures 9000x12000 and has been interpolated up "only" 150%, not the 300% that EVERY single shot instant capture image is interpolated up.

Your fifth point: judge a BL image by inspecting it or standing in front of it or printing it out at VERY large sizes.
Try any of the following with any other camera you can buy:
This was shot at 67% rez, NOT 150% or even 100%:
http://www2.iath.virginia.edu/panorama/watson_uva.html
OR any of these:
http://www.bohonus.com/pages/panoscan_vrs.php
OR any of these (I like Big Money because of the engraving work):
http://www.betterlight.com/fullRes_zoomifyLIST.html
OR this for comparison (BL capture at 50% rez) against 4x5 Ektachrome scanned at 4000ppi (potentially 3x the rez of the BL shot if the detail were there to resolve):
http://www.artisan-digital-services.co.uk/digcapture-3.html
Note the EXTREME differences in detail, dynamic range, clarity and subtle colors in this last example.

I notice that no one has pointed out that technically the 2x2 Bayer grid is actually the original full 3 color data point that is recorded from the scene. Since it contains 4 sensors that is 4 pixels (BL covers that same point in the scene with 12 sensors or 12 pixels - if you accept the industry standard). The manufactures could have used these "real" color values with no further processing for display or print but then the files could only have been enlarged to 1/4 of what we've now become used to without objection. So instead the manufacture process this data and "guess" at the 2 missing color value for each of those points so that the file may be better enlarged for display or print at significant sizes (double truck). They are very, very intelligent "guesses" but they are NOT REAL no matter how sophisticated the algorithms are (again, try shooting a diamond with any single shot instant capture then shoot it with a BL back). The manufactures had a choice, make smaller, more accurate (true to the color in the scene) prints or "cheat" and interpolate the files up to more significant sizes (and then compress them down). You too can "cheat" a BL file up to the same degree if you like by re-sizing and interpolating it up in Photoshop: then just imagine what a huge print you could make, maybe 60x80 or more, which would still be just as acceptable as whatever you are currently used to.

Finally... the 22MP Canon is sensing only one of three colors across a field measuring 3,744 x 5616 while the BL Super 6 is sensing all three colors across a field measuring 6000 x 8000. That, all by itself, should be enough to account for SOME perceptual degree of both quantitative AND qualitative difference in the results anyone can expect from a BL system.

All of the above is expressed in "lay" terms and is intended for thoughtful consideration only, to be reflected upon, to make a point. It is NOT a Ph.D. dissertation nor is it intended as such.

[Marko]

On my monitor the Motorcycle shot on Ron Finleys site has tons of banding, looks unprintable to me.

Of course it is unprintable - the image is greatly reduced in order to fit the computer monitor. If you tried to print it, it'd be the size of a postage stamp.

That's also where the banding comes from - reducing the size means throwing out pixels and it will be particularly noticeable in the gradients. There are ways to deal with this problem, but that would be way OT here.

It's been said so many times on this board, especially relating to the nuances of LF: web display serves only as an illustration and approximation of the image. You need to look at the print (or full size display) for effective comparison or judgement.

[jkuska]

By the way:
Anyone can go here:
http://www.flickr.com/search/?q=betterlight
and see examples of all sorts of BL photography OR, in some case BL's studio and other set ups.
Some of these images are not actually BL captures but most are.
AND some few of them allow you to look at 100% files.
Notice that some of them even have people in them.

AND

Anyone is welcome to visit the factory/studio in the Bay area of San Francisco, just give a heads up.

ps... there is a gentleman up in the Puget Sound who collects Stradivarius Violins (imagine that) who insists that only a BL back can capture the true subtle tones and detail of these masterpieces: http://www.sheilascorner.com/collectordave.html
It's likely he could afford any system of his choosing and he chose BL and taught himself view camera technique just to be able to use the scan back. His is acclaimed to be the finest violin collection in the world today.

[bob carnie]

Thanks for the education.
I just don't like banding, figured it could be done before putting on a site since I cannot see the large print maybe downsizing the file for web would be something to work on. just a thought.

Of course it is unprintable - the image is greatly reduced in order to fit the computer monitor. If you tried to print it, it'd be the size of a postage stamp.

That's also where the banding comes from - reducing the size means throwing out pixels and it will be particularly noticeable in the gradients. There are ways to deal with this problem, but that would be way OT here.

It's been said so many times on this board, especially relating to the nuances of LF: web display serves only as an illustration and approximation of the image. You need to look at the print (or full size display) for effective comparison or judgement.

[Marko]

Thanks for the education.
I just don't like banding, figured it could be done before putting on a site since I cannot see the large print maybe downsizing the file for web would be something to work on. just a thought.

Me neither. But that has nothing to do with printing. It's all about shoddy web development and production, which should not be the topic of this thread. Of course it can be done, nothing complicated about it. All one needs is a competent web developer supplied with a solid storyboard and firm and clear set of requirements.

Speaking of which, I fail to understand why so many photographers tend to skimp on one of the most important promotion tools at their disposal and instead opt to present themselves as cheap and/or technologically inept to anybody curious enough to look them up on the Web.

[Jeremy Moore]

jkuska,

Thanks for being a part of this thread and re-answering the same questions over and over again while being professional and polite.

I use an 8k at work and have compared it to medium format digital back shots (even the "latest and greatest") and there is still no comparison.

[mcfactor]

I just saw that Hasselblad is producing a 39 megapixel back with what they call "multi-shot" technology. If I read it correctly, the camera takes 4 shots, each with one color, then combines them. Would this be similar to what betterlight's do? and if so, could the hasselblad captures be called 156 megapixels (by betterlight's standards)?

[8x10 user]

Thats what Sinar's multi shot back does, it creates a final file with pure color just like betterlights system.

[Gordon Moat]

Well, Seitz Phototechnk AG of Switzerland make a real 160 MP camera:

http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f934.cfm

Check the file size of 7500 by 21250 pixels. If they used the Better Light "standard", then they could claim even higher Mega Pixels. Also, they are using a Dalsa sourced imaging chip, and not Kodak. The downside is that they are much more expensive than Better Light. I really have to wonder why they do not use the same "standard" of measurement as Better Light? Could it be they feel they do not need to overstate the capability?

[jkuska]

Well, Seitz Phototechnk AG of Switzerland make a real 160 MP camera:

http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f934.cfm

Check the file size of 7500 by 21250 pixels. If they used the Better Light "standard", then they could claim even higher Mega Pixels. Also, they are using a Dalsa sourced imaging chip, and not Kodak. The downside is that they are much more expensive than Better Light. I really have to wonder why they do not use the same "standard" of measurement as Better Light? Could it be they feel they do not need to overstate the capability?

No, Seitz could not claim a higher MP count because they don't have the MP to start with... the Dalsa chip that Seitz uses has only ONE row of sensors on it, not three like the Kodak chip. The sensors alternate Red, Green, Blue along the entire length of that single strip. This means that of the 7500 sensors along that strip there are 2500 EACH of R, G & B (note that there are NOT 2 green sensors for each pair of red and blue sensors), so Seitz's MP count is correct, 7500 x 21250 = 160 MP (1 sensor = 1 pixel). This also means that all Seitz images must be interpolated up 300% as well. Seitz is FAST and EXPENSIVE but it still does not compare to a BL capture: any way it's always a case of the right tool for the right job.

[jkuska]

Well, Seitz Phototechnk AG of Switzerland make a real 160 MP camera:

http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f934.cfm

Check the file size of 7500 by 21250 pixels. If they used the Better Light "standard", then they could claim even higher Mega Pixels. Also, they are using a Dalsa sourced imaging chip, and not Kodak. The downside is that they are much more expensive than Better Light. I really have to wonder why they do not use the same "standard" of measurement as Better Light? Could it be they feel they do not need to overstate the capability?

By the way... A Super 6 BL in Pano Mode can capture a field that measures 6000 x 65000 which by the accepted INDUSTRY standards is 1170 MP (or 390 MP by the way some here seem to prefer) and the Super 8 that becomes 8000 x 65000 or 1560 (or 520) MP by the same reasoning.

[Gordon Moat]

.....Seitz is FAST and EXPENSIVE but it still does not compare to a BL capture: any way it's always a case of the right tool for the right job.

Two out of Three claims there. I have seen prints from the Seitz, and prints from a Better Light, and there is a noticeable improvement from the Seitz. However, as you noted, they are more expensive.

Kudos to your persistent marketing efforts. I hope that convinces enough people to buy a Better Light that your company continues into the future, or at least until all the Kodak tri-linear CCDs disappear.

[jkuska]

Two out of Three claims there. I have seen prints from the Seitz, and prints from a Better Light, and there is a noticeable improvement from the Seitz. However, as you noted, they are more expensive.

Kudos to your persistent marketing efforts. I hope that convinces enough people to buy a Better Light that your company continues into the future, or at least until all the Kodak tri-linear CCDs disappear.

hmmm... the stating of technical specifications and industry standards are not simply "claims" nor do they constitute "marketing" (this forum does not permit ADVERTISING).

I would suggest that the prints you saw were not enlarged to equal degrees because there simply is no comparison. Seitz sacrifices dynamic range for speed AND, once again, is only recording 1/3rd of the color data in a rasterized pattern. I certainly don't see a "noticeable improvement" between these two images:
one from Seitz's site;
http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f1002.cfm
and one from a BL users site (NOT BL's site) at only 62% rez;
http://www2.iath.virginia.edu/panorama/watson_uva.html
Pay attention to the roof, brick and and other fine architectural detail and clock face found on the central dome here in the middle of campus. Also note the size of the little highlighted navigation indicator when zoomed all the way in. I do not see ANY comparison what so ever. I'd welcome links to other examples for a better example of a Seitz image but presume that Seitz would only post respectable examples showing their best capabilities.

BL has relentlessly been subjected to numerous shootouts, comparisons, analysis, discussions and articles for over a decade now and has always set the standard by which all other digital imaging is judged against. The only remaining competition in the market place that comes close are Cruse and Anagram, both of which effectively use identical technology and both of which, like Seitz, are remarkably more expensive.

[Nathan Potter]

Jim, have you surfed Anagramm (http://www.anagramm.com/produkte/scanback/david/index_david.html)'s site in a while? :D

The Anagramm scanback at 14,400 X 23,600 pixels for a 72 X 118 mm. format seems pretty formidable at first glance. But this translates to a 5 X 5 um xy pixel size. Now we're talking about a resolution of 100lp/mm. over that field of view for compatibility with the pixel size to contain a 5 um circle of confusion at best focus. Otherwise we're sampling a defocused spot.
That further requires about a perfect lens which is diffraction limited at f/3.5 (or wider) in order to achieve a compatible resolution. A tall order (not to mention the stability of the setup).

Nate Potter, Austin TX.

[jkuska]

The Anagramm scanback at 14,400 X 23,600 pixels for a 72 X 118 mm. format seems pretty formidable at first glance. But this translates to a 5 X 5 um xy pixel size. Now we're talking about a resolution of 100lp/mm. over that field of view for compatibility with the pixel size to contain a 5 um circle of confusion at best focus. Otherwise we're sampling a defocused spot.
That further requires about a perfect lens which is diffraction limited at f/3.5 (or wider) in order to achieve a compatible resolution. A tall order (not to mention the stability of the setup).

Nate Potter, Austin TX.

Large Format lenses use much more glass AND cover a much larger image circle, usually to allow for swings and tilts, than small and medium format lenses. Large format lens design has been about as excellent as it can be for decades, better than film was ever capable of, but not as good as digital sensors. For a real comparison with a BL 8000 at 50% rez against film scanned a 4000 dpi look here:
http://www.artisan-digital-services.co.uk/digcapture-3.html
Note the extreme differences in detail and dynamic range then page forward for 2 other examples.

When BL brought out it's 10K a couple of years ago the FIRST question asked of BL was when were they going to produce a 14K (because folks know of Anagrams 14K). BL answered NEVER because there is NO LF lens that can resolve that level of detail, the 10K was already pushing the questionable limits of possibilities here (note that Anagrams 14400 lines divided by 72mm divided by 2 = 100 lp/mm (a 5 micron sensor), WAY beyond lens resolution and diffraction limited around 5.6 (similar to a Nikon D2X) on top of everything else). Anagram has been selling the exact same lens BL recommends for repro, the Rodenstock Sironar S 180mm f5.6:
http://www.rodenstock-photo.com/mediabase/original/e_Rodenstock_Digital_Lenses_3-26__8236.pdf
(Note the MTF chart only goes out to 30 lp/mm)
BL knows that while Anagram is selling more pixels with their 14K, they are not resolving any more detail, they are just dividing an already "fuzzy" dot of detail (the airy disk - named after George Airy) into more pixels. As you stop down the airy disk gets larger and when it start spilling out onto other sensors you begin loosing even more detail. This is EXTREMELY limiting with small sensors, see the interactive example here, half way down the page under VISUAL EXAMPLE: APERTURE VS PIXEL SIZE:
http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm
Choose your camera (or sensor size) then stop down and see what happens, as the airy disk begins to spill over your loosing detail.

Since the narrow dimension of the high end Canon is 3744 lines AND the narrow dimension of a BL back is either 6000 or 8000 lines BL is certainly are capturing a finer record of the scene in the larger format at equivalent coverage, enlargement or magnification .

6000 & 8000 lines divided by 72mm divided by 2 = 42 & 55 lp/mm (with 12 and 9 micron sensors) at the imaging plane. Canon doesn't clearly post lp/mm used for their MTF charts but 3744 lines divided by 25.4mm divided by 2 = 78 lp/mm (6 micron sensors) but it's not clear if their lenses can actually deliver this level of detail either. Regardless, the larger image of the BL back has more total lines AND the larger sensor used in BL becomes diffraction limited much less quickly (at 81 and 144 µm² they have a diffraction limit around f16 and 11 for a far greater depth of field) than the smaller sensors in smaller cameras. Large sensors generally also have a greater "well depth" also which provides a superior dynamic range.

BL has tried to have custom lenses designed and manufactured that could deliver more lp/mm but the cost of the lenses were prohibitive ($5K plus) AND all the design gains were lost if you stopped up OR down 1/2 a stop off of optimum apertaure (f8). Additionally the gains were in that area of the long tail of the curve where all diminishing returns lay, insignificant. It other words it wasn't worth the cost for just a few more line pairs.

[Mike1234]

I rather like the FOVEON chips. It's utter BS that the unregistered... out-of-whack... side-by-side... Bayer sensors are better. Three colors are three colors whether they'er stacked or not and at least the FOVEON chips are REGISTERED and IN-LINE with the photons hitting the chip. WHAT DIFFERENCE DOES IT MAKE REGARDS TO "RESOLUTION" WHETHER THEY'RE STACKED OR SIDE-BY-SIDE? Just one difference... no Bayer funny stuff.

[jkuska]

I rather like the FOVEON chips. It's utter BS that the unregistered... out-of-whack... side-by-side... Bayer sensors are better. Three colors are three colors whether they'er stacked or not and at least the FOVEON chips are REGISTERED and IN-LINE with the photons hitting the chip. WHAT DIFFERENCE DOES IT MAKE REGARDS TO "RESOLUTION" WHETHER THEY'RE STACKED OR SIDE-BY-SIDE? Just one difference... no Bayer funny stuff.

You are right and by the industry standard Foveon can take it's 4.7 mega-senor chip with 3 distinct outputs from each sensor and count those as 3 pixels each, hence their 14 MP rating.

While I'm not sure about the technical limitations, wouldn't it be great if they came out with a full frame sensor? Since the Foveon sensor is about 14 x 21 mm that would increase the area about 3 times which potentially make them a 42 MP camera, with NO interpolation (no Bayer pattern sensor), basically blowing all the other small and medium format cameras off the market.

[Mike1234]

While I'm not sure about the technical limitations, wouldn't it be great if they came out with a full frame sensor? Since the Foveon sensor is about 14 x 21 mm that would increase the area about 3 times which potentially make them a 42 MP camera, with NO interpolation (no Bayer pattern sensor), basically blowing all the other small and medium format cameras off the market.

Okay... I just drooled on my keyboard. Dang... one could stitch some bad-A images with an adapter back on 4x5/5x7/8x10......

[8x10 user]

I use a sigma with a foveon for my eBay pictures. Its not a bad camera, I got mine for about $300. It preforms really well at ISO 50 (I use strobes) but I hear its not so good for high ISO's. The per channel image quality seems pretty pure to me. My lenses kinda suck (lowest end available) so I sometimes see some chromatic aberration at high contrast areas. The way the foveon chip works it shoots each color at a different layer. I have to wonder if there would be problems if it was used with a lens that is 100% apochromatic, and weather or not some of the channels are blurred by the layers in front of it. The images look good when opened with Sigma pro or camera raw but I remember opening them in regular photoshop and the images looked fuzzy. Maybe there is a trick to how the three layers of information are processed in camera raw or Sigma pro? Anyway thats my little blurb about the Foveon sensors.

I did hear something about 4x5" CCD chips that are being developed for the government and should reach the consumer level at some time in the future.

BTW Anagramm says they had special lenses made for the David scan back. My guess is that the scan back is best for Art Repo due to the limitations of depth of field and diffraction.

[Jim collum]

From experience, anything more than the Super 6k is a waste in the field... and with most lenses and shooting apertures, the 6K will suffice.

[Kirk Gittings]

Just curious Jim, how would you rank your BL back compared to a first class drum scanned 4x5 and your Leaf back (it is leaf right?)

[Jim collum]

Just curious Jim, how would you rank your BL back compared to a first class drum scanned 4x5 and your Leaf back (it is leaf right?)

the 4x5 film and the back is pretty close to a wash in my experience. You get a little more 'area' captured with the film, but you get a 'cleaner' image with the Betterlight. You can drum scan the film to a greater degree.. but in most cases, you're not capturing any more real data than with the Betterlight. Most lenses don't resolve more than the Betterlight captures using 'field' apertures. I'll still shoot film, but it's more for the look of a specific emulsion, rather than for added detail. I do like the immediate feedback of a finished image from the Betterlight. I can leave a site knowing that everything is 'right'

There are conditions where the Betterlight doesn't fare as well in the field.. Windy is the most obvious. It's not the wind blowing the subject, as much as the wind blowing the camera that's the issue.

The Leaf back is definitely more convenient than either the Betterlight or film. I use a Horseman SWD with it, so if necessary I can do a 9 shot stitch (the back moves, so there's no issue at all with parallax). The SWD can be used with a 35mm digital lens, and stopped down, covers the full stitched area well.... giving me a huge wide angle lens (great for interiors).

i find if i'm going to be shooting an older lens, wide open (Veritos, Cooke, Petzval).. then i'll use the Betterlight. Normal landscape, then it's a tossup as far as the mood i'm in. For architecture work, I have strobes, so for the cases where i need lighting, then I use the Leaf.

The only time I'll use film now, is if I want the look/feel of Tri-x and Pyro (using it either with 4x5 or 6x6 ).

From a customer perspective... I've had my Betterlight since 2001... and have dealt with the company since then. Betterlight has by far, the best support (technical and customer) than *any* other company I've ever dealt with.

[jkuska]

As an added note to the earlier comments on how BL's calculates its megapixel rating: Sigma (Foveon), Cruse and Anagramm ALL use the same industry standard that BL uses to determine their own MP rating.

The same standard was also used to calculate the MP ratings of the PhaseOne FX, Kigamo systems 6000 & 8000 & Jobos ProScan 10500; all 4x5 scan backs that went out of production some time ago.

[jkuska]

Just curious Jim, how would you rank your BL back compared to a first class drum scanned 4x5 and your Leaf back (it is leaf right?)

Hi Kirk... Jim

Did you check out the links I've provided to Richard Kenwards site, Artisan-Digital-Services, in England? He post clear comparisons of BL scans and scanned film here:
http://www.artisan-digital-services.co.uk/digcapture-3.html
and I was wondering what you (and Jim) thought of them?
Richard has his own drum scanner as some of his clients still like film, though I believe he shoots both now in those situations. Sometimes it is easier to just hold a sheet of film up to the light to see what you're talking about.

[Kirk Gittings]

Hi Kirk... Jim

Did you check out the links I've provided to Richard Kenwards site, Artisan-Digital-Services, in England? He post clear comparisons of BL scans and scanned film here:
http://www.artisan-digital-services.co.uk/digcapture-3.html
and I was wondering what you (and Jim) thought of them?
Richard has his own drum scanner as some of his clients still like film, though I believe he shoots both now in those situations. Sometimes it is easier to just hold a sheet of film up to the light to see what you're talking about.

I have not, but I will. I asked Jim simply because I know him and his broad experience, and greatly value his opinion.

[Mike1234]

I'd really like a large digital capture device but they're just too pricey for me... hence my decision to go with big film.

[Nathan Potter]

the 4x5 film and the back is pretty close to a wash in my experience. You get a little more 'area' captured with the film, but you get a 'cleaner' image with the Betterlight. You can drum scan the film to a greater degree.. but in most cases, you're not capturing any more real data than with the Betterlight. Most lenses don't resolve more than the Betterlight captures using 'field' apertures. I'll still shoot film, but it's more for the look of a specific emulsion, rather than for added detail. I do like the immediate feedback of a finished image from the Betterlight. I can leave a site knowing that everything is 'right'

There are conditions where the Betterlight doesn't fare as well in the field.. Windy is the most obvious. It's not the wind blowing the subject, as much as the wind blowing the camera that's the issue.

The Leaf back is definitely more convenient than either the Betterlight or film. I use a Horseman SWD with it, so if necessary I can do a 9 shot stitch (the back moves, so there's no issue at all with parallax). The SWD can be used with a 35mm digital lens, and stopped down, covers the full stitched area well.... giving me a huge wide angle lens (great for interiors).

i find if i'm going to be shooting an older lens, wide open (Veritos, Cooke, Petzval).. then i'll use the Betterlight. Normal landscape, then it's a tossup as far as the mood i'm in. For architecture work, I have strobes, so for the cases where i need lighting, then I use the Leaf.

The only time I'll use film now, is if I want the look/feel of Tri-x and Pyro (using it either with 4x5 or 6x6 ).

From a customer perspective... I've had my Betterlight since 2001... and have dealt with the company since then. Betterlight has by far, the best support (technical and customer) than *any* other company I've ever dealt with.

Jim, the practical "field aperture" that you mention above was the point in my original message. That is, the lens really is the limit in capturing detail for both film and scanning backs so I would guess that as you state, and Kirk asked, there should not be much noticeable difference between a scanned high quality image and a high quality scanning digital back as far as detail captured is concerned. Given, of course the best quality image capture in both cases.

In large format, practical apertures will generally be f/16 or smaller so in all cases film grains or pixels will be sampling defocused circles of confusion. Of course the rendering of defocused COCs will have suttle differences between film and scanning back. The mathematics of digitizing overlapping COCs is quite complicated and interesting but the visual results are what interest most of us.

Nate Potter, Austin TX.

[jkuska]

I'd really like a large digital capture device but they're just too pricey for me... hence my decision to go with big film.

You may have to start coating your own emulsions sometime here in the next decade after all silver halide based photography is relegated to the historical processes. At least we will still have Bostick & Sullivan and Photographers Formulary to fall back on for the foreseeable future.

You could shop eBay where you can often get a good value on high end digital for under $6K (typically BL but others as well). As Jim so kindly alludes to it doesn't matter where you buy a system, new or used, you will receive the same excellent support and service that every customer does. BL produces a very high end premium product that's built like a tank and it is the most upstanding and moral company I have ever worked for (and I'm 62).

[jkuska]

Jim, the practical "field aperture" that you mention above was the point in my original message. That is, the lens really is the limit in capturing detail for both film and scanning backs so I would guess that as you state, and Kirk asked, there should not be much noticeable difference between a scanned high quality image and a high quality scanning digital back as far as detail captured is concerned. Given, of course the best quality image capture in both cases.

In large format, practical apertures will generally be f/16 or smaller so in all cases film grains or pixels will be sampling defocused circles of confusion. Of course the rendering of defocused COCs will have suttle differences between film and scanning back. The mathematics of digitizing overlapping COCs is quite complicated and interesting but the visual results are what interest most of us.

Nate Potter, Austin TX.

hmmm... while Ektachrome has a Dmax of about 4.0, implying about a 12 stop range, only about 6 of those stops are printable with detail. BL has an 11+ stop dynamic range with the better part (if not all) of it being printable with detail. Of course different tone curves can alter this but the boss, for example, shoots all of his landscapes with a 10 stop "S" Curve that's very gentle and captures everything.

[Mike1234]

You may have to start coating your own emulsions sometime here in the next decade after all silver halide based photography is relegated to the historical processes. At least we will still have Bostick & Sullivan and Photographers Formulary to fall back on for the foreseeable future.

You could shop eBay where you can often get a good value on high end digital for under $6K (typically BL but others as well). As Jim so kindly alludes to it doesn't matter where you buy a system, new or used, you will receive the same excellent support and service that every customer does. BL produces a very high end premium product that's built like a tank and it is the most upstanding and moral company I have ever worked for (and I'm 62).

I can buy a BL system for $6K? Hmm... I wonder what it would cost to have an adapter/slider back made for a 5x12 Korona. :D

[Aahx]

Well for me at least the price is not my only consideration for not purchasing a scan back. But the wieght...not just of the scanner, but the laptop and cables as well. I use quickloads for that very reason to reduce wieght on my trips further afield. If I had to bring a laptop as well, i would have too drop a lens or two from my kit to make up for the added wieght/bulk. So a perfect solution for me would be a lightwieght large format back that was stand alone. This way I could also potentialy carry a "back up" one in case one went down in the inclement conditions here in Alaska. Untill such a device is invented, film will still be my prefered solution of choice.

[jkuska]

I have not, but I will. I asked Jim simply because I know him and his broad experience, and greatly value his opinion.

By the way Kirk... Jim and I met back in the late 80's when I had my own Carbon Lab in Santa Cruz (Limited Edition - Photo/Graphics: he walked in hoping to find a Dye Transfer lab but liked the Carbon process even better). I taught him everything he knows about photography. ;-)

I received my BFA in Photo from the Los Angeles Art Center, College of Design when it was still in Highland Park back in '76 and then worked for some of the big boys in town (like Carl Furuta, Jeff Nadler & Bill Robins); and a couple of great labs as well (like Hecht Custom Photographics); as well as making Ted Staidles acquaintance and picking his brain (a premier Cibachrome printer and masking/imposition artist/technician); all in Hollywood before moving up North to Santa Cruz and the Bay area and starting the Carbon project.

[jkuska]

I can buy a BL system for $6K? Hmm... I wonder what it would cost to have an adapter/slider back made for a 5x12 Korona. :D

well, the adapter may be pushing it but geez, have you seen all the good deals on 4x5 equipment, cameras and lens on eBay? I picked up two Cambo Legends for $250 plus shipping and an APO Sinronar S 180 for $450.

[tgtaylor]

You may have to start coating your own emulsions sometime here in the next decade after all silver halide based photography is relegated to the historical processes. At least we will still have Bostick & Sullivan and Photographers Formulary to fall back on for the foreseeable future.

You could shop eBay where you can often get a good value on high end digital for under $6K (typically BL but others as well). As Jim so kindly alludes to it doesn't matter where you buy a system, new or used, you will receive the same excellent support and service that every customer does. BL produces a very high end premium product that's built like a tank and it is the most upstanding and moral company I have ever worked for (and I'm 62).


Originally Posted by Kirk Gittings
Just curious Jim, how would you rank your BL back compared to a first class drum scanned 4x5 and your Leaf back (it is leaf right?)

the 4x5 film and the back is pretty close to a wash in my experience. You get a little more 'area' captured with the film, but you get a 'cleaner' image with the Betterlight. You can drum scan the film to a greater degree.. but in most cases, you're not capturing any more real data than with the Betterlight. Most lenses don't resolve more than the Betterlight captures using 'field' apertures. I'll still shoot film, but it's more for the look of a specific emulsion, rather than for added detail. I do like the immediate feedback of a finished image from the Betterlight. I can leave a site knowing that everything is 'right'


"...pretty close to a wash in my experience" sounds to me like Jim finds film still superior. That must be a poignant admission considering Jim has thousands invested in digital.

Glad to hear that you guys are doing alright. I'll probably be a BL customer one day.

Thomas

[Jim collum]

"...pretty close to a wash in my experience" sounds to me like Jim finds film still superior. That must be a poignant admission considering Jim has thousands invested in digital.

Glad to hear that you guys are doing alright. I'll probably be a BL customer one day.

Thomas

I didn't say that film was superior. Frankly.. they're all tools. I've always found just examining resolution a waste of time. Both film and the Betterlight will give me outstanding 40x50" prints .. and really.. how many 40x50" prints are needed :)

I shoot with an end result in mind.. and have a variety of tools to get me to what I want. Lately, my favorites have been the HBH Petzval lens (which I use mostly with the Betterlight.. since I can see exactly what i'm going to end up with before I leave .. and the look of that image can vary greatly with very minor tweaks of focus), and the Zeiss Planar 110/2, which I alternate using with the Leaf back, or trix film. Neither of those result in very detailed images

The Betterlight back at 6000x8000 will give an excellent 40x50" image with no visible signs of digital artifacts in the final print (and that's at 'nose-to-print' distance'). The back has performed flawlessly since 2001.. and Mike and company have been at the other end of email or the phone whenever i've had questions. There are compromises that are made if used in the field... but they're not as bad as one would think. Weight isn't a problem at all.. I can put camera (Ebony sv45Ti), 4 lenses, Betterlight, cables, battery and laptop in a f64 backpack (not their large one). It all fits easily in overhead storage on a plane.

[AFSmithphoto]

No, Seitz could not claim a higher MP count because they don't have the MP to start with... the Dalsa chip that Seitz uses has only ONE row of sensors on it, not three like the Kodak chip. The sensors alternate Red, Green, Blue along the entire length of that single strip. This means that of the 7500 sensors along that strip there are 2500 EACH of R, G & B (note that there are NOT 2 green sensors for each pair of red and blue sensors), so Seitz's MP count is correct, 7500 x 21250 = 160 MP (1 sensor = 1 pixel). This also means that all Seitz images must be interpolated up 300% as well. Seitz is FAST and EXPENSIVE but it still does not compare to a BL capture: any way it's always a case of the right tool for the right job.


All true, but to claim that the 160 megapixels on the Seitz produces an equivelent resolution to a betterlight's would be to claim that NO useful information can be gained from interpolating the three seperate channels into one single file. Is that your contention?

[AFSmithphoto]

hmmm... the stating of technical specifications and industry standards are not simply "claims" nor do they constitute "marketing" (this forum does not permit ADVERTISING).

I would suggest that the prints you saw were not enlarged to equal degrees because there simply is no comparison. Seitz sacrifices dynamic range for speed AND, once again, is only recording 1/3rd of the color data in a rasterized pattern. I certainly don't see a "noticeable improvement" between these two images:
one from Seitz's site;
http://www.roundshot.ch/xml_1/internet/de/application/d438/d925/f1002.cfm
and one from a BL users site (NOT BL's site) at only 62% rez;
http://www2.iath.virginia.edu/panorama/watson_uva.html
Pay attention to the roof, brick and and other fine architectural detail and clock face found on the central dome here in the middle of campus. Also note the size of the little highlighted navigation indicator when zoomed all the way in. I do not see ANY comparison what so ever. I'd welcome links to other examples for a better example of a Seitz image but presume that Seitz would only post respectable examples showing their best capabilities.

BL has relentlessly been subjected to numerous shootouts, comparisons, analysis, discussions and articles for over a decade now and has always set the standard by which all other digital imaging is judged against. The only remaining competition in the market place that comes close are Cruse and Anagram, both of which effectively use identical technology and both of which, like Seitz, are remarkably more expensive.

The fact that you prefer a print from a betterlight is not a technical specification or industry standard. That was the "third" point he was refering to.

What do you mean by "enlarged to equal degrees?" If you mean it as it sounds, I think you're being deliberately mis-leading.

"Enlarged to equal degrees" means an equal degree of enlargment. A 300 dpi print of a 6000 x 8000 betterlight file (48mp) would be a 20 x 27 inch print.

A 300 dpi print of a 7500 x 21250(160 mp) Seitz file would be a 25x70 inch print.

This is an equal degree of enlargment from each file, but the final print sizes are vastly different.

I'm sure the Non-interpolated Betterlight looks better inch for inch, but thats hardly a useful comparison. I'll make a a 1x1 inch print from the original Canon Digital Rebel, and it will look FANTASTIC inch for inch.

Print the Betterlight at a comperable size, not a comperable degree of enlargement. Then you can compare.

I'm not sure pointing people to a Seitz file of one subject and asking them to compare to a file of several stiched betterlight exposures of a different subject is any less mis-leading.

[percepts]

Choose your camera (or sensor size) then stop down and see what happens, as the airy disk begins to spill over your loosing detail.

but each sensor does not have one airy disc to itself. There are countless airy discs spread all over each sensor and all overlapping each other and overlapping sensors. They are not all centred on a sensor unless you have some magical powers over the lens optics which you don't.

[sanking]

the 4x5 film and the back is pretty close to a wash in my experience. You get a little more 'area' captured with the film, but you get a 'cleaner' image with the Betterlight. You can drum scan the film to a greater degree.. but in most cases, you're not capturing any more real data than with the Betterlight.

That is what I would have thought also, but the comparison posted at http://www.artisan-digital-services.co.uk/digcapture-3.html suggests quite a different story.

Is that a good comparison, or just marketing hype?

Sandy King

[Jim collum]

That is what I would have thought also, but the comparison posted at http://www.artisan-digital-services.co.uk/digcapture-3.html suggests quite a different story.

Is that a good comparison, or just marketing hype?

Sandy King

I suspect it's the real thing. Most of the Super 8K backs are sold into the art reproduction market, where you can control the environment, depth of field, and choose a specific high resolution lens for copy work. I suspect that back is an 8K. My comments were mainly used for my work.. I usually don't choose a lens for it's resolution as much as it's 'character'. If anyone would like to try a comparison with their own lens/camera, and are in the Santa Cruz, Ca. area.. let me know.. I'd be happy to supply the back for it.

Jerry will need to confirm the back type used in that example.

[tgtaylor]

I would suggest that what's being "compared" is the resolution capabilities of the two digital systems - the BL back and the "drum Scanner" - and not the film itself. I would further suggest that if the drum scanner took the picture itself without the aid of the negative, it would appear pretty much as in the comparison. In other words, the scanner is a crappy capture medium compare with the negative.

Earlier today I did a quick search for film grain size on this site and came up with a thread (11x14 is how many pixels) that stated that grain size in a negative can vary from sub micron to the typical 4 micron and larger. If that is true, then grain size seems to depend on what's being photographed and the associated exposure conditions. Pixel size, on the other hand, is fixed and non variable. Now ask yourself "What capture system is capable of the highest resolution: One whose sensors are variable in size depending on the conditions such as a sheet of film, or one whose sensors are fixed such as the digital/scanning back? Migrating grains of variable dimensions or static buckets? I would think that the system with the most variability would be superior to the fixed and static one.

But I admit that I am not educated in this. So I looked at a 11x14 print hanging in the hallway of San Francisco. This is a sharp print taken on a very clear afternoon (a Sunday) after a major winter storm blew thru that Saturday from the top of Twin Peaks with a clean normal (150mm) lens on a 4x5 camera. There was a tad, a hint, of blue haze in the far background so I used a clean B+W UV filter and shot the image on Fuji Acros, developed in Xtol, and printed on Oriental Seagull VC using a rented Beseler 45MX and glassless holder. The print is tack sharp edge to edge. You can see people walking on the street far below, drivers and passengers in their automobiles, and in the distance, say 3 miles, the hands of the clock on the Ferry Building at the foot of Market street. The clocks hands are distinct but looking hard you reckon the time is 2:30. A magnifying glass shows clearly 2:28. I think that this print is sharper than the BL school link shot and certainly sharper than the drum scan.

But then again, I'm no expert in this.

[Jim collum]

grain size isn't a consideration... at least in most cases. You'll find that most lenses in the field, at 'field' apertures, won't resolve down to the single grain level. In fact, you'll find that the camera/lens/aperture combination gives out well before the capability of a good drum scanner.

[Jim collum]

I think a nice post of a Martha Casanave pinhole image ( http://marthacasanave.com/coastal.htm ) should bring back a reminder of what's really important.... :)

Given the choice between a $30,000 digital capture setup, or her eye.... i know which i'd choose

[tgtaylor]

I think that there is two definitions of grain. One defines grain as the space between the silver halide particles, and the other as the silver particles themselves. The LF thread referred to in my post above was using the latter definition and comparing a pixel to a silver particle. In other words, a grain of silver is comparable to a pixel.

In Jerry's post earlier today he sates:

"Since the narrow dimension of the high end Canon is 3744 lines AND the narrow dimension of a BL back is either 6000 or 8000 lines BL is certainly are capturing a finer record of the scene in the larger format at equivalent coverage, enlargement or magnification .

"6000 & 8000 lines divided by 72mm divided by 2 = 42 & 55 lp/mm (with 12 and 9 micron sensors) at the imaging plane. Canon doesn't clearly post lp/mm used for their MTF charts but 3744 lines divided by 25.4mm divided by 2 = 78 lp/mm (6 micron sensors) but it's not clear if their lenses can actually deliver this level of detail either. Regardless, the larger image of the BL back has more total lines AND the larger sensor used in BL becomes diffraction limited much less quickly (at 81 and 144 µm² they have a diffraction limit around f16 and 11 for a far greater depth of field) than the smaller sensors in smaller cameras. Large sensors generally also have a greater "well depth" also which provides a superior dynamic range."

He appears to be saying that for digital systems larger pixels (81 and 144 um) coupled with large sensor platform is better than smaller pixels in a smaller platform. But what about film which has a much larger platform and a variable "pixel" (i.e., silver particle) size that varies from less than a micron to many microns depending on the energy (i.e., light) encountered?

Thanks for the Casanave link! Much appreciated.

[rdenney]

I think a nice post of a Martha Casanave pinhole image ( http://marthacasanave.com/coastal.htm ) should bring back a reminder of what's really important.... :)

Given the choice between a $30,000 digital capture setup, or her eye.... i know which i'd choose

Heh. You gotta love that infinite depth of field, and her amazing use of it.

Rick "sharpness is a state of mind" Denney

[tgtaylor]

Upon waking this morning I remembered that I took that image of San Francisco from Twin Peaks with a 300mm lens and not a 150mm as stated above. Specifically, a 300mm Nikkor-M.

Also, I took a peak in the data sheets for the resolving power and "diffuse RMS granularity value" for two of the film that I use. For Fuji Acros:

1.6:1 = 60lp/mm
1000:1 = 200 lp/mm
Diffuse Granularity = 7 um

For Kodak's TMax 100:

1.6:1 = 63 lp/mm
1000:1 = 200lp/mm
Diffuse Granularity = 8 um

A brief explanation from Wikipedia:

Film grain or granularity is the random optical texture of processed photographic film due to the presence of small grains of a metallic silver developed from silver halide that have received enough photons.

Contents [hide]
1 RMS granularity
2 Selwyn granularity
3 Grain effect with film and digital
4 See also
5 References


[edit] RMS granularity
Granularity, or RMS granularity, is a numerical quantification of film-grain noise, equal to the root-mean-square (rms) fluctuations in optical density,[1] measured with a microdensitometer with a 0.048 mm (48-micrometre) diameter circular aperture, on a film area that has been exposed and normally developed to a mean density of 1.0 (that is, it transmits 10% of light incident on it).[2]

Granularity is sometimes quoted as "diffuse RMS granularity times 1000",[3] so that a film with granularity 10 means an rms density fluctuation of 0.010 in the standard aperture area.

When the grains are small, the standard aperture area measures an average of many grains, so the granularity is small. When the grains are large, fewer are averaged in the standard area, so there is a larger random fluctuation, and a higher granularity number.

The standard 0.048 mm aperture size derives from a drill bit used by an employee of Kodak.[citation needed]

[Gordon Moat]

All true, but to claim that the 160 megapixels on the Seitz produces an equivelent resolution to a betterlight's would be to claim that NO useful information can be gained from interpolating the three seperate channels into one single file. Is that your contention?

I don't think there is much use in asking, though his claims that Bayer pattern interpolation vastly destroys data shows a lack of understanding. However, he works for Better Light, so I doubt he would ever concede that anything could be equal, or better. My opinion is that there is no point in getting him to honestly compare the Seitz to the Better Light, because all he does is type out the same tired responses over and over.
:cool:

[jkuska]

The fact that you prefer a print from a betterlight is not a technical specification or industry standard. That was the "third" point he was refering to.

What do you mean by "enlarged to equal degrees?" If you mean it as it sounds, I think you're being deliberately mis-leading.

"Enlarged to equal degrees" means an equal degree of enlargment. A 300 dpi print of a 6000 x 8000 betterlight file (48mp) would be a 20 x 27 inch print.

A 300 dpi print of a 7500 x 21250(160 mp) Seitz file would be a 25x70 inch print.

This is an equal degree of enlargment from each file, but the final print sizes are vastly different.

I'm sure the Non-interpolated Betterlight looks better inch for inch, but thats hardly a useful comparison. I'll make a a 1x1 inch print from the original Canon Digital Rebel, and it will look FANTASTIC inch for inch.

Print the Betterlight at a comperable size, not a comperable degree of enlargement. Then you can compare.

I'm not sure pointing people to a Seitz file of one subject and asking them to compare to a file of several stiched betterlight exposures of a different subject is any less mis-leading.

That is NOT a stitched Better Light file, that is a single, continuous panorama exposure, just as the Seitz shot is. It ONLY measures 3750 x 65000 (732mp) while the Seitz file likely measures 7500 x 21250 (160mp) - their standard 6 x 17 format. Both subjects have architectural details, grass, tree leaves and fine branches to examine. To my eyes the Seitz file falls apart at full zoom and the BL file holds together with clean crisp detail in comparison. Note that the pixel height of the BL file is HALF that of the Seitz file.

[jkuska]

I suspect it's the real thing. Most of the Super 8K backs are sold into the art reproduction market, where you can control the environment, depth of field, and choose a specific high resolution lens for copy work. I suspect that back is an 8K. My comments were mainly used for my work.. I usually don't choose a lens for it's resolution as much as it's 'character'. If anyone would like to try a comparison with their own lens/camera, and are in the Santa Cruz, Ca. area.. let me know.. I'd be happy to supply the back for it.

Jerry will need to confirm the back type used in that example.

You can clearly see the "Handle" of Richards system as he is inserting it into his Sinar, it's Blue in color which means it's a Super 8K-HS, the Super 6 is Maroon and the Super 10 is Green, non-super models are Black, be they 4's or 6's.

[Gordon Moat]

That is NOT a stitched Better Light file, that is a single, continuous panorama exposure, just as the Seitz shot is. It ONLY measures 3750 x 65000 (732mp) while the Seitz file likely measures 7500 x 21250 (160mp) - their standard 6 x 17 format. Both subjects have architectural details, grass, tree leaves and fine branches to examine. To my eyes the Seitz file falls apart at full zoom and the BL file holds together with clean crisp detail in comparison. Note that the pixel height of the BL file is HALF that of the Seitz file.

On your computer monitor? That's how you are comparing?
:rolleyes: