SillyComparison-Howtek-Eversmart-Epson

[sanking]

There has been in the past a lot of discussion re: the scan quality of various types of scanners. I thought the following comparison, which were all made by me with my equipment, might prove interesting to some of you. My motivation in carrying out this test was not to prove anything, but just to see for myself how the three scanners performed in practice with the same B&W negative, scanned by the same operator, with all factors equal. Note: I don't use color transparency film and have no interest in comparing color. I do use color negative film from time to time, but always with the intention of converting it to B&W, taking advantage of the RGB filters for conversion.

Equipment: Howtek D4000, Eversmart Pro, Epson V700

Negative: B&W Agfa Infrared, exposed with R72 filter. The crop area corresponds to an area of about .5" X .4" of the original negative.

Scan: Carried out at maximum optical resolution for the three scanners, then rezzed up or down to 4000 dpi. So, Howtek D4000 left as is, scan with Eversmart Pro at 3175 dpi rezzed up to 4000 dpi, scan wtih Epson V700 at 6400 dpi reduced to 4000 dpi. To makes this as apple to apple as possible I fluid mounted all three scans. I also used an Apple computer for all of the scans.

Sharpening: None done with any scanner at time of scan. Later, the same capture sharpening and creative sharpening routines were run with the scans from the three machines. This was done with the PhotoKit Sharpening software. Note that I did not try to match sharpness, just show what the scans look like at this point with the same sharpening routines applied.

Contrast: An attempt was made to normalize density and contrast. There is probably still some difference but I did the best I could.

Sandy King

[Peter De Smidt]

Interesting comparison. Thank you for doing it!

[Lachlan 717]

Which is which, Sandy?

[sanking]

Which is which, Sandy?

When you put your cursor over the file the name of the scanner should appear. It does with my comnputer. But if it does not, the crop scans from left to right are, 1) Epson V700, 2) Howtek D4000, and 3) Eversmart Pro.

Sandy

[engl]

Thanks for sharing the result of this comparison. Am I right in believing that the scans are left to right: Howtek-Eversmart-Epson? Or is matching the image to the scanner meant as a challenge :) ?

Two comments:
1. The contrast of the middle image seems a lot lower than the other two.
2. The images do not seem to be at 4000DPI. If the crop is a 0.5"x0.4" area I'd expect the image to be 2000 pixels wide.

Edit: I saw the answer to the "which is which" part had already been answered. I did not notice the text on mouse over, but I think it seemed unreasonable that the Epson had more detail than an Howtek :)

[sanking]

Thanks for sharing the result of this comparison. Am I right in believing that the scans are left to right: Howtek-Eversmart-Epson? Or is matching the image to the scanner meant as a challenge :) ?

Two comments:
1. The contrast of the middle image seems a lot lower than the other two.
2. The images do not seem to be at 4000DPI. If the crop is a 0.5"x0.4" area I'd expect the image to be 2000 pixels wide.

Edit: I saw the answer to the "which is which" part had already been answered. I did not notice the text on mouse over, but I think it seemed unreasonable that the Epson had more detail than an Howtek :)

Not sure how the image size appears for you, but the three image files I uploaded were about 2000 X 1600 pixel in size. I cropped them directly from the larger full size scan of the negative. Perhaps the image file was reduced in size for the site?

Sandy

[Lachlan 717]

When you put your cursor over the file the name of the scanner should appear. It does with my comnputer. But if it does not, the crop scans from left to right are, 1) Epson V700, 2) Howtek D4000, and 3) Eversmart Pro.

Sandy

Thanks, Sandy.

I was looking on an iPhone; no cursor!

[engl]

Not sure how the image size appears for you, but the three image files I uploaded were about 2000 X 1600 pixel in size. I cropped them directly from the larger full size scan of the negative. Perhaps the image file was reduced in size for the site?

Sandy

I believe they were, the three crops in your original post are about 50-60kB big each, and all 650 pixels wide.

[sanking]

I believe they were, the three crops in your original post are about 50-60kB big each, and all 650 pixels wide.

Well, I will be happy to upload the full size crops, (about 2000 X 1500 pixels in size in .jpeg format, around 3.5 mb) to another location if anyone wants to offer a temporary repository.

Seems ironic that on a LF format uploads are restricted to such a small size!!

Sandy

[Asher Kelman]

Well, I will be happy to upload the full size crops, (about 2000 X 1500 pixels in size in .jpeg format, around 3.5 mb) to another location if anyone wants to offer a temporary repository.

Seems ironic that on a LF format uploads are restricted to such a small size!!

Sandy

Sandy,

Impressive work! Using an S curve in Photoshop, how close can you get the v700 to the Eversmart scan?

Also, I'd love to know how you cam to buy all three? What's the logic?

Asher

[engl]

Well, I will be happy to upload the full size crops, (about 2000 X 1500 pixels in size in .jpeg format, around 3.5 mb) to another location if anyone wants to offer a temporary repository.

Seems ironic that on a LF format uploads are restricted to such a small size!!

Sandy

The problem is the amount of storage that would be needed for all those big files. Linking to big files, which get displayed at their full size, is possible.

I could put them in my Dropbox, or you could get your own. I believe the quota is 10GB per day transfer.

[engl]

Here are the full resolution images (feel free to edit and add them to the original message):

Full scene:
http://dl.dropbox.com/u/25632014/EversmartNoSharp.jpg

Epson V700:
http://dl.dropbox.com/u/25632014/EpsonV700.jpg

Eversmart:
http://dl.dropbox.com/u/25632014/EversmartPro.jpg

Howtek:
http://dl.dropbox.com/u/25632014/Howtek4000.jpg

[Bob McCarthy]

This shows the superiority of the better scanners, bark on trees is a fine place to see low res of Epson. The creo did well compared with the drum!

B

[sanking]

Sandy,

Impressive work! Using an S curve in Photoshop, how close can you get the v700 to the Eversmart scan?

Also, I'd love to know how you cam to buy all three? What's the logic?

Asher

Hi Asher,

First, thanks for uploading the full resolution crops.

No logic in acquisition. Scanning is a major part of my work flow so over the yeas I have compared, tested and purchased a number of scanners. The Epsons were purchased new, all of the high end ones were bought for less than a dime on the dollar re: original selling price. For example, paid $800 for the Eversmart Pro, $500 for the Howtek D4000, and also have a great Leafscan 45 that I picked up locally for around $300.

I have also had the opportunity to compare negative scans with a number of other scanners, including Howtek 6500, Howtek 4500, Colorgetter, Screen 1045, etc.

In terms of scan quality there is no question but that drum scanners are the best, followed closely by professional flatbeds like the Cezanne and Eversmart, and farther behind Epson consumer flatbeds. However, my personal work flow depends on testing and immediate feed back so all things considered I would rather scan my own work with a lowly Epson than have it done by someone else on a drum scanner.

That said, I work only in B&W, and print fairly small. Though I have made some very large prints, up to 40X50" in size, from a Leafscan 45, a CCD scanner made back in 1994.

Sandy

[Asher Kelman]

Where's the best place to find a used Eversmart Pro. Genesis adds a large premium price and says there's are refurbished. Can't say many of these floating around,

Asher

[sanking]

Where's the best place to find a used Eversmart Pro. Genesis adds a large premium price and says there's are refurbished. Can't say many of these floating around,

Asher

My advice for purchase of any high end scanner, drum or flatbed, is to search on ebay and Craigslist and find one that you can check out before purchase. These things come up all over the country and if you are prepared to make a road trip to check it out you should find a good one.

I strongly recommend against buying one on ebay that you can not check out locally before final purchase. If the seller will not provide an opportunity for you to check the scanner out locally I would pass on it.

BTW, top resolution of my Eversmart Pro is limited to 3175 dpi. Other Eversmart models (Supreme and Select) and some IQ Smart scanners, and Cezannes, have resolution up to 5080 dpi. Course, for large format scanning 3175 dpi is more than you will ever need since very few negatives of 4X5 or larger carry that much detail.

I bought the Howtek D4000 primarily for scanning medium format negatives. I had (have) a Leafscan 45 CCD scanner that gives more resolution (5080 dpi), but the work flow is slower with medium format since I have to scan in two passes and stich at that resolution.

However, for scanning large format film B&W film up to 11X14" in one pass the Eversmart is hard to beat. With any Eversmart you can scan up to the full optical resolution of the scanner (Eversmart -- 2400 dpi, Eversmart Pro (II) -- 3175 dpi, Eversmart Supreme, Select -- 5080 dpi), over the entire platen of 12X17". And I find fluid mounting with the Eversmart much easier than with a drum scanner.

The Fuji Cezannes are also very good professional flatbed scanners (and usually somewhat less expensive than Eversmarts) but I am less familiar with their qualities than with the Eversmarts so perhaps someone else might want to comment.

Sandy

[Ed Richards]

Sandy,

What is fairly small? I am thinking that the Epson is going to look about the same up to about 4x magnification print size, say 16x20. What do you think?

[sanking]

Sandy,

What is fairly small? I am thinking that the Epson is going to look about the same up to about 4x magnification print size, say 16x20. What do you think?

Ed,

I agree with you for 3X-4X, assuming B&W negatives, and that the negatives were not over-exposed or over-developed.

Sandy

[SCHWARZZEIT]

Sandy,

from my point of view all of your three scan samples have some problem or another.
I find the Epson the most pleasing as it has the most consistent rendering of film grain and therefore the tonality looks better to my eyes compared to the other two. It has an overall softness but the tonal transitions look somehow right. The shadow detail should be better.
On the Eversmart the problem of shadow clipping is even more severe. The largest tree trunk and the wooden disc look sharply cut out from the dark backdrop. As the drum scan shows there's obviously some shadow detail on the film. Extracting low frequency shadow detail from negative film shouldn't be a problem for any transparency scanner.
The Howtek offers the sharpest rendering of the details in the tree bark but the darker grey areas look awfully soft with a very muddy tonality unlike what I would expect from a drum scanner. While the grain is somehow resolved in the lighter wood it's absent in the darker areas which looks wrong. As far as I remember I have no experience with this Agfa infrared film myself. So the problem could be in the film itself. But then it looks too digital.
Did you open the aperture for this scan?
Did your sharpening routine cause any selective grain softening?

Scanner resolution results tested with a high contrast scanner target cannot necessarily be transferred to scanning photo film because the contrast of the high spatial frequencies is much lower on film and thus much harder to detect for the scanner. Even though I'm working with a high end ICG drum scanner that is capable of resolving up to 175 lp/mm on a high contrast scanner target the useful real world film resolution I can extract is somewhere between 80 and 120 lp/mm depending on the film. It's almost always less than what can be seen on film under a microscope. Scanning is a lossy process even with high end machines. But the losses are negligible with larger film formats and in most photographic real world images anyway as these high frequencies get compromised so easily by sloppy technique, film grain and other factors.
On the other hand a high optical scanning resolution leads to smoother transitions of detail. If we are talking about lines these are often not just hard black and white lines but there's a transition with some tonality in between. This is where optical super sampling fills the gabs with grain texture to retain the analog quality of an image even if the real level of detail is much lower. Of course it's only relevant for high magnification output or in case you're scanning for archive where you try to get most of the information stored on film into the image file.

-Dominique

[Brian K]

There has been in the past a lot of discussion re: the scan quality of various types of scanners. I thought the following comparison, which were all made by me with my equipment, might prove interesting to some of you. My motivation in carrying out this test was not to prove anything, but just to see for myself how the three scanners performed in practice with the same B&W negative, scanned by the same operator, with all factors equal. Note: I don't use color transparency film and have no interest in comparing color. I do use color negative film from time to time, but always with the intention of converting it to B&W, taking advantage of the RGB filters for conversion.

Equipment: Howtek D4000, Eversmart Pro, Epson V700

Negative: B&W Agfa Infrared, exposed with R72 filter. The crop area corresponds to an area of about .5" X .4" of the original negative.

Scan: Carried out at maximum optical resolution for the three scanners, then rezzed up or down to 4000 dpi. So, Howtek D4000 left as is, scan with Eversmart Pro at 3175 dpi rezzed up to 4000 dpi, scan wtih Epson V700 at 6400 dpi reduced to 4000 dpi. To makes this as apple to apple as possible I fluid mounted all three scans. I also used an Apple computer for all of the scans.

Sharpening: None done with any scanner at time of scan. Later, the same capture sharpening and creative sharpening routines were run with the scans from the three machines. This was done with the PhotoKit Sharpening software. Note that I did not try to match sharpness, just show what the scans look like at this point with the same sharpening routines applied.

Contrast: An attempt was made to normalize density and contrast. There is probably still some difference but I did the best I could.

Sandy King

The Creo did well despite being rezzed up and having it's contrast a bit higher. I think if it were done with a Supreme or a IQSmart3 at optical resolution it would have been a real toss up between the creos and the Howtek. But where a drum scanner does best compared to a pro flatbed is with a very poor quality negative, something too dense, too thin, too flat. There the PMT versus CCD aspect will shine. But with a good quality negative it's awful close.

BTW I like the original image, looks like a good one.

[sanking]

The Howtek offers the sharpest rendering of the details in the tree bark but the darker grey areas look awfully soft with a very muddy tonality unlike what I would expect from a drum scanner. While the grain is somehow resolved in the lighter wood it's absent in the darker areas which looks wrong. As far as I remember I have no experience with this Agfa infrared film myself. So the problem could be in the film itself. But then it looks too digital.
Did you open the aperture for this scan?
Did your sharpening routine cause any selective grain softening?

-Dominique

The main issue I have with the Howtek scan of this negative is that the 6 micron aperture really accentuates the grain. I believe in retrospect that the scan would look a lot smoother opening up the aperture to 13 microns.

Sandy

[onnect17]

...
Seems ironic that on a LF format uploads are restricted to such a small size!!
Sandy

About time a big Kahuna mention the irony. For the rest of us the post will be deleted.

[SergeyT]

...
Scan: Carried out at maximum optical resolution for the three scanners, then rezzed up or down to 4000 dpi. So, Howtek D4000 left as is, scan with Eversmart Pro at 3175 dpi rezzed up to 4000 dpi...

Could have scanned on Eversmart Pro at 4000 and let the scanner's SW do the "half"-interpollation for you...

[sanking]

Could have scanned on Eversmart Pro at 4000 and let the scanner's SW do the "half"-interpollation for you...


That is true. In fact, I could have scanned at 6400 or 8200 dpi and gotten a much better scan because of the way the software interpolates the greater optical resolution, up to 8200 dpi, in the direction of the stepper motor. If you test a high resolution target there will really be a difference in resolution in dpi between a scan ad 3175 dpi and another at 8200 dpi, but of course you will only see the difference in the direction of movement of the stepper motor.

But there are aslo things that I might have tried with the D4000 and V700 to get a better scan.

Ultimately, the best way to test any two or three scanners is to use the scan files, work them as needed in PS, and then and make an actual print at the maximum size you plan to print. Now compare the prints.

Sandy

[Ken Lee]

About time a big Kahuna mention the irony. For the rest of us the post will be deleted.

The forum runs on a donated server. Image files take up a lot of storage compared to text.

If the forum charged a membership fee, then perhaps it would make sense to let people upload large files. ;)

[mdm]

There is no need to since you can include an image linked from elsewhere. Running data centres is a job for the Googles and Yahoos of the world because of the economy of scale they generate and security they can provide.

[dim.ka_1]

Sandy thanks for the comparison here, from my side - here is a small test of Nikon LS900 Scitex ProII (raw DTfile developed in CS4) and Drum scanner ICG365

http://album.foto.ru:8080/photos/or/41852/1615928.jpg


and i got Howtek D4500 a week ago, so here is quick test for deep shadows from Howtek and Scitex ProII - a bit different colors, but this is not a case in this comparison

Howtek D4500 silverfast 6Ai
http://album.foto.ru:8080/photos/or/41852/1829192.jpg

Scitex ProII (DT file developed in CS4)
http://album.foto.ru:8080/photos/or/41852/1829191.jpg

[onnect17]

I do not mind to budget up to $100 or 10GB of space a year for it if the moderators were a little more careful with the "delete" key.


The forum runs on a donated server. Image files take up a lot of storage compared to text.

If the forum charged a membership fee, then perhaps it would make sense to let people upload large files. ;)

[SergeyT]

... here is a quick test for deep shadows from Howtek and Scitex ProII - a bit different colors...

Howtek D4500 silverfast 6Ai
http://album.foto.ru:8080/photos/or/41852/1829192.jpg

Scitex ProII (DT file developed in CS4)
http://album.foto.ru:8080/photos/or/41852/1829191.jpg

It mostly comes down to the ICC profiles accuracy and the workflow.
ProII is quite capable of seeing through a DMax of 4. But an average quality profile can kill the shadow details captured by the scanner.
I assume a DTView.icc was used to "develop" a DT in CS4...It is not color accurate, in my opinion, and I wonder if it was ever intended to be used for anything but simply viewing the content of a DT file...

[dim.ka_1]

It mostly comes down to the ICC profiles accuracy and the workflow.
ProII is quite capable of seeing through a DMax of 4. But an average quality profile can kill the shadow details captured by the scanner.
I assume a DTView.icc was used to "develop" a DT in CS4...It is not color accurate, in my opinion, and I wonder if it was ever intended to be used for anything but simply viewing the content of a DT file...

Concerning color with DTfiles - personally i am not fully satisfied with all these convertions with color profiles. May be someone who is more skilfull with DTfiles could point or correct my workflow with raw DT files?
for now it looks like this - i open raw DTfile in CS4 - assign DTview color profile (clearly could see color shifts) - converting to Lab - start initial editing with curves - get the final look.

[SergeyT]

DT format is the best for archival scanning, since it records what CCD captures with no processing applied. Once scanned into a DT at max optical, the scan can be re-purposed into any RGB or CMYK output format without the original ever being scanned again.
For now I have created some custom ICC profiles based upon few IT8 targets.
I open my DT scans in oXYgen Scan (File->Open)and re-purpose them into 8-bit RGB TIFFs applying the custom icc profile as input profile, ProPhotoRGB as output one and setting up the endpoints.
Once the RGB TIFF is generated I open it in photoshop, check and make sure the endpoints were set correctly, clean it up from dust and resize/convert for printing.

I found a reference to a "oXYgen Smart Tips: Opening and Retouching DT Files" on web but could not locate the file itself. If someone can share the file it would be appreciated.

[Peter De Smidt]

I'm not sure that using an 8-bits per channel and ProPhoto is a good idea. It's such a large space that the steps between colors will be quite large. Bruce Frazier, for example, only recommended using such a large color space with 16 bit per channel images.

[SergeyT]

That all is mostly lyrics…Result is what matters.
Large is not small and will guarantee that all the color info available during the scanning process is preserved.
All the conversion from the input to output profile as well as tone correction is done by the oXYgen in 16 bit. Given a quality input profile was used, the final 8-bit RGB usually does not require any further color or tone corrections, so I do not see how usage of ProPhoto can affect the workflow in a negative way.
Usually I either like what I see on the light table or not. If I like what I see I try to reproduce it on paper as close as I can...and a color\tone accurate scan is the key.
If I do not like the image on the light table (for whatever reason) I just do not bother scanning it...I'm not into heavy manipulations , not with color images...

[Asher Kelman]

I'm not sure that using an 8-bits per channel and ProPhoto is a good idea. It's such a large space that the steps between colors will be quite large. Bruce Frazier, for example, only recommended using such a large color space with 16 bit per channel images.

Peter,

For the folk that don't do anything much with the file then 8 BIT might stay looking good. However, when printing to a particular paper, the curve correction might then cause gaps and a "tooth-comb" effect in the histogram as one jumps between values. One might do better converting to 16 BIT before doing any alterations.


That all is mostly lyrics…Result is what matters.
Large is not small and will guarantee that all the color info available during the scanning process is preserved.

Not so! The 8 BIT has already lost 90-97% of the color information from that which was captured in 42 or 48 BIT.



I'm not into heavy manipulations , not with color images...

So I can see how you and other photographers might find no problems as they don't stretch the profile in any way! Still, it doesn't make sense to work in such a large color gamut in 8 BIT when one is going to remap the colors anyway to an entirely different gamut in the printing process. Today's printers have a huge gamut compared to what we see on the monitor, (even an Adobe RGB monitor). During the remapping, especially in the saturated hues beyond those shown on the monitor, you really need the best files possible to get the best results. If you have no dense hues and you don't fiddle with the curves, then you're likely home free! Perhaps having 8 BIT files will never be an issue!

Asher

[rdenney]

I have found that with VueScan and the LS8000, using the ProPhoto color space will prevent the scan software from clipping the image. With that output space, I can capture the entire histogram of a negative or (especially) a transparency, with a bit of unused tail on each end. I will still have to reduce the space somewhat for any given output device, but at least I get to choose where I do the compressing.

With the sRGB color space (for intended web display), VueScan would clip one or the other end of the histogram, even if I set the white point and black point percentages to zero.

This was true irrespective of whether I used 8-bit or 16-bit output. I prefer 16-bit output so that I can make larger moves with the tone curves without posterization. I always write output files in TIFF format.

But if I planned to compress the color space anyway for my output device and did not plan any big corrections, I don't see an incompatibility with a wide-gamut space for scanning and 8-bit color depth.

Rick "acknowledging that this might be different with other scan software" Denney