300 ppi does not give you best prints

[CaptJR]

This thread was actually brought about because of another thread called 'DPI vs JPEG compression'. http://www.s2pro.com/forums/showthread.php?s=&threadid=2356 That thread prompted me to do a test and this thread is the results.

I don't plan on speeding a week doing rebuttals so I'm going to explain the test I did as thoroughly as I can.

I used a picture I took at my daughters birthday party, held at a hotel, so they could use the indoor pool.

I applied what I thought was the correct amount of USM then saved the file in jpeg compression level 11.

In deciding how to do the test I had to decide what resolutions to use. Since I was going to use 4x6 prints, I thought the highest resolution should be 510 ppi. This gave me a 4x6 as close to the original resolution as I could come. I also used 250 ppi and 300 ppi because it was mentioned in the other thread and because it is said that these resolutions, in particular 300 ppi are optimal.

tiff pictures are entirely to large for uploading for printing or for here so I'll be using jpeg compression. That was part of the original question anyway. The original question that prompted this was, To reduce upload time for printing, I wondered if it would be better to reduce the file by reducing resolution or increasing jpeg compression.

I usually use jpeg compression level 11 so I used that. Then I wanted also to use a greater compression level that reduce file size significantly enough without degrading the picture to much. I picked level 8.

So I made four pictures. For each picture I started with the original picture. I put a distinguishing mark on each picture so I'd know which picture was which when I got them back from the printing.

Picture 1. resized to 510 ppi and 4x6 inches and saved at jpeg 8
Picture 2. resized to 300 ppi and 4x6 inches and saved at jpeg 11
Picture 3. resized to 300 ppi and 4x6 inches and saved at jpeg 8
Picture 4. resized to 250 ppi and 4x6 inches and saved at jpeg 8

I uploaded the pictures to have them printed on good non-inkjet printer. I won't be mentioning where or what printer, although I think I could, I don't feel like defending this in court.

When I got the pictures back, I scanned them all at 600 ppi.

How I made the picture posted here -

1. Brought up one of the scans in PS and used the zoom tool to zoom into where I could see any differences in the scans.
2. Hit the print screen button to capture the zoomed image.
3. Pasted the capture into a 1024x768 blank.
4. Highlighted a small portion with some distinct lines and marks.
5. Pasted to a new image.
6. Increased the canvas size to make room for text indicator and added identifying text.
7. Increased the canvas size 200 percent right and 200 percent down to make room for the other 3 captures.
8. Repeated 1-5 for the other 3 scans, pasting them into the expanded canvas.
9. Expanded the canvas enough to paste a size down of the original so you could see how much I zoomed.

My conclusion which I'm not going to go on and on about. If I can see it, you can see it.

Looking at the glasses rim and the small birth mark even thought the 510 ppi-jpeg 8 was a smaller file than the 300 ppi-jpeg 11, the 510 ppi-jpeg 8 was a clearer print. WOW
The 300 ppi- jpeg 11 and the 300 ppi-jpeg 8 were very similar, but I'd give the edge to jpeg 11.
The 250 ppi-jpeg 8 show just how bad it can get.

Was this noticeable to the naked eye. Comparing the 510 ppi-jpeg 8 to the 250 ppi-jpeg 8 it definitely was. To see the difference between the 510 ppi-jpeg 8 and the 300 ppi-jpeg 11, I had to put on a pair of 'old folks reading glasses'.

To the naked eye you have to really look to tell the differences but I figure some of my customers do. So I'll continue to upload higher resolution pictures even though it takes a little more time.

CaptJR

[cthornhill]

I can only give my opinion here, so take it for what it is worth (IE not much - every person has to decide for themselves)...

As for compression in JPG. I feel that JPG compression has a more dramatic effect on the quality of the image than resolution does, after a certain minimum resolution is reached. At the same size in pixels less compression is almost allways more pleasing, since the artifacts of heavy JPG compression (cosine transform artifacts) are pretty ugly to many people. In color images the transform artifacts tend to create problems in the skin tone transition areas and lead to blotchy looking areas. This is compounded by the nature of the digital capture device and the digital printing systems in many cases. For my own work, I use a setting of either 11 or don't use JPG for print work. I just think the solution to uploads is bandwidth not compression (well up to a point, of course other file formats offer better compression quality and size than JPG, but are not very standard yet).

Quality - I would like to clarify my statements in another post. I said I do not seek the 'best possible output' but try to get prints that 'are comparible to lab prints'. Here is what I mean by that:

1. Best is not allways the best choice - I have access to a few printers. My main in-house printer is a 2200. It is capible of 2400 dpi (well Epson says so anyway...). I don't ususally run it at more than 1440 since the results at 2400 dpi are not allways visable or relavent to the print in question. On some stock for some purposes I run it at 720 dpi. Best is best for the purpose, and that can change. On some stock you will see no difference in 2400 dpi or 1200 dpi output. On some stock the lower dpi actually prevents some issues with blockup in shadow detail.

2. Cost is allways a consideration. I am a business, and I need to provide my clients high quality output at a cost that I can bear. All commercial labs have to maintain a certain number of feet per minute from their printers. Most do this by balancing the resolution and the print stock with the RIP settings to achive a high quality print at the greatest throughput. The print quality is judged against a reference test standard. That is the method I use too. I have tested higher res modes, and found ways to get a bit more out of the printer, but the trade off in speed was not worth the minimal benift in fine detail, and it made the system subject to much more variation in shadow detail issues print to print. Like most labs, I choose to balance my output for cost/quality at a point where the output prints were visually matching results from reference sources.

Visual inspection of print quality is not done with a microscope most of the time. While I have been know to use a microscope to inspect film and prints, I use controlled lighting and normal viewing distances for most print inspection, as per the standard methods first documented by Kodak. Inspection of prints by instrument is useful and tells us lots, but for normal value judgements of print quality, the standard is human viewing from normal distances. There are lots of texts on this in the graphic arts trade, which is how I learned to judge prints.

My reference prints use a grey and color scale, so I can use a refelective densitomiter as well, but I do not use one as a matter of course for my prints at this point. I 'print to taste' and have spent some time and money calibrating my monitor so I can soft-proof for my printer. When I can afford it I will add more measurement tools for output, but they will be color and density tools, not microscopes. I find resolution is not the major issue for me any more with current output systems. Tonality and color are more relevant to my work day to day.

In general, I believe you will find that most textbooks recomend that Excellent prints begin at about 11 line pairs per millimeter. This requires a little more than 550 lines per inch (you need two lines for a line pair). This is well matched to pratical obsevation, since most people can not distinguish detail below 1/500 of an inch even at high contrast (from a normall viewing postion in standard lighting for an acuity test). Good and Acceptable prints (each is less than Excellent in descending order on the chart) can run as low as 4 or 5 line pairs per millimeter (as low as 200 lines or 1/200 of an inch). These standards were for back and white prints, and color prints will often have much less data than Excellent black and white prints. That is considered acceptable due to the lack of color accuity in human vision.

For most of my in-house work I try to produce Good prints. I do not feel the systems I have can achieve Excellent print results, though they can get close. It is possible using specialized RIP software and ink systems to do so, but I have to get that sort of work out of house.

That is why I say I don't seek the 'best' my printer can do. I seek a commercial product of good quality at a reasonable cost to me. I can get that from a desktop system, and I do so. Best quality is something I would love to have, but frankly the cost is prohibative to bring this gear in house.

Print quality overall from my in house system has not yet ever been an issue for any customer. I don't know of any pro in my area that has had any issues with the current generation of digital color printer. I do know people who prefer wet black and white work. Out of house prints have had wider variation in my experiance than those i do for myself. When I need larger prints I do get the benefit of a better RIP. It offers me a bit more resolution in fine detail, and better tone scale. That (a better RIP) is the first thing I would add when I can - not for the resolution, but for the dynamic range and color control quality.

As for why many people decide to set up their systems for certain resolution levels, I would recomend anyone review the detailed information at sites like Norman Koren's:
http://www.normankoren.com/Tutorials/MTF3.html

It offers a pretty detailed example of the how and why many labs and photographers use the settings they do, and why 300 ppi is so common. You may or may not agree with all the methods and conclusions, but it is a good source of information.

I encourage anyone to do test for themselves, and to use the settings they find are best to them. I would also suggest that everyone remember that the final judgement of print quality is their own judgement - if you feel something is better for your prints, go for it. They are, after all, your prints. If you like the look of 500 ppi prints sent through any print process better than 300 ppi prints, then I suggest using 500 ppi.

Personally, I see little difference if any in small prints (8 x 10 or less) under normal viewing conditions in output samples I have seen from input files much above 300 ppi. For me, above 8 x 10 the extra pixels are good to provide more data for larger prints. What I do see is that if I start with an original of higher resolution (from a DCS back or a multi shot S2 image), and I downsample, I have much nicer results even at the lower ppi. To my eye the resulting prints look more like what I got from film on a medium or large format camera. I find this is because of the way the fine detail is averaged in the downsample. This may be just me...

I am sure that if I had access to a printer with greater resolution and more tonal range it would be nice, and I would like the pictures beter. Of the two, I would like better tonal range first, then I'd like to see resolution climb a little. Personally, I think the current generation of ink-jets is the first acceptable for photo pringing, but I don't expect it to be the final word. I actually prefer light-jet output myself, but can't see the cost as worth it on most prints.

Cecil

[Eddie the Gnat]

Please forgive me if I've completely missed the point here (:) ), but surely if these files are all 6" x 4" then the 510 ppi version contains more than four times as many pixels as the 250 ppi one. This is of course going to lead to a clearer print.

Similarly, even though you saved a 300 ppi version with less compression it was still beaten by the more heavily compressed 510 ppi file because that image simply has a greater number of pixels.

This will hold true (I believe) no matter what printer is used.

Apologies in advance in case I have the wrong end of the stick!

Regards,

Eddie

[cthornhill]

The number of pixels in the file has nothing to do with the resolution of the output device. If you make a one inch print on a device of 300 dots per inch (assume full color dots like a Frontier or Lamda laser device for now) the output resolution will show no more than 300 distinct points no matter what the input resolution was. You either loose data, or you print a bigger piece of paper. You can never get more dots per inch into the output device than it supports no matter how much data you feed it.

It is true that if you downsample a 600 pixel per inch file with fine detail to a 300 dot per inch device (again assume continuouse color/greyscale for now on the pixels and dots) you might see an improvment in the way the fine detail is reproduced at the lower resolution. You still won't have any more data, but you might like the way the lower resolution device reproduces the high resolution file. A good example is you screen. Many people like to work on screen graphics at a higher resolution and then downsample for the final product. It just tends to appear crisper and cleaner to most people. You have to be careful thoough, since many fine detail elements in a design can get lost this way.

As for a 4 x 6 or any other print, the system (and person) doing the printing has only got two options. They can discard the extra pixels you give them or they can do a larger print. If they discard the pixels by throwing the extras away without consideration, you don't get any benefit from them. If they average the extra pixels into the lower res file, you might get some benefit.

Neither method will do anything to the output resolution since it is fixed by hardware. The only way to see if sending more data than can be printed helps you is to try it. This is very dependant on the way the data is handled by the operator and software of the printer system. If the operator has the option of processing the file before it reaches the printer (say in Photoshop) they can downsample. If the workflow does not allow pre-processing you get what the machine gives you - most of the time this results in a cropped photo - or a rejected file and no print.

I can tell you that I like the look of downsampled images. I do this all the time. I shoot as many pixels as I can as sharp (in focus good exposure) as I can, then reduce the final file to the needed size for use in a print piece (halftone), photo print, or on line. For my part, I try to send the printer only the minimum datastream needed to get the job done. I do any processing before I send the file to the rendering device. This keeps the workflow to the printer as efficient as possible, and is flat out required by most labs I work with (they just won't take any files not at the required PPI/DPI/colorspace setting).

There are a lot of tricks you can try to make your prints look better. None of them alter the physical limits of the output device. They all try to take advantage of ways of making the limited data you will get back look as good as it can. All of the Fuji Frontier output units I know of are built to output (and can only output) 300 dots per inch (full color and tone dots). Durst now offers a machine that outputs a bit over 400 dots per inch.

Ink jet printers don't offer full color dots. They make pattern of color based on the RIP's screening software that simulates the original pixel. There is considerable argument over what the equivelent resolution the popular ink jets can produce. Most are able to generate vector prints that rival 300 - 500 dpi systems on the right paper. Color output is similar, and in some cases with special software B&W output can be very high too. It is generally accepted that the current round of 'photo quality printers' offers output of similar quality or better than most minilab systems - i.e. they compete with 300 dpi continuouse tone systems like the Fuji Frontier, and many printers offer even larger output (like the Epson 7600 9000 series and 10000 series, or HP and Roland's similar units).

Dye sublimation is the other popular technology. It clearly offers 300 dpi output and provides a form of continuose tone dot. It is not able to provide as wide a range of tone as laser systems, nor can it print on the wide range of paper ink jet can.

As far as tonal range, the laser systems offer the best results of any system now in the market. There are proponets of B&W ink jet systems that claim to rival laser systems for tone, but they are special software and ink combinations, not common for consumers. As for dye sub vs ink jet, everyone needs to make up their own mind. For high speed output in commercial situations most pros I know choose dye sub. For large format, or variable paper, most I know choose ink jet. Each format has its own best use and cost is a big issue.

[cthornhill]

Just a thought...

If you are going to compare input files using the same output system, and the output resolution is fixed at 300 dpi (like, oh maybe a Frontier...), then it might be good to test the following input data sizes:

200 ppi (well below the output max)
250 ppi (a bit below output max)
300 ppi - in theory as good as it gets on output (what we are testing)
350 ppi (a bit better than what we can hope for in the output - we want to see if it helps)
400 ppi - (should not matter, but hey, that is what we are testing...)

If I understand the original test here the source was at 250ppi and 510 ppi. I am pretty sure the device (if it is a minilab printer) can show a difference between 250 ppi and 300ppi, but I just can't say what it can do above 300ppi. To be a really accurate test, compare a 300 ppi source to a 400ppi or 510 ppi souorce and see if there is a difference. I don't think there is any question that a source well below 300 ppi is going to show less data than a source at 300 ppi... I think the interesting question is how much better (if any) a source above 300 ppi looks than a source at 300 ppi.

I am not saying this to upset anyone! I am just noting that a source that is 5/6 ths the resolution of the output might indeed look different from any source that is 6/6ths or above the output resolution. What I can't say is how much better a 6/6 ths file looks than one that is oversampled.

I suspect that oversampling and data reducing help the quality of output if that service is offered. How much is subjective, and hard to judge without the print in front of you.

Cecil

[CaptJR]

cecil

I'm glad you said up there somewhere that I can agree or not agree. LOL For I have to disagree completely on most points. Mostly I'll just let the initial test speek for itself. It proves that the pro labs that say 'you can't do better by using a higher resolution than 300 ppi' are wrong, completely wrong. The extra data is not discarded it is evidently used by the software to make a better 300 dpi pattern for the printer to print.

I used the zoom to make this obviouse in this post, but I can see the difference with the print right in front of me. The birth mark is the give away, but you can also see it in the glasses frame.

JR

[CaptJR]

Theo I have to redu this reply after reading yours more thoroughly. Doing it now but wanted to substatute this right away
JR

[cthornhill]

OK - No big deal - Everyone has to have opinions and that is OK. For me, I honestly don't see much difference in the originals. I did download and subtract them and found tone differences, but no big detail differences aside from compression artifacts (between the 500ppi and the 300ppi). That is just how I see those two images - both are about equal to me in actual resolution, one is just printed a bit better with regard to contrast.

I understand completely if you don't see it that way, but that is how I would call it.

As for data submission to digital printers, the manufacturer recomends 300 ppi on the Fuji Frontier. As I said, if your lab is willing to take higher res files and downsample them for you great - I don't have that option, nor do I see much point in it, but if it works for you go for it.

Good discussion, the bottom line for me is that you have a way of getting results you like better, and if others can use it that is a good thing.

Cecil

[crabby]

I'm going to confuse the issue a little more in saying that everyone is correct, to varying degrees. CaptJR-No one is disputing your eyes. If you are viewing two or more prints side by side and say x print looks better then y print who's to say that you are wrong. That's where it ends in the final analysis. Your images, your eye. I must agree with Cecil that it is hard to judge without the print in front of you.
Cecil-I think you have hit the nail on the head and explained it more clearly then my attempts. The one thing that I missed was the RIP. I'm not sure when you were talking about downsampling a 400 or 510 ppi file if you ment in photoshop or letting the RIP interpolate down to its fixed printer size. I understand that the frontier has a pretty amazing RIP. The problem is for this discussion that there are a lot of different Fuji Frontier printers. The older ones and/or the cheaper ones probably don't have such a sophisticated RIP. I think some will interpolate down an oversized file and some don't. The lab I used for cheap C-prints has a Kodak Pegasus which is the same type of printer as the frontier. I tried sending them 300 ppi files and got back cropped prints. They needed 240 ppi or the would charge extra to have a tech resize the file, but if you didn't ask they would just run it through and give a cropped print.
Lumens-is correct too. For his printer as well as alot of IJ printers. You are better downsizing you file to the printers optimum resolution. Upsizing is a different story sometimes the ptinter software can make a better print from a 100ppi file then to interpolate up to say 300 ppi in photoshop. But usually to my eyes they look identical so why bother. I also must agree with him here on jpg compression. I think I would see more degradation from higher jpg compression then I would from downsizing a file from 500 ppi to 300ppi.
So a lot of the magic is in the RIP. Photoshop is an excellent RIP by the way. And the judgement is in the eye of the beholder. Labs are just trying to find a happy medium that pleases the most customers. Some people prefer higher contrast, some like more color saturation. Most like a nice sharp print.
IMNSHO

[cthornhill]

I agree - Capt is the best judge of his prints! All the theory is fine, but he should do what looks best to him. Getting god results is the bottom line.

I am trying to get more detail on how the new Frontier downsamples. I was able to get in touch with a person who owns one, and they do agree that they like the way it downsamples to small prints.

The example they gave was sending in an 8 x 10 and also getting some 4 x 5's and walletts. They feel the smaller prints look extra good due to the way the data is processed in the software of the Frontier. This is very similar to what the Capt did sending a 500ppi file, so I have to believe the downsampling to the print head (still 300 dpi) is giving the effect.

I used to produce RIP software at my firm (yeah ancient history now), so I am a little interested just for historical reasons...I would like to know more about the design decisions made in this RIP, but getting that data will most likely take a while. In the meantime, it is just good to know that the new software on the Frontier is going to make lots of labs more flexible with file submissions. I think this system is so popular and so good that the labs with more painfull workflows (the ones that reject un-sized prints) are going to have to change. That is a good thing.

My correspondent also reminded me that (as the Fuji sales sheets stress) the Frontier is also great at interpolating up from a small file. They (Fuji) make a big deal out of this, and are very proud of how well the new Frontier does when making big prints from small files. This is also a good thing.

Like I said, good thread, worthwhile discussion. This stuff changes all the time so it is good to keep looking at it.

Cecil

[cthornhill]

Theo,

I understand. I have had good and bad experiances with the Frontier, and as far as I can tell it is very dependant on the operations staff.

I used to work down the street from a facility that did great scans and prints with a Frontier, and now I am very close to another well regarded lab that I feel does not do as good a job with the Frontier. Same lab system, different operational standards/staff.

I have gotten good prints from my local place - and that is just what they are, good prints, but not great prints. I have also gotten prints where the color is just not what is should be. After a few such experiances, I choose to print my own work or use a different lab.

I still feel that the people did OK work, and I have received excellent Kodak prints from them, but I just was not happy with the Frontier work. I can run my own prints up to super B size and then I have only myself to blame. The cost is about the same (I save money on prints over 8 x 10) for most prints, but I like the control of printing for myself.

In general I would have preffered to go dye sub myself, but I need the range of papers ink-jet offers for many projects. I did a lot of work with Kodak when the current generation of dye subs were being developed and I am pretty fond of the results from those print engines. I expect the printer you use does great stuff too - I have only see a few samples, but owners really like them.

What I really miss is the old Cibachrome glossy prints. You know, produced by the big bad sulfuric acid printers that took a room and would kill you if you made any ventilation errors. I have Cibachromes I still love from almost 20 years ago. Ah well, I miss the smell of Dektol too, but that is just a way of showing how old I am.

Nowadays the best we can hope for is a Lightjet. Nice, but pricey...

In general I find variations from lab to lab and from run to run at a lab are often very great. I keep hearing this complaint from other photographers about most labs. I know a few I never hear of problems with, but they are all West Coast firms and a bit hard to work with interactivly from the East Coast.

Good prints are as and where you find them. There are seveal places here that do good work, but you still have to evaluate every job that comes back to be sure there are not issues. That is just the nature of the beast as far as I can see. I find it a lot like press work. Some runs just go easier than others....

I just wanted to let anyone know that apparently there are some interesting software tricks in the new Frontier front end, and they may indeed offer oportunities to adjust the images...

Yours,

Cecil

[laue]

As about printing..I am not sure if this can answer some of our questions..
The Qimage site has some explaination about interpolation involve when you print...!

http://www.ddisoftware.com/qimage/quality/

I think by sending out a higher dpi file to a lab for printing can be better if the interpolation software they use in the printer driver is better then what you use at home, or we are dealing with the same issues..! The native resolution of the printer is an important factor for getting better result.

Check out the site and tell me what you think!

Eddy