Help with TIFF Exporting Please

[TestTools]

Hello; I have read as much as I can, and experimented quite a lot, but I am still getting what I think are odd artifacts while exporting ...though I don't know what level of perfection is possible. (I need perfect.)

 

My files, one included, is 4096 x 2160, which is also the size that the projector wants. I also need to export to TIFF.

 

Understanding enough to barely get my feet wet, I thought that I would hit Export and a pixel of the same color and size would show up at the right point on the matrix. Now I understand that is illogical for diagonal lines and curves in circles in fonts. I've also learned that for a graphic like this that Nearest Neighbor gives the best results...but not uniformly.

(See, for an interesting explanation: <http://gis.stackexchange.com/questions/10931/what-is-lanczos-resampling-useful-for-in-a-spatial-context/14361#14361>)

 

As you can see in the drawing, there are 20 single pixel white lines separated by 4 pixels width of black on the sides and top and bottom. When exported the top and bottom turn out nice (maybe a stray extra pixel at the ends in certain experiments), but the sides turn distinctly grey. There is also a line that goes around the 10% boundary that is white top and bottom but grey on the sides – you can see it where they join that it isn't an illusion. 

 

There are little bugs of fine lines all around – on the layers they're called MTF. They are single lines separated by single black at different angles. They were all derived from the same drawing, then flipped and rotated. After exporting some of them are completely smeared, some are ok on the horizontal, not many ok on the vertical. If there is a logic to why the ones on the top right side are better than the ones in the center, I'd like to learn it.

 

Cranking up the dpi to 6400 seems to improve some of the finer fonts, though some sites seem to say that it shouldn't. 

 

I just learned the "Turn off Embed Metadata" trick and that maybe is improving some smearing items, so I'll make a complete workflow test tomorrow. But the white lines still turn grey so that isn't optimum. 

 

I've tried variations with every resample option.

 

1) How close to perfect should I expect?

1a) Should I be doing some pre-treatment before exporting perhaps?

 

2) When exporting to the same size as the drawing, is there a way to turn 'Resample' OFF? as I really don't want resampling in this case...or is that just a different name for the function of how the rasterization is down.

 

3) Is there some logical reason that the horizontal lines are doing better than the vertical? 

 

4) Is it ridiculous to get white and black lines as close as a pixel away and expect them to export that way?

 

5) I also need to take this down to 2048 x 1080 for other projectors. So far they look a mess when I just export this drawing to the lower size. Is there some way to do this vector resizing easily or do I have to redraw things?

 

Many thanks.

Alignment Flat 4096.afdesign

[MikeW]

No drawing or image.

[Fixx]

To me the file exported to TIFF looks good. I could spot colour variation in said large rectangles where verticals are dimmer than horizontals. I think 1 pixel lines do not snap here perfectly to pixel grid and therefore become contaminated with background colour. There is a snapping control for that somewhere..

I used default bilinear which seemed to work nice. Nearest neighbor may lose antialiasing.

If you resample your design to 1080 you will lose a lot of details (1 pixel lines end to something else), you may want to prepare separate document for that.

[TestTools]

Good idea on the snapping – thanks. I'll give it a try and report back.

[TestTools]

I changed the file to put the straight lines onto exact integers, with snap and by typing. And I changed the MTFs to include some lines that are separated by 2 pixels in addition to the pairs separated by only 1. 

 

I have made several exports and get improved results, but not great white lines. The MTFs are good, maybe even fine enough, but the long lines that were OK horizontal top and bottom are now as grey as the sides are. And the fonts next to them are all broken up on rendering...is there a different solution for rendering fonts. ...or do I need bigger fonts?

 

And the 10% line is white on the horizontal, but not on the vertical – which means that I still don't have the right position, or that the algorithm is not looking both ways correctly perhaps.

 

I still would like to know why there can't be a direct 1 for 1 pixel render instead of a resample. My drawing is the same size as the output. I don't need a resample.

 

I've tried the drawing as Photo and as Print...perhaps one time as Devices...sd

 

Ideas? I can't settle for 'good' and 'good enough' on this project? 

 

Thanks

Alignment Flat 4096Photo.afdesign

[anon1]

if you don't rescale on export, the resampling does not make any difference 

 

I tried your image with nearest neighbour and langszos 

put them both into AD/ or AP on top of each other and selected "difference" blend mode and everything is black which means everything is the same 

 

in apples preview I get slightly varying results zoomed out, but they also look the very same zoomed in

they also don't have the same checksum which is probably because of some metadata differences in the header 

 

so guess you are good to go 

 

would be interesting to know what this "drawing" if used for?  :)

[Fixx]

looks like monitor test image..

[TestTools]

MBd, thanks.

"if you don't rescale on export, the resampling does not make any difference"

Are you certain of that? I really want it to be true. I mean, I'm not calling you on it, I'm just a newbie with AD and trying to figure this out.

 

What I know is that on a big screen it looks extremely bad, but there are several points of potential failure that I have to double check. Thank you for giving me some clues. 

 

Speaking of clues, any idea how to improve the smaller fonts? If I could use PDF, I would since it scales so well...but it is only 8 bit and I need 12.

 

Indeed, it is one of several slides for checking projection and monitors. It gets matched up with audio test sounds as well.