Help judging performance of my FE55z, FE35f2.8z horizon/infinity

Started Jul 5, 2016 | Questions thread
l_d_allan Veteran Member • Posts: 5,087
Why stacked layers in TIFF's? (and other ?'s on evaluating results from "good copy vs bad copy")

JimKasson wrote:

l_d_allan wrote:

JimKasson wrote:

The format in which I find it easiest to do these tests is to export the images as full-sized TIFFs from your raw developers to Photoshop as separate layers. Then label the layers with the settings, write the file out as a PSD, host it on Dropbox or something similar, and post a link here.

I'm curious what kind of info you would include in the "separate layers".

Lens, aperture.

My observation is that the EXIF has most of the relevant info (EFCS?, temp?, IBIS?).

You lose that when you stack all the captures into a single image.

I'm being slow on why it is essential to stack images ... for easier A/B examination?

And I use TIFF with layers all the time to annotate targets for making printer profiles with 4000+ patches for my i1iSis. However, a set of 4 layered tiffs is only 800KB x 4 = 3.2 MB.

Would the layer have details about non-EXIF such as tripod usage? How much care taken for issues such as atmospherics?

I really try to stay away from TIFF's unless there are compelling reasons .. and I have TB's of inexpensive storage. I get impatient with the sluggishness of huge .tif files.

Try striping (RAID 0) with a fast disk controller.

It's been years ago, but my experience with RAID was negative. I would be hesitant to trust RAID 0 (no redundancy and multiple points of failure?) on rotating rust, but apparently it works for you.

Anyway, my WIP (Work in progress) is done on a relatively recent TB Samsung SSD.

My experience is that a .tif is about 5x larger than a cRAW, at least for my a7Rii. I almost always use uRAW and then DngConverter to get .dng's with "pragmatic lossless compression" ... so a uRAW goes from 85 MB to 45 MB. But then it would bloat to about 230 MB for the .tif.

Don't forget, you're going to throw away the TIFFs as soon as you create the stackled layer image, and you're going to throw that away as soon as you reach conclusions about the test you're performing.

Well, uploads of multiple files would still be slow. And once I'd have gone to the trouble of putting together results for a LUT (lens under test), I'd be inclined to keep it around for future reference ... and DPR discussions.

But for the number of files involved, I'll concede that having a few TIFF's around isn't going to be enough of a problem to worry about. My "push back" was (and still is) why TIFF's are necessary in the first place.

BTW, my practice is to buy 3 packs of better quality 2+ TB USB-3 drives at NewEgg shell-shocker prices, and backup pretty much everything. They're really gotten quite fast ... not that much slower than SATA-3 rotating HDD.

Then anybody doing the reviewing can just click on the layer visibility icons to directly compare any images in the stack.

Is that so you can quicker / easier do A / B checking of the center vs the individual corners vs the extreme corners?

Is the upper left about the same as the upper right as the bottom right as the bottom left? Are the corners only somewhat less sharp than the center? How does the center of the LUT compare to a "known to be good copy" of my FE55 on a7Rii?

BTW: I'd appreciate having your push-back on errors I may be making on how I evaluate results from brick-wall testing. If I were to write up my approach (attempting to be concise) ... would you be willing to do this grasshopper the favor of letting me know about the error of my ways? I've learned a lot when you've pointed out my misconceptions and ignorance about a variety of topics over the years (and thx for your patience).

The downside of this is that it means that everybody involved has to have Ps, though.

Unless I'm unaware of lots of info that should be in the "separate layers", I'd speculate that a simple uRAW or cRAW or .dng ... WITH more or less detailed annotations about exceptions for standard practice... might suffice, and also avoid the need for PS.

Then it's harder to click back and forth. Also, with layers, you can align images that weren't perfectly captured.

I looked at your raws. They look fine to me.

I would very much appreciate learning how you went about evaluating the OP's raws.

Perhaps you could "think out loud with your keyboard" to explain what you look for, what you observed with the OP's raws, and what it meant.

I imported all the images into Lr. I exported them to PS as layers (down at the every bottom of the "open in" menu) thus avoiding the TIFFs. I zoomed to 2:1, and looked at and compared corner sharpness, comparing the tilts in one direction as a group and the tilts in the other direction as a group. If I really wanted to get fancy, I would have flipped all the images that were tilted in one direction, so I could easily compare them all.

Have you done something like that as a TLW blog article?

Nope. At this point, I'm not sure I want to encourage people to do this kind of testing,

I think you have enough "earned credibility" that it might be helpful to have something of a check-list of people worried about if their lens was ok:

  1. Did you read the TLW blog article at link [fill-in-blank] about how to test for decentering? Any questions?
  2. Did you read the TLW blog article at link [fill-in-blank] about how to do brick wall or far horizon "good copy vs bad copy" testing? (less involved than MtfMapper involved protocol). Any questions?
  3. Did you read the TLW blog article at link [fill-in-blank] on how to evaluate the results from the brick wall or far horizon tests? Any questions?
  4. Have you uploaded RAW's or TIFF's to a publicly accessible site like DropBox?
  5. If not ... expect to be ignored ...

BTW, my impression is that you are underwhelmed / unenthusiastic about using a brick wall for "good copy vs bad copy" lens testing. If so, I'm curious why you seem to prefer something like a "slanted horizon at near-infinity".

Does that reflect your current situation with typically dry clear air with usually little or no wind in the morning?? Not handy to use a brick wall?

Maybe I'm one of the few with 3 nearby schools with very good brick walls within walking distance?

after all the examples of how to do it wrong I see on the web (not the images in this thread, however.).

Since being in the market for pricey Sony lenses with their earned rep for so-so QC,

I'm not sure that the rep is well earned. Decentering happens will all vendors. It's never happened to me with Leica, by Lloyd Chambers has reproted on several decentered Leica lenses. I do respect what Roger C has done on the subject, and his results don't put Sony miles away from other vendors, at least the way I read them.

Sorry for being argumentative (moi ?), but that's not my reading AT ALL of LR articles from Roger C. on Sony q/c, at least for the non-GM lenses.

Au contraire mon frère , my take-away was something like "if Canon can have stellar q/c on their entry level $150 cheapo lenses, why is the q/c on very pricey Zony lenses such a crap-shot (his term)?"

My impression: he was shocked / horrified at the crap shot q/c for the FE90, one or several zooms, and especially the notorious FE35 f/1.4 ... 10 of 10 which were unacceptable. Did we read the same article? Not just about the recent GM's?

I've done perhaps more than my share of brick wall testing. However, I have my doubts I really know how to evaluate the results of this tedious and error-prone "good copy vs bad copy" testing.

I never got close to getting valid, objective MtfMapper numbers, so it all seems rather subjective "by eyeball".

Most people's difficulty in getting good MTF Mapper numbers is, I believe, the result of several issues:

  • Alignment
  • Getting a high-resolution target
  • Alignment
  • Getting the target big enough.
  • Alignment
  • Getting the target far enough away
  • Alignment
  • Lighting the target properly
  • Alignment
  • Vibration
  • Alignment
  • Field curvature

Oh, and alignment is important, too.

Mostly agree, but my impression from the series of articles on your proposed Protocol was that the razor edge approach made alignment less critical. However, I admit to my shame that much of that series was over-my-head .

 l_d_allan's gear list:l_d_allan's gear list
Panasonic Lumix DMC-TZ3 Panasonic Lumix DMC-FZ47 Canon PowerShot S110 Canon EOS 5D Mark II Canon EOS 600D +27 more
Post (hide subjects) Posted by
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow