Pages: 1 ... 5 6 [7] 8   Go Down

Author Topic: Not worth it ?  (Read 28495 times)

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #120 on: November 25, 2013, 03:34:56 pm »

Hi,

What is crap with this:


or this:



or this (note this image is not HDR):



or this



or this


or this


ort this


Best regards
Erik

Myth legend or whatever, digital is still crap with highlights and near highlights :-)
« Last Edit: November 25, 2013, 03:46:36 pm by ErikKaffehr »
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #121 on: November 25, 2013, 03:53:20 pm »

Hi,

I expose so that I don't get clipped histogram and no clipped highlights. The histograms on the Sony Alpha do a decent job. I also use lowest ISO whenever I can.

Best regards
Erik

IMO 8 stops of DR ain't bad at all when most of the color photos you've (I've) taken have been on Kodachrome with its 5 or so lovely stops. My approach when using a camera that crushes highlight tones is to "underexpose" enough to move those tones into linear territory. As with transparency film let the shadows fall where they will. Use HDR technique (with subtlety, please) if/when desired.

I think having a background in transparency film gives a person a better appreciation of both the benefits and drawbacks of electronic photography. You can see technological limitations as boundaries to be leaned against rather than as barriers blocking your way.

-Dave-
Logged
Erik Kaffehr
 

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: Not worth it ?
« Reply #122 on: November 25, 2013, 04:08:14 pm »

Erik - sorry if I'm standing on the line so to say:

What I believe I do understand well what aliasing is.

But what I still do not understand:
When the object is rendered on the sensor at a certain size, but would be sampled by the same amount of pixels - lets say that star in your example would cover 5x5 pixels - let it be on 6.9µ or 4.8µ pixels - I still don't understand why the pixel size would matter here.
I understand well, that a larger resolution with smaller pixels on the same projected image causes less aliasing, thats pretty obvious,
but on a larger format sensor with larger pixels you would use a wider lens to get the same image and the image should
cover the same amount of pixels, if focal length was chosen accordingly.
Why should this situation cause more aliasing?

I somehow understand that the MTFs of MF and FF lenses differ and that f-stop plays a role (as AA Filter so to say), but if we could eliminate this variable - would the pixel size still matter?
And if yes why?

Sorry again - I have a scientific academic background and usually grab things pretty well - but here I just don't get it.

And yes - in the images of the feather - did you have exactly the same number of pixels per image feature, like lets say 10 pixels per streak of the feather or so?

Cheers and thanks for taking the hassle to answer
~Chris
« Last Edit: November 25, 2013, 04:10:06 pm by Christoph C. Feldhaim »
Logged

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: Not worth it ?
« Reply #123 on: November 25, 2013, 04:16:55 pm »

Hi,

When a lens project an image on the sensor the projected image will have a certain resolution. If you photograph a star, the image will not be a single spot, but a small disk. If that disk is smaller than a pixel and it falls across a pair of pixels the different situations are possible:

1) Both pixels see it
2) Neither pixels see it
3) Either pixel sees it but not another

So details smaller than pixel size can not be resolved by the sensor. The response to the same disk can be different if the pixel is moved a tiny little bit.

Check out the enclosed image, it was taken with 150 mm at f/8 alt f/9 at 3.5 m distance on a:

- Top left P45+  (MFD, 6.9 my pixels, not AA filtered)
- Bottom left Sony Alpha 99 (6 my pixels, AA filtered)
- Right Sony Alpha 77 (3.9 My pixels AA filtered)

Neither sensor on the left can reproduce detail correctly. We can see cross hatch detail and incontinous strains. In the image on the right the 3.9 my pixels of  the SLT make a better job. The ideal combination would be the surface area of the P45+ and the pixel size of the Alpha 77, corresponding to 120 Megapixels.

Another way to see it:

I have measured MTF for my Zeiss Sonnar 150/4 that was used on the Hasselblad, it looks like this:


Now, if we check the black curve, we can see that at Nyquist limit (120 lp/mm) on the SLT 77 the lens has an MTF of 4% (or so). This is to low to cause a lot of aliasing.

The P45+ has a pixel size of 6.9 microns, corresponding to 72 lp/mm, here the MTF is around 22%, more than enough to cause significant aliasing.

Best regards
Erik



This doesn't seem right to me. What I see in the image is a larger feather from the A77. Therefore the lines of the feather are not near the size of the pixels, therefore there is not aliasing. You are relying on the magnification (FL crop multiplier) to get the detail. If the width of the feather was the same in all shots, say for example 1000pixels across, you would have higher contrast from the bigger sensors as well as the same detail.

Your pictures on the next post are quite nice BTW.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
convert some "shadow DR" to "highlight DR" by reducing exposure or ISO setting
« Reply #124 on: November 25, 2013, 04:21:53 pm »

The DR of modern the dslr I don't see as a bonus, good in the shadows, bloody lawful with highlights or near highlights.
This simply means that the camera is overexposing, which unfortunately is a hazard with the common practice of exposure metering and analog gain levels that place metered midtones at around 12% of maximum raw level.

There is a simple solution: for example, to buy one stop of highlight headroom protection with one stop of that surplus shadow handling excellence:
1. when there is enough light to use base ISO speed, reduce the exposure by one stop
2. in lower light, where a higher ISO speed is needed, keep exposure (aperture and shutter speed) the same, but reduce that ISO setting by one stop
(Note that either of these can be achieved with -1 exposure compensation in A, S, or P mode).
3. in either case, compensate with a +1 levels adjustment in raw conversion (or just use auto-levels as a starting point.)

In fact some cameras [like the Olympus EM5] do this more or less as default, producing raw files with four or more stops between typical mid-tone placement and maximum raw level. (They are the cameras whose DXO measured sensitivity (what DXO calls "ISO speed") is significantly lower than the camera's ISO speed setting.)
« Last Edit: November 25, 2013, 04:24:24 pm by BJL »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #125 on: November 25, 2013, 04:25:03 pm »

Hi Cristoph,

The aliasing comes from "excess resolution". You always get aliasing if the resolution of the lens exceeds that of the sensor. Now, MF lenses use to be pretty good, so at optimal aperture they easily outresolve the sensor. (I would rather not use that term, but I use it for simplicity.)

You can get rid of that aliasing by stopping down. Stopping down to f/16 - f/22 reduces resolution, so aliasing goes away.

The image below was shot with a Distagon 40/4 at f/8 and shows color aliasing


A second image shot at the same time with f/16-22 doesn't show color aliasing:


What I really say is that small pixels are beneficial to image quality.

Best regards
Erik

Erik - sorry if I'm standing on the line so to say:

What I believe I do understand well what aliasing is.

But what I still do not understand:
When the object is rendered on the sensor at a certain size, but would be sampled by the same amount of pixels - lets say that star in your example would cover 5x5 pixels - let it be on 6.9µ or 4.8µ pixels - I still don't understand why the pixel size would matter here.
I understand well, that a larger resolution with smaller pixels on the same projected image causes less aliasing, thats pretty obvious,
but on a larger format sensor with larger pixels you would use a wider lens to get the same image and the image should
cover the same amount of pixels, if focal length was chosen accordingly.
Why should this situation cause more aliasing?

I somehow understand that the MTFs of MF and FF lenses differ and that f-stop plays a role (as AA Filter so to say), but if we could eliminate this variable - would the pixel size still matter?
And if yes why?

Sorry again - I have a scientific academic background and usually grab things pretty well - but here I just don't get it.

And yes - in the images of the feather - did you have exactly the same number of pixels per image feature, like lets say 10 pixels per streak of the feather or so?

Cheers and thanks for taking the hassle to answer
~Chris
Logged
Erik Kaffehr
 

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...

This simply means that the camera is overexposing, which unfortunately is a hazard with the common practice of exposure metering and analog gain levels that place metered midtones at around 12% of maximum raw level.

There is a simple solution: for example, to buy one stop of highlight headroom protection with one stop of that surplus shadow handling excellence:
- when there is enough light to use base ISO speed, reduce the exposure by one stop
- in lower light, where a higher ISO speed is needed, reduce that ISO setting by one stop
- in either case, compensate with a +1 levels adjustment in raw conversion (or just use auto-levels as a starting point.)

In fact some cameras [like the Olympus EM5] do this more or less as default, producing raw files with four or more stops between typical mid-tone placement and maximum raw level. (They are the cameras whose DXO measured sensitivity (what DXO calls "ISO speed") is significantly lower than the camera's ISO speed setting.)

I always thought ETTR at lowest possible ISO would be the solution to this.
And if you don't like the highlight rendering compress in post by applying curves later to get a shoulder.

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: Not worth it ?
« Reply #127 on: November 25, 2013, 04:28:55 pm »

Hi Cristoph,

The aliasing comes from "excess resolution". You always get aliasing if the resolution of the lens exceeds that of the sensor. Now, MF lenses use to be pretty good, so at optimal aperture they easily outresolve the sensor. (I would rather not use that term, but I use it for simplicity.)

You can get rid of that aliasing by stopping down. Stopping down to f/16 - f/22 reduces resolution, so aliasing goes away.

The image below was shot with a Distagon 40/4 at f/8 and shows color aliasing
..image..
A second image shot at the same time with f/16-22 doesn't show color aliasing:
..image..

What I really say is that small pixels are beneficial to image quality.

Best regards
Erik


Okay - the excess resolution thing I understand - I also understand that a simple blur in post cannot fix this.
So - basically in comparable situations you'd just have to stop down in MF to get diffusion or very slightly defocus?

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
convert some "shadow DR" to "highlight DR": ETTR? or just bracket!
« Reply #128 on: November 25, 2013, 04:37:40 pm »

I always thought ETTR at lowest possible ISO would be the solution to this.
I suppose that is a solution in cases where (a) you have enough light to use base ISO speed, and (b) you can reliably check the exposure (histogram preview?) to know that your brightest highlights are placed just below clipping.

But often that is not possible; for one thing the preview histogram is based on a conversion to JPEG. I instead have decided that
1. At base ISO sensitivity, the sensor's DR is usually far more than my images need, so that shadow noise is not a problem unless I underexpose horribly, and then I am happy to leave some comfort zone at the right end of the histogram.

2. If there is time to check and review histograms in pursuit of a perfectly right-justifed histogram, it is usually also possible to bracket exposures and choose the best one later, which is typically a much quicker and easier approach!
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #129 on: November 25, 2013, 04:39:01 pm »

Hi,

Stop down.

One thing to keep in mind is that diffraction is benign to sharpening.

The enclosed image shows some artefacts I was very unhappy about.

Regarding diffraction, I think that these figure illustrates it pretty well:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=1

But, there is also sharpening: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=2

I sort of plan to write two articles pretty soon now, one on sharpening and one on aliasing.

Best regards
Erik

Okay - the excess resolution thing I understand - I also understand that a simple blur in post cannot fix this.
So - basically in comparable situations you'd just have to stop down in MF to get diffusion or very slightly defocus?
Logged
Erik Kaffehr
 

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: Not worth it ?
« Reply #130 on: November 25, 2013, 05:00:23 pm »

One thing to keep in mind is that diffraction is benign to sharpening.

Yeah - since I learned that I became more courageous to stop down.
And Topaz InFocus, which I use does an incredible job at estimating PSF if you chose the right estimation area.

Christoph C. Feldhaim

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2509
  • There is no rule! No - wait ...
Re: convert some "shadow DR" to "highlight DR": ETTR? or just bracket!
« Reply #131 on: November 25, 2013, 05:01:07 pm »

I suppose that is a solution in cases where (a) you have enough light to use base ISO speed, and (b) you can reliably check the exposure (histogram preview?) to know that your brightest highlights are placed just below clipping.

But often that is not possible; for one thing the preview histogram is based on a conversion to JPEG. I instead have decided that
1. At base ISO sensitivity, the sensor's DR is usually far more than my images need, so that shadow noise is not a problem unless I underexpose horribly, and then I am happy to leave some comfort zone at the right end of the histogram.

2. If there is time to check and review histograms in pursuit of a perfectly right-justifed histogram, it is usually also possible to bracket exposures and choose the best one later, which is typically a much quicker and easier approach!

Sounds very reasonable to me.

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686

I always thought ETTR at lowest possible ISO would be the solution to this.
And if you don't like the highlight rendering compress in post by applying curves later to get a shoulder.

In my experience it depends on the sensor and what the camera does to the RAW data before saving it. I agree re. creating a shoulder (and toe) via curves when you want greater tonal compression. This is useful particularly when printing. But what do you do when an ETTR exposure results in mashed highlights because you've pushed them beyond the sensor/camera's linear range? They're not overexposed per se, just turned into Cream Of Wheat.   :o  Granted, I don't currently own any cameras that do this, or at least I don't use 'em in a way that reveals such behavior. (These days I expose for a useful JPEG even though I use the RAWs for printing. This likely mitigates any sensor non-linearity. The Pentax 645D with its well-behaved CCD is an exception...there I push the histograms TTR and rely on in-camera RAW conversion to give me lovely JPEGs.) I guess what I'm getting at is that techniques like ETTR should be used contextually rather than unilaterally.   :D  Do what works.

-Dave-
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #133 on: November 25, 2013, 05:11:29 pm »

Hi,

There is not a lot of difference in MTF between MF and FF lenses, some are obviously better than others. The two lenses I used were quite close in MTF, with a slight edge to the Sonnar 150/4. So, each strain covers more pixels on the SLT 99 than on the P45+.

In the feather shots, everything was kept constant, focal length, distance and aperture, with small deviations. So the difference is down to pixel size. You can say it shows what an image from a 149 MP MF sensor with crop factor 1.0 would look like.


Best regards
Erik


Erik - sorry if I'm standing on the line so to say:



I somehow understand that the MTFs of MF and FF lenses differ and that f-stop plays a role (as AA Filter so to say), but if we could eliminate this variable - would the pixel size still matter?
And if yes why?

Sorry again - I have a scientific academic background and usually grab things pretty well - but here I just don't get it.

And yes - in the images of the feather - did you have exactly the same number of pixels per image feature, like lets say 10 pixels per streak of the feather or so?

Cheers and thanks for taking the hassle to answer
~Chris
Logged
Erik Kaffehr
 

Phil Indeblanc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2017
Re: Not worth it ?
« Reply #134 on: November 25, 2013, 05:22:47 pm »

Quote
basically in comparable situations you'd just have to stop down in MF to get diffusion or very slightly defocus?

Is this to say that it is OK as long as we introduce diffraction or expect some sort of AA filter effect to control this? Just wondering as....

One area of MF legitimacy in file quality is in the scientific and product photography that is at macro levels and these details become the center stage of the problem you are dealing with. making them slightly "unsharp" would to some degree defeat the use of MF. Take an image of a faceted stone and you will see how soft the separations get.
Logged
If you buy a camera, you're a photographer...

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #135 on: November 25, 2013, 05:33:02 pm »

Hi,

Yes, MFDBs will induce some artefacts, unless stopped down so diffraction acts as AA-filter. Images below Sonnar 150/4 at f/5.6 (left) and f/22 (right). Note color fringes at top of left image.


Please note, I am not in the forgery business and shooting 1$ bills is not my favorite subject. I like the dollar bill because:

1) It contains a lot of fine detail

2) Most people who can buy MFDs have access to 1$ bills, so they can easily reproduce my test and post some nice raw files, like I do. Some people survive on that money each day, but they don't buy MFDBs.

Best regards
Erik

Is this to say that it is OK as long as we introduce diffraction or expect some sort of AA filter effect to control this? Just wondering as....

One area of MF legitimacy in file quality is in the scientific and product photography that is at macro levels and these details become the center stage of the problem you are dealing with. making them slightly "unsharp" would to some degree defeat the use of MF. Take an image of a faceted stone and you will see how soft the separations get.
« Last Edit: November 25, 2013, 05:42:08 pm by ErikKaffehr »
Logged
Erik Kaffehr
 

tho_mas

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1799
Re: Not worth it ?
« Reply #136 on: November 25, 2013, 06:04:40 pm »

MFDBs will induce some artefacts, unless stopped down so diffraction acts as AA-filter. (...) Note color fringes at top of left image.
The color fringes may or may not be pronounced (or suppressed/leveled out) in the raw software. May also have to do with the "calibration" of the DB. I've looked through some of your captures (raws). Although your captures seem to be somewhat soft they show quite a lot of color artefacts. I do shoot almost the same back (P45 non-plus) on Contax and while my captures (raws) are typically sharper/more detailed they also show much less color artefacts. I also shoot the very same back on a tech cam with very sharp lenses and while the captures are sure even more sharp/more detailed than shot with Contax lenses (at least at the edges...) the back still doesn't show that much color artefacts.
I don't want to say my back does not produce color artefacts ... it does. But overall much less... and much less pronounced.
Moiré, of course, is another story...
So your findings may very well apply to your kit... but not necessarily to MFD in general...??
« Last Edit: November 25, 2013, 06:07:12 pm by tho_mas »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #137 on: November 25, 2013, 06:15:24 pm »

Hi,

What sharpening are you using?

I am using LR 5, and it yields more color artefacts than Capture One, but C1 is not free from color artefacts either.

It is interesting that your images are sharper than mine, as the Contax lenses are said to be very similar to the Zeiss lenses on Hasselblad V. I shoot on a decent tripod with a decent head and MLU, so I don't think it is vibration. Focusing errors are possible. I think I have a tendency to front focus, I will check more into that.

Best regards
Erik
The color fringes may or may not be pronounced (or suppressed/leveled out) in the raw software. May also have to do with the "calibration" of the DB. I've looked through some of your captures (raws). Although your captures seem to be somewhat soft they show quite a lot of color artefacts. I do shoot almost the same back (P45 non-plus) on Contax and while my captures (raws) are typically sharper/more detailed they also show much less color artefacts. I also shoot the very same back on a tech cam with very sharp lenses and while the captures are sure even more sharp/more detailed than shot with Contax lenses (at least at the edges...) the back still doesn't show that much color artefacts.
I don't want to say my back does not produce color artefacts ... it does. But overall much less... and much less pronounced.
Moiré, of course, is another story...
So your findings my very well apply to your kit... but not necessarily to MFD in general...???

Logged
Erik Kaffehr
 

jjj

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4728
    • http://www.futtfuttfuttphotography.com
Re: Not worth it ?
« Reply #138 on: November 25, 2013, 06:20:09 pm »

Hi,

Yes, MFDBs will induce some artefacts, unless stopped down so diffraction acts as AA-filter. Images below Sonnar 150/4 at f/5.6 (left) and f/22 (right). Note color fringes at top of left image.
But what do the images look like after post processing to get the best result? Or is this after work has been done?
As it stands one is sharp with artefacts and the other soft with no artefacts. Personally I would choose the image that looks in focus over the artefact free shot.
Come to think of it, did you adjust for focus shift when you stopped down?
Logged
Tradition is the Backbone of the Spinele

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11296
    • Echophoto
Re: Not worth it ?
« Reply #139 on: November 25, 2013, 06:35:01 pm »

Hi,

I don't think there is significant focus shift on a Sonnar 150/4 when stopped down from f/5.6 to f/22. Both images are shown with same post processing. The idea was to demonstrate the effects of diffraction, the reduction of color artifacts is just a side effect.

I didn't vary sharpening (which I usually do when stopping down) because it is easy to get into a sharpening race. If I sharpen the right image a bit more, I can also sharpen the left one a bit more.

This is an older test showing the effects of quite extreme sharpening.

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=2

Best regards
Erik

But what do the images look like after post processing to get the best result? Or is this after work has been done?
As it stands one is sharp with artefacts and the other soft with no artefacts. Personally I would choose the image that looks in focus over the artefact free shot.
Come to think of it, did you adjust for focus shift when you stopped down?
Logged
Erik Kaffehr
 
Pages: 1 ... 5 6 [7] 8   Go Up