Pages: 1 2 [3]   Go Down

Author Topic: Kodak Sells Image Sensor Business!  (Read 9847 times)

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Kodak Sells Image Sensor Business!
« Reply #40 on: November 14, 2011, 10:26:45 am »

Ray,

What more are you testing than DxO?  They have results for all the major MFDB's; and a related website

http://www.sensorgen.info/

converts DxO results into more standard figures of merit for sensor performance (this site only analyzes a few MFDB's but gives details on how to extract the numbers from DxO data).
Logged
emil

ondebanks

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 858
Re: Kodak Sells Image Sensor Business!
« Reply #41 on: November 14, 2011, 11:48:07 am »

Emil,

First of all, DxOmark have actually only tested a small minority of recent/current MFDBs/MFDSLRs. And they show no inclination to test any older models; only some new releases. But people keep using the old ones; they have greater longevity, both physically and in terms of desirability, than DSLRs. And second-hand older ones are the usual entry point for MFD among amateurs, etc. We deserve to know how these units perform, too.

I am indeed aware of sensorgen.info. It's a brilliant idea, with tremendously useful information (I go to it sooner than I go to DxOmark). But it is a bit limited in what it tells you, because it inherits DxOmark's limitations.

Here are a few extra things that my tests will address:

- DxOmark/Sensorgen do not tell you about how many bits of data are actually used.
- They do not tell you about raw manipulation, like zero clipping.
- They do not give the gain/ADU at the different ISOs [this could probably be worked out from their data, however]
- They do not give any indication of dark current / long exposure performance (I'm not addressing that yet, but intend to do so in time, as it's hugely important in MFD)
- They do not give any indication of sample variance from unit to unit.
- They do not give any indication of count/percentage of defective pixels (again, that won't be my focus initially, but it's there to be seen in the data)
- There are also strange anomalies in some DxOmark/Sensorgen measurements, which need further independent investigation. For example, look at the Pentax 645D in Sensorgen: why does the readnoise suddenly jump up, apparently almost double, in going from ISO 800 to 1600?! (And the less said about the Nikon D3100's readnoise of "0.0 electrons", the better!)

Ray
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Kodak Sells Image Sensor Business!
« Reply #42 on: November 14, 2011, 03:18:38 pm »

Hi,

I absolutely agree that there is much misunderstanding about focusing accuracy, especially regarding image space. I wanted to write an article on the issue but I lack the necessary drawing and pedagogic skills.

Regarding Mark Dubovoys shimmming of the digital back to ten microns, he has posted sample images on the issue and they clearly indicate the importance of shimming. The digital back he is using is the Phase One IQ180 (I think) having a sensor pitch of 5.2 microns and lenses from Alpa, probably based on  Schneider Digitars and Rodenstock HR Digarons. Both series of lenses being very good. I agree with your view that 10 microns of misalignment would need a 2.5 micron pitch sensor to detect using f/4.

I have made some series of exposures with different amounts of defocus on an Sony Alpha A55 with 4.77 micron sensor pitch and found that a CoC of 6.3 microns on that sensor causes clearly visible loss of sharpness.

Correct focus

Defocused  
Camera moved 3 cm at object distance of 3.00 m, f=100 mm aperture = f/5.6.

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

I'd also like to point out that you may perhaps confuse some gentlemen. The author shimming the back on the Alpa was Mark Dubovoy  while 645D lens/sensor alignment issues were discussed by Nick Devlin and Mark Segal.

I admit that examining images at actual pixels is pixel peeping, but it is the only way to actually judge image quality. Why would you spend tens of thousands of Dollars on pixels if you are throwing away 75% of them?! Admittedly, often those pixels may not be needed, sometimes it's hard to see differences between 12 MP and 24 MP in A2 prints (16"x23").

I'm mostly shooting with a Sony Alpha 900, and I normally don't see focusing issues when using AF in a careful way, but I mostly use f/8 and that may mask some focusing errors.

Best regards
Erik







But actually not that much. He has an article here on LL where he states focus tolerance is a product of focal length. Which if you understand the problem, focal length has nothing to do with it. Tolerance is a product of the f-number of the system--regardless of focal length, if focus is off x mm, the amount of defocus is based on the f-number of the system. (Talking about front focus or back focus misses the entire problem as it is an image space issue, not an object space one.) Also in the Mamiya/Pentax shoot out article, he suggests (actually, claims) the the problem with edge sharpness is unaligned sensors--it could not be user error (after all it was two guys doing informal testing) or even the wall might not be flat--not that old buildings are not level and plumb. Occam's razor would throw that conclusion into doubt. And he shimmed his back 10 microns? Exactly what back is he using that would have a 2 micron CoC, assuming he is shooting at about f/4.5.

I would not consider this gentleman a great authority on focus.

From what I understand, Mr. Chambers is a pixel peeper. There are real flaws with evaluating images at 100%.

The problem I find with many reviews/claims is that they are extreme. That every slight flaw or perceived flaw (found in single samples most of the time) are grossly exaggerated.
« Last Edit: November 14, 2011, 03:49:27 pm by ErikKaffehr »
Logged
Erik Kaffehr
 
Pages: 1 2 [3]   Go Up