1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

    Any content, information, or advice found on social media platforms and the wider Internet, including forums such as AP, should NOT be acted upon unless checked against a reliable, authoritative source, and re-checked, particularly where personal health is at stake. Seek professional advice/confirmation before acting on such at all times.

The death of ISO (cover date 30 April 2011)

Discussion in 'AP Magazine Feedback & Suggestions' started by Learning, Apr 24, 2011.

  1. Learning

    Learning Ethelred the Ill-Named

    Am I the only person who is too thick, or too ignorant, to understand Bob Newman's article? I thought that when we changed the iso setting of a digital camera then we were changing the analogue gain prior to A/D conversion so that the range of charge on the sensor is matched to the range of pd allowed by the 14 bit converters. It seems that I was wrong but cannot see what really happens. I feel that a further article, starting with the basics and then explaining this week's article might be helpful. If this doesn't merit another article on the subject then perhaps Profbob might be puersuaded to write something for the AP web site.
  2. Fen

    Fen Well-Known Member

    Dunno... my sub copy hasn't arrived yet :(

    Want to see the APOY results!
  3. AlexMonro

    AlexMonro Old Grand Part Deux

    I had another read through the article, and I think I understood what Prof Newman was saying - that the read noise and quantum efficiency of the new Sony sensor are so good that you could get the effect of shooting at high ISO just by boosting levels in post production.

    However, this seems to ignore the effects of quantisation noise, or error, in the A/D converter. If you've only got 14 or 16 bits of A/D, and you effectively throw away the top 6 bits by underexposing 6 stops (assuming linear sensor response), then you're left with only 10 or 12 bits. By the time you've done the linearisation processing for the Bayer colour filtering, you've only got 6 or 8 bits of image dynamic range left. So your final result would have the blockiness of a cheap phonecam. It would be a bit like a simple film camera with fixed aperture and shutter speed, and all the exposure adjustment is done at the printing stage.

    This assumes the readout and A/D conversion on the sensor works how I think it does - I haven't exhaustively researched it yet. I used to design electronic instrumentation systems in industrial process control, so I'm fairly familiar with A/D conversion, but I can't think of any other way that it could work.

    I'd be very interested in a future article which explored in full detail the complete image capture system of an "ISOless" camera.
    Last edited: Apr 25, 2011
  4. Terrywoodenpic

    Terrywoodenpic Well-Known Member

    I read it three times ... but I am still missing something.
    No sensor can capture all possible brightness ranges in one go at its native sensitivity. There needs to be a selection stage that tops and tails that brightness range, in a way that the sensor can both cope with, and is wanted for the formation of the the Desired image. ( or boost the sensitivity in low light situations so that reasonable shutter speeds or small apertures are available. )

    All photography is limited by the final output used to view the image. in the days of film, we could capture a surprisingly wide range of tone levels but they could never exceed the ratio of clear film to fully developed silver content. The paper print could never even reproduce this range in its entirety.
    The same must be effectively true of sensors, and final outputs, however they are viewed.
  5. beejaybee

    beejaybee Marvin

    Explained at great length and with many examples in "Astronomical Image Processing", get a copy from your library as it's very expensive to buy unless you really need it.

    Fact of the matter is, an incoming photon liberates the same number of electrons whatever the ISO setting. Multiply this value by analogue or digital means and you're introducing a reduction in signal to noise ratio ... and possibly clipping high values as well. The chip has a "native" ISO value - usually the lowest in the set - and that's the setting it produces its best results at.
  6. Learning

    Learning Ethelred the Ill-Named

    Agreed. What I am not seeing is how small numbers of electrons liberated in a weak exposure can give rise to pds suitable for A/D conversion on the same converter as is suitable for strong exposures.
    I agree that the sensor should give its best performance at base iso setting. But while the sensor may be performing at its best S/N ratio its output may not be adequate to drive the converter over more than the lowest of its 14 bits.
    In the illustrations for the article the claims about noise are demonstrated but even with magazine printing they also show that the D7000 image shot at iso 6400 has shadow detail not visible at iso 100.
    I am not saying that the article is wrong. I am saying that I need more explanation before saying it is right.
  7. mark_jacobs

    mark_jacobs Retired

    Noted, I will see what I can do :) It's good to have the opportunity to discuss though, perhaps lacking elsewhere, is it not? :)
  8. daft_biker

    daft_biker Action Man!

    Perhaps only because the cart comes before the horse currently?

    If there's enough in-camera processing power the amplification could be worked out after exposure rather than before, and still provide a useful JPEG for immediate use.

    As a guide to the amplification that is required ISO number is probably just as useful to me as using another number but before exposure it would just be an estimate so you can see how the exposure compares to light levels.

    TTL flash though:confused: Without telling it how much amplification will be used how can flash output be controlled automatically?
  9. beejaybee

    beejaybee Marvin

    The same way that the OM-2 used to do it? Or, open the shutter, fire off a trial flash, sample the readout, compute the power required, flush the CCD, fire off the exposure flash & finally close the shutter. With a small sample this would not take more than a few milliseconds & the trial flash need only be a few percent of the available charge in the flash condenser.
  10. Alex1994

    Alex1994 Well-Known Member

    He seemed to be saying you don't need to worry about the exposure (if you're shooting raw), just choose the aperture/shutter speed combination you want and fire away!
    I bought a D7000 recently, and will test this out over the weekend ...
  11. El_Sid

    El_Sid Well-Known Member

    Which is fine until you're 5 stops underexposed and your raw converter can only pull back 4...:rolleyes:
  12. TimF

    TimF With as stony a stare as ever Lord Reith could hav

    Having had a quick read of the Prof's article in the supermarket this morning, it strikes me that his proposition is not dissimilar to the way in which the late Garry Winogrand is said to have worked.

    In his case he used an ISO 400 film (probably Tri-X) and a constant exposure setting on the camera, don't know what this was - a guess would be 1/125th at f/? When the roll was finished Winogrand would attach a label describing the weather conditions to the tub and then developed the film accordingly (or didn't, as at his death he left something like 2,500 undeveloped rolls of film!)
  13. MickLL

    MickLL Well-Known Member

    I found this article and the other about sensor comparison (see the other thread) both very fascinating although, like others, I didn't entirely 'get' the prof's article.

    I rather suspect that it's a combination of trying to write within certain word count and also trying not to confuse the readers too much. However some of us (and I really really hope that there's enough ) want more. A series of basic articles leading to the more advanced stuff - maybe just a page a week - would be great.

    I used to find some of the late GC's articles quite impenetrable I do hope that you don't allow the Prof's contributions to go the same way and that you can allow him the space and latitude to really teach us. if you do I could be persuaded to subscribe.:D:D:D

  14. Learning

    Learning Ethelred the Ill-Named

    Our editor's thread, started 2 days after this one of mine is now covering much the same ground. I suggest that it would be helpful to keep further discussion of this topic on his thread only.

Share This Page