Advantage SMaL? Dynamic Range Comparison!

Discussion in 'Mirrorless Digital Cameras' started by rayfraser, Apr 1, 2005.

  1. The photos below were taken with a 1280x860 SMaL imager and a
    1280x960 Micron imager (both in cameras made by Pure Digital).
    Please comment on whether SMaL's claimed 120db dynamic range helps
    provide better zones in a single exposure. Is it obvious which one
    is SMaL? I find it odd how the duct shadow on white wall near red
    valve handle has disappeared. While I can clearly see duct shadow
    and red valve with my eyes, I cannot see swirls in bulb. If more
    detail is desired, full images are in my Advantage SMaL folder<p>I
    would welcome any suggestions on how to accurately measure a 120db
    range. Hopefully my basement photos will not draw any comments
    like "There are guys in their basement working on perpetual motion
    machines, everlasting batteries, gasoline engines that get 1000mpg
    and contacting aliens." Note I am still trying to remove bayer mask.
  2. Yup. If Ansel Adams were alive, he'd throw out all his Zone stuff. With 120dB dymanic range, who needs film?

    Nice shot of the bulb, but the loss of the shadow is a bit troubling....
  3. This isn't an April Fool's, right? If so, you're 44 minutes late. ;>

    It's a little weird - the top photo looks like a combination of flash and slow shutter speed was used to retain highlight detail in the fluorescent lamp while brightening the shadows.

    Not that it's bad or anything, just odd looking.

    I wonder whether something similar could be accomplished with a fairly ordinary dSLR and some very carefully tweaked custom curves. I won't claim I get anywhere close to that dynamic range from my D2H, but the right custom curve setting significantly improves shadow and highlight detail while keeping the midtones reasonably natural looking.
  4. Hopefully my basement photos will not draw any comments like "There are guys in their basement working on perpetual motion machines, everlasting batteries, gasoline engines that get 1000mpg and contacting aliens."

    Hmmm... the fill flash in the 1st shot is so obvious it's hard to know what to say. Where did the edge illumination come from on the front edge of the plenum strapping? The hard shadow off the immediate left of the plenum?

    It's April 2nd, so no humour can be assumed. Just dishonesty.
  5. Lex - Fully perpendicular that time! You nailed it 100%!

    The upper shot is illuminated mostly (or maybe entirely) by an on-camera flash. You can see the shadow of the "swirl" lightbulb to the left of, and below, the lightbulb. There's no "duct shadow" on the wall because the flash has filled it in totally.

    There's still flash visible in the lower shot, but the lighting ratio is a lot different. It favors the bulb (which is why the bulb itself is blown). The bulb still has a shadow from being illuminated by the flash, but it's a warm shadow, because light from the bulb is getting into it. The duct shadow is blue, because the duct is blocking the warm light from the bulb, but the cool blue light of the flash is filling in the duct shadow a bit.
  6. Ray - to accuratly measure 120dB (20 full stops), you first need a real, honest 120dB input signal to the camera system. In your example, you obviously haven't done this, because the flash reduced the dynamic range of the first picture dramatically, but the second had no such advantage. The first shot, in optical engineering terminology, is "contaminated".

    I do such testing with backlit targets in a darkened, black walled optical lab. In a more open environment such as yours, you need to use "black holes", targets with black shrouds to keep room illumination off them, as the low illumination parts of your test. 120dB is, literally, a million to one lighting ratio. You'll need some specialized gear to measure that: a highly directional, close range light meter (not your typical Minolta or Sekonic photographer's light meter).

    Avoiding contamination is basically the most difficult and most important part of the test. You'll come a long way in what you're doing, if you simply turned off the flash on both cameras.

    Then, you need a 120dB (or better) signal path from the test target to the sensor. As I mentioned, 120dB is a million to one lighting ratio. Imagine you have a dirty smudge on a lens that scatters just 0.01% of light. That's 1/10,000 of light. So if there's a bright object in the scene, its light focuses where it's supposed to focus, but 1/10,000 of the light scatters and illuminates all parts of the sensor. This scattered light is now filling in and obscuring all the deep shadow detail that's less that 1/10,000 as bright as the bright object. So the image reaching the sensor only has a 10,000:1, or 80dB dynamic range. So it doesn't matter if the sensor has 90dB, 120dB, or 150dB of dynamic range, the image only has 90dB of range.

    Now, it isn't just dirty smudges that limit the dynamic range of an optical system, I just used that example because it was easy. It's the quality of the lens coatings. The quality of the optical material itself. Even the finest optical glases from from Schott, Hoya, Nikon, etc. have sufficient levels of contamination to scatter some light. Plastic lenses scatter a lot more, just by the nature of photon propagation through polymers. The lenses on the disposable cameras that you're using simply cannot pass 120dB signals.

    Most general purpose photographic lenses cannot pass 120dB signals without contamination, either. There are a few that might be able to do it. They would be simple designs, with a limited number of elements. Slower lenses, so the elements are not only limited in numbers, but they're also thinner (so the optical material itself doesn't have as much opportunity to scatter light).

    You'll need to modify your cameras, because no one in this field is going to believe your results unless you use the same lens for both sensors. Not the same "kind of lens", the very same chunk of glass.

    And, even if your lens was beyond 120dB, and your scene is a real 120dB, you then need to preserve the 120dB in the optical box of the camera. Real cameras, even our technical cameras (which are designed to specs way beyond general photographic gear) have problems with light hitting the sensor, scattering around the camera's chamber, reflecting off the chamber walls (or the rear element of the lens) and scattering back to the sensor, limiting the optical range.

    The sensor itself scatters light internally. Multiple reflections between the optical glass window of the chip package and the surface of the chip itself.

    That's the main reason that I consider SMaL to be a fraud. Even if they really had a 120dB sensor, there's no way they can get a 120dB signal to it. Their literature shows contrived examples involving security cameras. Well, welcome to the real world, guys. Security cameras don't have lab grade lenses. And they sit behind flat glass windows or plastic domes. These are not cleaned by trained optical technicians using precision techniques and optical grade refined cleaning solutions. They're occasionally wiped off by a janitor with a filthy rag and a squirt of windex. You'd be lucky to get 40-60dB (7-10 full stops) of dynamic range through their optical systems.

    I see Detroit pictures in your portfolio, and 5 years at Ford. I'm "that" Joe Wisniewski, Henry Ford Technology Award winner (the Ford equivelant of the Nobel Prize), as well as an optics designer, and the technical photography teacher over at Midwest Photographic Workshops. I'm quite well versed in how to conduct these types of experiments, as well as the characterists of the entire optical system that will affect the outcome. I can help you set this up, but it won't be easy, and it won't be cheap. But I cannot for the life of me see why you'd want to pursue this.

    To be blunt, you seem to be on some sort of crusade here, and I can't for the life of me understand why. Why is proving that there is what you keep referring to as "advantage SMaL", so important to you?
  7. Thanks for a very valuable but winded response Joseph. I was nominated twice for the HFTA as part of a group but in my own my mind did nothing equivalent to a Nobel Prize and believe most of the "winners" are chosen based on non-technical aspects (monetary benefits). Do you still work for Ford and what do you know about their relation to SMaL autovision systems?<p>
    I have been fascinated by Keith Fife's several patents and am always interested in technological advances. While SMaL was a very small MIT company until bought by Cypress and has only produced low-end imagers, the possibilities when varying voltage pixel by pixel intrigue me (I wonder whether Philip Greenspun has any SMaL comments). I just can't seem to get the idea of wearing sunglasses at a movie theater out of my mind. I do not care whether it is SMaL or someone else that produces higher dynamic ranges in a single exposure. Sorry, I must work cheaply or else I would gladly enlist your services Joseph. This was my first attempt and each camera's flash, lenses, slightly different distances to produce same frame ..., clearly prevent an accurate comparison. I am now thinking of performing an outdoor comparison with no flash and lenses removed using a common pinhole instead.
  8. You didn't answer why you used flash. If you didn't realize that it totally screws up and invalidates any claims you may make about dynamic range, then you're lacking essential knowledge. Either that are you're deliberately trying to fool people.

    While you may not have been trying to invent a perpetual motion machine, everlasting battery, gasoline engine that gets 1000mpg or contact aliens, you were still in the basement...
  9. The top photo looks like it was made with flash, the bottom photo without.
  10. These SMaL threads seem to be more like advertising; where the 120dB talk is like household vacuums with 6HP decals; and wimpy 18AWG cords. In astrophotography; one can have a faint star; a few seconds or arc minutes away from visible "bright" star. Here the flare of the lens; spikes from the mirror diagonals; bleed from the sensor or film swamp out the radically fainter star. Large brightness ratios physically close to each other is a problem; 120dB is abit absurd..
  11. There is enough dirt; grim; crud; fog; dust in and on an average camera's lens that a 120db ratio in lighting will have a radically less transfer function in dynamic range when hitting the film or sensor. This is like listening to classical music from a CD while driving in a noisy old convertable; with the top down; at freeway speeds; with the cars speakers..; with the car next to you booming out rap music.
  12. Sorry about flashes, but cheap $10 digital cameras do not have an easy option to disable flash other than getting out of the basement. Note both cameras did flash. I did not intentionally do anything to weight the results. I appreciate all comments and agree with most of them. I will try again without a flash and just a pinhole for a lens using a scene similar to my Hart Plaza water fountain Sun shot. At this point, I still do not think SMaL has lied and exaggerated their imager claim.
  13. Yes, the HFTA can be pretty political. On our project, the engineers had to put up a heck of a fight to keep from being pulled off the award and replaced by markiting folk. I had a similar incident many years ago when my boss decided he would be the one to present a n SAE paper that I authored.

    There are a lot of possible ways to achieve HDR (high dynamic range) imaging. SMaL is non-linear, and hard to linearize, which means its colors are never going to be quite right. It makes all the things you need to do to make a high quality camera very difficult. Even trying to map your flat field and gain normalize all the pixels in a non-linear domain is difficult. With the SMaL design, the actual non-linearity varies on a per-pixel basis, so you need to apply a different non-linear curve to each pixel. Instead of doing the flat-field with a simple linear gain table as big as the sensor pixel size, you're suddenly dealing with a table of third or higher order functions.

    I'm rather fond of the way Fuji is doing HDR with their SR sensors. Right now, they're only using a 2 or 3 stop sensitivity difference between the S and R pixels, but they could go considerably higher. Four stops, and we should be running much more frequently into the limits of the optical systems.
  14. Haven't heard anything about Ford and SMaL. Sounds like a horrid mismatch, a HDR sensor in an automotive application. Shoot a couple of pictures through the window of a car, and you'll see what I mean about dirty glass knocking the dynamic range way down...
  15. I think auto application might be closer to road and able to distinguish between bright headlights and dark yellow median lines - until you hit a muddy puddle.
  16. Ray, It is interesting that you claim that you are trying to remove the bayer filter but could not manage to put a masking tape in front of those tiny flashes to block out the flash light!
  17. So, Ray, you've just discovered fill flash. Better patent it before someone else steals the idea....
  18. Vivek beat me to it, durn his ornery hide. The only possible drawback I can see to simply taping over the flash is if the camera's exposure system (presumably all AE since it's a cheap camera) expects to "see" flash and will go whacko if none is fed back. OTOH, nothing ventured... I don't see any reason to criticize Ray for these discussions. Heck, even if he is pulling our collective legs, I'm a joker myself. I tried to fool my niece and nephew tonight with a faked night photo of "spirit orbs", so popular among too eager wanna-believers. The believers seem convinced that only digital cameras can be used to photograph spirits. The old Sony Mavica with 3.5" sorta floppy drive was a popular ghost hunting tool for a while. Anyway, I just set the shutter to 8 seconds, had my friend point to a tree and used slow sync flash. I squiggled the camera around to form a circular blob from a neighbor's yard light. After a few tries we managed to get one just right, where it appears that my friend is pointing to a "spirit" climbing down a tree. My nephew wasn't the least bit fooled and my niece smacked me when I finally told her it was faked. But I didn't want her to be mocked at school after telling kids she'd seen photographic "proof".
  19. I agree, Lex, you're quite right - there really isn't a need to get angry with Ray and his mysterious experiments. We should just pull his leg in a jocular manner, which is much more fun anyway :) I'll remember that next time!

    Truth be told, I'm genuinely interested to see what happens when you manage to scrape off that Bayer filter, Ray. It sounds something akin to shaving one's tongue....
  20. Lex, that ties into your other thread. The pulsing light is "stuttering"...
  21. Joseph, are you trying to tell me that my shaky hands are actually ... shaky? ;>

Share This Page