Heh. believe what you want. Clark is one of those people who asks, "who are you going to believe? me and a bunch of math and graphs, or your lying eyes?" I'd like to know where I can buy one of his 14-stop 5Ds, since mine only does 9 stops. there's a lot of misinformation there (check out his film vs digital comparisons to learn how to handicap a desktop scanner to look as bad as possible). But if it floats your boat, that's fine with me. However the first ten pages here debunked that kind of thing, which is why I posted what I did. Whatever, I'm out
Pondria wrote:
The 9 stop figure is consistent with various measurement efforts - DPReview, Imaging Resource, Imatest and Ourselves, FMers.
The discrepancy is simply that Roger is measuring the maximum possible dynamic range of the sensor itself, and not of the sensor+raw converter+gamma adjustment, and possibly even using Photoshop which is actually using 15 bit math instead of 16 bit. Furthermore, since different raw converters could affect dynamic range, that variable must be entirely omitted from the evaluation, otherwise there's no way to determine whether it was the sensor or the raw converter that made the difference. His numbers may be higher than 'real world' tests, but they are equally valid - he is just measuring things a little differently.
It is analogous to measuring the fuel efficiency of a vehicle on a dynamometer versus real world driving conditions. Drivers may never obtain the EPA mileage numbers, however, a standardized test is important because real world driving conditions can vary quite a bit depending on traffic, driver technique, engine tuning, hills versus flatland, etc.
I've been playing with a Hoya 072 filter for some near IR photography, with post processing for false colors, and I'm hitting problems with quantisation quite often in PP. I get to a point where just boosting saturation by 5% will result in color artefacts. The extra headroom of having 16 more discrete levels would definately help here...
Tentacle wrote:
I've been playing with a Hoya 072 filter for some near IR photography, with post processing for false colors, and I'm hitting problems with quantisation quite often in PP. I get to a point where just boosting saturation by 5% will result in color artefacts. The extra headroom of having 16 more discrete levels would definately help here...
That's a good case for exposing to the histogram. You may be blowing out one color channel or another from exposure. Again, this is an instance where the problem is not the intermediate tones, but rather the upper limit of responsiveness.
Yes, by virtue of removing the IR-blocking filter over the sensor.
But the sensor doesn't know what's IR and what's not.
All it knows is the relative amount of energy hitting a red photosite, a green photosite, or a blue photosite, which are only red, green, and blue by virtue of filtration.
A grossly overexposed image will be blown in all three channels. But overexposing in a certain wavelength, especially with highly saturated reds, may only blow the red channel.
This is actually one argument used for editing in wide color gamuts like ProPhoto RGB and LAB color (forget Adobe RGB, which is much smaller than ProPhoto).
But you might want to make sure you aren't blowing out your red channel with these shots -- or operating very close to the right side of the histogram. You might be getting your color artifact by oversaturating that channel, which means that the affected pixels will always have a 255 value in R.
20Da doesn't actually lack an IR sensor - it has a different IR filter stack that passes the just-outside-visible H-alpha (656.3nm), but blocks longer wavelengths of near IR. minor point
The sensor itself still doesn't differentiate between wavelengths. Whether you put a filter on the front of the camera or right over the sensor, the sensor itself is not measuring wavelengths -- it's only measuring energy.
At any rate I think the important point is to expose such that none of the channels are blown. That can be a challenge with long wavelength IR, because some things are very brightly reflective and others are very dark. So you might be really overexposing the red channel, but really underexposing the other channels. If you're trying to edit color with even one channel blown out, then you're basically turning a 36-bit image (12-bits per channel x 3 channels) into a 24-bit image for some colors but not for others. In other words, you may have 256 or 65,536 or whatever values per channel, but if your red channel is locked at 255 for all iterations, that's a dramatically lower latitude that will lead to color artifacts, oversaturation, and banding.
Reviweing the thread, multiple times it's been mentioned that 14bit has more head-room in in-camera or post processing.
I'm not convinced that this is the case. Embedded processors or RAW covnerters have all minimum 16/32 bit data width. 12 or 14 bit data need to be padded to 16 bit or 32bit anyway.
