In rec.photo.digital Floyd L. Davidson <floyd@[EMAIL PROTECTED]
> wrote:
> Chris Malcolm <cam@[EMAIL PROTECTED]
> wrote:
>>In rec.photo.digital Mr.T <MrT@[EMAIL PROTECTED]
> wrote:
>>>> I suspect you may be confusing resolution with range. The problems
you
>>>> refer to are problems of range.
>>
>>> Did anyone mention resolution, can't see it in this post myself.
>>> (other than your latest "input")
>>
>>The connection between number of encoding bits and resolution of the
>>resulting encoding in analogue to digital conversion is probably
>>something else your photographic education omitted. If so then you
>>wouldn't have realised that you were the one who brought it up by
>>suppposing that the number of encoding bits in the camera sensor
>>affected dynamic range. It doesn't. Any number of bits can encode any
>>dynamic range you like. What the number of bits affects is the
>>resolution with which that dynamic range is encoded.
> Your description is incorrect. The sensor is analog, it
> has no such thing as a "number of encoding bits". The
> digital to analog conversion is not integral to the
> sensor,
In some sensors it is, but that doesn't really matter. I was referring
to the combination of sensor and ADC as an integral unit since in a
digital camera there is no way of accessing the sensor data in any
other way than via the ADC. They are functionally integrated.
> and of course the analog output from a sensor
> has a known dynamic range; however, while the number of
> bits, e.g. 12 or 14 are now common, does not determine
> the dynamic range of the sensor output, it can and
> sometimes does limit the dynamic range of the resulting
> digital signal (e.g., when the dynamic range of the
> digital format is less than the dynamic range of the
> analog output from the sensor).
You're perfectly correct. We've had this argument before. I note that
our last argument taught you something, since you now no longer claim
a necessary connection between the number of bits and the dynamic
range, merely a contingent one.
> I.e., no matter what
> the dynamic range of the sensor, the *maximum* possible
> dynamic range for a digital signal is limited by the
> number of bits/sample to (6.02 * N) + 1.76 where N is
> the number of bits.
That begs the question, because the normalising assumptions behind
that equation specify what kind of dynamic range we're dealing
with and how that dynamic range should be encoded. You should be able
to find those assumptions from the textbook you got the equation from.
> Suggesting that "any number of bits can encode any
> dynamic range you like" is indicates that you *still* do
> not understand the significance of digital technology.
I've been working with it since 1964. I was writing software to
convert RAW digital camera images and making my own cheap digital
cameras from dynamic RAM chips in 1980. I was teaching digital
technology at postgrad level in 1986. But you're right, digital
technology has quite revolutionary significances which I'm still far
from fully comprehending.
You haven't however yet convinced me that you have a better grasp of
it. :-)
> "The resolution with which that dynamic range is
> encoded" is no different than saying "the distortion
> generated by encoding", which is exactly what *limits*
> the dynamic range. In essence, quantization distortion
> increases if the step size increases.
Of course it does. That's where the assumptions behind your equation
come in. You're treating perfectly good engineering compromises as
though they were scientific principles.
>>There is a tendency for technology to progress on a number of fronts
>>at the same time. Hence there has been a tendency for improvements in
>>dynmanic range to accompany improvements in encoding resolution, just
>>as improvements of dynamic range have also been accompanied by
>>improvements in battery capacity and optical zoom ranges. But there's
>>no necessary connection between them.
> The fact is that as the dynamic range of the analog
> sensor has increased, camera manufacturers have
> progressed to using more bits per sample in order to
> retain, in the digital signal, the dynamic range that
> exists in the analog range.
The assumption you're not stating is that they were trying to encode
the new increased dynamic range at the same or higher resolution than
the old dynamic range. There is NO connection between number of
encoding bits and dynamic range without making assumptions about
resolution. The equation you like to quote to "prove" your point
embodies such assumptions.
> You are suggesting that 14
> bit ADCs were not available 10 years ago when digital
> cameras were moving from 10 bit to 12 bit ADCs.
No I'm not. 16 bit ADCs have been available for decades. Camera makers
haven't been using the state of the art stuff you find in research
labs, they've been using what could be packaged and sold to a mass
market.
> That is
> not true. While certainly there have been improvements
> in 14 bit ADCs over the past 10 years, it is a fact that
> if CCD sensors had required 14 bit ADCs ten years ago,
> they would have been used. 14 bit ADCs were not used
> until the dynamic range of the sensors improved to
> greater than what a 12 bit ADC can provide.
You're perfectly correct, and the point you're missing which I'm
insisting on making is the engineering compromises and normalising
assumptions that lie behind that fact.
I think the dispute between us is simply that I'm not allowing you to
state engineering compromises and normalising assumptions as though
they were scientific principles or mathematical theorems. I'm not
disagreeing with them as perfectly good engineering compromises.
--
Chris Malcolm cam@[EMAIL PROTECTED]
DoD #205
IPAB, Informatics, JCMB, King's Buildings, Edinburgh, EH9 3JZ, UK
[http://www.dai.ed.ac.uk/homes/cam/]
|
