[filmscanners] RE: Dynamic range

["Austin Franklin" <darkroom@ix.netcom.com>]

Hi Paul,

> > "Distance" is only one part of the equation, how finely one can
> > discern over that distance is the other part.
>
> I don't think that's an issue, because it's trivially easy to use enough
> bits in the A/D converter that you're limited by noise _throughout_ the
> range, not just at the dark end.

That's true, noise is the limiting factor, and A/Ds are plenty good enough,
and fast enough, for resolution of the A/D to not be the problem...but that
doesn't negate my statement.

> > The dynamic range changes based on the minimum discernable
> signal (noise)
> > and/or the overall signal range.  If one changes, the dynamic
> > range changes.
> >
> > The dynamic range of 16 bits = 10log2**16 or 48dB
> > The dynamic range of 14 bits = 10log2**14 or 42db
> >
> > A sort of "counter" example...Now, if you designed two scanners
> > that had the
> > same noise for both a density range of 2.2 and then a density
> > range of 3.9,
> > the scanner that had the density range of 3.9 would have the
> > higher dynamic
> > range...remember, the dynamic range is based on the OVERALL
> range AND the
> > noise, so if you increase the overall range, and have the same
> noise, you
> > increase the dynamic range.
>
> You can't really base dynamic range specifications on the numbers, because
> there's no guarantee that the numbers bear any particular relationship
> (linear, log, gamma) to the light power.

Well, they don't have to, really, it depends on what you are trying to
characterize.

> Dynamic range really should be
> measured as the (log of) the ratio between the strongest light power that
> the scanner can read without clipping, and the weak light power that is
> equivalent to the electrical noise in the sensor and subsequent
> electronics.

Correct, but that's a different characterization than I cited above.  You
can characterize the dynamic range of many things...

> If you then took the log of the signal level, and converted that to a
> digital value, the number of bits you have would no longer have any
> relationship to the dynamic range, only to the resolution.

That's not true.  You NEED so many bits to "contain" a particular dynamic
range.  It's quite generally publicized...do some web searches for "number
of bits" and "dynamic range".

> That is, if the
> electrical noise corresponds to a light level that's one millionth of the
> light level that clips the sensor and converter, you've got 60db
> of dynamic
> range, regardless of whether you use eight bits (60/256 db per step) or
> sixteen bits (60/65536 db per step).

No, that's wrong.  You are talking about DENSITY range, not dynamic range,
they are different.  If you divide the overall range by the noise, you get
the dynamic range, and...how many bits you need to "contain" that dynamic
range.

> All the stuff about rulers is meaningless, because dynamic range is only
> meaningful when measuring the ratio of power levels, not linear quantities
> like length.

Dynamic range can be used to characterize anything that has the particular
properties that can be used to calculate it.  Where is it written that it
can ONLY be used for power levels?  Yes, dB is a power measure, but so what,
the concept still applies.  BTW, the rulers are simply an example meant to
simplify the discussion for some.  Dynamic range has also been used to
characterize mechanical systems too...

Austin

----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body