Filmscanners mailing list archive (firstname.lastname@example.org)
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[filmscanners] Re: IV ED dynamic range... DYNAMIC RANGE!
This is composed into a single post because I know that this topic is
overexposed and frustrates many people. It frustrates me too, but it would
be wrong not to try to correct misinformation which is propagated with such
authority that it has succeeded in hijacking the moral and technical high
ground on this authoritative list. The purpose of this list is to allow
all of us to discuss and get a handle on exactly this kind of question. I
know that Austin has a deserved great reputation amongst list members,
partly because he is a prolific and unflagging contributor and has obvious
technical knowledge. Just the same, for whatever reasons, the view he puts
forward on dynamic range is not in accord with any textbook, paper or
standard of which I am aware, and this "definition" misleads and distorts
and confuses much consequential discussion. Worse than that, it has
succeeded in stifling a lot of the useful discussion we should be having on
a pretty basic topic because people have realised they don't understand
this most basic aspect of scanning - in large part because they are
confused by unnecessarily difficult and incorrect constructions of what
dynamic range is.
For those of you who have assumed that Austin's view is correct and
therefore not attempted to read my earlier post in any detail - I beg you,
please read what I wrote (first post headed "RE: IV ED dynamic range...
DYNAMIC RANGE!") and try to follow the logic of it, don't just assume that
any particular person has the natural authority here. I tried hard to make
this post short-ish and non-engineering. If it makes any difference to
you, I have at least as much experience with using Dynamic Range in my
career as Austin so don't make any assumptions about level of knowledge
based on presentation or style. Go and look at every definition of dynamic
range you can find for yourself on the web or in textbooks - you will not
find one which says that dynamic range is a resolution. It is a range.
Answers to many posts below:
At 23:06 08/08/02, Austin wrote:
> > DYNAMIC RANGE on the other hand, is the smaller range within the Density
> > Range that the scanner can capture AT ONE TIME i.e. dynamically i.e in one
> > scan. It is the instantaneous range the scanner can handle.
>Absolutely not correct. Where on earth did you get that? Please please
>provide any credible source that says anything to the such. The ISO spec
>doesn't define dynamic range that way...nor do any of the resources I have
Austin - in my other post you'll see that the draft ISO spec does support
my assertion. You need to get over this mental block as to what the
"dynamic" means in "dynamic range". Here is another very simple example to
illustrate the distinction between Dynamic Range and the non-dynamic kind
of range - a very simple distinction that people need to
understand. Consider a basic analog 3-range voltmeter. It has a graduated
scale, a needle, and you can switch between 3 ranges, 1v, 10V, 1000V. We
can measure on this meter from max = full scale deflection, down to a min =
the smallest graduation on the scale (let's say). The meter is divided
into 100 graduations, this is equivalent to saying the *resolution* is
1/100th of full scale. So, on the 1V range we can measure from 0,01V to
1V. On the 10V range we can measure from 0.1V to 10V. On the 1000V range
we can measure from 10V to 1000V.
The "Dynamic Range" of this meter is max/min = 100 in each case. BUT, and
here is the rub, this meter can - overall - measure voltages from 0.01V to
1000V. This is the total range or just the "range", the kind or range we
talk about without the word "dynamic" in front of it. In this example, the
"range" is 1000/0.01 = 100,000 to 1.
So for this meter:
total range = 100,000:1, and
dynamic range = 100:1
Engineers might say that total range = 100dBV and dynamic range = 40dBV.
The difference between these two figures is EXACTLY analogous to the
difference between the Dynamic Range and "Density Range" of a
scanner. Dynamic refers to "at one instant", it means the signal range of
the thing without changing it's configuration. Same in radio, same in
audio, same in signal theory, same in light.
Notice that in above example the resolution is 1/100th of full scale. You
could express this resolution if you wanted to as a number of
distinguishable levels, i.e. 100. The number of distinguishable levels
(i.e. loosely, the resolution) is the same *number* as the Dynamic
Range. But they are not the same thing! And under different assumptions
even the numbers would not be the same. More on that below.
>Absolutely incorrect. Dynamic range is absolutely NOT a "range" as you
>believe it is. It is what happens WITHIN "A" RANGE...or it would not
>contain the word "dynamic".
You are arguing against everything I have ever seen written... please save
us from going on for days on this in hopeless spirals - just post your
authoritative sources to support this.
> A range is simply a range. I'll repeat that
>again. IT IS NOT A RANGE. Dynamic means "change" basically, and it is the
>amount of changes that can be discerned WITHIN a particular range. Plain
Hmmm. That is not plain or simple, it is just unnecessarily complicating
something that IS plain and simple! Pls show supporting info.
> > The ***Density Range*** is a larger range than the dynamic range
>The two are NOT related. They do NOT represent the same thing at all, and
>therefore can NOT be compared. Anything can be represented logarithmically,
>so simply because they can be represented logarithmically does NOT mean you
>can compare them. Your claim of "larger" is like saying 100 volts is larger
>than 100 pounds.
Density range and dynamic range are absolutely related, they are both
ranges. They can be compared, and they often are. They are both measured
in dB or ratios. The dynamic range can never exceed the density range, but
the density range is often greater than the dynamic range. The dynamic
range fits somewhere inside the density range.
>You can have a density range of .1D units, and have a dynamic range of
No you can't. And don't confuse us by using different units, you can
measure both in dB, or both in D units if you want, although there is a
convention to use one for the other. If the density range is .1D (that
is 1dB) then the maximum possible dynamic range is 1dB.
>You can also have a density range of 4.0, and a dynamic range of 10
>dB. Clearly you can not say the Density Range is larger than Dynamic Range
>simply by association, as this example shows your statement to not be true.
No you show this by considering what both these quantities are, see my
example in earlier post. Or consider my voltmeter example.
> > PRACTICAL EXAMPLE
> > *************************
>A scanner that can capture 14 bits of valid data has a higher dynamic range
>than a scanner that can only capture 12 bits of valid data.
You are literally correct, but deliberately misleading. The trick is in
your use of the word "valid". This implies that the noise level is lower
than the ADC step size in both cases. Which is another way of saying that
scanner dynamic range is determined in this case only by the ADC step
size. But in general for a scanner, the dynamic range at the bottom end is
determined by either : a) number of bits in the ADC, or b) the noise
level. Whichever of these two things predominates will determine the
All your statement is saying is that the 14 bit scanner has a lower
effective noise level than the other (effective noise = max(analog noise,
This encapsulates exactly what is wrong with your idea as to what DyR is -
DyR is in fact and in general determined by the MIN DISCERNABLE SIGNAL, and
this may be determined by many different things. It MAY be determined by
number of bits, it MAY be determined by signal noise level, and if the
latter, it will be affected by how the user decides to define the min disc
signal viz a viz noise level. Noise level is not included within the
general definition of Dynamic Range for the good reason that it may not be
relevant. But it MAY be defined within a standard in order to provide a
repeatable and comparable means of measurement, if someone decides that for
the purpose under discussion noise is the predominant effect, and if they
decide how the min disc signal is related to noise in this case. The
general defintion of DyR as you read it everywhere - as it has been posted
on this list many times from many different sources - is based on the
MINIMUM DISCERNABLE SIGNAL. This is so for a reason, and that is that the
dynamic RANGE may be limited by different things in different situations.
>for dynamic range is exceptionally simple for bits. It is log10(2**n) and
>has absolutely NO bearing on the density range, they are NOT related, except
>that with a larger density range, your "resolution" or steps represent a
>larger density per unit.
This exceptionally simple equation is only applicable if it is the digital
step size that is limiting the min disc signal! You are as usual making a
big assumption which removes us from the DyR definition, and results in
confusing, incorrect pronouncements. Above a certain point (determined by
the noise level), you can increase the number of bits all you like and it
won't change the dynamic range one jot. Do you understand this?
>A 12 bit scanner with a Density Range of 4 has a resolution of 0.0009765625
>density units, and with a dynamic range of 3.6. A 12 bit scanner with a
>Density Range of 3 has a resolution of 0.000732421875, but STILL has the
>same dynamic range of 3.6.
Good grief Austin. If the scanner truly has a Density Range of 3, then its
max possible dynamic range is 3. You cannot have a dynamic range greater
than the density range it can handle. Show me a scanner with a Density
Range less than its dynamic range. Please, show me. Read the proposed
ISO standard for measuring DyR and tell me how you could possibly measure
DyR higher than the scanner's density range.
If the scanner above has 12 ***meaningful*** bits, that is they are valid,
then this means that the noise level is such that you can measure density
range of 3.6. You can't go arbitrarily changing all these quantities, they
>A 14 bit scanner with a Density Range of 4 has a resolution of
>0.000244140625 density units, and with a dynamic range of 4.2.
>Obviously, the scanner with 14 bits has a higher resolution over the same
>range than the 12 bit unit, and it also has a higher dynamic range, but the
>same density range.
>To make this REALLY simple. 14 bits has a higher dynamic range than 12
>bits, that's undisputable...as the dynamic range that can be represented by
>14 bits is 4.2, and of 12 bits is 3.6...and the last time I checked, 4.2 IS
>larger than 3.6.
ONLY if the number of bits is the limiting factor on min signal! As long
as scanner noise level is higher than the smallest bit, then the number of
bits is irrelevant. You keep making this mistake, equivalencing "number of
bits" with "number of meaningful bits". Are you saying that a scanner of
16 bits will have a DyR of 4.8 just because it has an ADC with 2 more
bits? Of course not.
David J. Littleboy said:
>Ouch. Sigh. Dynamic range:
>1. The difference, in decibels, between the overload level and the minimum
>acceptable signal level in a system or transducer.
>5. The difference between the maximum acceptable signal level and the
>minimum acceptable signal level.
>(Modern Dictionary of Electronics, 6th ed.)
David, I don't understand what you are saying in your post. I agree exactly
with these definitions you quoted here, they are like every other
definition of Dynamic Range, and unrelated to resolution, and do not
support Austin's statements.
>You are right. The max optical density of our ss120 scanner as an example is
>about 3.6~3.7. We measure this we a slide we made in house on Velvia film.
>Each step on the gray scale is .1 density units different and we look at the
>point of clipping as the maximum density.
Yes this MAY be so if there is no manual or automatic means of adjusting
gain It is not true for Nikon scanners. Your test slide is exactly the
slide I described in my simple example. (0.1 density units is the same
thing as 1dB of course)
Laurie Solomon wrote:
>The whole damn thing turns on the phrase, "acceptable signal level."
>Austin, if I read him correctly, holds that "acceptibility" is defined as
>being above the noise level at the low end; Julian probably implicitly
>accepts this also. However, he as not explitly stated as such and has not
>defined acceptibility in term of noise but presupposes it as a give in the
>notion "minimium acceptible signal." Thus, if you substitute minimum signal
>for minimum acceptible signal in the equation, you would need to duivide by
>noise to get the minimum acceptible signal.
It is much more basic than a discussion of what I see as "min discernable
signal" but this is part of it. The more basic difficulty is that Austin
sees Dynamic Range as a resolution and not a range, and this pushes every
thing off kilter. etc...
>...What others, who do not believe my understanding of
>dynamic range, believe is that dynamic range is not based on noise, but on
>the largest value of density the device can detect, minus the smallest
>density value the device can detect, basically, dMax - dMin, and there is no
>noise in the equation... This is simply the density range, and is not the
Austin, the "largest value of density the device can detect" is determined
entirely by the device's noise! That is the relationship. Of course the
dynamic range is based on noise, as is the Density Range that the scanner
can handle. (That is unless it is determined by digitisation steps, it
will depend on which is bigger). What do you think Dmax is determined by
if it is not noise? You seem to be being obtuse here, I hope not
>Again, what does "minimum acceptable signal level" mean? Is that the
>minimum measurable change in signal, or the lowest "voltage" the signal can
>attain.... What does "maximum acceptable signal level" mean? The largest
>amplitude the signal can attain, or the highest voltage the signal can
It is very simple, the "minimum acceptable signal level" is whatever you
decide is required for the signal to be usable to you. That is precisely
why the DyR definition uses the language it does. In fact if I decide that
the Min Disc Signal (to use our previous terminology) is where is signal is
*twice* the noise level, and someone else decides that their MDS is where
signal is *equal* to noise level, then we will end up with different
dynamic ranges. This is a fact, and is important to understand. That is
why standards want to define precisely how to measure the MDS in a
particular case. Eg in the proposed ISO scanner spec that I have been
quoting, they say that the Dmin (the MDS in this case) is when "Dmax =
Density where the Signal to noise ratio is 1", and then they go to some
trouble to show you how to measure Signal to Noise.
The result is a derived specification of how to measure dynamic range in
one particular circumstance for one particular purpose. It is not a
DEFINITION of dynamic range, it is a specification of how to measure it.
Peter and Austin wrote:
> > Some time ago you promised us a paper setting out your definition,
> > derivations and sources.
>I HAVE provided definitions, clear, concise definitions. I have also
>clearly provided my assertions etc. I said I would write-up something, I
>never made any PROMISE to do so, nor stated any time frame for doing so.
Please Austin, even if you don't write a paper as we hoped, please provide
the sources that support your definitions. You keep saying you have, but
you haven't yet. What Peter is asking for seems pretty reasonable given the
extent to which you promote this otherwise unsupported theory on this list.
All we need is a couple of sources saying that Dynamic Range is not a range
but a resolution, and a couple to say that the word dynamic refers to the
fact that it is happening within a range, and a couple that say that
dynamic range is defined as (something - something) / noise. Not derived
equations, but as a definition.
>Julian, if I read him correctly seeks to define the dMax-dMin range as being
>that which already accounts for the base noise on the lower dMin end and the
>full saturation level on the dMax end; thus he is implicitly accounting for
>the basic noise level without making it an explicit part of the equation.
>Thus, his definition of dynamic range is the readable or measurable range
>and for all practical purposes identical to the dMax-dMin range. The
>phrases, "...in one scan. It is the instantaneous range the scanner can
>handle," I take to be throw away phrases where he is for purposes of
>simplicity and consistency with how specs are typically derived trying to
>eliminate the possibility that multiple passes could very well alte5r and
>extend the dynamic range by lowering the basic noise level on average.
Dmax is just the film version of Min Discernable Signal. A scanner will
struggle to reproduce very high densities. If you have a slide with steps
of variable density, different levels of density will be more or less
visible above the scanner noise. It is up to the individual to decide what
is his definition of MDS for this situation, that is, how you measure
Dmax. The proposed ISO standard describes one method - obviously if you
want to compare one measurement with another you want everyone to have used
the same method. That is what these standards are for, they INTERPRET a
definition in one particular way, so as to be reproducible and useful.
As for my "in one scan" phrase, it is not throw away, it is the very
definition of what makes this measurement the DYNAMIC range and not the
density range or some other range. I have tried to give some simpler
parallel examples of this elsewhere. I know Austin does not agree with
this, but in this, he is wrong.
I am indeed eliminating the possibility of multiple passes to alter the
noise level, although this is arguable. But more especially I am
eliminating the more obvious possibility of doing the measurement of Dmax
under different conditions from measuring the Dmin. If you do this, you
get the Density Range.
David J. Littleboy wrote:
> > 1. Dynamic range is a range, not a resolution
>No, it's a ratio; a value measured in dB. As such, it implies a resolution,
>namely the number of divisions it makes sense to divide (quantize) the range
Good grief David. A ratio defines a range. 1W to 50W is a range, OK? The
ratio is 50:1. 50W is 17dB greater than 1W. The range is 17dB.
What the hell do you think Dmax - Dmin is if it is not a range? That, if
you read all my post, is the proposed ISO specification for Dynamic Range.
The more interesting point is that what dynamic range is NOT is a
resolution. Yes it (often, not always) IMPLIES a resolution. It is not a
> > 2. Dynamic range is the range that the scanner can capture
> > AT ONE TIME i.e. dynamically i.e in one scan
>No. Dynamic here is meaning 4. in Random House "4. of or pertaining to the
>range of volume of musical sound." Dynamic ranges are basically volume
>measurements; that is, a dynamic range is a ratio of volumes, loudnesses,
>or, more generally, signal amplitudes.
Dynamic range is the ratio of signal amplitudes, that much we are agreed
upon. Can you please go back and read my two examples on the difference
between Dynamic Range and Density Range, or total range, or "range". Look
at the example, a simple one, of the voltmeter above. The parallel is exact.
In the scanner world, the dynamic range translates to a single scan in
>I am trying to find my cyanide capsules... ;-)
I seem to have dropped mine, but I am struggling to fish them out of the
keyboard as fast as I can, I just want to finish this post first.
Unsubscribe by mail to email@example.com, with 'unsubscribe
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or