On Tue, 10 Oct 2017 11:46:45 -0400, nospam
wrote:

In article , Eric Stevens
wrote:

put simply, the dynamic range in stops can't ever be more than the
number of bits in the adc.

Only if you code one stop per bit. There is no requirement that it be
a simple linear scale. If you have (say) a dynamic range of 16 stops
it is still possible to code it with 14 bits: i.e. 1.14 stops per bit.
This can be variously ignored or tidied up by the raw decoder.

that's true, except that sensors are linear devices and therefore that
does not apply.

Dynamic range is being compressed all along the way. A classic example
is HDR.

hdr is done with multiple exposures.

So?

so it doesn't count.

Of course it counts. When the subject is dynamic range compression you
can't go around excluding the product of some techniques while
excluding the product of others. The point is that any wide dynamic
range has to be compressed if the full image is to be viewed by either
screen or print. Otherwise you get burned out highs and plugged solid
shadows.

you're so lost.

I'm not lost. I'm ahead of you. :-)

With a wide range source it is compressed to enable it to be viewed on
a scren or monitor.

displays are non-linear.
sensors are linear.

With almost any source it is compressed when it is printed.

printers are non-linear.
sensors are linear.

So?

you're so lost.

Please explain why.

Why on earth should it not be compressed (a little or a lot) when it
is encoded in a raw file?

because sensors are linear devices.

And linear devices can't be compressed?

you're so lost.

No. It's just that I am not stuck with one particular way of looking
at things.

i explained this already.

feel free to design a non-linear sensor. until that time, they remain
linear.

i've said this several times. why do you ignore it?

Because it is not binding.

math and physics are as binding as it gets.

Only as binding as the axioms which lie behind them. And no axiom is
binding.

you're so lost.

And you are in a rut.
--

Regards,

Eric Stevens