Re: OT: Rock and Chips



On 10 Feb, 21:04, Kennedy McEwen <r...@xxxxxxxxxxxxxxxxxx> wrote:
In article
<0224a205-9eb3-4f5b-8ac8-d1bf0f9ab...@xxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"davidrobin...@xxxxxxxxxxxxxxxx" <davidrobin...@xxxxxxxxxxxxxxxx> writes



You've contradicted yourself - "20x less blurred", but "not
significantly different from the 100% shutter".

No contradiction at all - read what I did: 1kHz frames at 100% shutter
downsampled to 50Hz by sinc-based weighting.  That means each of the
adjacent *contiguous* blurred* images in the downsampled image *is* 20x
less blurred than a single 50Hz 100% shutter image would be. 20x less is
certainly not "slightly less blurred" as you put it.

That's what I assumed you meant. See below.

 Stop being a
complete twat!

So your ultimate arguing technique is to revert to insults - good -
glad we confirmed that. It'll save time later.


Other than that, I don't disagree. You can use a non-sinc filter that
still satisfies Nyquist by removing everything above fs/2, and yes
this implies some temporal overlap. It's _not_ "ideal" because you've
sacrificed the top of the available bandwidth (just below fs/2) to
reduce the ringing - but this is a good compromise.

It isn't a case of "making them overlap with a non-sinc filter" - they
*must* overlap, since the weighted average extends much further than the
20 samples required to get a 50fps frame from 1kHz.  More precisely, if
the movement is constant speed then the peak in one frame *must* fall
exactly on the first null of the next, that is what sinc (and anything
derived from it) filtering ensures.

Yes, I know, it was me who first mentioned sinc filtering, remember?

So now _again_ in your first paragraph you say the bluring is 20x
_less_ than a 100% shutter, and in your last paragraph you say the
bluring is _more_ than a 100% filter - make your mind up!

And yet you claim you're not contradicting yourself. If you're not,
you're making a pigs ear of explaining yourself.


 If you knew what you were talking
about rather that just arguing for the sake of it then that simple fact
would have been obvious to you.

That part (the temporal extent of ideal sinc filtering) was _always_
obvious to me.


I've been disagreeing (all along!) about the terms you've misused -
"ideal" being the one that's left here. Ideal sampling has a very
specific meaning - you need a sinc filter.

Wrong - as I said, I ignored that error because I expected you would
argue about that as well due to your ignorance.  Once again, read
Shannon's original work (or any of the repetitions of it by numerous
authors.  Understand the difference between amplitude and intensity or
voltage and power.  Nothing is darker than no photons, at least in this
universe - who knows what its like in yours.  In this universe sinc
weighting of any signal transduced linearly from intensity is the wrong
weighting to obtain a flat brick-wall filter.

The fundamental papers I have, e.g...
http://www.stanford.edu/class/ee104/shannonpaper.pdf
....don't cover this.

Do you have something more recent in mind?

Cheers,
David.
.