RSS feed reader

98 Guy wrote:
Bill in Co wrote:

Just out of curiosity, why are you still using FF 2.0? You can
run FF 3.5 with KernelEx

I tried running several versions of FF 3 about 1.5 to 2 years ago and
found that when scrolling a web-page up and down that a white line would
appear across any bit-mapped images on the page where they were cut off
by the frame before being scrolled up or down.

I think this problem is described in the following links:

https://bugzilla.mozilla.org/show_bug.cgi?id=477236

http://www.zen-cart.com/showthread.p...ot-in-Explorer

https://bugzilla.mozilla.org/show_bug.cgi?id=201198

http://www.neebar.com/2008/05/mozill...izontal-lines/

Some people have mentioned seeing it in the win-98 forum on msfn as
well.

But honestly, I find that FF 2.0.0.20 works VERY WELL on 99% of the web
sites that I browse to.

Also regarding FF 3.x, I recall problems with the menu icons doing
strange things (disappearing, turning into noise, colors of the menu
bars becoming inverted). This was back in early 2011 and I've since
changed the video card in this computer (I had Nvidia MX440, now have
Nvidia 6200) and maybe that change combined with the newer version of
KernelEx it might make a difference - but trying FF 3 is not high on my
priority list right now...

OK, but you might give it consideration (ver 3.5.x, not ver 3.6, and above,
which have the java issue), now that you've changed things around.

Plus I expect you'll find it renders some web pages better, and hopefully
with fewer of those annoying "your browser is out of date" warnings (and
some "deficient" web page renderings).

"Bill in Co" wrote in
m:

As you said, you might have to wiggle lines by
lots of repeated trial and errors in the TD to get it to "sound better".

Two or three is nearly always enough. Usually just the one. All it's doing is
adding low level noise in a rough simulation of the surrounding noise so the
brain doesn't register discontinuity. It IS a kind of FSE, it's just
converted from frequency to time domain by hand. Remember that on the 1-pixel
zoom scale, we can SEE signals that are in the 10KHz+ range, but it's usually
enough to manage up to 2K or so to give enough info for a brain to fill in
based on the longer-term signal. You know as I do, that if the noise floor
appears unchanged, any remaining small discontinuity is perceived as part of
the original signal.

I imagine this goes into psychoacoustics, rather that FSE in terms if actual
process, but either way I will stay with simple time domain tools because
they work for me. The only way they'll ever stop working is if my hearing can
no longer judge them, in which case I won't be able to judge results of FSE
editing. I realised a long time ago that I needed to find methods that
worked, and stay with them so I used skill and practise to maintain high
quality and consistency. If I do things the same way I know it's having the
same kinds of effect on the sound, whether I can hear it or not. THis is true
in general, but there is a very specific help too, eyesight. I can SEE time
domain. I cannot see frequency domain, because the plots are 2 dimensional
and complex. As you say, we cannot edit. We have to trust the machine. I
won't, when I know I can trust my own brain and body.

Lostgallifreyan wrote in
:

As you said, you might have to wiggle lines by
lots of repeated trial and errors in the TD to get it to "sound
better".

Two or three is nearly always enough. Usually just the one. All it's
doing is adding low level noise in a rough simulation of the surrounding
noise so the brain doesn't register discontinuity.

Actually I'm often wiggling a detail into the signal that isn't noise, it may
be a crude emulation of some strong harmonic. Doesn't matter what exactly,
the point is that if it strong enough for me to see in time domain, it is
strong enough to psychoacoustically mask other absences in frequency domain!
Neat, no? It's amazing how much wiggle room we get there.

Lostgallifreyan wrote:
Lostgallifreyan wrote in
:

As you said, you might have to wiggle lines by
lots of repeated trial and errors in the TD to get it to "sound
better".

Two or three is nearly always enough. Usually just the one. All it's
doing is adding low level noise in a rough simulation of the surrounding
noise so the brain doesn't register discontinuity.

Actually I'm often wiggling a detail into the signal that isn't noise, it
may
be a crude emulation of some strong harmonic. Doesn't matter what exactly,
the point is that if it strong enough for me to see in time domain, it is
strong enough to psychoacoustically mask other absences in frequency
domain!
Neat, no? It's amazing how much wiggle room we get there.

But isn't it possible that you may not be able to see everything clearly
enough in the time domain OR know how to reconstruct it well (like for
larger discontinuities)? In which case, the FSE method could potentially
do a better job (due to the computer algorithms in a FSE editor being able
to analye the immediately adjacent material)?

By larger T-D "discontinuities", I'm thinking of perhaps 50-100 ms duration
or more. (NOT just 10 or 20 ms. :-)

From my experience, it seems in several cases I've been able to see things
better in the FD, at least for a quite a few of these glitches. But not
always; sometimes the TD view has proven more useful. But with a good
audio editor, you can switch back and forth rapidly between the two views
(even when zoomed in) which is really helpful. The one I've found best in
that regard is Adobe Audition 1.5 (descendant of Cool Edit Pro), which
introduced the FSE capability. But unfortunately, it won't install on
Win98SE - even with KernelEx. Cool Edit Pro will, but lacked FSE. The
newer versions of Adobe Audition suck (they dumbed down the interface). :-)

"Bill in Co" wrote in
:

By larger T-D "discontinuities", I'm thinking of perhaps 50-100 ms
duration or more. (NOT just 10 or 20 ms. :-)


Even 20ms is a lot, that's 882 samples at CD quality. No single click or pop
will be that long. Bursts of interference from AC sources with bad
suppression might be (and often long enough to destroy all hope). When I find
gaps long enough to need full replacement I don't think of audio at all, but
of music, as it's usually music I'm working with. It's usually possible to
know from the rhythm when to drop a marker at beginning and end of the
smallest temporal subdivision affected by the problem, then fine tune it by
eye and testing by ear. As music is big on repetition as a structural device,
there's nearly always some chunk somewhere nearby, similar enough to provide
a graft. I've rarely found a break so bad I had to do this, but whenever I
have, it's worked.

If you're dealing with much more 100ms, likely all bets are off. By then you
can miss whole beats at most tempos. At gap sizes less than that, it may even
be acceptable to close the gap and not fill it, IF the music isn't
metromically accurate in tempo to start with. Voices and solo instrument
recordings are especially easy to fix that way, and doing it creatively has
often been the basic technique of radio comedy. Reporters do it too, if they
screw up they just read into the mic again, exactly as if nothing had gone
wrong, and the resulting similarity makes it easy to cut and splice.

In short, it's usually easier to replace witha patch or graft from real
content, than to try to synthsize it, which is what FSE fill-in is. Either
that, or it's just doing what we can already do anyway.

"Bill in Co" wrote in
:

But with a good
audio editor, you can switch back and forth rapidly between the two views
(even when zoomed in) which is really helpful.

That is true, but as with light, we can train ourselves to do it without
tools. Once you've seen the spectrogams of various lamps (a CD is a good
cheap spectrogram if you play with viewing angles a bit.) Once you get
familiar with the kinds of spectra associated with various types of light
source, you can get a mental image for a non-standard light source, to enough
extent to sense where the gap is, or how narrowband a light source is when
you look at it. I find that the fewer tools, the better, so long as SOMEHOW
we get that impression needed to make the right call.

Lostgallifreyan wrote:
"Bill in Co" wrote in
:

By larger T-D "discontinuities", I'm thinking of perhaps 50-100 ms
duration or more. (NOT just 10 or 20 ms. :-)


Even 20ms is a lot, that's 882 samples at CD quality. No single click or
pop
will be that long. Bursts of interference from AC sources with bad
suppression might be (and often long enough to destroy all hope). When I
find
gaps long enough to need full replacement I don't think of audio at all,
but
of music, as it's usually music I'm working with. It's usually possible to
know from the rhythm when to drop a marker at beginning and end of the
smallest temporal subdivision affected by the problem, then fine tune it
by
eye and testing by ear. As music is big on repetition as a structural
device,
there's nearly always some chunk somewhere nearby, similar enough to
provide
a graft. I've rarely found a break so bad I had to do this, but whenever I
have, it's worked.

If you're dealing with much more 100ms, likely all bets are off. By then
you
can miss whole beats at most tempos. At gap sizes less than that, it may
even
be acceptable to close the gap and not fill it, IF the music isn't
metromically accurate in tempo to start with. Voices and solo instrument
recordings are especially easy to fix that way, and doing it creatively
has
often been the basic technique of radio comedy. Reporters do it too, if
they
screw up they just read into the mic again, exactly as if nothing had gone
wrong, and the resulting similarity makes it easy to cut and splice.

In short, it's usually easier to replace witha patch or graft from real
content, than to try to synthsize it, which is what FSE fill-in is. Either
that, or it's just doing what we can already do anyway.

You're right - I shouldn't have called it a "click" (my mistake).

These "glitches" are often more complicated than simply being a click. And
in such cases, I've often been pleasantly surprised at just how well the FSE
can fill it in. No, it's not perfect, but nothing can be, and it sure is a
LOT better afterwards!

See, I often don't have patches or grafts available with some of this
material, in which case the FSE works pretty darn well. And manually
patching in these wider glitch cases from adjacent regions is too much of a
blind hit and miss proposition in such cases - the computer algorithms can
do better than me in these cases by analyzing the adjacent material and
making appropriate compensations.

"Bill in Co" wrote in
m:

See, I often don't have patches or grafts available with some of this
material, in which case the FSE works pretty darn well. And manually
patching in these wider glitch cases from adjacent regions is too much
of a blind hit and miss proposition in such cases - the computer
algorithms can do better than me in these cases by analyzing the
adjacent material and making appropriate compensations.



Ok, but I have to weigh the need to change a whole system to get a tool that
I haven't needed so far. Unless I find a problem that is harder to solve that
changing tools and methods, AND there is also certainty that the new system
will solve the problem, the incentive to change, while not infinitessimal, is
very small. Even 'basic' FFT or DCT, while being vital to translate from time
domain in many processes, is rarely something I need to see. Actually, when I
do look at the Sound Forge (FFT-based) spectrum analyser, it is usually a
very crude way to tell me something I can already sense by ear (so long as
it's within hearing range). The ability to move in time domain is limited
only by sample-to-pixel accuracy, so the limit is also that of the digital
medium itself. My brain has to make the conversions I need for some edits,
but that's ok so long as it's getting all the info it can get. FSE is like a
complex image of the world that the machine can use, but we can't. It makes
us into pilots who have no choice but to trust their instruments. In the
specific situation that system is built for, it will perform better than the
unaided pilot, but ONLY in that situation. This is why I prefer to use
methods closer to what we evolved with, they're vastly more adaptable.

Lostgallifreyan wrote:
"Bill in Co" wrote in
m:

See, I often don't have patches or grafts available with some of this
material, in which case the FSE works pretty darn well. And manually
patching in these wider glitch cases from adjacent regions is too much
of a blind hit and miss proposition in such cases - the computer
algorithms can do better than me in these cases by analyzing the
adjacent material and making appropriate compensations.



Ok, but I have to weigh the need to change a whole system to get a tool
that
I haven't needed so far. Unless I find a problem that is harder to solve
that
changing tools and methods, AND there is also certainty that the new
system
will solve the problem, the incentive to change, while not infinitessimal,
is
very small. Even 'basic' FFT or DCT, while being vital to translate from
time
domain in many processes, is rarely something I need to see. Actually,
when
I do look at the Sound Forge (FFT-based) spectrum analyser, it is usually
a
very crude way to tell me something I can already sense by ear (so long as
it's within hearing range). The ability to move in time domain is limited
only by sample-to-pixel accuracy, so the limit is also that of the digital
medium itself. My brain has to make the conversions I need for some edits,
but that's ok so long as it's getting all the info it can get. FSE is like
a
complex image of the world that the machine can use, but we can't. It
makes
us into pilots who have no choice but to trust their instruments. In the
specific situation that system is built for, it will perform better than
the
unaided pilot, but ONLY in that situation. This is why I prefer to use
methods closer to what we evolved with, they're vastly more adaptable.

I wasn't sure what you meant by "having to change a whole system", unless
you were referring to your mindset and not the computer? :-) But you can
still find some tools with FSE or glitch patching capabilities that will
still run on Win98SE, but granted, the choices are more limited.

I'm not sure how many of these tools still run on Win98SE, but I know some
do. Programs such as: Wave Corrector Pro, Wave Repair, Vinyl Studio, and
at least some older versions of Magix Audio Cleaning Lab, just for starters.
(one can often find older versions on Amazon or on eBay, etc). But there
are pros and cons to each of these. Incidentally, I think the first three
are British in origin. :-)

"Bill in Co" wrote in
m:

I'm not sure how many of these tools still run on Win98SE, but I know
some do. Programs such as: Wave Corrector Pro, Wave Repair, Vinyl
Studio, and at least some older versions of Magix Audio Cleaning Lab,
just for starters. (one can often find older versions on Amazon or on
eBay, etc). But there are pros and cons to each of these.
Incidentally, I think the first three are British in origin. :-)




Once a computer has been paid for, modified, built to solve certain tasks,
fitted with hardware that depends on the OS running on it, there's no useful
distinction between it and the mind of its operator. Most of the reason for
bloat and ever faster machines to do the same basic tasks is precisely
because people want to change computers like they change shoes, always
afraid to get deep into what they have, instead preferring to take every new
promise instead of building their own life no matter how small and futile it
might seem to 'limit' themselves that way. Not so long ago, most people
couldn't even change shoes that readily either. As most of our serious
problems come from the same things that bring us so many 'solutions', I
prefer to be cautious and use what I have if it's a choice between that and
giving most of it up. We've had this discussion several time, and the one
thing I won't do is turn over my machines, or way of thinking, for one task
at the expense of all the others. It's not going to happen (and likely
wouldn't even if I had a few thousand bucks to throw at the problem because
there are other things I'd want to do more). I have explained in extensive
detail how and why I manage, in ways many professional sound engineers don't
know so I think I've stated my case well enough. Just let me live with my
choices. That doesn't hurt, but going round in the same circles again and
again despite explaining very good reasons not to, CAN hurt, so let it rest.