🚀 go-pugleaf

"Ulrich Grassberger" <grasso@my-deja.com> wrote in message
news:3b0542da.6177497@News.CIS.DFN.DE...
> >Mr. Grasso you are way ahead of yourself. You've presumed that the
> >ringing is audible, but you've neither cited a relevant well-known
> >psychoacoustic authority nor have you provided reliable evidence
> >based on a listening test. Anybody can make unproven assertions
and
> >many people have.

> Calm down!

Say what? Mr. Grasso, you started this pie-throwing party with the
following gratuitous insult:

"Yeah, he (Arny) seems to mess up some things some times. If the
results are
strange he doesn�t bother to investigate further, to investigate if
HE
HIMSELF might have made a mistake instead of the soundcard being
bad."

> I gave a possible psycho-acoustic explanation, based on
> facts, on my own listening experience, and on the opinion of some
> audio researchers.

Mr. Grasso you made the following very specific claim:

"So you use a linear-phase filter (in the samples at
http://www.pcabx.com/technical/low_pass/index.htm)   which leads to
symmetrical ringing."

So far so good.

Then you said:

"Now, the pre-ringing can be detected by our ears"

I've said I'm perfectly willing to accept that if you can show a
psychoacoustic text to support your claim, or provide results from a
DBT proving you've reliably heard the ringing. Seems fair enough, no?

Then you said:

" If you used a filter latter than 100dB/oct but with a lower limit
frequency and different coefficients you might be able to reduce
pre-ringing. That means that a lower limit frequency can sound
better."

At this point I'm beginning to think that you've missed the point.
The point is that other than people who've been listening via
monitoring systems with lots of high frequency nonlinear distortion,
*nobody* can hear the difference these filters make until the corner
frequency is below 16 KHz.

There is BTW a well-documented psychoacoustic reason for this which
I've mentioned on several newsgroups several times and I don't think
there is any informed controversy over. Above about 12 KHz human
adults have one critical band. It tends to be captured by the lowest
frequency and strongest sounds in its frequency range. The nature of
musical program material is that there is plenty of content below 16
KHz to do that. Since the critical band is dominated by the content
below 16 KHz, the presence or non-presence of content above 16 KHz is
not perceived.


>I explain it at grassomusic.de/english/digital.htm#96khz in full
length.

Lets take that at face value.

http://www.grassomusic.de/english/digital.htm#96khz  (BTW folks
that's what a proper URL looks like...)

Mr. Grasso writes:

"Besides, if I�m already at it, 96KHz is a sensible thing: Not
because there was important musical information at 20KHz, or that
most people could hear that, no, a tweeter with an upper limit
frequency of 15KHz is just fine."

The thesis is controversial but the fact presented is pretty close to
how things are.

"The reason is that: If aliasing will occur not until 48KHz the
lowpass filters can be made much simpler and softer. The way in which
signals over 20KHz are cut off in today A/Ds, D/As and resampling
filters is brute: Linear-phase filters lead to pre-ringing, an effect
which is absolutely impossible in the analogue world (in the nature
physics that our ears are made for), that is alien stuff. "

Pre-ringing seems to happen with musical instrument sounds, not just
test signals. Musical instrument sounds in the last audible octave
come from lightly-damped resonators that take a decent number of
cycles to build up their resonance, but the still seem to be able to
stimulate pre-ringing, albeit at pretty low levels.

Anybody who wants to can download this file:
http://www.pcabx-pro.com/technical/reference/triangle-2_2496.wav and
take a look at it with a 24/96 wave file editor. This file was
created using a pair of B&K 4007 measurement mics, an Audio/Technica
48volt phantom power supply, a pair of Benchmark Media MP2 mic pre (r
esponse to 200 KHz), and a CardDeluxe running at 24/96.  There is
about 4 milliseconds of background noise before the striking of steel
on steel.  The left channel peaks at 0.00375 seconds from the
beginning of the file and the right channel peaks about 0.00390
seconds. OK the triangle was not perfectly centered between the two
mics.

OK, now lets downsample this file to 1644 with some nice brutal
brickwall filtering. That file is called
http://www.pcabx-pro.com/technical/reference/triangle-2_1644ds.wav  .
If you compare the first 4 milliseconds of the two files you can see
that all sorts pre-ringing is taking place in the downsampled files.

OK so here is the challenge. Prove the  pre-ringing is audible by
just listening to the two files, no peeking!

I don't think anybody is going to walk way with the prize on this
one. The reason is clear to anybody who understands backwards
temporal masking.  Here's a reference:
http://www.umiacs.umd.edu/~desin/Speech1/node11.html

Moving right along...

"Interpolating from phase detection in 3D perception, the quantity of
that effect can be detected, and actually hearing tests prove it."

What hearing tests?

"Analogue tape machines are very sensitive for high amplitudes at
high frequencies, the dynamic range drops down dead if 20KHz is to be
recorded with -20dB on a cassette recorder."

That would be pretty irrelevant to 15 ips or 30 ips master tapes,
wouldn't it?

"Therefore, when CDs were invented twenty years ago the minimum
filter steepness was standardized with 50dB fall in a few half-tones.
(...approximately - I don�t have the papers handy now.)"

Yes, brick wall filters have been around for quite a while. Aliasing
is pretty nasty sounding stuff. I don't think anybody intended to
idealize CD so that transcriptions could be done from cassette tapes,
though.

"In CD players of the first generation filtering was done with
analogue elliptic 9th order filters, leading to very very long
post-ringing, difficult to calibrate and unstable over time and
temperature."

OK.

"Pre-ringing was plainly accepted by audio industry because
oversampling technology diminishes the total amount of ringing (looks
much nicer in the oscilloscope) and eliminates stability problems,
and it�s a rather easy thing: Most times FIR filters are harder to
build than IIR filters, but oversampling is an exception, just right
for cheap silicon chips."

True, but why avoid pre-ringing if the pre-ringing is not heard?

"To show you another point on that field: The HF modulation frequency
in analogue tape machines is higher than 80KHz, simply because it
needs to be that high for a smooth upper end."

The major benefit of high bias is that it minimizes intermodulation
products that might otherwise fall in the normal audible range, not
smoother high frequency response.

>The digital audio engineers in 1980 should have calculated the same
way, but the urge to tweak playing times was overwhelming.

Or maybe someone knew about critical bands...

>And yet another point: In a frequency splitter as needed for
multi-way loudspeakers linear-phase filters are great because both
signals become summed again later, but if the bandwidth ends at the
filter frequency you get pre-ringing.

Still not addressing the audibility issue.

>The issue about electromagnetical wave reflections on audio cables
due to lacking impedance fitting is High-End Voodoo, self-induced
("induced", ha ha ha!) imagination, at least with normal cable
lengths (and even if it does strike it�s still less audible than
pre-ringing), but 96KHz is not.

How about a rewrite that makes a little more sense?

>I don�t have any references handy now because as I
> stroll thru life I don�t always have in mind that I might need to
> legitimate my opinions to people like you.

The place to keep your references would be your web pages, no Mr.
Grasso?

If you check out the "techtalk" pages at www.pcavtech.com you'll find
references for many of the things I thought might be controversial.