Over The Top
 

In article , Arny
Krueger
wrote:
"Serge Auckland" wrote in message



Consequently, we never experienced this sort of problem, and it was
then acceptable for digital players to clip at just over 0dBFS, albeit
for the wrong reason.

Thing is, these FS overs are very rare.

Depends on the context. I have found various commercial CDs that generate
excursions above 0dBFS. In some case, on many occasions during the CD.
However these are almost all rock/pop music CDs. Much rarer with classical
or jazz in my experience. The difficulty here is checking enough CD issues
to results that have any statistical meaning, though.

As Jim says, this might account for why some CD players sound
different to others.

I doubt that. First, you have to find the good CD players that actually
do sound different from others.

At present that is the unanswered question which would need testing in this
context. There are CDs that generate the excursions, and others don't. I
have also see test results which show that different players can give
visibly different results on something like a scope when given 0dBFS
material. But have no evidence that this leads to audible differences.

There have been comparisons where the players seemed indistinguishable.
That might mean this doesn't matter. But because of the above we'd now
really need to check what CDs were used as source material, and if they
produced any 0dBFS excursions or not. And if they did, the effects might
vary from case to case, which makes this difficult to track down, except in
terms of a generalisation like that is 'usually' has no audible effect, or
whatever...

FWIW I don't yet feel any urge to change my DACs despite having some reason
to suspect they can't cope with the possible excursions without some amount
of waveform distortion. Simple reason being that they produce fine results
to my ears, and I've not yet noticed any problem when listening.

But we could null out all such concerns if the bods at the commercial music
companies simply avoided clipping the CD material and kept the level down
so that no 0dFBS excursions could be required on replay.

There is a supreme irony here that all CD players *would* sound the
same if CDs were mastered properly. Ah well....

Since I generally record with lots of headroom, the problem of FS
overs takes on a different form. I end up with recordings that have
very infrequent excursions that are 3-4 dB above the norm. If they
weren't there, I could up the average level of the recording by 3-4 dB
without clipping. Since my distributed recordings are typicaly
uncompressed, I'm already producing recordings that sound far softer
than most commercial releases.

I find much the same when doing home recordings, but it does not matter as
I am not trying to get anyone else to listen to the results. So can just
turn up the replay volume control a few steps. But as you say, this does
tend to make even more obvious just how loud many commercial CDs are!

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html

Over The Top
 

In article ,
Jim Lesurf wrote:
It's interesting (at least to me!) that although this problem has
existed for as long as digital audio has, in the early days of digital,
we always left a few dBs headroom, even as much as 10 dB if the levels
couldn't be precisely controlled. (1)

FWIW when recording anything from an analogue source onto CDR I always
use a peak headroom of over 6dB - typically more like 10dB. Until
recently I wasn't sure how much was needed for dealing with this
specific problem, but I would now certainly recommend avoiding allowing
the peak samples above about -6dBFS. 10dB seems wise as an aiming point
to allow for operator error, etc.

10dB over peak is still used by broadcasters. Standard line up tone which
is 8dB below peak is set at -18dBFS on video recorders, etc. And used in
the main for lining up things like FreeView and DAB transmitters. Several
of the radio stations on there peak to 0 dbFS, though. Leave any headroom
anywhere and someone will exploit it. ;-(

The reason for this 'headroom' is easily explained. The standard measuring
device in the UK for broadcast is the PPM - a device invented before WW2.
It reads peaks - but not absolute ones. It can't react fast enough to
measure them all. Of course in this day and age you can make one which did
- but then it wouldn't be such a useful tool for those balancing actual
sound. In analogue days these fast edged peaks didn't matter much - the
tape or whatever just absorbed them without audible effect. But digital
systems without a brick wall limiter will make *very* nasty noises if over
modulated.

--
*In "Casablanca", Humphrey Bogart never said "Play it again, Sam" *

Dave Plowman London SW
To e-mail, change noise into sound.

Over The Top
 

"Jim Lesurf" wrote in message

In article
, Arny
Krueger
wrote:
"Serge Auckland" wrote in
message


Consequently, we never experienced this sort of
problem, and it was then acceptable for digital players
to clip at just over 0dBFS, albeit for the wrong reason.

Thing is, these FS overs are very rare.

Depends on the context. I have found various commercial
CDs that generate excursions above 0dBFS. In some case,
on many occasions during the CD.

Many = 10s, 100s, thousands,???

However these are almost
all rock/pop music CDs. Much rarer with classical or jazz
in my experience. The difficulty here is checking enough
CD issues to results that have any statistical meaning,
though.

I know how to make them, and I know how to keep them from happening.

As Jim says, this might account for why some CD players
sound different to others.

I doubt that. First, you have to find the good CD
players that actually do sound different from others.

At present that is the unanswered question which would
need testing in this context. There are CDs that generate
the excursions, and others don't. I have also see test
results which show that different players can give
visibly different results on something like a scope when
given 0dBFS material. But have no evidence that this
leads to audible differences.

I can explain why it is highly unlikely that they will.

There have been comparisons where the players seemed
indistinguishable. That might mean this doesn't matter.
But because of the above we'd now really need to check
what CDs were used as source material, and if they
produced any 0dBFS excursions or not. And if they did,
the effects might vary from case to case, which makes
this difficult to track down, except in terms of a
generalisation like that is 'usually' has no audible
effect, or whatever...

It's just a matter of knowing what's audible, and what is not.

Over The Top
 

In article , Arny
Krueger
wrote:
"Jim Lesurf" wrote in message

In article , Arny
Krueger wrote:
"Serge Auckland" wrote in message



Consequently, we never experienced this sort of problem, and it was
then acceptable for digital players to clip at just over 0dBFS,
albeit for the wrong reason.

Thing is, these FS overs are very rare.

Depends on the context. I have found various commercial CDs that
generate excursions above 0dBFS. In some case, on many occasions
during the CD.

Many = 10s, 100s, thousands,???

Varies. IIRC The largest number I found was one track on a 'remastered'
sic Hendrix CD where there were there were thousands of samples in
flat-top-runs of clipping. (See the webpage on this on the audiomisc.couk
website which gives more info.) In effect, the waveforms though most of the
track were clipped.

Hard to do reliable statistics forCDs in general as I only had time to
choose a couple of dozen test tracks as examples and see if they showed
clipping. Some did not. But IIRC overall I found a number of tracks of
pop/rock which showed multiple clipping events involving hundreds of
samples. This problem was almost absent from classical or jazz CDs.


However these are almost all rock/pop music CDs. Much rarer with
classical or jazz in my experience. The difficulty here is checking
enough CD issues to results that have any statistical meaning, though.

I know how to make them, and I know how to keep them from happening.

Yes. I see no reason to doubt that those who produce commercial rock/pop
CDs *could* easily avoid this problem if they so chose. Hence my view that
it is the result of either incompetence or a cynical delusion on their part
that the buyers won't notice or care, and that their blind faith in
'loudness sells' trumps all considerations of sound quality. However I
can't know if this is ignorance on their part, or a cynical view that it
doesn't matter. You may have noticed that I have had disagreements with
Iain about this. :-)

I do not accept the view that the fact that people don't refuse to buy any
of the resulting CDs, or return them all, shows people are 'happy' with the
result. Just that they are given no comparison or choice, and not told that
what they are being sold is 'damaged goods'. Also that most people (in the
UK at least) feel it is a waste of effort to try and complain to a record
company.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html

Over The Top
 

"Jim Lesurf"

I have now produced a new webpage which looks at the way signals from CD,
can exceed the nominal 0dBFS level. The page is at

http://www.audiomisc.co.uk/HFN/OverTheTop/OTT.html

The material is an expanded form of what was previously published in the
August 2007 issue of Hi Fi News. It shows some examples of the waveforms
reconstructed from commercial CDs exceeding 0dBFS, and also shows a test
waveform that should (if reproduced without distortion) reach to above
+5dBFS.


** No exotic test waveform is needed to show the differences in peak
output capability between the various DtoA reconstruction filters. Just a
white noise signal that has been digitally generated by a random number
program ( M - sequence dither) does it fine. The Denon " Audio Technical
CD " has such a track (track 79).

When played on a variety of machines, the peak level as seen on a CRO varies
from the expected 2.8 volts on most modern CD and DVD players, up to 4 volts
peak on ones that use analogue reconstruction filters - like the Sony CDP
101.

Arny might like to include this interesting test in his pages.

I doubt it has the slightest audible effect on music program.



....... Phil

Over The Top
 

"Jim Lesurf" wrote in message

In article
, Arny
Krueger
wrote:
"Jim Lesurf" wrote in message

In article
, Arny
Krueger wrote:
"Serge Auckland" wrote
in message


Consequently, we never experienced this sort of
problem, and it was then acceptable for digital
players to clip at just over 0dBFS, albeit for the
wrong reason.

Thing is, these FS overs are very rare.

Depends on the context. I have found various commercial
CDs that generate excursions above 0dBFS. In some case,
on many occasions during the CD.

Many = 10s, 100s, thousands,???

Varies. IIRC The largest number I found was one track on
a 'remastered' sic Hendrix CD where there were there
were thousands of samples in flat-top-runs of clipping.

But how do you know that they weren't simply too-loud, as opposed to being a
consequence of only samples at or below FS?

I'm speaking about just the latter kind of waveform that goes over FS even
though all the samples that make it up that ever existed are at or below FS.

Over The Top
 

In article , Arny
Krueger
wrote:
"Jim Lesurf" wrote in message


Varies. IIRC The largest number I found was one track on a
'remastered' sic Hendrix CD where there were there were thousands of
samples in flat-top-runs of clipping.

But how do you know that they weren't simply too-loud, as opposed to
being a consequence of only samples at or below FS?

Two reasons.

1) In many of the cases I examined I have more than one version (different
CD issues and/or LP) and the stats of their dynamics clearly differ with
the clipped version showing clear signs of clipping. e.g. a secondary peak
at 0dBFS. In itself this is not conclusive. But when combined with...

2) The data shows runs of identical (or in some cases near identical)
values at each of the clipping points. The chances of this occurring
without clipping are statistically remote. Given the other issues of the
same source material without the effect it seems clear it wasn't part of
the source waveform.

Visual inspection (cf below) of the waveforms also tends to confirm the
presence of flat-top clipping.

FWIW I also found examples of 'soft clipping; where the waveform on the CD
can be seen to squash up against a max level, but where the sample runs
were similar, not identical. So in some cases you can identify runs which
seem to have been flat-topped, and others where the waveform has been 'soft
clipped'.

Have a look at the Henrix and 1812 examples I show as plots. They look
like clear cases of flat-top clipping to me. Afraid I didn't have the space
or time to do all the examples I found. If you examine the waveforms
the diagnosis seems fairly clear to me. However if you have other evidence
I'd be interested to see the details.

I'm speaking about just the latter kind of waveform that goes over FS
even though all the samples that make it up that ever existed are at or
below FS.

If you look at my HFN articles on the audiomisc website you can see that I
have dealt with three distinct problems in different articles. One is
flat-top clipping, another is excessive level compression, and the third is
where sets of samples indicate out of range intersample values.

However if you doubt what I report, then please do a similar analysis of a
range of pop/rock CDs of recent manufacture and report your results. I'd
genuinely be interested in seeing what you (and anyone else) gets as I was
only able to do a small number of examples for lack of time, etc. And the
more people pay attention to this, the more chance we have of getting the
producers to cease clipping the CDs. As I indicated previously, I can't
say how many CDs have the effect to a given extent as I was unable to
examine a large enough selection to get results that would have
statistical reliability. So can only say that I found it easy to
find examples from trying just a small number. I'd welcome more
data from a bigger range of CDs.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html