In article , Arny
Krueger
wrote:


We've been in a similar place around here just lately. People have
lately gone on record as demanding sufficient headroom to play back
fairly improbable but rarely observed signals whose samples are all
less than FS but in combination can trick a reconstruction filter into
creating an analog signal that significantly exceeds FS.

The distinction between that and clicks on an LP is that there is no need
to 'trick' the reconstruction filter and nor does it require the disc to be
physically damaged as in the case of 'clicks' from an LP. The excursions
above 0dBFS arise because the recorded waveforms include them. They are
part of the recorded waveform, not a 'trick' or the result of physical
damage.

So the word 'trick' in this context is misleading and inappropriate, and
the LP and CD cases are not similar in cause. One behaviour arises due to
physical damage and makes the required waveform unrecoverable. The other
can be a result of the required waveform defined by the series of samples
recorded on a correctly produced and physically undamaged disc. The
required waveform may still be defined by the series of samples even if a
given DAC can't cope with reproducing that.

The behaviour (requirements for some peaks above 0dBFS) arises as a
consequence of the Sampling Theorem and basic Information Theory. The
series of samples correctly define the waveform to be reconstructed, as per
the Sampling Theorem. Although some test waveforms show extreme examples
of this, required excursions above 0dBFS can occur for commercial
recordings of music. This is not due to any damage equivalent to an LP
being scratched, but a part of the data on the CD. if you check some
pop/rock CDs you can probably find examples without trying very hard.

I'm not buying either situation as being one that needs to be pandered
to.

The main way I "don't buy it" is to avoid buying rock/pop CDs that have
been affected. :-) Alas, you may not know until you have one of the CDs.
So I (and I assume many others) do have CDs where the provided samples
require the reconstructed waveform to exceed 0dBFS between some samples.
What is then unclear is what individual CD players make of this when they
try to reconstruct the waveforms from the data.

Since this is being xposted to a group I don't normally read it may be
helpful for readers there if I give a couple of URLs that examine these
matters in case they haven't encountered this issue previously.

http://www.audiomisc.co.uk/HFN/Clipp.../clipping.html

Shows some examples of one way this can occur with real commercial CDs

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

Examines the effect in detail. Examples show around +2dBFS for commercial
CD, and then an extreme case test waveform.

The important points here a

1) That the effect examined on CD arises a consequence of using sampled
data. So is an aspect of digital information methods for purposes like
recording audio waveforms.

2) That it can arise as a result of what was recorded, not because of a
situation like an LP being scratched.

3) If the DAC and reconstruction filter are correctly designed they would
reproduce the defined waveform correctly, including the parts above 0dBFS,
whereas a scratched LP would be reproducing the effects of damage.

Note also that this has been discussed in AES papers, etc. And I know that
the BBC (sound radio) people tend to keep down the max sample levels on R3,
etc for similar reasons.

FWIW I'd agree that this *should* be a non-issue in practice, *if* CD
makers ensured they kept down the recorded sample levels to avoid
intersample peaks above 0dBFS and/or if DAC makers ensured they played
waveforms above 0dBFS without distortion. However I'm far from sure that
both these provisos are universal. Certainly I know of both CDs and DACs
that fail to meet these conditions.

Given the large dynamic range of 16bit CD it would be easy for pop/rock to
avoid peaks above 0dBFS. But alas the pop industry obsession with
'loudness' seems to drive them to the maximum they can manage.

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html