In article , John Phillips
wrote:
I often see texts that imply or state outright that ringing in
time-domain audio waveforms, usually associated with time-domain
discontinuities, is to be avoided. I wonder what is the basis for this.

The argument seems to be primarily about 'pre' ringing. See below.

For example, if you take a perfect square wave with all of its
harmonics, and mathematically remove all of those above a certain point
you will see a waveform with ringing. I think this is referred to as
Gibbs' Phenomenon.

The distinction I think people have latched onto is that if you use one of
the familiar types of analogue LP filter, then any ringing tends to occur
after the transitions of the square-wave. This tends to be interpreted as a
'causal' requirement. That the filter can't react to the transition until
it occurs. Although this interpretation is rather shakey since it is quite
possible for the entire pattern to have been delayed in transit though the
filter, so any 'pre' transition ringing may be entirely causal - even with
an analogue system.

However far from being a problem, assuming you do have to band-limit a
signal, the presence of ringing (in this case, anyway) seems to show:

- perfect removal of all frequencies above a certain point; and

- perfectly preserved amplitude and phase relationships amongst the
remaining frequency components.

Well, ringing visible on something like a scope can be removed if you allow
the slope/shape of the LP filter to be fairly gradual. See, for example,
the work by Peter Craven in JAES. The results can remain time symmetric,
but the cost is a fairly gentle filter that then either cuts the HF inband
and/or passes rubbish out of band.

I note though that Meridian have now used a filter that gets all the
ringing post transitions. Suspect this was based on Peter Craven's work.
The downside it a tendency to phase spread, I assume.

As long as a filter perserves all human-audible frequencies I cannot see
an objection (to this form of ringing, at least). The only argument I
can think of for objecting is the possibility that the non-linear
behaviour of the ear may result in specific audibility issues which
wouldn't be heard with fully linear hearing. Indeed, I do see articles
that cast doubt on the audibility of ringing. For example
http://www.stereophile.com/reference/106ringing/.

See also the articles on this on audiomisc in the 'hearing' section which
also considers hearing nonlinearity in this context.

FWIW I am less than convinced ringing is audible in time symmetric filters.
It seems to be another of the popular ideas in audio for which I've not
personally found any reliable evidence. But given hearing nonlinearities,
it might be audible. So I would not totally dismiss the idea from what I
know at present.

Slainte,

Jim

--
Change 'noise' to 'jcgl' 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