Forums

Sound gaps

Started by Unknown December 9, 2008
Hello,
       A while ago this group and comp.compression were very helpful
in pointing to some websites to help develop my software sound driver
(note: aside from the DAC all processing/filtering is done in
software.).

Now I have a new problem where I don't know where to look for the
answer(s) so I am begging for a little help.

After lots of work in improving sound quality I just ran into a new
problem.  It seems no matter what settings (priority) the driver is
set for BeOS has a service that at semi-random times (.1-3 seconds)
takes over a leaves a gap of non-playing (changing) output.

By reading the system_time() I can keep the sound playing properly,
but if the gaps are too large (also semi-random) you can hear it as
jiggles in the sound quality.

Note: This changes depending on the speed/caching for a computer, so
on some machines there is no problem, and others it is very clear.
Since I have access to only a very fast machine and two very slow
machines I don't know where on the power curve  the problem shows up.

I would like the driver setup to test and identify the frequency and
sizes of gaps on a machine before installation, but I have no idea at
what range the average human ear really can start to hear this
problem.  I could create test files for myself, but I have no idea how
close to average I am or how it changes with age/sex/speaker hardware.

Any clues on what to google?
Please note:  I am trying to search with Google, but all I keep
finding is complaints about hardware with gaps in their playback, out
gaps between songs when playing albums.  Clearly I am not using the
right search terms.
On Dec 10, 1:12 am, earlcolby.pottin...@sympatico.ca wrote:
> Please note: I am trying to search with Google, but all I keep > finding is complaints about hardware with gaps in their playback, out > gaps between songs when playing albums. Clearly I am not using the > right search terms.
Changed my question and found the answer: Percentage of people who can hear a 2ms gap is very close to zero (4%). 3ms is about 20% can hear it , and 4ms about 70%. Testing 5ms & 6ms about 90% can hear the gaps and 100% can hear 8ms and greater. Thanks everyone, just being able to post here cleared my thinking enough to change my search to: human ear detection gaps in sound And Google found it in the very first entry/report.
On Dec 9, 8:23�pm, earlcolby.pottin...@sympatico.ca wrote:
> On Dec 10, 1:12 am, earlcolby.pottin...@sympatico.ca wrote: > > > Please note: �I am trying to search with Google, but all I keep > > finding is complaints about hardware with gaps in their playback, out > > gaps between songs when playing albums. �Clearly I am not using the > > right search terms. > > Changed my question and found the answer: > > Percentage of people who can hear a 2ms gap is very close to zero > (4%). �3ms is about 20% can hear it , and 4ms about 70%. �Testing 5ms > & 6ms about 90% can hear the gaps and 100% can hear 8ms and greater. > > Thanks everyone, just being able to post here cleared my thinking > enough to change my search to: human ear detection gaps in sound > And Google found it in the very first entry/report.
depending on the waveforms instantaneous values of the waveform at the "splice" even a 1ms gap could still sound lick a CLICK. Mark
On Tue, 9 Dec 2008 18:46:18 -0800 (PST), Mark1 <makolber@yahoo.com>
wrote:

>On Dec 9, 8:23&#2013266080;pm, earlcolby.pottin...@sympatico.ca wrote: >> On Dec 10, 1:12 am, earlcolby.pottin...@sympatico.ca wrote: >> >> > Please note: &#2013266080;I am trying to search with Google, but all I keep >> > finding is complaints about hardware with gaps in their playback, out >> > gaps between songs when playing albums. &#2013266080;Clearly I am not using the >> > right search terms. >> >> Changed my question and found the answer: >> >> Percentage of people who can hear a 2ms gap is very close to zero >> (4%). &#2013266080;3ms is about 20% can hear it , and 4ms about 70%. &#2013266080;Testing 5ms >> & 6ms about 90% can hear the gaps and 100% can hear 8ms and greater. >> >> Thanks everyone, just being able to post here cleared my thinking >> enough to change my search to: human ear detection gaps in sound >> And Google found it in the very first entry/report. > >depending on the waveforms instantaneous values of the waveform at the >"splice" even a 1ms gap could still sound lick a CLICK.
Even a single sample misplaced at the rail during a quiet passage, or a 'gap' (zero-value sample) of one sample during the peak of a bass note, is audible as a click. To the OP, what you've apparently found through Google is something on the audibility of these gaps, but I got the impression from the first post that's not what you want, you just want to know how to stop the gaps from happening (audibility is subjective, and musicians and recording studio pros can hear many things an average listener would miss, but the 'average listener' can often be taught what to listen for). If I've misread you, my apologies, but regardless, I would want to eliminate the gaps completely, not just know when they're 'inaudible.' I'm not familiar with BeOS (isn't it essentially a dead OS, didn't the company that made it go out of business years ago?), but I'm familiar with audio recording and playback on MS Windows. For comparison, you can do recording and playback (perhaps both simultaneously but I don't recall having tried it) on a 200MHz Pentium running WinNT4 or Win98. But to do so you may have to remove a lot of processes, including all other running applications and background tasks such as 'findfast' file indexing, antivirus scanners, anything that isn't absolutely needed while doing audio recording or playback. Audio is a real-time task and Windows is NOT a real-time OS. It only works at all because the buffers in the soundcards and drivers are large enough to make up for the lack of response time in Windows, and more recent hardware is fast enough to do most tasks (even including those that mess up audio on older systems) with some breathing room. If you can control and remove other tasks in Beos, much like it's done in Windows (ctrl-alt-delete brings up a little window with tasks in 95-98-ME, or the Task Manager in the NT-XP series, in which you can delete tasks), then you can stop the offending process(es) in the slower systems.
> >Mark