G
Guest
Guest
Archived from groups: rec.audio.pro (More info?)
"TonyP" <TonyP@optus.net.com.au> wrote in message
news:4118a8e8$0$6942$afc38c87@news.optusnet.com.au
> "Scott Dorsey" <kludge@panix.com> wrote in message
> news:cf80hm$7r9$1@panix2.panix.com...
>> As I pointed out to Phil, it's entirely possible to get insanely low
>> top octave THD values with standard measurement procedures because
>> all of the harmonics being generated are outside of the passband of
>> the measuring device.
Agreed. The third harmonic of 7.4 KHz and up, as well as the second
harmonic of 11.1 KHz and up get lost in the reconstruction filter, if not
the anti-aliasing filter.
> This is a very common way for manufacturers to cook the numbers.
Agreed, but people who want to accurately characterize nonlinearities can
uncook the books in the same context with multitone measurements.
One of the most flexible tests I know of is what I call an empty well test.
One drives the UUT with a multitone (or music) that has groups of missing
tones or filter notches that create empty wells. The wells will start
filling up in the presence of a wide variety of different kinds of
nonlinearities. Nevertheless, not all kinds of nonlinear distortion will add
significantly to the wells - see below.
> Who, apart from you, mentioned only measuring the 10k-20k band. It is
> very selective to choose ONE octave and ignore the other NINE. Treble
> starts below 10kHz in my book.
Focus on the top octave is, IME not about misapprehensions.
(1) Audio gear tends to be more nonlinear as frequency goes up - in the
analog domain open loop gain goes down, and digital filters are more prone
to clipping near the stop band.
(2) Over the years musical program material seems to have tended towards
stronger content in the top octave(s).
(3) Because of Fletcher-Munson effects and masking, signals in the top
octave are less likely to be heard. This means that they are less likely to
distract listeners from any cross-modulation effects they stimulate when it
intermodulates down into the lower octaves.
When I talk about reliably hearing 0.1 % nonlinear distortion, I'm
specifically thinking of natural sounds with lots of energy in the top
octave, and less, little or no energy in the range where the ear is most
sensitive - 2-6 KHz. If there are common kinds of audible nonlinearity in
the upper octave, it will be heard with this test. This is especially true
of clipping - relatively tiny amounts of clipping can be heard under these
conditions.
Since I'm flogging Doppler distortion in another thread another group, I'm
compelled to point out that all of the *wonderful* tests I've described are
pretty much blind to Doppler distortion. The good news is that while Doppler
distortion is real and present in loudspeakers, it's also almost always
quite a bit less than the AM distortion that generally dumps quite a bit of
energy at exactly the same frequencies. This AM distortion is generally
heard and measured with the listening and technical tests I've been talking
about.
"TonyP" <TonyP@optus.net.com.au> wrote in message
news:4118a8e8$0$6942$afc38c87@news.optusnet.com.au
> "Scott Dorsey" <kludge@panix.com> wrote in message
> news:cf80hm$7r9$1@panix2.panix.com...
>> As I pointed out to Phil, it's entirely possible to get insanely low
>> top octave THD values with standard measurement procedures because
>> all of the harmonics being generated are outside of the passband of
>> the measuring device.
Agreed. The third harmonic of 7.4 KHz and up, as well as the second
harmonic of 11.1 KHz and up get lost in the reconstruction filter, if not
the anti-aliasing filter.
> This is a very common way for manufacturers to cook the numbers.
Agreed, but people who want to accurately characterize nonlinearities can
uncook the books in the same context with multitone measurements.
One of the most flexible tests I know of is what I call an empty well test.
One drives the UUT with a multitone (or music) that has groups of missing
tones or filter notches that create empty wells. The wells will start
filling up in the presence of a wide variety of different kinds of
nonlinearities. Nevertheless, not all kinds of nonlinear distortion will add
significantly to the wells - see below.
> Who, apart from you, mentioned only measuring the 10k-20k band. It is
> very selective to choose ONE octave and ignore the other NINE. Treble
> starts below 10kHz in my book.
Focus on the top octave is, IME not about misapprehensions.
(1) Audio gear tends to be more nonlinear as frequency goes up - in the
analog domain open loop gain goes down, and digital filters are more prone
to clipping near the stop band.
(2) Over the years musical program material seems to have tended towards
stronger content in the top octave(s).
(3) Because of Fletcher-Munson effects and masking, signals in the top
octave are less likely to be heard. This means that they are less likely to
distract listeners from any cross-modulation effects they stimulate when it
intermodulates down into the lower octaves.
When I talk about reliably hearing 0.1 % nonlinear distortion, I'm
specifically thinking of natural sounds with lots of energy in the top
octave, and less, little or no energy in the range where the ear is most
sensitive - 2-6 KHz. If there are common kinds of audible nonlinearity in
the upper octave, it will be heard with this test. This is especially true
of clipping - relatively tiny amounts of clipping can be heard under these
conditions.
Since I'm flogging Doppler distortion in another thread another group, I'm
compelled to point out that all of the *wonderful* tests I've described are
pretty much blind to Doppler distortion. The good news is that while Doppler
distortion is real and present in loudspeakers, it's also almost always
quite a bit less than the AM distortion that generally dumps quite a bit of
energy at exactly the same frequencies. This AM distortion is generally
heard and measured with the listening and technical tests I've been talking
about.