Re: Test report MBP built-in audio device
Re: Test report MBP built-in audio device
- Subject: Re: Test report MBP built-in audio device
- From: Richard Dobson <email@hidden>
- Date: Tue, 26 Aug 2008 14:27:44 +0100
Mikael Hakman wrote:
He understands this very well since long, long time back because he is
working with digital signal processing for decades, long before there
were digital audio and Mac computers around. Digital signal processing
had its applications long before anybody used it for audio. These are
hardly new topics; some were known and used for centuries, some others
for decades, even if they may appear new to people in digital audio
because digital audio application itself is relatively new. Used
methodology isn't.
If you look at the next page entitled "windowing II" you'll see a signal
reminiscent of a guitar or piano note decaying (top picture). After
windowing function you get the signal depicted on the next row. Which
one do you listen to when you listen to a guitar? Then, which one should
be analysed?
??? But you are not analysing a signal, you are measuring (or claiming
to be) a ~system~ (ADC-something-DAC loop), which we all hope behaves as
a(n almost) perfect LTI system. OF COURSE if the goal is to get the most
accurate and detailed analysis of a finite ~signal~ such as a guitar
note, many methods are appropriate (not least - partial-tracking using
overlapping windows over time, as in everything from SNDAN (which uses
classic MQ analysis) to SMS, ATS, Lemur, etc, to say nothing of
wavelets, neural nets etc. Even "hetro" in Csound can do a very good job
with things like that). You are presenting the exercise as measuring an
LTI system to obtain a frequency response, and noise and distortion
figures. The techniques for doing have also been known for a long time.
It is the difference between hifi listening tests where golden-eared
people gush about the musicality of a cable (where of course the
direction of the oxygen-free connection and the phase of the moon matter
immensely), and an engineering analysis, which will use dull things like
square waves, sine waves, white and pink noise, and yes, the good old
windowed DFT.
In short, to measure an LTI system, you don't inject guitar notes, you
inject time-invariant signals such as sinusoids (with precisely known
levels of harmonic distortion), square waves, pink noise - all signals
with very precisely known properties invariant over time. The square
wave will reveal much about DC, bandwidth, phase and transient response
(even a simple visual inspection of the output can reveal much about the
behaviour of the system).
There is a place for musical listening tests; but however "warm" a
filter may be declared to be, "warm" may mean different things to each
reader. The one thing it is not is an objective scientific measure.
In any case, all that page is explaining it that the centre of the
window should ideally be aligned with the transient peak. This is
entirely conventional, as by mathematical convention the centre of a
window is located at t=0 (because that is how the DFT is defined as well
, between +-PI). Of course this is impractical in a real-time streaming
system, as it involves looking into the future. You could equally use
those diagrams as illustrations of the action of a compressor; which is
of course not an LTI process.
There are indeed known problems employing the DFT when analysing
time-varying signals such as transients (and even the chirp signal). The
trick in analysing a system is not to use time-varying signals!
Richard Dobson
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden