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: "Stephan M. Bernsee" <email@hidden>
- Date: Thu, 28 Aug 2008 17:07:05 +0200
Hi all,
I think that this debate might be due to a misunderstanding because to
me it looks like you're all basically saying the same thing but are
actually thinking of different scenarios.
As it has been stated by Brian and Richard, the DFT requires windowing
if you are looking at a signal with a period that is not an integer
multiple of the period of the sinusoids that are used to calculate the
Fourier transform. Why? The DFT looks at the signal that you are
analysing as being a stationary, periodic signal with a period equal
to the DFT size, which is to say that it sees the signal as repeating
outside the analyzed interval, forever. Clearly, for most signals this
is not the case, and for signals that do not fall into this category
it will "see" high frequencies due to discontinuities that are not
there in reality since the signal snippet that you are analysing is
not actually repeating. Take a sinusoid and repeat a small snippet of
it and you will immediately hear these frequencies yourself. You know
that they're not part of your signal and you have no interest in
measuring them.
You could even say that they are an artifact introduced by the DFT due
to its very (circular) nature. So, in oder for the DFT to still be
useful in a realword context you need windowing, because most signals
are not stationary, periodic signals with a period equal to the DFT
size and you will be introducing noise into your mesurement that is
not actually there.
There is one case where this scenario doesn't apply and where
windowing is actually harmful and this is when the period of your
measuring signal is exactly an integer multiple of the DFT transform
size. In this special case I would have to agree with Mikael that
windowing will pollute the measurement and you will lose resolution.
So it all depends on how the DFT is being used if windowing is
actually desired, or not. In my opinion it is necessary to include
information on how these measurements were obtained. IOW: if you're
not using a window you should mention this so people could interpret
your graph in the right context.
Also, to go back to the original application, taking these
measurements to judge the quality of an audio algorithm may be
misleading depending on the nature of the algorithm. A time stretching
algorithm may have a perfect frequency response but may still sound
horrible to human ears due to transient smearing and other factors
that will not show up in a DFT graph.
Kind regards,
Stephan
--
Stephan M. Bernsee
CTO, The DSP Dimension
http://www.dspdimension.com
_______________________________________________
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