Hi Evert,
Yes, I'm afraid too, that not all of the artifacts shown in this
spectroram are caused by ATRAC. Therefore, it would be good if
someone else could repeat that test with a different SPDIF interface.
http://www.avisoft.de/compression/compressed441MD.gif
What I could imagine is, that the additional frequency-shifted parts
at about 20 kHz (e.g. at t=3D0.57 ... 0.69) are a result of some other
signal processing in your system (the Kenwood recorder has a sample
rate converter at its digital input too). However, because I have
some experience with sample-rate conversion algorithms (I
implemented a nearly distortion-free re-sampling feature in Avisoft-
SASLab Pro), I know that simpler algorithms (interpolating only
linearly) would add overtones (multiples of the original signal)
only. But what we see here are no overtones. It is some kind of a
copy (shifting) of the original signal to a higher frequency. Such
effects can not be caused by straightforward distortion as we now
that from analog signal processing. Also, it would be difficult to
explain the short 'invented' sine signals at 19 kHz (t=3D0.26 ...
0.31). A poor re-sampling algorithm would produce a more steady
signal. Thes effects we see here must have been caused by some kind
of more 'intelligent' signal processing. So I have the strong
feeling, that these effects are also produced by ATRAC. At least,
the extremely noise at the steep frequency slopes (e.g at t=3D1.27 or
t=3D3.33) are clearly caused by ATRAC. We have seen similar (but
weaker) effects in the MP3 example. It should also be noted, that
the original constant sine signal at 17 kHz (absolute amplitude -38
dB) disappears completely at more complex passages (t=3D0.85 ...0.87).
The constant sine signal at about 10 kHz is not much affected in
this example. However, it should be noted, that I used a very fair
spectrographic display with low gain and low contrast (because such
displays are often used in bioacoustics). Nevertheless, you will
find some distortion also to the 10 kHz signal if you look closer (I
found amplitude drops of up to 8 dB at the worst-case situation).
Regards,
Raimund
> Hi Raimund,
>
> I am not sure if it is all ATRAC, I do think the SoundBlaster Live
I
> have isn't that great a soundcard. Because when I am transferring
> from my MD deck to my SPDIF input, I have to put the inputlevel to
59
> %
> In other words, the data is changed in real time ; the volume
level
> is reduced with 41 %
>
> Not a bit-to-bit transfer... Stupid Soundblaster Live :(
>
> A question from me, can you indicate roughly at which frequencys
the
> ATRAC is 'lost' in following the original? Are we talking above 15
> kHz for instance (i.e. not audible by my ears) or are we talking
> above 10 kHz or something like that?
>
> Regards, Evert
>
> --- In "Raimund Specht"
> <> wrote:
> > Evert just conducted another test (many thanks to him for doing
> > that). This time he used the original full-bandwidth test file
at a
> > sample rate of 44.1 kHz (no resampling at all), recorded it
> > digitally onto his KENNWOOD DM3020 MD recorder (that device has
a
> > fixed sample rate of 44.1 kHz) and played it back via SPDIF. The
> > spectrogram created from this sound file still shows some
dramatic
> > artifacts:
> >
> > http://www.avisoft.de/compression/compressed441MD.gif
> >
> > So I guess, that this is the normal behavior of ATRAC 4.5.
> >
> > The low-pass filtered test file which were recorded and played
back
> > in the same way as the above example is still acceptable:
> >
> > http://www.avisoft.de/compression/compression.htm.
> >
> > If I interpret these results correctly (and if no other faults
were
> > involved), then ATRAC 4.5 (like MP3) would do a good job as long
as
> > no major high-frequency energy is present in a recording. In
more
> > extreme situations as the more demanding full-bandwidth test
> signal,
> > ATRAC 4.5 will reach its limitations and will clearly lose
against
> > MP3.
> >
> > Regards,
> > Raimund
________________________________________________________________________
________________________________________________________________________
|
