naturerecordists
[Top] [All Lists]

Re: Only a question

Subject: Re: Only a question
From:
Date: Sun, 18 Jul 2004 22:14:35 +0200 (MEST)
Ok, I cannot exclude that there was some kind of saturation (the recording
was made by someone else who has both a DAT recorder and a pro-level MD
recorder). At least it was no simple digital clipping. The waveform does not
indicate that: 

http://www.avisoft-saslab.com/comression/waveform.gif

In any case, it would be an additional drawback of MD when it is impossible
to safely identify clipping. In an uncompressed DAT recording, clipping can
be clearly identified:

http://www.avisoft-saslab.com/comression/waveform2.gif

As you know, I also made some tests on synthetic signals in order to exclude
any potential clipping artifacts. These tests clearly indicate that ATRAC
may sometimes distort sounds that have large bandwidths (or rapid frequency
modulations)

Raimund




> From: 
> 
> > 
> > Hi Julie,
> > 
> > I would like to explain a different perspective. I know that the
> majority of
> > the Nature Recordist group emotionally favors the MiniDisk system and I
> > don't want to re-launch the previous passionate discussion on that
> subject
> > ;-) However, there are also a few arguments that suggest to use a
> recorder
> > that is able to record without compression (e.g. the Marantz solid state
> > recorder)
> 
> I've waited and waited, it appears there are no hordes of wild eyed 
> minidisc supporters. Just one old geezer trying to keep the facts 
> straight. Wasting his time. I'd suggest you go sit with the novice 
> recordists while I give a different perspective on your different 
> perspective.
> 
> > It is true that the potential ATRAC compression artifacts are often less
> > dramatic than expected. However, the ATRAC artifacts heavily depend on
> the
> > frequency content of the recordings. Large signal bandwidths (e.g. noisy
> > sounds or high-pitched whistles with harmonics) may push the ATRAC
> > compression algorithm to it's limits. This is often not a problem
> because
> > there is usually only little energy above about 8 kHz if you are
> recording
> > at larger distances from the sound sources (higher frequency components
> will
> > be attenuated while traveling over large distances). Things will change
> when
> > you get closer to your subject or if you use a parabolic reflector.
> > Consequently, the energy of the harmonics will rise. Unfortunately, the
> > ATRAC algorithm does not only distort these high-frequency components -
> it
> > then also degrades the more important lower frequency components. As you
> > might imagine, the opportunities for approaching a subject very close
> are
> > usually rare. So, I would be very disappointed, if my recorder failed to
> > capture those close-distance sounds appropriately.
> 
> So, if I get it right the recording of one bird calling, if recorded 
> close, will overwhelm the capabilities of ATRAC? Amusing
> 
> Then one can assume it won't have enough capability to record any more 
> species calling simultaneously. Now how is it that I and quite a few 
> other recordists do this with high quality frequently? With frogs I've 
> had a maximum of 8 species calling at a site, with sometimes thousands 
> of individuals. Quite a few of these species use multifrequency calls, 
> not just harmonics, but combination calls with multiple frequencies 
> simultaneously. Yet the sonograms of all that is just as clear with 
> minidisc as it is with DAT, maybe clearer in some cases. And I've not 
> even touched on the bird and insect species that are in those recordings 
> simultaneous with the frogs. The Georgia breeding bird study was very 
> happy to mine out night calling birds from my Georgia Herp Atlas 
> recordings. As a result I've county records for birds that I was not 
> even trying to record. Just incidental behind all the mobs of calling
> frogs.
> 
> The point is to say that ATRAC can be overwhelmed by a single call at 
> any distance is a joke. As those of us with extensive experience with 
> minidisc in real recording know.
> 
>   An example of such
> > artifacts can be seen on the following spectrogram of a MiniDisk
> recording
> > (made with a pro-level MD Recorder):
> > 
> > http://www.avisoft-saslab.com/compression/MDtest2MD.gif
> > 
> > The following spectrogram shows the same (undistorted) signal that has
> been
> > simultaneously recorded onto a DAT recorder (a CF recorder running
> without
> > compression would produce the same result):
> > 
> > http://www.avisoft-saslab.com/compression/MDtest2DAT.gif
> > 
> > One of the artifacts is a temporarily higher noise floor that can be
> much
> > higher than the self-noise level of the microphone or the recorder
> > electronics. 
> 
> Everyone reading this, if anyone is, set up these two gifs so you can do 
> a blink comparison. In netscape just put them on two tabs in the browser 
> and click back and forth between the tabs. Look at the waveform at the 
> top, not the sonogram, as you do this. Note how much fatter that 
> waveform is in the MD version. It was recorded at a higher signal 
> strength than the DAT version. Note, particularly in the second call how 
> the time period that contains the anomaly on the sonogram has peaks that 
> are hitting the 0dB cutoff, particularly at the negative end. Since this 
> waveform display is a extreme compression of all the waves involved 
> there are most certainly many more points than the averaging shows. Note 
> also that the peaks are higher on the negative side than positive.
> 
> Now what do we make of this? First off, it's not a equal test but a 
> biased one. In fact the MD was fed a high enough signal to clip at the 
> loudest points. In addition the offset toward negative indicates the 
> input signal contained a DC bias.
> 
> What will this do. Well, the A/D clipped. Then that flawed digital 
> signal was fed to ATRAC, which tried to faithfully reproduce those 
> flaws. A clip introduces random changes in the instantaneous changes in 
> sound levels, ie the slope of the waveform. It is this which results in 
> producing random signals all across the frequency spectrum as it's the 
> slope that gives the frequency. It really did not matter that the 
> damaged digital went through ATRAC, although it was changed by it. The 
> D/A at the other end would have produced messed up sound even without
> ATRAC.
> 
> Now, remember that DC offset? Well, it's also a nice way to introduce 
> extra noise in a recording. Digital equipment is designed for signals 
> that are symmetric around the zero line, and will produce noise from the 
> DC offset. It's why good sound editors have the ability to recalculate 
> the samples to remove the offset.
> 
> The conclusion I get from these samples is that the recordist committed 
> a novice mistake. The gain going into a recorder is something we set. 
> It's very common for novices to set that gain too high. Because they 
> don't hear overt clipping they think it's ok. But, this fine type 
> clipping does exactly as Raimund said, it sounds like a self noise 
> problem from mic or pre. And it's not a error of the compression.
> 
> We have had this discussion many times, the metering on recorders is 
> still that designed for the more forgiving analog tape. It's not good 
> enough to accurately run a digital recorder. It gives us a hint, but we 
> have to learn the rest by using the recorder and checking what happens. 
> With time and experience with a recorder we will be able to generally 
> avoid clipping by setting a offset. I typically use about 15 dB of pad, 
> in other words the high indications on the meter are down about 15 dB 
> >from top. Some may consider this excessive, but even at this level I do 
> still get clipping sometimes.
> 
> > The overall quality of a recording system is always limited by the
> weakest
> > component within the entire signal chain. In certain recording
> situations
> > (at close distances), the ATRAC compression system might become the
> weakest
> > part (instead of the microphone or the preamplifier).
> > 
> > This is of course my personal opinion...
> 
> I had thought better of your technical abilities. You lowered my 
> estimate of your ability with this novice error. Try to become familiar 
> with the equipment you are testing before testing.
> 
> Please, everyone note, I'm not saying ATRAC compression does not change 
> the digital stream. In fact I'm saying very little about it because the 
> case here does not pertain to it. Every single recorder there is does 
> not fully reproduce the signal it's given and that includes minidisc. 
> I'm only saying that on balance a quality minidisc will do a excellent 
> job of recording. In fact, in some ways, by more intelligent allocation 
> of memory ATRAC has advantages over uncompressed. That comes from years 
> of experience using one and I'm not the first one to have that opinion.
> 
> And, if you get closer, lower the gain appropriately to avoid clipping. 
> It is actually the point of getting closer, to lower the gain so less of 
> the background noise is picked up. It's not a argument against ATRAC 
> that you can force a minidisc to clip, you can force any digital 
> recorder to clip. It has a long history in anti-minidisc, it's Cornell's 
> main argument to this day. And is as wrong from them as here.
> 
> I'm a little ashamed of the group that none of the technically adept in 
> the group commented on this obvious error. Of course maybe everyone 
> assumed it was so obvious it did not need to be pointed out.
> 
> Walt
> 
> 
> 
> 
> 
> "Microphones are not ears,
> Loudspeakers are not birds,
> A listening room is not nature."
> Klas Strandberg 
> Yahoo! Groups Links
> 
> 
> 
>  
> 



________________________________________________________________________
________________________________________________________________________

<Prev in Thread] Current Thread [Next in Thread>
Admin

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU