gibson@unc.UUCP (Bill Gibson) (12/09/84)
Comments and questions on dbx breathing :
Works I have noticed it on : Fresh Aire solo piano pieces (Cathedrals, I think)
I would guess that this problem is mostly noticeable when program material is
quiet or soft. For instance, in the piano piece mentioned, a solo piano is
played softly at some points, and the piano's harmonic content is small enough
so that the music does not mask the breathing noise. Note that breathing noise
can be masked by program material just as all tape hiss can be masked. The
attempt that dbx makes is to allow *all* program material to mask underlying
hiss by amplifying the (hiss+music) only when the music amplitude is high
enough to mask hiss. Therefore, it seems to me that any breathing noise which
is noticed is due to the program material 'mask' ('mask' as in a CPU's
interrupt mask, but made of the music's spectrum combined with its overall
amplitude) not matching the spectrum of tape hiss.
I don't have any data on what the frequency plots look like for
1) tape hiss spectrum on the same plot as
2) plot of the level of signal it would take to mask the tape hiss spectrum.
I would suspect that 2. would not be a simple addition of (n)dB to 1., or
anything else so simple, but that 2. would also depend on an ear's sensitivity
plots. Note - 2. would probably be different for any two given ears (e.g. my
left and my right, or my left and your left). Difficult to design for.
As has been mentioned, the multi-band type of dbx compression/expansion
would allow the program material to be encoded (recorded) in such a way that
the best 'mask' would be used on playback (i.e. only bands with signal would
be expanded, so frequency bands containing little signal and much hiss wouldn't
expand very much -> lowest possible noise). The more frequency bands the better.
If the professional dbx system uses only two bands, where is the split
frequency? Is it at the average frequency at which hiss is highest for all the
different kinds of tape?
Works I haven't noticed breathing on : Chick Corea, _My_Spanish_Heart_.(Spanish
flavored jazz; most of the program material has a pretty full spectrum,
what with trumpets, strings, drums, & forceful piano). There are some quiet
sections, but I can't hear any breathing noise there.
I think this program material is conducive to noise-masking, since the
spectrum is so full that the tape hiss range is probably occupied by music
much of the time. But I wouldn't say that the dbx is a waste on this tape,
since there are sections which are probably quiet enough not to mask the
noise normally.
Also, I have dbx-encoded Art Tatum (piano jazz) and Tangerine Dream and
Jean-Michel Jarre (both electronic-type music) which I don't hear the breathing
on.
I think the kind of the music being encoded/decoded with dbx has plenty
to do with the acceptability of the result. For instance, quiet solo piano has
a low register (for many pieces) and therefore doesn't mask the hiss band well
when expanded; however, I suspect that solo violin, which has lots of highs,
would acceptably mask hiss when expanded. Also, music which includes lots
of dynamic level changes is a prime candidate for 'heavy breathing'. Lots of
orchestral works have a pretty constant amplitude, so they shouldn't breathe.
As for rock music, any pieces with a strong beat which is accented by drums
probably has the dynamics mentioned above and would allow noise except for
the fact that most rock seems to have a full spectrum (guitars, keyboards,
vocals) which could easily mask noise. Lots of rock has a high enough average
amplitude that it probably masks hiss well enough without dbx. It depends how
sensitive you (and your stereo) are to noise in the tape hiss region.
Non sequiter:
One application of dbx that I've never seen mentioned : portable monophonic
tape players. Since dbx compresses all the sound toward a single amplitude,
the dynamic range (of the encoded signal) is squashed. So a 80-dB dynamic range
record produces maybe a 40-dB dynamic range encoded tape. If you play back
the tape without decoding it (i.e. no dbx), you get a signal that a cheap
tape player can amplify without overloading. Also, if you listen in a limited-
dynamic-range environment (read car), you can hear all the music, including
the soft parts. Of course, you have to equalize for it to be acceptable since
the encoded signal has strong high-frequency. (I think this is because if the
encoded signal is (original - (1/n)original), the encoded power is
(original - (1/n^2)original) and so the high frequencies keep a lot more
power in the encoded version than do low frequencies ( since power~freq.)
Does anybody know about this?)
Bill Gibson ...Waiting for writable CD's.
gibson@unc ...[akgua,decvax,philabs]!mcnc!unc!gibson