I wanted to transfer a tape to CD (for personal use to preserve the
tape) so ran the hard-drive recorded tracks through Audio Cleaning Lab
to take out some of the tape hiss. I find that with a conservative
application of noise removal, I get a good result without noticeable
amounts of the "alien babble" artifacts that occur with overly
aggressive application of noise removal. I find you don't have to
completely remove every last vestige of hiss. If the original track is
hot enough, at normal listening levels the noise is for all intents
and purposes gone.

However, on careful listening, I notice there's an effect on certain
high frequency sounds - brushes, cymbals and other metallic percussion
sounds, the high frequency portions of steel string guitars etc. It's
hard to describe but they just sound "funny". Sort of flat,
2-dimensional, and like some of the sound is gone, even though the
vocal and other lower frequency elements sound fine. Admittedly, a
non-musician casual listener probably wouldn't even notice it but I
can definitely hear it, especially after comparing with the original
unprocessed track.

Can *any* sound cleaning program of any price range take out noise
without affecting anything else in the sound? Or is the effect just of
a lesser magnitude on pricier programs like Cedar?

Doc,

Can *any* sound cleaning program of any price range take out noise without
affecting anything else in the sound?

I've never tried the really expensive programs or devices, but I've tried
several of the affordable ones. As far as I know they all work on the same
basic principle - a many-band noise gate with the threshold set
automatically to just above the residual noise level in each band. So I
can't see how a $5,000 program will do appreciably better than a $100
program. Since they all basically notch out extremely narrow bands, they all
give a swishy flanging effect sound when applied is substantial amounts.

That said, the instructions that come with Sonic Foundry's NR plug-in, which
is what I use, say that for minimum artifacts you should do a few passes
with less reduction each, versus one pass with a lot of reduction. I have
found this to be good advice.

--Ethan

Doc wrote:
I wanted to transfer a tape to CD (for personal use to preserve the
tape) so ran the hard-drive recorded tracks through Audio Cleaning Lab
to take out some of the tape hiss. I find that with a conservative
application of noise removal, I get a good result without noticeable
amounts of the "alien babble" artifacts that occur with overly
aggressive application of noise removal. I find you don't have to
completely remove every last vestige of hiss. If the original track is
hot enough, at normal listening levels the noise is for all intents
and purposes gone.

How did yoou do the transfer? What was the tape and what speed and how
did you make sure the azimuth was right? Your first job here is to get
the best possible playback from the original tape.

However, on careful listening, I notice there's an effect on certain
high frequency sounds - brushes, cymbals and other metallic percussion
sounds, the high frequency portions of steel string guitars etc. It's
hard to describe but they just sound "funny". Sort of flat,
2-dimensional, and like some of the sound is gone, even though the
vocal and other lower frequency elements sound fine. Admittedly, a
non-musician casual listener probably wouldn't even notice it but I
can definitely hear it, especially after comparing with the original
unprocessed track.

Well, try processing differently. Don't be as aggressive next time.

Can *any* sound cleaning program of any price range take out noise
without affecting anything else in the sound? Or is the effect just of
a lesser magnitude on pricier programs like Cedar?

CEDAR has a lot of settings. Everything will affect the sound, but the
operator has the ability to make the decisions between how much noise is
too much and how much artifacting is too much. However, Cedar does seem
to have fewer artifacts decrackling than anything else I have tried.
--scott


--
"C'est un Nagra. C'est suisse, et tres, tres precis."

On 23 Nov 2003 05:12:59 -0800, (Doc) wrote:

I wanted to transfer a tape to CD (for personal use to preserve the
tape) so ran the hard-drive recorded tracks through Audio Cleaning Lab
to take out some of the tape hiss. I find that with a conservative
application of noise removal, I get a good result without noticeable
amounts of the "alien babble" artifacts that occur with overly
aggressive application of noise removal. I find you don't have to
completely remove every last vestige of hiss. If the original track is
hot enough, at normal listening levels the noise is for all intents
and purposes gone.

However, on careful listening, I notice there's an effect on certain
high frequency sounds - brushes, cymbals and other metallic percussion
sounds, the high frequency portions of steel string guitars etc. It's
hard to describe but they just sound "funny". Sort of flat,
2-dimensional, and like some of the sound is gone, even though the
vocal and other lower frequency elements sound fine. Admittedly, a
non-musician casual listener probably wouldn't even notice it but I
can definitely hear it, especially after comparing with the original
unprocessed track.

Can *any* sound cleaning program of any price range take out noise
without affecting anything else in the sound? Or is the effect just of
a lesser magnitude on pricier programs like Cedar?

Oh yes, "Underwater Violins" as I call it ) -- I think you've gone
too far with settings. Here, less is more.

I think that there's no a "Magic Box" exists, whatever price may be
-- the magic is in the implementation. There's so many tricks one
could write (and people are writing sometimes) books about. Try
leaving the hiss at a smallest level possible without sacrifying any
program material, pay attention to sharp transients, esses, cymbals
etc. They shouldn't be "worn" or distorted more than the original.
Maybe give the hiss a more unobtrusive colour so it roughly matches
with, say, snare drum resonances or ambient noise. The worst things
are well in the audible area, often 2-3 KHz, mainly it's distorsion
and "worn" signal. You shouldn't go that low in filtering so leave it
at it is. You can cope with his at high frequencies but be careful
with that in the 4-7 KHz range. And distorsion is alas always
broadband.

Edi Zubovic, Crikvenica, Croatia

"Doc" wrote in message
om...
I wanted to transfer a tape to CD (for personal use to preserve the
tape) so ran the hard-drive recorded tracks through Audio Cleaning Lab
to take out some of the tape hiss. I find that with a conservative
application of noise removal, I get a good result without noticeable
amounts of the "alien babble" artifacts that occur with overly
aggressive application of noise removal. I find you don't have to
completely remove every last vestige of hiss.

Yes. Even a reduction of half is a lot. Don't aim for "noiseless", just make
it so it's not bothersome.


However, on careful listening, I notice there's an effect on certain
high frequency sounds - brushes, cymbals and other metallic percussion
sounds, the high frequency portions of steel string guitars etc.

Most NR software allows you to adjust how much removal you want done at each
frequency. In Sonic Foundry's NR for example you just move the nodes up or
down depending on if you want less or more reduction at those frequencies.
Just move them around so not as much removal is being done at the high
frequencies. Also, try moving the nodes for things we can't really hear
(especially that would not have been captured well on the source tape) so
that no reduction is happening at all. For example, set 12K and up so that
no reduction is being done whatsoever. (I know we can hear about 12K but you
know what I mean; no use screwing with stuff that won't matter).

"Doc" wrote in message
om...
I wanted to transfer a tape to CD (for personal use to preserve the
tape) so ran the hard-drive recorded tracks through Audio Cleaning Lab
to take out some of the tape hiss.
Snips here and below
However, on careful listening, I notice there's an effect on certain
high frequency sounds

Why not just leave out the processing completely if you prefer the sound of
the tape to the processed CD? Otherwise, try following the other good advice
you've gotten already, or cough up the money to have someone with Cedar do
the noise reduction. Unfortunately, I don't think there is any guarantee at
all that you will prefer the Cedar output to the original.

Jerry Steiger

Doc wrote:

I wanted to transfer a tape to CD (for personal use to preserve the
tape) so ran the hard-drive recorded tracks through Audio Cleaning Lab

?? ... must be some software

to take out some of the tape hiss. I find that with a conservative
application of noise removal, I get a good result without noticeable
amounts of the "alien babble" artifacts that occur with overly
aggressive application of noise removal. I find you don't have to
completely remove every last vestige of hiss. If the original track is
hot enough, at normal listening levels the noise is for all intents
and purposes gone.

However, on careful listening, I notice there's an effect on certain
high frequency sounds - brushes, cymbals and other metallic percussion
sounds, the high frequency portions of steel string guitars etc. It's
hard to describe but they just sound "funny". Sort of flat,
2-dimensional, and like some of the sound is gone, even though the
vocal and other lower frequency elements sound fine. Admittedly, a
non-musician casual listener probably wouldn't even notice it but I
can definitely hear it, especially after comparing with the original
unprocessed track.

Don't do anything if you worry about such, and if there are such
qualities then any need of noise reduction is improbable.

Can *any* sound cleaning program of any price range take out noise
without affecting anything else in the sound? Or is the effect just of
a lesser magnitude on pricier programs like Cedar?

Audition seems to do better than CoolEdit 2000 in this respect. Again
and with all respect for various even better solutions: if you worry
about the sound of a brush on a cymbal, then digital noise reduction is
in its current forms not for you. Allow me to reiterate that if such
qualities are available, then thd+noise is likely to be low and there is
thus no need for reducing anything.

Some time ago we had a noise reduction contest in this newsgroup. At a
double blind listening test in a tape recordists club I am a member of,
the untreated sound of the quite crappy 78-rarity in question won as
"overall best".

/* somewhat unscientific explanation ahead */

You have to understand what noise is, it is a zone of uncertainty within
which the actual signal value is. Noise reduction tries to optimize that
precision, usually by trading precision in time to precision in level,
but there is no way to get precision into a dataset that is not there
from the beginning. Consequently a computers guess at what a waveform
originally was sounds like "computervoice" from a sci-fi movie. Noise
reduction is just that: a guess at what the signal would have been
without the added noise, signal + noise only defines a zone within which
the signal must be.


Kind regards

Peter Larsen

--
************************************************** ***********
* My site is at: http://www.muyiovatki.dk *
************************************************** ***********

Some time ago we had a noise reduction contest in this newsgroup. At a
double blind listening test in a tape recordists club I am a member of,
the untreated sound of the quite crappy 78-rarity in question won as
"overall best".

/* somewhat unscientific explanation ahead */

You have to understand what noise is, it is a zone of uncertainty within
which the actual signal value is. Noise reduction tries to optimize that
precision, usually by trading precision in time to precision in level,
but there is no way to get precision into a dataset that is not there
from the beginning. Consequently a computers guess at what a waveform
originally was sounds like "computervoice" from a sci-fi movie. Noise
reduction is just that: a guess at what the signal would have been
without the added noise, signal + noise only defines a zone within which
the signal must be.


Kind regards

Peter Larsen


I've got that CD and none of the finished noise reduced product was all that
great.

A de popping for the big noises is about all that I do anymore.

Richard H. Kuschel
"I canna change the law of physics."-----Scotty

Doc wrote:


Can *any* sound cleaning program of any price range take out noise
without affecting anything else in the sound? Or is the effect just of
a lesser magnitude on pricier programs like Cedar?

I've not used Cedar - however I sometimes find that a simpler approach
works well. I use the Waves C1 in freqency sensitive mode to reduce the
high frequency level when things are quiet - a little like the old
Philips DNL system and probably similar to other single ended noise
reduction system.

I remember Gordon Reid from Cedar saying that they found that things
sound better if you leave a little hiss remaining.

Cheers.

James.