I've been wrestling with cleaning up an old tape and have been trying
various approaches. The recording is of a jazz combo - Trumpet, electric
guitar, piano, drums, bass and was made on an old mono tape recorder many
moons ago. The primary software tools I have a

Soundforge 5.0
WaveRepair
DartPro32

I also have Magix Audio Cleaning Lab but have found this to come up somewhat
short compared to the other software and haven't used it in a while.

Running sound through an Audigy2 on a PC.

DartPro and WaveRepair both have noiseprint based noise reduction,
Soundforge has Graphic, Parametric and Paragraphic eq.

Something I've noticed is that running the same settings for ex. on the
graphic eq vs parametric (say centered on 11k -5db .5 octave) don't seem to
yield the same results. Is there a fundamental difference in the way they
work?

I initially tried recording the untouched tape and tweaking it inside the
computer, but I seem to get better results boosting certain elements of the
sound - the cymbals, the trumpet, by running the sound through a 10-band
graphic eq before it hits the soundcard. I've got it to where you can hear a
surprising amount of sound, such as the shimmer of the cymbals. The problem
is, of course this enhances the tape noise even more.

Problems I run into using the noiseprint solution is that if you apply it
lightly, when the music peaks, you hear a marked emergence of hiss, if you
crank it up to where this goes away, it takes out a lot of high end detail.

So far, I actually seem to get the best results using either parametric or
paragraphic eq in Soundforge, but I still seem to lose some of the shimmer
in the cymbals. I've been playing around in the vicinity of 10k to 15k which
seems to be where most of the more obnoxious hiss lives, but apparently so
do the cymbals. Is there any way to have my cake and eat it too, kill the
hiss without dulling that nice cymbal sound, or is that a compromise I'm
going to have to live with?

Any other strategies or suggestions?

Many thanks.

First and foremost, have you tweaked the playback head azimuth for maximum
treble output?

Peace,
Paul

there's also a trick which might or might not work ... some "aural
exciter" processing can take one band of frequencies, and use it to
create higher harmonics of the same sound, or something like that.
Aural Excitation is a subtle form of distortion, so you might be able
to reduce the hiss, and then put back some of those frequencies derived
from the rest of the sound.

Others on this list might be able to give more details of how this
works, and how to model the relevant type of distortion in other
software.

Chris

On Sun, 10 Apr 2005 02:18:17 GMT, Doc wrote:

DartPro and WaveRepair both have noiseprint based noise reduction,
Soundforge has Graphic, Parametric and Paragraphic eq.

Something I've noticed is that running the same settings for ex. on the
graphic eq vs parametric (say centered on 11k -5db .5 octave) don't seem to
yield the same results. Is there a fundamental difference in the way they
work?

There is a fundamental difference: Procedures that allow you to take a
noise print generate some statistics of the noise signal and are later
able to adapt the filtering based on both the statistics and the
volume of the signal that is to be cleaned. Thus the filtering
constantly changes with the volume of the recording - hopefully in the
same way as the ear adapts to the sound of the instruments or to the
noise.

IOW, a good noise reduction program removes more hiss when the volume
is low and - if setup properly - doesn't change the recording when the
volume is high as noise is covered by the music.

I use Adobe Audition for noise reduction. It has several parameters in
the plugin to tweak the procedure depending on the volume, the instru-
ments used in the recording and more. It required some try and error
and sometimes leaves you still with unpleasant results, such as
artefacts from the filtering.

Norbert

Doc wrote:
I've been wrestling with cleaning up an old tape and have been trying
various approaches. The recording is of a jazz combo - Trumpet, electric
guitar, piano, drums, bass and was made on an old mono tape recorder many
moons ago. The primary software tools I have a

Soundforge 5.0
WaveRepair
DartPro32

I also have Magix Audio Cleaning Lab but have found this to come up somewhat
short compared to the other software and haven't used it in a while.

What about the tape machine? How did you get the original transcription?
Was the playback done on a machine with the exact same head configuration?
Was the azimuth set to match the original tape?

Getting a good transcription is the hard part. If you can get that, noise
is much less of a worry.

Running sound through an Audigy2 on a PC.

DartPro and WaveRepair both have noiseprint based noise reduction,
Soundforge has Graphic, Parametric and Paragraphic eq.

Something I've noticed is that running the same settings for ex. on the
graphic eq vs parametric (say centered on 11k -5db .5 octave) don't seem to
yield the same results. Is there a fundamental difference in the way they
work?

Well, the graphic shouldn't allow you to set the Q and you're stuck having
the filter on a fixed frequency, right?

I initially tried recording the untouched tape and tweaking it inside the
computer, but I seem to get better results boosting certain elements of the
sound - the cymbals, the trumpet, by running the sound through a 10-band
graphic eq before it hits the soundcard. I've got it to where you can hear a
surprising amount of sound, such as the shimmer of the cymbals. The problem
is, of course this enhances the tape noise even more.

You will get a lot more from cutting than boosting in most cases. And you
will find the outboard graphic is probably _not_ a good solution for
anything. If you're having to boost the top end to hear the shimmer on the
cymbals, something is probably wrong with your transcription job.

Problems I run into using the noiseprint solution is that if you apply it
lightly, when the music peaks, you hear a marked emergence of hiss, if you
crank it up to where this goes away, it takes out a lot of high end detail.

Right. Doing it in dozens and dozens of light passes can help, but NR is
not in any way a magic cure-all.

So far, I actually seem to get the best results using either parametric or
paragraphic eq in Soundforge, but I still seem to lose some of the shimmer
in the cymbals. I've been playing around in the vicinity of 10k to 15k which
seems to be where most of the more obnoxious hiss lives, but apparently so
do the cymbals. Is there any way to have my cake and eat it too, kill the
hiss without dulling that nice cymbal sound, or is that a compromise I'm
going to have to live with?

This is one of the places where the Aural Exciter can actually be useful,
for salvaging old recordings with top end problems. Before doing anything
else, though, make sure your transcription job is good. Azimuth errors are
really easy to get, especially with a full-track mono tape.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

wrote in message
ps.com...
there's also a trick which might or might not work ... some "aural
exciter" processing can take one band of frequencies, and use it to
create higher harmonics of the same sound, or something like that.
Aural Excitation is a subtle form of distortion, so you might be able
to reduce the hiss, and then put back some of those frequencies derived
from the rest of the sound.

Others on this list might be able to give more details of how this
works, and how to model the relevant type of distortion in other
software.

Hmm.. Actually, the Magix Audio Cleaning Lab has an aural exciter function.
That's not what they call it but it does what you outline. I don't know how
well it works in comparison to other programs. I never found a real use for
it. I played with it a little, found it seemed to add a "funny" sound to the
high end. Maybe I wasn't applying it properly.

Doc wrote:

Something I've noticed is that running the same settings for ex. on
the graphic eq vs parametric (say centered on 11k -5db .5 octave)
don't seem to yield the same results. Is there a fundamental
difference in the way they work?

Taking what you say literally, it seems you're asking whether dialing
in say a 4 dB lift at 11KHz on a 1/2 octave graphic eq should sound
the same as dialing in a 4 dB lift at 11 KHz with 1/2 octave bandwidth
on a parametric eq.

A mjaor potential discrepancy relates to the fact that different
graphic eqs implement equalization in different ways. This results in
different frequency response curves for the same front-panel setting.
One example of this kind of discrepancy is shown in this paper:

http://www.rane.com/note101.html

Compare figure 3 with figure 4 to see how two different graphic eqs
can have vastly different frequency response curves, given the same
nominal characteristics (number of bands, range of adjustment) and
front panel settings.

"Paul Stamler" wrote in message
...
First and foremost, have you tweaked the playback head azimuth for maximum
treble output?

Can this particular playback head be adjusted?


http://home.mpinet.net/~docsavage20/.../Document.html

"Doc" wrote in message
nk.net...

"Paul Stamler" wrote in message
...
First and foremost, have you tweaked the playback head azimuth for
maximum
treble output?

Can this particular playback head be adjusted?


http://home.mpinet.net/~docsavage20/.../Document.html

The picture's too murky for me to make out, but somewhere on or near any
playback head, there's a screw that adjusts the azimuth.

Peace,
Paul

"Doc" wrote:

I've been wrestling with cleaning up an old tape and have been trying
various approaches. The recording is of a jazz combo - Trumpet,
electric guitar, piano, drums, bass and was made on an old mono tape
recorder many moons ago. The primary software tools I have a

Soundforge 5.0
WaveRepair
DartPro32


Uninstall Dart from your machine RIGHT NOW. If you have any disks for
it, bury them where even archaeologists won't find them. Nasty, awful,
evil sounding stuff it makes. Ick. Poo.

I've never heard of WaveRepair so I can't comment on that, but it sounds
like maybe you should be careful with it. Nothing wrong with Forge.

A colleague of mine told me he tried a freeware audio tool called
Audacity that did a decent job of noise removal. I haven't tried it
myself, in fact, I haven't even heard what he did with it, but he ain't
stoopid and he says it works okay.



Something I've noticed is that running the same settings for ex. on
the graphic eq vs parametric (say centered on 11k -5db .5 octave)
don't seem to yield the same results. Is there a fundamental
difference in the way they work?

Yes. No two EQs sound the same. Differences in the shape of the curve,
centre frequencies (it ain't usually more than "kinda close" to what the
front panel says it is), type of filter, etc. all contribute to any two
EQs sounding different even with seemingly identical settings.



Problems I run into using the noiseprint solution is that

....is that it's noise reduction and it's tracking dirty footprints all
over the nice clean music (take your shoes off before you listen!).

Before you do anything, ask yourself: Do you really HAVE to "clean it
up?" Is it ESSENTIAL that you apply noise reduction at all? How bad
would it be to leave it the way it is?

Steady noise like hiss tends to disappear after a few seconds of
listening -- the listener tunes it out and doesn't notice it after a
while. Noise reduction often leaves blemishes that are MUCH more
obvious to the listener, like lost reverb trails, hiss that pumps in and
out, instrumental timbres that change from moment to moment, and other
unsightly stains and embarrassing odours. Which is worse, the disease
or the cure?

--
"It CAN'T be too loud... some of the red lights aren't even on yet!"
- Lorin David Schultz
in the control room
making even bad news sound good

(Remove spamblock to reply)

Doc wrote:

I've been wrestling with cleaning up an old tape and have been trying
various approaches.

You've written all about the cyber part of this mess. What about the
analog portion of the recipe? What you got going on there?

--
ha