Whoa! Stop! Danger! These are very old settings from a previous
version of the program and even there they could have been better; I
was -- still am -- learning. My current advice for v1.2 is highlight
only the recorded material -- in case you have recorded the Ghost in
the Machine at the beginning and/or end of the wave (that is the noise
made by your electronics when there is nothing playing, useful for a
noise profile of the most harmless nature) -- after reversing the
whole wave. I think best results are had with 32-bit resolution but
it is time and space devouring. Start with the Old Record - Quiet
Audio preset; change the 45 in Find Big Pops to 100, or uncheck it.
Enter 50 for Sensitivity and 100 for Discrimination, then use Auto
Find All Levels. When the program has filled in the blanks, run the
transform. If too many clicks are left to be removed in person, do it
again. Yields great results and hurts the music not at all.

Try it, MG, you'll love it.

-----------------------

And I think a Smoothing Amount of 1.25 is better in the noise
reduction transform these days. If I read the instructions correctly,
that should cause it to leave a little more noise in the wave but have
less of an effect on the music than a setting of 1.0. What you use
for a noise print is a very important aspect of this transform. If
the level of the content is sufficient to mask noise you might want to
avoid using it in those areas of the wave, perhaps only using it at
the beginning and end.

That's my guiding star: try to hurt the music as little as possible
while getting rid of the vinyl sounds. The documentation for the
program doesn't always fit it. The click/pop eliminator dialog says
that detection levels range from 6 to 60 and rejection levels from 1
to 100. That's where the earlier numbers you found came from. But
that is wrong; the Auto Find All Levels puts in numbers in excess of
those limits, at least for the detection level. The last time I used
it, all three detection values were in the 90s somewhere. I think 150
is the highest value -- the higher the value, the less detected, the
more rejected, in other words, the less done -- for detection and
rejection and the detect big pops goes up to 200. Those numbers I
gave Micro Guy for Sensitivity (50) and Discrimination (100) are the
limits, the lightest setting. Even set like that, the program cleans
a lot of clicks. If it leaves too many, I will try another pass and
only go to a lower value if that doesn't work well.

When you use this transform, you sometimes end up with distorted
clicks -- abrasive things that often hang out with percussion
transients, maybe not really clicks. I think the program causes them
by "cleaning" _more_ than it should. Another artifact it can create
is a sort of a whumf sound. I believe that is made when a small area,
a few cycles, of the wave has been cleaned by the transform and it
leaves a quiet valley between two relatively higher-level areas. That
noise is a pain to remove because it is often hard to decide just
which peaks should be reduced to remove the encapsulation. I hope
that the Pulse Train Verification reduces these sorts of problems by
causing the transform to do less.

A thing to watch for is a very low detection level value; 6, say.
Auto Find can give you that if you selected some of the Ghost In The
Machine; I think it keeps turning up the power seeking something that
is not there. Yeah, I just ran an experiment, recorded 15 seconds of
nothing and did AF on it: the columns read:

max threshold -90 6 10
avg threshold -90 5 9
min threshold -90 4 8

You don't ever want to process a file with settings like those; it
will detect everything and reject that detection hardly ever. It will
run for a long time and severely alter your music.

Some people have the conceit that they won't do editing on a decent
vinyl album, leaving it to the downloader to do any post-processing
after reconverting the mp3 into a wave. That is uninformed, I
believe, because the encoding-decoding process does something to a
wav. I'm not sure, maybe it changes close to silence into silence.
But it produces those same type settings in CEP's Auto Find. Much
much harder to work on a reconstructed wave than an original.

All this is just opinion, probably wrong, not the FAQ.

I've been asking and searching for ways of cleaning my vinyl wavs and
never found anything worthwhile (somewhat). So I went perusing my old
archives of messages I had saved. I ran across a nfo file posted by
lamer back in 7/4/99;
Subject: 160S - vinyl - live - Axton, Hoyt - Boney Fingers.mp3 (00/18)

Contained just what I was looking for and thought I'd share with
others who are in the need. Hope you don't mind lamer;

Back to Top

start lamer's message
----------------------------------------
Side two, track one of the 1979 Fantasy double-lp Bread & Roses,
Festival of Acoustic Music - Greek Theater, U.C. Berkeley. Recorded
with CEP v1.2, cleaned with same; encoded with MP3Enc31 -br 160000
-no-is -qual 9.

The album is in pretty decent shape but not without record noise. The
procedure:
Recorded with Monitor Record Levels Option having been used so that
peaks were between -3 and 0 dB.

Save the wave right now, either track by selected track or all at
once. After saving the album side track by track, I closed the wave. 
Then I opened this track. I have CEP set up to
Options>Settings>Data>Auto-convert all data to 32-bit upon opening. 
Noise reduction works better that way.

Transform>Amplification>Amplify>Center Wave preset. 
Transform>Reverse. Transform>Noise Reduction>Click/Pop Eliminator
with these settings:
Detect Column: 60 60 60 Reject Column: 100 100 100 Second Level
Verification is checked Pulse Train Verification is checked Detect
Big Pops is checked and its value is 60 Link Channels, Smooth Light
Crackle, Multiple Passes are NOT checked FFT Size Auto is checked 
Pop Oversamples 12 samples Run Size 14 samples After all these had
been set, Max, Avg, Min Thresholds were filled in with Find Threshold
Levels Only. Auto Find All Levels was not used and so its Sensitivity
and Discrimination values are not important.

After one pass, a 2.5-second area of "silence" from the lead-in was
selected for the noise profile. Transform>Noise Reduction>Noise
Reduction with settings: Log Scale NOT checked Live Update is
checked Noise Reduction Level 100 FFT Size 6000 points Remove
Noise radio button selected Precision Factor 11 Smoothing Amount 1
Transition Width 0 dB Snapshots in profile 300. The record was
slightly warped, causing a periodic noise in the lead-in waveform and
2.5 seconds, just my standard time, is sufficient to capture the noise
of a complete revolution of the lp. Click Get Profile from Selection
then Close (not Cancel) the dialog. Deselect the noise print area by
clicking anywhere on the wave; click on the Zoom Out Full button so
the whole wave is transformed. Return to Transform>Noise
Reduction>Noise Reduction and click OK.

When it is finished Transform>Reverse the wave. Listen to it through
headphones for noises not removed by the transforms. In general,
narrowly select the noise and use Transform>Click/Pop Eliminator>Fill
Single Click Now. This feature is effective on most noises of short
enough duration but finding a single little bitty piece of wave that
represents the noise is a problem. Read the Help for the Click/Pop
Eliminator dialog and use the Spectral view to help you find them. 
Here's some of what it says about that:

Use the Spectral View feature with the spectral resolution set to 256
bands and a Window Width of 40% to see the clicks in a program. See
/Options/Settings/Spectral to adjust these parameters. Clicks will
ordinarily be visible as bright vertical bars that go all the way from
the top to the bottom of the display.

When you have taken the very few clicks remaining from the wave,
sculpt the ends of the wave so that it does not make noise when it
starts playing. Leave about 0.35 second of lead-in before the music
begins. Transform>Silence the first 0.2 second. Select the 0.15
second after the silence to just before the music starts and choose
the
Transform>Amplification>Amplify>Fade In preset, click OK. At the end
of the wave, leave a few tenths of a second after the music stops and
delete everything after that. After selecting the area between the
end of the music and the end of the wave, choose
Transform>Amplification>Amplify>Fade Out preset, and click OK.

Choose Edit>Convert Sample Type with these settings: Sample Rate
44100 Stereo Resolution 16 bit Enable Dithering is checked 
Dither Depth (bits) 0.54 p.d.f. Triangular Noise Shaping
(44.1KHz), and click OK.

Check to see if the ends are still proper after the conversion; you
may have to silence and fade in again if the wave makes a noise at its
start. Save the wave. Encode it to make the mp3.

Work your fingers to the bone,
What do you get?

L:)
---------------------
End lamer's message

I tried it with a fav album of mine and like the results. I recorded
it with an older Kenwood turntable and Stanton cartridge. It's Stanley
Clarke's "School Days" lp. I've done the first song so far (School
Days) and am posting it in the main group.

Take a listen and tell me what ya think. All feedback welcomed. Always
looking for perfection.

The only thing I did that he/she didn't mention was to
Transform/Amplitude/Normalize to 98% before converting the wav back to
16bit.

Before I start I wash the vinyl in warm soapy water. I've had a lot of
comments about long term damage this might cause to records since I
first mentioned it, but I've found it helps with some surface noises,
and once i've got them onto cd the original record doesn't interest me
any more. If you're worried about this, don't do it. I keep before and
after versions of the wav files on cdrom so that I can easily start
again when I learn something new.

I use cool edit pro for everything, and record at 44k/16bit stereo. If
it's an album, I'll play a typical track through once without
recording, watching the recording levels and adjusting them with the
soundcard controls. I try to aim for an average of -3db. 

When I've got the levels right I record the whole album and save it as
2 wavs (one for each side). Doing it on a song by song basis, or not
recording it all in the same session would make it more difficult to
get the same levels on each song. I split the big wavs into individual
songs by selecting each song and using paste to new. When I've saved
the new wav I close it and delete it from the big wav.

Working on each song in turn, I zoom in on the wav in spectral view so
that I've got about 2 seconds worth on screen. I'll listen to this
several times using the play looped control (a figure 8 on its side,
or infinity symbol), noting where the clicks are, then zoom right in
on them, select just the click (if it's just in the left or right
speaker I'll select it in that channel only), then use the click/pop
eliminator to 'fill single click now'. Finding the click in spectral
view is a lot easier, because it shows up as a thin red rectangle. If
you wanted to, you could even find all the clicks without listening to
the music at all (unless it includes a live audience clapping or a
drum machine, which would look very similar to a click).

I'll carry on like this until I reach the end of the song, then I'll
play it from the beginning in waveform view at a x4 zoom factor, watch
the wave form scroll across the screen as it plays, listening for
anything I've missed. When I hear something I'll stop the playback, go
back to spectral view, fix it, and then play from the beginning again.
It's very easy to get carried away with this, hunting down clicks that
you can barely hear through headphones, and definitely wouldn't hear
through normal speakers. 

When I'm happy with the body of the song, I'll concentrate on the lead
in and lead out. Back in waveform view I'll zoom right in on the
beginning of the song, and zoom in vertically (the little + magnifying
glass on the far right bottom of the screen) to find the precise start
of the music. I find the point about half a second before this and
delete everything to the left. Then I'll select the half second of
'silence' and fade it in. if it's a song that fades in I'll extend the
selection to include some of the song too (usually up to where it gets
up to full volume). The end of the song gets the same treatment, but
with a 1 second lead out and the obvious reversal of the stages above.

I don't use noise reduction filters any more, even on very low
settings they tend to take something out of the music along with the
noise you're trying to remove. But on really duff pressings I'll use
the 30 band graphic equaliser filter to try and lift the mid ranges a
bit. I use my main speakers for balancing the sound to, since these
are what I'll be listening to it through most often.

All this takes a long time, but I think the results justify it --
after all, it only needs to be done once. But I can understand people
wanting a 'quick fix' way of doing it.