Cleaning Up A Cassette Rip. HELP.

Special Interest Groups Audio Processing
1

I ripped my first cassette tape yesterday and have been fiddiling around with the recordings all day but I don’t seem to be getting anywhere.

I amplified the volume to normal levels but I’m hopeless at clearing up the background noise. There is also an unfortunate skip on the tape, is there anyway to make it less noticiable?

I really want to make these recordings as good as possible so I’m hoping an expert on here can help me out. If it’s eaiser I could send over the WAV’s.

Any help would be great.

Thanks,
Josh

2

Please, anyone…

3

Calm down. The forum has to ripple around the earth once. That takes time.

Which computer have you got and which Audacity?

Koz

4

Haha sorry.

Seems to be version 1.2.6.

5

<<<Seems to be version 1.2.6.>>>

On a Windows XP PC?

Cassette is hard. Most of the noise is “pink nose” hiss and that’s really hard to get rid of. You can’t send any of it on the forum, but can you post a little bit somewhere else and point to it? Pick a piece with a silent lead-in and then somebody talking. Ten seconds or so.

You may be better off in Audacity 1.3 instead of 1.2. 1.3 has better noise removal tools–or more versatile. Once you have a place to leave stuff, send a segment with the gap in the middle. Use the WAV files. Don’t bother sending MP3. MP3 distorts sound and makes the performance much harder to manage.

Koz

6

I’m on Vista (I hate it haha)
I’ve uploaded the first 30 seconds of the first track and PM’d you. (I hope that’s ok?)

It’s the hiss that is really bugging me, but like you said, it’s really hard to get rid of. At least without taking some of the track with it.

7

If you find that the noise reduction in 1.2.6 works better on this particular example than the noise reduction in 1.3.x, perhaps you could let Dominic know: http://audacityteam.org/forum/viewtopic.php?f=28&t=7010#p30888

8

One at a time…

Get good strong coffee. I can put you to sleep in no time flat.

The sample posted (perfectly) has three problems. The obvious one is the hiss typical of the oxide layer of a tape that just isn’t moving that fast. We can probably deal with that–or at least suppress it. Much more of a problem are the two hums. Yes, two. I measured them and I have the analysis around here somewhere, but it isn’t that important. If you listen very closely to the lead-in to the first performance note, it goes from almost dead silence to hiss with a subtle background musical note component and then to more hiss with a second note about a fifth down from the first. So this is a very serious multiple dub with damage from each pass.

Those two music notes make the rescue effort challenging. Because they are actual musical tones, although subtle, I can’t get rid of them without damaging some of the quieter performance notes.

http://www.kozco.com/tech/audacity/1a.wav

The correction kicks in just before the first note.

I took the profile sample in the intense noise just before the first performance note. That means Audacity will try to suck out the hiss plus both of the hum components. If you sample further and further toward the beginning, you may get cleaner results, but that last hum will get left in the show.

The settings are Attack=0, Frequency Smoothing=1000, Reduction=14dB. It was performed in Audacity 1.3.5 on a Mac. The earlier noise reduction tools didn’t have those adjustments and I don’t like those tools very much.

So you can experiment with this, too. Oh, I boosted the level.

You can increase the Reduction for less and less hiss, but because of the two hums, some of the quieter notes start to fall apart. Ears act funny and this is a time when you need to pay attention. Signal level drops by half every 6dB. The ear hears half volume or loudness every 18 dB. So the 14 dB Reduction setting reduces the trash multiple times, but doesn’t quite reduce the hiss by half to your ear. I tried the 18 dB number and the hiss was much improved, but the lead musical notes started to fall apart.

So there you have it. Your turn–or anybody else who wants to take a shot at it.

Koz

9

I wouldn’t mind giving it a shot if someone can post the original WAV sample.

10

Sorry. Not paying attention…

Same address. 1.wav is before 1a.wav is after.

Koz

11

Since this is the “Audio Processing” section rather than “Windows” or “Mac”, I’ll use Linux.
This is from using the stand-alone audio cleaner in Ubuntu (Gnome Wave Cleaner 0.21-05).
I amplified it a bit first, then de-clicked it, then de-noised. I wasn’t sure if the “second tone” that starts at around 7 seconds was part of the music or not, so I left it in:
http://easyspacepro.com/audacity/examples/1b.wav

12

Thanks a lot for the help guys.
Well the second attempt is a lot clearer but I’m not sure how the rest of the song would sound with those settings, I have a feeling some of the song would be cut out.

Is it just a low quality recording or is the tape damaged from wear?

13

The “tone” that is present from the beginning was probably there on the original recording.
The pink noise is pretty typical tape hiss and is present to some degree on all cassette recordings, though some types of tape (chrome tape and metal tape) generally have better signal to noise than standard ferric tape (and are more expensive). The signal on tapes gradually fades over time and the amount of noise gradually increases, though again more expensive high quality tapes tend to last a bit better. It’s a bit like how old photographs fade over time. Keeping tapes well away from magnetic fields (for example in metal cases) helps to preserve them.

14

Well this particular demo tape was released in 99 so I guess that explains the quality. Then again it is just a demo tape.

15

by kozikowski » Sun Dec 28, 2008 5:20 am

Those two music notes make the rescue effort challenging. Because they are actual musical tones, although subtle, I can’t get rid of them without damaging some of the quieter performance notes.

Suggestion: Before subtracting the hiss sample from the track, (e.g. using Noise Reduction on Audacity), try notching the hiss a little at the frequency of the “two musical notes” (about 1500Hz).
When this notched hiss is subtracted from the intro, less of the “two musical notes” will be removed, thus improving the signal to noise ratio.
The silence at the start will now have a very faint tone corresponding with the notch frequency (1500Hz), this faint tone can eliminated using a gate.

[Subtract notched hiss from the intro to rescue the “two musical notes”, subtract un-notched hiss from the rest of the track]

16

<<<(about 1500Hz).>>>

It’s two tones, though. Like G and the C below that. It’s a musical fifth. If you listen on headphones next to a keyboard you can figure which notes they are. They match some of the performed notes which is what’s so entertaining about this. You can’t just notch them out because two of the notes on the keyboard later will vanish.

I have another note on the forum about souping up the accuracy of the Noise Suppression Profile tools. Fast Fourier Transform is our friend.

Also, there is the limit that the performance is taken from a cassette tape. Only the top end machines had controlled servo motors. Everybody else turned free-running motors on and hoped for the best, so the note pitch is going to wander. I can do better than my sample, but I’m betting I can’t make it stick over the course of the show. Eventually, it will be taking out the wrong thing.

Koz

17

We don’t have near enough useless information in the thread, so allow me.

The two “motor noise” tones at the top of that sample are A# below middle C followed by D# below that. A musical fifth. But they’re not perfect. To get a pitch perfect match (so to speak), I have to pull the pitch shifter down ever so slightly, so the actual tones are in the gutter between those two black keys and the white keys just below.

There. That should clear everything right up.

Koz

18

Most of this is going right over my head haha.
I applied your (Koz) settings to the whole song and it does quieten the noise slightly. I’m not sure if you want to do this but I can upload the whole track somewhere and let you have a crack at it if you want?

19

This is an interesting example for “audio restoration” as it contains several problems:

There is “broadband noise” (“pink” noise containing frequencies right across the audio spectrum).
There are a couple of distinct tones, that could be “notch filtered” out, but that will also remove those tones from the music that we are trying to restore.
Fairly low signal level, so the SNR (Signal to Noise Ratio) is quite poor (In places the music is at a similar level to the noise).

When trying to restore this kind of thing, there are some questions that you need to ask yourself;

There needs to be a balance between how much noise to remove and how much erosion of the music can be allowed. The more noise you remove, the more damage will be caused to the music, so you need to strike a balance. It’s like cleaning a picture, if you scrub too hard at cleaning the dirt off, some of the paint will come off as well.

Which types of noise do you find most distracting? You can choose to deal with the most distracting noise more severely than the less obtrusive noise.

What types of damage to the music are most/least acceptable? For example, the high frequency hiss can easily be removed by filtering out the high frequencies (low pass filter) but this will also remove the high frequencies from the music and make it sound dull - is that acceptable or not?

How much time/trouble/expense are you prepared to give on this project? I guess that you do not have Linux on your computer, but there is nothing to stop you from installing it - it’s free, and you can set it up to “dual boot” so that you still have access to your current operating system - but can you be bothered to go down that route?

I would certainly suggest that you experiment with this yourself All For Nothing as you will learn a lot from it (experience is a huge part of the game with audio processing).

Personally I find the mid/high frequency hiss to be the most annoying, so using just Audacity, I would copy a section of noise, then amplify it a little and filter it so that the high frequencies are more pronounced, then use that as my “noise sample” for the “Noise Reduction” effect.

I’ve only got an ancient computer so I would not like to download the whole track, but if you could upload a couple more samples I’d be happy to experiment and make suggestions about settings based on my preference.

20

I suppose I just want it as clear as possible but with no noiticable loss of the actual music. I know that’s the whole point, but I’m not too fussed about removing little thing I probably won’t even notice, if it also takes some noise with it.

I’ve PM’d you three small parts of the song for you to fiddle around with.