Soften Harsher Sounds

Special Interest Groups Audio Processing
21

Did you try to open up the raw capture files in Windows Media or QuickTime rather than Audacity?

A quick note. Start talking into the unit and then slowly make it do a backflip in front of your face. At one point, you will be talking into the bottom of the unit. Then do the same thing and slowly spin it like a top. Take about ten seconds to do each.

I’ve seen people trying to talk into the end of one of these…

http://www.coutant.org/2.html

…instead of the flat side which is normal for this microphone. Speaking into it wrong will get you some very odd sound.

Koz

22

Just what I was thinking :smiley:
http://audacityteam.org/forum/viewtopic.php?f=20&t=4805&p=18902#p18902

On Linux it is:
home/<username>/.audacity-data/EQCurves.xml

So I’m guessing for Windows it’s:
C:Documents and Settings<username>Application DataAudacityEQCurves.xml

23

Exactly, I have XP and I have it here:
c:Documents and SettingsxxxxusernamexxxxApplication DataAudacityEQCurves.xml

24

Thanks, you guys are working really hard.

I downloaded EQCurves and applied Drumbeat3 to a WAV Recording…

http://www.mags-scorey.pwp.blueyonder.co.uk/MP3Temp/Drumbeat%20Fire%20WAV24bit%20Level27%20b%20PART.wav

Won’t have a lot of time Tuesday to get back to this, maybe Weds, but not Thurs … but will be back.

Mags

25

It was a juggling act curve between getting rid of the teeth-drilling sibilance sounds and sucking all the life out of the performance. There’s still an occasional zing that goes through, but it should be livable until we figure out something else.

Anybody figure out how to use a de-esser in Audacity? I need personal experience only. Use exclusively the “I” pronoun, not my sister in Schenectady knows somebody…

Koz

26

<<>>

Better. Being compulsive, I can certainly see where I could massage it a little more, but it’s much better than the teeth-drilling from earlier.

Now you need to work on the room. I can practically reach out and touch the faded, rose-cluster wallpaper. I can hear that picture of Brighton Beach over there by the window and I personally would not have chosen that particular shade of mauve for the hard, wooden trim, I guess it will have to do.

I can tell you how big the room is by analyzing the echoes, but it’s too close to bedtime. You have two choices; much bigger or much deader. Anything in Prince Albert Hall sounds grand. Anything in a soundproof room can be made to sound grand. This room can’t be any more than 3M in any one direction, right?

http://www.kozco.com/tech/soundtests.html

The Left-Right thing was recorded in my bedroom with the sound equipment spread out on my goose feather duvet. I have a California cottage cheese ceiling and tatty, but wall-to-wall carpeting, so it would be hard to tell what I was really doing.

Koz

27

Thanks again, Koz …er…your link doesn’t work …can we…? (You select Preview then you test the link then you post, smile) Just proved that, cos one of my links didn’t work.

The Lounge is 12’ 11" x 11’ 9" and is extremely quiet … I close all doors…
http://www.mags-scorey.pwp.blueyonder.co.uk/ImagesTEMP/Back%20Room%20Window.jpg

This is Drumbeat Fire, recorded with Edirol on a Cushion on an easy chair. I stood in front of the chair.
Bought into Audacity and Amplified to 20. Eq Drumbeat3 applied

http://www.mags-scorey.pwp.blueyonder.co.uk/MP3Temp/Drumbeat%20Fire%20WAV24bit%20RecLevel27%2020080610%20b%20PART.wav

I’m slightly deaf, which helps a lot … but I still think there’s background noise. A friend of mine changed software because Audacity didn’t get rid of the background noise created by his Edirol-R09. I bought mine because of him … they’re professional microphones …

I’m not going mad … this is fun … I’m not going mad … this is fun and it will come good in the end and will be worth it, because this poem and recording is going to be sent to the band, Tinariwen, because it was their performance that inspired it…this is important stuff …the world could be changed … er…but not my lounge…

Thanks, Mags

\

28

Mags,

I’m gonna jump in on this one as well.

This is Drumbeat Fire, recorded with Edirol on a Cushion on an easy chair. I stood in front of the chair.
Bought into Audacity and Amplified to 20. Eq Drumbeat3 applied

You’re right that there is still a considerable amount of noise on this recording and the source of the noise is the “Amplified to 20” bit. There exists something called a noise floor. Any time you have to amplify a signal you’re going to end up raising the noise floor along with all the stuff you want to keep. The upshot is that if you have a very quiet recording that is not much louder than the noise floor, you will not be able to amplify the sound to be loud enough without the noise being too loud as well.

The only way to avoid this noise is to record louder (or use a microphone with a much lower self noise rating, but I wouldn’t worry about that for now). My suggestion is to find a mic stand with a clip that will hold that microphone and to use a pop filter as others have suggested. It will be up to you to experiment with placement in order to avoid any sibilance problems. You may also benefit from hanging some heavy drapes over the windows in that room to cut down on the room tone, but placing the microphone closer to your mouth will also have the effect of reducing room tone, so do this scientifically and make only one change at a time.

29

I patched that link. I left one of the bracket symbols off, so the system couldn’t see it as a link.

I would have to really lean into “Drumbeat” with headphones and the volume turned up. It doesn’t sound “natural,” but it’s so close I can’t tell what’s actually wrong with it–other than I can clearly hear the room slapping in the background.

I’m fixated on room echoes because that’s usually the first thing that separates a primary-school “YouTube” performance from a higher quality show. There was one skit where the kids did an amazing job of duplicating the set and lighting of a major news show. They were fine until the first words. They were using the little microphone atop the camera and the sound was all but unintelligible. It completely destroyed the illusion.

Doing illusion is harder than it looks.

Koz

30

<<>>

I was dangerously close on the rose-cluster wallpaper, though, wasn’t I? Are you sure you don’t have a picture of Brighton Beach on one of the other two walls?

That explains the slightly “live” presentation. Two sets of parallel walls and; is that linoleum on the floor?

I’d still be in the bedroom using the bed as a desk for the performance. I bet it makes a world of difference.

Is that wav file after filtration? That’s better than I thought it would do. I’m going to see if I can suppress the hiss in the background. That and a test on the duvet to get rid of the echoes and you’ll be golden.


Oh, why all the fuss? We don’t often get a very good live performer with a laundry list of easily preventable errors. It’s live recording’s version of catnip.

When cats sense the bruised leaves or stems of catnip, they may roll over it, paw at it, chew it, lick it, leap about and purr, often salivating copiously.
–WikiPoedia–

Koz

31

Right then.

Here’s the performance with a gentle noise reduction laid on.

http://www.kozco.com/audacity/Drumbeat%20Fire%20WAV24bit%20RecLevel27%2020080610%20b%20PART2.wav

Koz

32

Thanks for all advice …

Leave my wallpaper alone, can’t change that !! All doors are ALWAYS closed. That is CARPET on the floor. House is over 100 years old, maybe it’s echoes from the past. Would rather not use bedroom to record, stairs are steep. Well perhaps I will, but not today, too much other stuff going on…

Right, I recorded this one with Edirol on cushion on top of set of drawers and fluffy cat to keep me company/inspire. So I was standing right in front of the mike.

Into Audacity, shoved ‘Silence’ into all the gaps, Drumbeat3 applied

Then, because you’re still working on the previous WAV file and I don’t have enough website space, exported this as an MP3 and increased to 85db, using something called MP3Gain - which I can hear!

http://www.mags-scorey.pwp.blueyonder.co.uk/MP3Temp/Drumbeat%20Fire%2020080611%20d%20PART.mp3

Won’t have time to do any more today … or Thu. Be seeing the musician on Thursday … will strangle her for starting all this fuss … but haven’t we learnt a lot?

Thanks, Mags

ps. Posted this before I read your last post … thanks for that … chasing tail here …how did you do that? Can’t stay on PC any longer, but will be back, maybe Friday now … or Saturday

33

No need for that - As koz said in his inimitable way: “It’s live recording’s version of catnip.”
I’ve just got home from work, made myself a nice cup of tea, and logged in to catch up with the latest developments with this issue. (252 views of this topic so far, so I think a few people must be following this particular story :slight_smile: )

That is a huge amount of amplification, and it still peaks below -10dB. If I understand correctly what you are saying. then the original peak amplitude of the recording will have been around -95 dB.

In case the above sentence sounds like technical mumbo-jumbo, I’ll try and explain in layman’s language:

With digital recording, there is a maximum level (loudness) that the recording can be. If you attempt to record louder than this, then the recording will distort rather than getting louder.
In Audacity, the recording shows up as a blue shape along the track, or if you zoom in on it, a blue wiggly line. This is the “waveform” of your recording.

Here we can see the waveform of the words “drumbeat fire”:
(Click on the image to expand and remove the scroll bar)
drumbeat.png
And here I have zoomed in on the first part of the word “drumbeat”:
dr.png
Looking at the first picture you can see the vertical scale that goes from +1.0 to minus 1.0. This represents the maximum vertical height (loudness) that the recording can go.

In Audacity, the maximum is (somewhat confusingly) +1 to -1, which is equivalent to 0 dB (zero dB).
Signals (waveforms) are usually measured in dB, with reference to 0dB as the maximum. Any sound that is quieter than this is said to be so many dB’s below zero - that is, a signal will have a negative number to represent the size (amplitude) of the signal. Total silence will show up in Audacity as a horizontal line at a level of 0.0, and this es equivalent to “minus infinity dB”.

Here is a section of one of your recordings, where I have changed the view to show a “dB” scale:
dB.png
You will notice in the above picture, all of the audio is in the range from about -70dB (the quietest parts) to about -10 dB (the loudest parts).

Ideally, your original recording should go up to a maximum level that is close to zero dB. In practice, you must leave a bit of “headroom”, so a maximum peak somewhere around -6dB is generally good. However, it seems that your original recordings had a maximum peak of only -85dB which is very quiet indeed.

Here is a picture of a recording from my computer when it is recording “silence”. Again I have selected the “dB” scale and stretched the track so that you can see the measurement in dB.
silence.png
You can see that even though I was attempting to record silence, there is random electrical “noise” at around -65dB". This is (as mentioned by alatham) what is called the “noise floor”. Anything and everything that I record will have this noise added to it.

If I record my voice, and my original recording goes up to near the 0dB level, then this noise is almost insignificant and will just sound like a very, very quiet background hiss.
However, if my voice is only registering up to say -50dB, then the noise floor is relatively much higher, and when I turn up the volume, or amplify the signal to be able to hear it clearly, then I also turn up this background hiss.

In order to get a louder recording, and thus keep the noise floor relatively low, you need to get closer to the microphone, and you may also need to turn up the recording volume.

I think that even when we have a good volume level in the recording, there may still be a certain amount of sibilance, but it will be very much less of a problem.

34

Thanks for all the advice, I only increased the db because Koz said he couldn’t hear a previous version.

Am taking a break from computer for a while, spending some time with the musician. So will get back to this when I can.

I think one of the great things that has come out of this, is the Drumbeat3 sibilant reducing Equalisation. I couldn’t have done that.

Am retired, not a sound engineer, more of a poet. So have been glad to air this.

This isn’t the right forum, but you may like to see the whole poem. Be back in due course. Mags.

Drumbeat Fire

I listen to a one-stringed fiddle
And consider those with one gift. Let that gift burn.
I close my eyes …
and am given the words, ‘Drumbeat Fire’.

Then upon the screen, a tiny fire,
bursting forth into flames.
I am amazed!

A strong beat stirring my soul,
Leaning forward in my seat,
Glancing from one to the other,
trying to capture every moment
Words, just a few, hitting my notebook

My gaze is captured
by the man in turquoise,
showing only his eyes.
the eyes of his soul,
burning with fire.

Drumbeat Fire.

Rhythm sways my body,
Drums pounding… guitars… singing,
Images of Africa onscreen, swirling in and out
Describing the agony, the fight, the eternal longings.

Vibrancy such as I have seldom experienced,
Woman who knows Africa… sitting beside me
Touching my soul,

People dancing, some afire with the beat
and the message.
Reaching out to touch musician’s hands.

Two women singing and dancing on stage,
The man in turquoise plays to and dances with one,
He is mysterious.

Drumbeat Fire

The fire has been burning too long in our lost slumbers,
Have we forgotten there is a war to wage?
Is there energy inside us, left to give to it?

Drumbeat Fire.
What can I do with this eternal longing?
’63 has gone, but will return’
So stir me from the slumber,
in which I have been for many years.

Drumbeat Fire…. Drumbeat Fire

© Margaret Scorey 28th May 2008

Notes
Written after Concert by Tinariwen,
plus Justin Adams and Juldeh Camara
at Birmingham Town Hall on 24th May 2008
’63 has gone, but will return’
refers to the brutally crushed 1963 Touareg uprising.

35

<< I only increased the db because Koz said he couldn’t hear a previous version.>>
Not that I am shure you would not understand the moral of Steve’s post, but this allows me to retell:

If the level is increased after the recording (whatever way), also various noises are increased
(both the noise of analog circuits and noise of digital quantisation and …)

So we have to increase(or, set) the level of sound that is recorded in the right way.
We have two boundaries: one of the two is the maximum, the clipping level, which is denoted “0dB”.
We are not allowed to clip, but we want to stay close to it (about -6dB, or, say, -15dB if we are not confident what would be the level of loudest sound) because there is noise on the other side.
Therefore the level meters are so important.

Not only that noise is not nice thing to hear, it can sometimes make small surprises or unsatisfactory results with various tools like EQ, deEsser,… (BTW there is one demo deEsser listed in compatible plugins, but I would not expect much from it.)

So that is as regards noise.
Besides noise, there is also other thing to cope with: echoes of your room and sounds from outside.
To ‘separate’ the voice from them, you need to get closer to the microphone.
(This also positively changes the volume level.)

36

<<<ps. Posted this before I read your last post … thanks for that … chasing tail here …how did you do that? Can’t stay on PC any longer, but will be back, maybe Friday now … or Saturday>>>

That’s the noise reduction tool in Audacity 1.3.5b. I told it to “sniff” a little bit of the non-voice performance at the beginning of the show (noise only, more or less) and then told it to subtract what it found from the rest of the show. It’s a two step tool. If used very gently, noise reduction can simulate using a much quieter microphone.

I can send you the settings when I get home. I don’t know what I’m doing completely with this tool like I do with the equalizer, but I know enough to improve things a little.

Can I guess you didn’t much like the editing where you reduced the spaces between phrases to silence? That’s back to Price Albert Hall once again. You can’t do a performance in a completely sound-proof room, It sounds funny–sterile and other-worldly. Or if you do it with a sound editor, clipped and mechanical. That’s been part of the enjoyable challenge with this. Clean the performance without mucking it up–or without seeming to clean it.

Thanks for the lyrics. This will be like following along with the words that come in the little booklet in the CD.

I never said I didn’t like the rose cluster wallpaper. I just said I could hear it in the show. It’s unmistakable, really.

Koz

37

Wow, what a wonderful day in the countryside we had yesterday … everyone ought to do it.

Thanks for all your advice… the poem (words only) has now gone off to Tinariwen and …

Sent: Thursday, June 12, 2008 10:07 AM
Subject: Re: Birmingham Town Hall Concert 24th May 2008

Hi Margaret,

Thank you for sending your poem - and I’m glad that Tinariwen/Justin and Juldeh inspired it! I’ll forward your email to Andy Morgan, Tinariwen’s manager, as I’m sure he’ll be interested in reading it.

With very best wishes,
Sally Reeves

I have not sent a recording.

Take 2, 3, 4, 5 … you will love this, I realised in the last recording, with the Edirol on the cushion, it was facing the wrong way, so I turned it upside down so it faced my mouth … then Drumbeat3 Equalisation then into an MP3…and put silence in the gaps before I read Koz’s latest reply …oops! As I said I’m no sound engineer… so may concentrate on writing. This is the whole poem, but no room in webspace for a WAV file, so finalised as an MP3.

http://www.mags-scorey.pwp.blueyonder.co.uk/MP3Temp/Drumbeat%20Fire%2020080611%205pm%20c.mp3

I don’t think I’ll be recording it again and also think the topic has been aired enough … but thanks everyone a great deal.

38

See, then? Not everybody’s topic goes to four chapters. You’re an Audacity Celebrity.

Thank you for giving me the opportunity to stretch my skill-set with more advanced tools. I had never used Audacity 1.3 before and I had no experience with the newer noise reduction tool at all. I’m still not completely comfortable at the steering wheel yet, but I’m getting there.

Do you still need the noise reduction numbers? I’m pleased to have helped with the sibilance problem. I’ve never authored a curve before. Any idea how I’m going to squeeze that into my résumé/CV?

Koz

39

Dear Koz,

Erm…yes I was getting a bit embarrassed at the length of this … no idea why it should have attracted so much interest, except that many, many people have problemsssssss with their …sssssibilants.

Yes, please I would like your noise reduction numbers … I still haven’t got it right…must make notes for future reference.

And you haven’t authored a curve before … yes, how does one word that on one’s CV. You’ll just have to say it was an Audacious Curve … no that doesn’t sound right …erm…a Curve in Audacity … no, that sounds like you’re audacity’s taken a tumble … erm… ohh…doh…

Maybe it’s the banter from various people that has kept a rather dull and boring thread going…

See ya in a bit … Mags

40

Not at all, it has been most interesting.

In fact, I am hoping that you will allow me to use a recording of your voice for testing purposes. Partly inspired by this thread, and the fact that there seems to be no “de-esser” currently available for Audacity, I intend to have a go at making a de-esser plug-in, and will obviously require some samples for testing. Ideally the recording (which only needs to be a few seconds long) would have lots of s’s and sh’s, and would be recorded nice and loud (not amplified to make it loud, but actually recorded loud by speaking nice and close to the microphone). Also it would be best for my purposes to be completely unprocessed - straight off the recorder with nothing done to it.

If you are willing to assist in this way, I can send you my e-mail address as I understand that you are running short of web space, and of course I will provide you with a copy of the plug-in if I manage to get it working.