Normalization nightmare.

JonathanAnon

Just gonna throw this one out there. Not sure if anyone has the same issues, but I get tracks sometimes from different sources, i.e. download from karaoke sites, made up tracks from cubase, midi conversion, tracks just handed to me on a disc.

I've been using a program called MP3Gain, which is free, to normalise the tracks so that they have the same volume level. But I find that it is not really working. It tends to normalize tracks from the same source to the same relative volume level, but when I normalize all tracks from all sources there is a noticable difference between the tracks that come from different sources... if you know what I mean.

Anybody have any suggestions. I'm gonna try Audacity today and see if it will do a better job.

MilanPan!c

JonathanAnon wrote: »

Just gonna throw this one out there. Not sure if anyone has the same issues, but I get tracks sometimes from different sources, i.e. download from karaoke sites, made up tracks from cubase, midi conversion, tracks just handed to me on a disc.

I've been using a program called MP3Gain, which is free, to normalise the tracks so that they have the same volume level. But I find that it is not really working. It tends to normalize tracks from the same source to the same relative volume level, but when I normalize all tracks from all sources there is a noticable difference between the tracks that come from different sources... if you know what I mean.

Anybody have any suggestions. I'm gonna try Audacity today and see if it will do a better job.

not surprised.

Normalisation raises to volume so that the peak in the track is as loud as desired level.

It's not compression.

If you want all the tracks to be about as loud, compress them, like a radio station does.

I think you are probably just using the wrong tool.

JonathanAnon

hmm.. think you might be right..

http://en.wikipedia.org/wiki/Dynamic_range_compression

Audacity seems to have a dynamic range compressor built in .. gonna try it now.

MilanPan!c

JonathanAnon wrote: »

hmm.. think you might be right..

http://en.wikipedia.org/wiki/Dynamic_range_compression

Audacity seems to have a dynamic range compressor built in .. gonna try it now.

I'm def right.

Don't overcompress though, it'll sound like ****.

google: compression tutorial

Using a compressor is kinda of an art into itself.

studiorat

A good rule is never to use any processing on something on it's own. Compress and EQ when the tracks are playing together, then you get a sense of perspective in what you are doing.

A simple gain plug-in should work fine. You were probably trying to normalize all the tracks together, so it looks at the loudest track and calls that the maximum. If you are going to be using all the tracks together just get them all in the DAW together first. You might not need to bring all the tracks up to digital 0dB.

MilanPan!c

studiorat wrote: »

A good rule is never to use any processing on something on it's own. Compress and EQ when the tracks are playing together, then you get a sense of perspective in what you are doing.

A simple gain plug-in should work fine. You were probably trying to normalize all the tracks together, so it looks at the loudest track and calls that the maximum. If you are going to be using all the tracks together just get them all in the DAW together first. You might not need to bring all the tracks up to digital 0dB.

I'm not I would say that's a good rule of thumb, but I am sure that we can't give you actual advise until we actually know what you're trying to do.

Compression will do it. Gain will do it.

Just depends on your actual goal.

JonathanAnon

studiorat wrote: »

A simple gain plug-in should work fine. You were probably trying to normalize all the tracks together, so it looks at the loudest track and calls that the maximum. If you are going to be using all the tracks together just get them all in the DAW together first. You might not need to bring all the tracks up to digital 0dB.

What I really need is something to apply on the MP3, cos I dont have the cubase, protools, midi etc etc source for most of the items... The normalization tool MP3Gain used to bring all the tracks to -89db which worked for most of the back tracks that I produce myself, but not for the commercial ones..

Not having much luck with the dynamic compression on Audacity.. I'll keep searching.

MilanPan!c

JonathanAnon wrote: »

What I really need is something to apply on the MP3, cos I dont have the cubase, protools, midi etc etc source for most of the items... The normalization tool MP3Gain used to bring all the tracks to -89db which worked for most of the back tracks that I produce myself, but not for the commercial ones..

Not having much luck with the dynamic compression on Audacity.. I'll keep searching.

-89db?

JonathanAnon

MilanPan!c wrote: »

-89db?

Sorry 89db.. From the MP3Gain help file

"The default is 89.0 dB because most mp3s will not have clipping at this volume level.("Clipping" means that when the mp3 file is decoded by your player, some of the sound samples will be too loud. The player will "clip" these samples so that they do not exceed the maximum allowable value. This clipping creates a sort of rough, "scratchy" sound during loud parts of the song.)

This usually comes out as being quite low, and I end up having to turn up my mp3 player or whatever I'm playing it back on.

madtheory

What is 89dB? SPL? That doesn't make any sense.

MilanPan!c

madtheory wrote: »

What is 89dB? SPL? That doesn't make any sense.

No it doesn't.

krd

Something like Winamp has a 'look ahead' automatic gain adjustment.

Normalising to peak doesn't always work very well. it might make some stuff louder than it should be, and other stuff quieter.

JonathanAnon

krd wrote: »

Something like Winamp has a 'look ahead' automatic gain adjustment.

yeah theres a couple of different plugins available for WinAmp, I spotted them on the link that I put above for the dynamic compression on wiki.

What is 89dB? SPL? That doesn't make any sense.

I'm not sure what you mean .. that seems to be the standard that they use. I've kind of become used to 89 decibels from using the software. A few guys who also use the same software (on the net) say that they set it to 92 decibels cos 89 is too low.. It's kind of a moot point anyway I would think, as I'm gonna have to abandon MP3Gain in favour of a compression.

MilanPan!c

JonathanAnon wrote: »

yeah theres a couple of different plugins available for WinAmp, I spotted them on the link that I put above for the dynamic compression on wiki.



I'm not sure what you mean .. that seems to be the standard that they use. I've kind of become used to 89 decibels from using the software. A few guys who also use the same software (on the net) say that they set it to 92 decibels cos 89 is too low.. It's kind of a moot point anyway I would think, as I'm gonna have to abandon MP3Gain in favour of a compression.

89db is the volume of a two-stroke engine or say a lawnmower.

So that's not typically how sound engineers talk about volume of tracks.

(maybe speaker weirdos)

We'd think more like -2 db, as in 2 db below when it distorts (in extremely simple terms).

I mean if I said I recorded a lawnmower, you could turn down the volume when playing it back to it was, say, 40db.

Do you see?

Actually, think of meters. When the meter is green, it's below 0 db, anything in the red is above 0 db.

89 db must mean this sound will be 89 db if played through these spec speakers, etc.

So that's what the SPL reference is all about...

JonathanAnon

http://www.youth.hear-it.org/forside.dsp?area=455

Did you know... Rock concerts can be a threat to fragile ears with noise levels up to 120 dB. During a concert featuring the British rock band Oasis the noise level was as high as 143 dB. Did you know... Many people were moved by the blockbuster film Saving Private Ryan. Not surprising considering the sound level. The soundtrack reached 118 dB. Did you know...When listening to music for long periods at a time, the sound level should not exceed 85 - 95 dB.

I heard that in reference to Oasis concerts before saying that they measured it and said that they had reached levels of over 100 decibels.. ??

MilanPan!c

JonathanAnon wrote: »

http://www.youth.hear-it.org/forside.dsp?area=455



I heard that in reference to Oasis concerts before saying that they measured it and said that they had reached levels of over 100 decibels.. ??

there's really two different usages for db.

One is Sound Pressure. We might call that ACTUAL volume.

The other is measuring the RELATIVE volume of sounds to other sounds (or more specifically volume relative to background noise/sound floor).

It's all related to the magical 1 volt level, below which sound won't distort due to hardware gain.

But again, what you should care about it that if software says 89db, that doesn't actually mean 89 db SPL, because you're controlling the final volume with a big knob/slider.

What you need to use is the other db, the one used to measure relative volume.

So if two recordings both peak at -2 db and you don't touch the volume knob between them, they should sound about the same volume, relative to each other.

Another reason this is important is that if you go into a piece of actual music software and choose to normalise to 89db you will just get a huge wall of crude.

Most plugins max out at about +12db over 0db anyway (and unless you're doing something crazy, turning up something by 12 db is a little nuts...do your initial recording with a higher volume or it'll sound like ****).

MilanPan!c

Just went into my DAW of choice and found a much more succinct definition of relative DB:

The reference level (0 dB) usually corresponds to the current loudness of the sound.

PaulBrewer

MilanPan!c wrote: »

there's really two different usages for db.

One is Sound Pressure. We might call that ACTUAL volume.

The other is measuring the RELATIVE volume of sounds to other sounds (or more specifically volume relative to background noise/sound floor).

It's all related to the magical 1 volt level, below which sound won't distort due to hardware gain.

But again, what you should care about it that if software says 89db, that doesn't actually mean 89 db SPL, because you're controlling the final volume with a big knob/slider.

What you need to use is the other db, the one used to measure relative volume.

So if two recordings both peak at -2 db and you don't touch the volume knob between them, they should sound about the same volume, relative to each other.

Another reason this is important is that if you go into a piece of actual music software and choose to normalise to 89db you will just get a huge wall of crude.

Most plugins max out at about +12db over 0db anyway (and unless you're doing something crazy, turning up something by 12 db is a little nuts...do your initial recording with a higher volume or it'll sound like ****).

Are we getting our Decibels mixed up here Milan ? I think we might be ...

MilanPan!c

PaulBrewer wrote: »

Are we getting our Decibels mixed up here Milan ? I think we might be ...

I wouldn't be surprised if I said something stupid... But the basic point is accurate.

I say without rereading what I wrote.

Paolo_M

The DeciBel is a referenced fraction, albeit a fairly complex logarithmic one.

There are no absolute flavours of it (ie. you can never that music is playing at 40dB because that would be like saying that music is playing at 4, though most people would probably understand that you're refering to SPL in that case).

There are many, many "flavours" of it though, just to make life interesting.

It was developed originally to describe signal degredation across telephone lines

In sound engineering a VU meters 0dB is a max reference, and everything on the gauge is a fraction of that, including "intrument" and "line" levels, -10db and +3dB. 0bD references a signal of .775Vrms which is the voltage level that delivers 1mW into a 600 ohm load resistor.
MilanPanic, I think this is what you mean when you say the majic 1 volt. .775 Vrms is approximately equal to a peak voltage of 1 volt.


When talking about SPL then 0dB is the threshold of human hearing or minimal SPL detectable by the average human ear. Don't ask me how "they" standardised on that 'cos I though we all have different hearing sensitivities which vary with age and health...
Everything else SPL-wise is a fraction of that. Remember that 10 can still be expressed as a fraction ie. 10/1.

There are also other dB references that use thing like energy produced by 1 mW across a specified load, dBm. These load references even vary for different industry segments.

The wiki atricle is pretty good at explaining all of this.

Geeky former Electronics Engineering student starting out in home recording if you're wondering.

MilanPan!c

Paolo_M wrote: »

The DeciBel is a referenced fraction, albeit a fairly complex logarithmic one.

There are no absolute flavours of it (ie. you can never that music is playing at 40dB because that would be like saying that music is playing at 4, though most people would probably understand that you're refering to SPL in that case).

There are many, many "flavours" of it though, just to make life interesting.

It was developed originally to describe signal degredation across telephone lines

In sound engineering a VU meters 0dB is a max reference, and everything on the gauge is a fraction of that, including "intrument" and "line" levels, -10db and +3dB. 0bD references a signal of .775Vrms which is the voltage level that delivers 1mW into a 600 ohm load resistor.
MilanPanic, I think this is what you mean when you say the majic 1 volt. .775 Vrms is approximately equal to a peak voltage of 1 volt.


When talking about SPL then 0dB is the threshold of human hearing or minimal SPL detectable by the average human ear. Don't ask me how "they" standardised on that 'cos I though we all have different hearing sensitivities which vary with age and health...
Everything else SPL-wise is a fraction of that. Remember that 10 can still be expressed as a fraction ie. 10/1.

There are also other dB references that use thing like energy produced by 1 mW across a specified load, dBm. These load references even vary for different industry segments.

The wiki atricle is pretty good at explaining all of this.

Geeky former Electronics Engineering student starting out in home recording if you're wondering.

This is the long version I was not wanting to get into.

Didn't know the telephone line thing though. Interesting.

studiorat

Paolo_M wrote: »

When talking about SPL then 0dB is the threshold of human hearing or minimal SPL detectable by the average human ear. Don't ask me how "they" standardised on that 'cos I though we all have different hearing sensitivities which vary with age and health...
Everything else SPL-wise is a fraction of that. Remember that 10 can still be expressed as a fraction ie. 10/1.

The threshold of hearing is something like 0.000002 pascals, I could be a zero out there. It's 10* pow -5 anyway. The threshold of pain is about 20 pa. Pascals are units of barometric pressure, (air pressure). Because the difference is so big between the bottom and top of the scale we use the dB scale because it is logarithmic, in that it grows in multiples rather than by addition. Although it doesn't seem like it it makes it easier to work with!!

Talking about digital metering in your DAW or whatever, peak is 0dB fs (Full Scale) which means decibels BELOW the point of clipping, this will be different with a 16 or 24 bit system. Of course it doesn't take into acount intersample peaks, drone drone, prime number frequency to avoid sub mutliples of the sample frequency blah, blah, blah zzzzzz.....

Suffice to say always give your self half a dB or so headroom, and your test tone should be 997Hz instead of 1kHz. And sine waves will be different to square waves and all the rest of it.

Hows that for a nightmare?

madtheory

...so talking about 89dB as a "level" for mastering is nonsense, because the Sound Pressure Level depends on how loud your speakers are.

So again... can you explain what this 89dB business is?

PS 0dBSPL, the threshold of hearing is based on the sensitivity of a young undamaged ear, in the ear's most sensitive range, which is between 1kHz and 4kHz, which is where the ear is most sensitive (see Fletcher Munsen Curve).

For more, see Yamaha Sound Reinforcement Handbook, Chapter 3.

PPS studiorat, the bit depth has no bearing on the clipping point.

studiorat

madtheory wrote: »

PPS studiorat, the bit depth has no bearing on the clipping point.

confused myself there. I was thinking about dynamic range, the level difference is at the opposite end of the system.