I am concerned about my tweeters in relation to signal coming from the computer.

I have machine LINE OUT fed into my preamplifier (hardware audio component) all the time. The max levels of tuner, tape recorder, turntable, etcetera are all "equalized". That is, they all more or less deliver the same output level to the preamp. The difference between a CDDA level and that of another one is always minimal. The same goes for vinyls (not "so minimal"). As to radio broadcasts, one may be xmiting with great power and make differences larger, but still no harm for the tweeters, relatively speaking.

Mine are 50w four-way loudspeaker systems (forgive the word, I used to call them baffles). The power amplifier is 30w per channel so, theoretically, it cannot damage the speakers. But the preamp (Yamaha C-2) can deliver a very large signal. When this signal is unduly large, the woofer and mid-range speakers won't suffer. But the enormous resultant distortion can damage the tweeters.

Let us now come to the computer. Some audio files have the signal recorded at very high levels. And I could launch aplayer, play or mplayer, or any GUI player, or I could receive an email audible notification when amixer's output controls are near the top and put my tweeters in danger. I think the core of the problem is the huge variation in level between different audio files.

This has been the problem exposition. But it has to be the problem of every Linux user who connects his audio equipment to his computer. And hence, there must be a well known solution. Maybe you guys know one.

I would use 'normalize' to normalize the volumes of all the audio files. You can even set decibel limits.

Thank you for replying. It seems a great solution. I could come to some conclusions as to the applications of 'normalize' by trial and error, but first I'd like to fully understand the manual. I have read it, and it seems there are only two audio formats it can normalize: WAVE and MPEG, the latter by modifying the metadata. Is this true?

Yeah, it's made for normalizing wav files, then you can convert them to other formats and it should be ok.

Not entirely true.

Giving speakers a distorted signal less then 100% of their maximum power can damage them while good speakers can handle more then 100% of their power as long as it's clean.
I have driven systems for small and large venues with up to 200% of clean undistorted power for 1-2 hours and not damaging a single component while I have seen speaker break at 50% of their maximum power rating.

Having a amplifier with more power is the best choice in most cases unless your goal is to have distortion.
This goes for all components from source until transducers : you need 'headroom' on all components.

Here is a article that might be of interest : http://www.jblpro.com/catalog/suppor...=246&doctype=3

PS : I know it has nothing to do with Linux, but just wanted to point that out

And of course, if I have downloaded a file which is not wav, I can convert it to wav, run normalize on it and reconvert to the original format. Am I right? This is question #1.

Question #2: what about the fidelity (quality) of the final product respect of the original? Does the signal not loose some quality with each new conversion?

Regards.

There might be some loss but I think you should ask yourself if you will be able to hear it or will your system be able to reproduce it.
If you take a 128kpbs MP3 (which is pretty low) convert it to standard wav you shouldn't have any loss because it's converting it to a uncompressed format, it's the action of converting it back to a compressed format that will be critical.
Reconverting it to 128kpbs might result in some quality loss but the original was already not that great, so it's up to you, just listen to it and judge for yourself.
I would opt to always increase the kbps rate one level after such operations if it's under or equal to 224kbps.

PS : when you listen to a track you converted try to not convince yourself of hearing quality loss, lots of people tend to convince themselve to hear thinks that aren't there

I should have underlined the word theoretically. I quite understand what you say and am looking forward to read your link. Notice how I did special mention of the tweeters, who are specially sensible to distortion, according to what I have heard. The reason is easy to understand, if you consider that a saturated stage in the amplifier will produce high frequency harmonics. So the power delivered to the speakers is mostly high frequency power, and the crossover will take care of routing it to the tweeters. Thanks for your post, Dani1973.

It would be interesting to run such an experiment, keep converting between mp3 and wav with the same settings every time, will loss accumulate and how quickly ? I'll have to run the experiment. I think there will be loss as Dani1973 suggests, but maybe not as quickly as one conversion.

I always use VBR to encode mp3s or 160 kbps.

And then one should measure loss. And I think this would not be very difficult. Using cmp between the original and a nearby replica, and them again between the original and, say, the 50th replica, within a script able to perform some adittional tasks, or better yet, by means of an ad hoc program that would take the mean of the differences.

I'll perform the experiment too. P.S.: why do you not fall back on sox?

Yes, the same effect can be accomplished with sox, so you can try that.

I don't quite understand. I take a 224kbps mp3, convert it to wav, normalize with the normalize command and convert back to mp3. If I understood well, I now would increase the kbps rate one level. And what would the benefit of this last operation be?
EDIT: I misunderstood you. I convert to wav, normalize and then reconvert to mp3 but using a higher bitrate now. Still, the question remains. Can conversion to a higher bitrate improve the degree of fidelity of the sound to the sound captured by the microphones during the recording session?

You will not improve the quality but avoid more loss then the original conversion.

Once you convert audio to let's say 192k you will loose some quality (especially in the high frequency range).
When compressing occurs the loss is mostly on places where the magnitude changes very fast (that's high frequency).
Since you will be normalizing tracks this means you will be most probably increasing those magnitudes thus making it more sensitive to compression.

If normalising results in a magnitude drop (track was too loud) you should ask yourself if the track wasn't distorted in the first place.

PS : converting a track to a higher bitrate without modifying it (plain conversion to wav and back to a better format) will not improve the quality since 'damage' has been done (info is lost and it will not magically reappear)

Using cmp will just tell ou files a different but not give you any information on the sound quality (I wish it was that simple).
Complicated RTA analysis would be required here but it still would be 100% clear if one could hear the difference or if his speaker would be able to reproduce the difference.

Try this:

I did this for a song, and you can notice distortion at i=15, and extreme distortion at i=100, I mean really really bad.