Exporting as MP3 creates reverb/echo in recording

[bfuhr]

I haven't resolved it just yet and I keep backing the compression down. I'm starting to go nuts after about ten renderings. I've never had a problem with too much bass before.

I know all voices are different, but maybe one of the audiophiles here could suggest a compression setting to treat a raw voiceover of reasonable volume that would:
a) give sufficient warmth and strength
b) be friendly at say, 96kbps.

I can't find any documentation (that I can interpret) on the internet that details what frequencies the MP3 compression algorithm will truncate.

[elove]

do you think this has anything to do with Vista?

[kozikowski]

<<<what frequencies the MP3 compression algorithm will truncate.>>>

That's because it doesn't. MP3 maintains a wide bandwidth audio channel almost to the bitter end. That's what's so crazy about your problem. This isn't an MP3 problem.

The common explanation of how MP3 works is that it suppresses or eliminates the quiet parts of a performance that you wouldn't be able to hear anyway, leading you to believe that in a five violin concert, you only get the three loudest violins. It's much more subtle than that.

Let's say in the same five fiddle concert, three of the violins were excellent, top quality instruments run by the best musicians in their field and the other two were good musicians driving Sears and Roebuck violins. In the real performance, even I would be able to tell who was who because the differences would be striking. After overcompression in MP3 (intentional damage) all the instruments would be able to hit all the original notes in the piece exactly correctly, but they would all be Sears and Roebuck. All the delicate overtones and tiny variations between the instruments would be carefully filtered out for greater compression efficiency. If you kept going, you would find some of the complex bowing, strike, tail and emphasis would be missing leaving all the musicians sounding pretty much the same--good, but not great.

At no time would any of the notes in the performance be missing--even though the show would start to sound like garbage.

Since I'm no longer sure who is having the problem because of thread hijack, I would be Exporting the show as a WAV file, playing it in Windows Media (not Audacity) to make sure it sounds OK and then convert it to MP3 in Something Else.

Koz

[kozikowski]

Another note. As the Linux users could tell you, lame is a stand-alone program and you can use it on the linux command line with nothing else running.

>> lame piano2.wav piano2.mp3 -h

That will produce my piano recital as a high-quality MP3 file 1/10 the size of the original WAV file.

I'm looking for a version of lame that you can use inside Windows like that. You should look too.

Koz

[Norma]

Hi again
Sorry for taking so long to get back to you.

I went in and changed the MP3 Export bit rate to 128 (under edit-preferences-file formats) and the play back became clear. no more echo and tunnel-like sound. Thanks so much for help.

[bfuhr]

Since I'm no longer sure who is having the problem because of thread hijack, I would be Exporting the show as a WAV file, playing it in Windows Media (not Audacity) to make sure it sounds OK and then convert it to MP3 in Something Else.
Koz

I'm not so much a hijacker as a thread reviver. The one prior to mine ended over a year ago with no peace and I asked if there had been any conclusion.

When I exported the original to WAV it sounded excellent. Something in the compression, and I won't blame it on audacity as it also sounded crummo when rendered in Adobe Audition. But anyways, today I'm rereading a third time and I'm going to apply the following filters manually as advised by a recording friend.

"I’d suggest a 2.5:1 ratio and use EQ to remove the excess bass. Put the EQ in the chain first, then the compressor. Try a low shelf filter around 150hz and try a cut filter at various locations (i.e., -3 to –5 dB at 400hz, and a +1 or +2 dB boost at 2K) to see how it works. There may also be presets."

[steve]

I'm not so much a hijacker as a thread reviver. The one prior to mine ended over a year ago with no peace and I asked if there had been any conclusion.

I don't think koz was specifically criticising you bfuhr, but it does get confusing when there are several enquiries in the same thread. In this case, the original question was from Norma on Mar 02 2008, followed by your "revival" on Mar 04 2008, followed by elove's "revival" on Mar 06 2008, followed by Norma's solution on Mar 07 2008, and now back to you - so I think you can see how it gets confusing.

"I’d suggest a 2.5:1 ratio and use EQ to remove the excess bass. Put the EQ in the chain first, then the compressor. Try a low shelf filter around 150hz and try a cut filter at various locations (i.e., -3 to –5 dB at 400hz, and a +1 or +2 dB boost at 2K) to see how it works. There may also be presets."

That sounds a pretty reasonable starting point.

The first thing should always be to get as good and clean a signal in the original recording - allow a good amount of overhead to avoid clipping (even with as much as 12 to 15dB overhead it is easy to get occasional peaks that are close to 0dB).

Using a spider or some other means to isolate the microphone from vibrations should be used.

Record in a quiet room with nice acoustics (not in the bathroom or a bare cellar unless you want it to sound hollow and echoey)

Avoid blowing on the microphone - a pop filter can help.

Adjust the microphone placement to achieve a nice tonal ballance. Many microphones (those with directional pick-up patterns) will exhibit a "proximity effect" that will raise the level of the bass as the distance between the sound source and the microphone gets smaller.

Adjust the recording level so that it peaks close to 0dB, but NEVER reaches 0dB.

Once you have your recording:

Apply Eq before compression (be careful if you are boosting any frequencies that it does not cause clipping).

If you need to use noise reduction, use it after Eq and before compression or any other dynamic processing.

In Audacity it is often easiest to amplify to 0dB before using compression.

A compression ratio of 2.5:1 should be sufficiently gentle to sound "transparent". The "threashold" level will depend on the actual material (this is why amplifying to 0dB first helps), but about -12dB is a reasonable figure. The LADPA SC4 compressor offers much finer control than the standard Audacity compressor, but is less easy to use as there are more controls.

Finally amplify up to about -0.1 dB. This will allow a little headroom for the mp3 encoder.

Try a low shelf filter around 150hz and try a cut filter at various locations (i.e., -3 to –5 dB at 400hz, and a +1 or +2 dB boost at 2K) to see how it works. There may also be presets."

Those are good ball-park settings for recording a voice, and are similar to the response of microphones that are specifically designed for voice. Audacity does not come with Eq presets for voice, but in v.1.3.4 you can save your own settings.

I'm looking for a version of lame that you can use inside Windows like that. You should look too.

I use Lamedrop in Windows and find it very easy and convenient.