The MP3 files here are recorded for the purpose of CD burning and online listening. Here's a general overview of how.

I. Hardware and Software used
=============================================
Hardware used:
iRiver iFP-390T (256mb) or 380T (128mb)
A computer (the faster the better)

Software used (free):
CDex Version 1.51 (http://cdexos.sourceforge.net)
MP3Gain v1.2.2 (http://mp3gain.sourceforge.net)
Mega Mp3Splitter 1.1 (http://digilander.libero.it/mega27/megasoft.htm)
Mp3 Tag Tools v 1.2 Build 008 (http://massid3lib.sourceforge.net/)
dBpowerAMP Music Converter (http://www.dbpoweramp.com/)
Windows Media Encoder (http://www.microsoft.com/windows/windowsmedia/download/default.asp)

Software to consider (also free):
LAME version 3.93 MMX (http://lame.sourceforge.net/, see also http://mitiok.free.fr)
all2lame v1.9.1 (http://home.wanadoo.nl/~w.speek/speek.htm)
RazorLAME (http://www.dors.de/razorlame/)
Audacity (http://audacity.sourceforge.net)
mp3DirectCut (http://www.rz.uni-frankfurt.de/~pesch/)
DeepBurner (http://www.deepburner.com/)
CDBurnXP (http://hem.bredband.net/cdburnerxp/)

II. Acquiring the source audio
=============================================
The iRiver iFP-380T/390T is a flash MP3 player and voice recorder. It has a stereo line-in jack which can be used to record MP3 audio. I record the audio on the line-in jack at 96kbps at mono. The iRiver iFP-390T has a capacity of 256mb and cost (at this time) $199 U.S. dollars. The iFP-380T has a capacity of 128mb and costs $139. At 96kbps, I can record about 5:30:00 of audio on the 390T, which allows me to record at least two, if not three, events. The 380T gives me 2:45:00, which should be sufficient if you are diligent in archiving your audio files.

Why not record at 128kbps? This is due to a firmware issue on the iRiver which caused clipping and other distortions if audio was encoded at 128kbps. As of the date of this writing, January 2004, iRiver's UMS (USB Mass Storage) firmware version 111 addresses the issue by limiting the line-in jack to 96kbps and using FastENC for MP3 encoding.

Why record in mono instead of stereo? The final output file I desire is mono, so recording in stereo is a waste bandwidth. Also, note that a mono 96k MP3 has the same quality as a 192k joint stereo stream recoded to mono. I get a better quality source file by encoding in mono.

Since I am recording sermons from a live church service, I am plugging directly into the sound board. I am using the Phone Out jack, which is pre-fader, in order to avoid someone turning down the pastor's mic for the house and killing my recording level. I record the audio file with the dial at unity (0 gain). Ideally, I would use a pre-fader mono aux send, but my setup at church limits me. Also, I find that disabling all effects, such as reverb, is best for the speech portions of the recording.

III. Splitting and compressing the audio
=============================================
I am using Windows XP, although most of this software if available for free on Linux as well, with the exception of Mega Mp3Splitter. If you are not on Windows, you can substitute Audacity for Mega Mp3Splitter.

A. SPLITTING THE MP3 FILES
I use Mega Mp3Splitter to (duh) split the source MP3 into smaller chunks. First, I listen to the MP3 file and note the times where the sermon begins and ends. Then, I load the file with Mp3Splitter and create the smaller audio files. The interface is intuitive, so I will not spend time here explaining it. This usually results in a 40mb file for me, which we'll call SERMON.MP3.

B. ADJUSTING THE VOLUME
Run MP3gain, and analyze SERMON.MP3. I use a level of 92 db for my baseline, although the suggested level is 89 db. Adjust the gain with MP3gain, you'll be glad you did. The human ear cannot detect changes smaller than 1.5 db, so changes in increments of 3 db is suggested.

C. TAG THE FILES
Using "Mp3 Tag Tools" I label the tracks with the speaker, our church info, and so on. Tagging the split up files allows all transcoded files to also have the same tags, thus saving you work later.

D. PREPARING MP3 FILES FOR UPLOAD
I use CDex, with Encoding Option set to "Windows MP3 Encoder" with a Codec Bitrate Selection of "16kBit/s, 11k Mono" and "High Quality" checked. This results in a FastEnc (FhG) encoded MP3 file, which sounds better for low bit rates than LAME encoding does. The file is about 17% the size of the original.

E. PREPARING WMA FILES FOR UPLOAD
I used dbPowerAmp Music Converter and the WMA redistributable from the dbPowerAmp. I also take the tagged SERMON.MP3 and compress it using "Windows Media Audio 9 Voice" at 12k. The resultant file is smaller than the compressed MP3, and sounds a little tingy but the quality is quite good. dbPowerAmp will copy the tag information from SERMON.MP3, whereas the Microsoft encoder does not. The file is about 13% the size of the original.

F. PREPARING CDS
If I am making an audio CD, I take the adjusted and tagged SERMON.MP3 and load it with Mp3Splitter again. This time, I break it up into 10-minute chunks. The resultant files are what I use to burn to CD. The track splitting is only to make finding where you left off less tedious. (Have you ever fast-forwarded through 45 minutes of audio?)

IV. A few final comments
=============================================
Why MP3? Why not Real Audio, or Vorbis? I selected MP3, not for its quality or size, but for the ubiquity of the format. Everybody has a MP3 player/decoder on their computer, and it will "just work." Although I am using FastEnc, I am not be required to pay royalties, from what I understand, see the URL http://www.mp3licensing.com/help/enduser.html#4 for details. If you are distributing the files commercially, you could use LAME, and then you do not need to concern myself with royalties. Of course, if you're making enough money (at least $10,000 annual) on this, you can afford a real encoder.

Why WMA? Windows Media Encoder (WMA) encodes voice at passable quality as low at 8kbps, and 12kbps is significantly better. WMA has size as an advantage, the encoder is free, and there are no licensing fees for redistribution (http://www.microsoft.com/windows/windowsmedia/create/licensing.aspx#audio).< /p>

Why not OGG? I considered Oggs Vorbis, but while a quality of -1 does save 30% in size (a 6mb MP3 becomes a 4.2mb OGG), the Vorbis codec is simply not intended for voice, and does not sound as well. Without a size advantage, it doesn't make sense to transcode to Vorbis. Royalties are also not an issue here, but there is no compelling reasons to use OGG. (http://www.vorbis.com/)

Finally, many folks, just as myself, have portable MP3/WMA players. Vorbis and other formats I have considered are usually not supported, or not supported at low enough bit rates for me to worry about them.

I initially used LAME to produce the smaller, lower-quality MP3 files to save space. I use all2lame (although any LAME front-end will do) and us the following encoding parameters: --short --athtype 3 --lowpass 9.1 --cwlimit 7.2 --substep 0 --ns-bass 6 --ns-alto 13 --ns-treble 21 --ns-sfb21 6 --strictly-enforce-ISO --vbr-new --abr 24 --verbose -q0 -b 8 -B 40 -mm
For notes on which version of LAME to use: http://www.hydrogenaudio.org/index.php?showtopic=478
I am using LAME 3.93 because 3.90.3 doesn't work with All2LAME & tag.exe.
The file is about 20% the size of the original.

I'm not an audio engineer or anything like that. I've done my homework and a lot of reading, but I'm just a computer geek that loves Jesus and wants to further His kingdom. I'll try to help you if you are trying to setup digital recording at your ministry, just don't expect a pro. You can contact me via email at
lee DOT jones AT fccfontana DOT org

TO GOD BE THE GLORY,

Lee

Recording History

44khz, 128kbps, stereo line-in

12-07-2003 - recording volume was too low to be usable, was recording off the "tape out" but the faders affect that.
Switching to "Phone Out" instead, which is pre-fader.
12-14-2003 - sounds just fine
12-21-2003 - OK until about 25 minutes into sermon, then clipping and corruption.
I thought it was the battery, and tried a new battery.
12-24-2003 - OK, but muffled because mic was under costume. Tests still indicate that the clipping issue is persistent.
12-28-2003 - clipping and corruption began before sermon (should re-record from tape).
The fresh-battery approach didn't make a difference.

44khz, 96kbps, stereo line-in, UMB firmware version 111

01-04-2004 - sounds just fine. I had upgraded the iRiver firmware.
Subsequent 3.5 hour recording test was also OK. This UMB firmware limits line-in to 96kbps.