Mpeg encoding notes:

File used for comparisons = "MintyQ2.wav" from the MSU Library's writers series, 4/23/99.
This 12.5 second wav file is of high quality, originally mastered onto DAT at 44K/16bit/stereo. 

MP3:

Used BladeEncoder for first 64k/s pass:  Quality is very good.  No noticable or meaningful  
					 differecnces from original.  File size = 98K  

All remaining passes used NexEncode:
		    	      48k/s:	 Still quite good response with some artifacts.
					 Quality very acceptable for voice recordings.
					 File size = 74K

			      40k/s:	 Good response.  Slightly noticabe difference in 
					 some speech sounds, consonant attack.  Some 
					 "whiring" artifacts in background.  File size = 62K

			      32k/s:	 Noticable speech distortions and undesirable 
					 background artifacts. File size = 49K

Mpeg layer 1 and 2:

Using Cool Edit 96, the following files produced results as noted:

minty48L2M1.mp2 
MP2 Model-1 Musicam @ 48k/s:  	Result is muffled with highs noticably effected.
				No notiable artifacts.  File size = 74K
minty48L2M2.mp2
MP2 Model-2 AT&T @ 48k/s:	Result is clear and natural, but highs are noticably effected.
				No notiable artifacts.  File size = 74K
minty64L1M1.mp1
MP1 Model-1 MUSICAM @ 64k/s: 	This result using layer 1 is clearer than 
				the MUSICAM layer 2 at lower bitrates.  File size = 98k
minty32L1M1.mp1
MP1 Model-1 Musicam @ 32k/s:	Result is very distorted and unusable.
				File size = 49K	
minty32L1M2.mp1
MP1 Model 2 AT&T @ 32k/s:	Result is too distorted and undesirable.
				File size = 49K
minty32L2M1.mp2			
MP2 Model-1 Musicam @ 32k/s:	This layer 2 32k/s version is an improvement over the same 
				model-1 and bitrate of layer 1 above.  
				However, there are still noticable and unacceptable 
				muffled distortions.  File size = 49K
minty32L2M2.mp2
MP2 Model-2 At&t @ 32k/s:	This 32k/s file provides the best response at this bitrate.  
				It is even better than the 48k Musicam version providing a 
				clear result with no distortions.  Even with the highs muted, 
				this scheme is preferred over the 32k/s mp3 that has noticable artifacts.
				With file size = 49K, this would be a good choice for very large files.
minty64L1M2.mp1
MP1 Model-2 At&t @ 64kb/s:      Similar in sound to the mp3 32kb result with a little more 
				distortion and larger size.  File size = 98K
minty64L2M1.mp2			
MP2 Model-1 Musicam @ 64kb/s:	Not as bright sounding as the Mp2 32kb result using model-2.
				File size = 98K
minty64L2M2.mp2
MP2 Model-2 At&t @ 64kb/s:	Very Similar in quality to the mp3 48kb version, 
				but uses more space.  File size = 98K
minty56L2M2.mp2
MP2 Model-2 At&t @ 56kb/s:	Very Similar in quality to the mp3 40kb version, 
				but uses more space.  File size = 86K
minty56L2M1.mp2
MP2 Model-1 Musicam @ 56kb/s:	Not as bright sounding as the Mp2 32kb result using model-2.
				No significant difference from the 64L2M1.mp2  File size = 86K.


Conclusions:  

Not surprising, the brightest, most accurate or best overall sound is obtained 
by using mp3 in most cases.  The layer 2, model-2 32kb/s bitrate is preferred over
the mp3 of the same size and bitrate.  In fact, the only acceptable option for mpeg at 32kb/s
is the layer 2, model-2 algorithm.  Model-2 is the best choice for all non mp3 schemes under layer 1 or 2.
Best results can be optained by combining layer 2 with model-2 (AT&T) at all bitrates.

For smaller files or most accurate reproduction for voices, use mp3 at 48kb/s.
For larger files or sound with a little less "edge" or brightness, mp2(model-2) @ 32kb/s works well.