4. Motion Picture Experts Group
โข MPEG-1:Video CDs, MP3 Audio
โข MPEG-2: Digital TV, Set-Top Boxes
โข MPEG-4: Online Multimedia (Video)
โข MPEG-7: Audio and Video Search
โข MPEG-21: Multimedia Framework
5. MPEG-1 for Geeks
โข Layer 1
โข Simple 32-Band Algorithm
โข Philips DCC (Digital Compact Cassette)
โข Layer 2 (a.k.a. MUSICAM)
โข Also 32 Bands
โข International Standard for Broadcasting
6. MPEG-1 Layer 3 (MP3)
for Geeks
โข Psychoacoustic Masking
โข 32 Bands Divided into 576 Subbands
โข More Accurate Masking Thresholds
โข Redundancy Reduction
โข Lossless Huffman Encoding
โข Bit-Reservoir Buffering
โข Joint Stereo
7. Sample Rate for Geeks
โข The Nyquist Theorem
โข Sample at 2x the Highest Frequency
โข 22.05kHz Sample Rate for 11kHz Audio
โข Sample Rate Is aSource (WAV or AIFF)
Property of
Uncompressed
8. Sample Rate in Practice
โข Standardize on 44.1kHz Sample Rate
โข Flash & Other Players Require n*11.025kHz
โข Resample if Source is 48kHz from DVDs
9. Bit Rate for Geeks
โข Independent of Sample Rate
โข Specifies Encoder Output File Size (CBR)
โข @64kbps, 1 hour โ 27MB
โข Variable Bit Rate (VBR)
โข For Higher Bit Rates Only
โข Not Universally Supported (Avoid It)
10. Bit Rate in Practice
โข โUse Higher Bit Rates for Music?โ
โข Itโs a Myth!
โข Human Voices Are Complex
โข Music Masks Its Own Artifacts
โข 64kbps is Most Common Today
โข 96kbps is Gaining
11. Podcasting Bit-Rate History
โข June 2003: 32kbps. โFiles too largeโ
โข April 2004: 48kbps. โNo problemโ
โข September 2004: 64kbps. โQuality is lowโ
โข Today: Still 64kbps.
โข Tomorrow??
12. Stereo Encoding
โข โStereo MP3s are twice as large as mono.โ
โข Itโs a Myth!
โข Only Bit Rate Specifies Output File Size
โข You May Want to Use Higher Bit Rates for Stereo
13. Stereo Encoding for Geeks
โข Dual Channel or Independent Channel (IC)
- Entirely Separate Left and Right
โข But Most L/R Information is Redundant
โข Intensity Stereo (IS)
โข Mid/Side Stereo (MS)
โข Joint Stereo (JS) Allows IS/MS Combination
14. Stereo Encoding
(Even Geekier)
โข JS Encodes L+R and L-R
โข If L=R then L-R=0
โข SinceUsesRate is ConstantStereo Information
Bit
L=R Fewer Bits for
15. Stereo Encoding in Practice
โข StereoReason to (not Music vs.Voice) is a
vs. Mono
Good Use Higher Bit Rates
โข Greater Separation Suggests Higher Rates
โข If Mostly Speech, Consider 100% Mono
โข If Mono, Make L&R Digitally Identical
โข Always Encode in Stereo for Compatibility
16. Mastering for MP3
โข Help the Encoder: Eliminate Unnecessary Data
- High-Pass Filter at 80Hz
- Low-Pass Filter at 11kHz (@64kbps encoding)
- Normalize
17. Which is Louder?
โข Itโs Not the Height of the Peaks (voltage)
โข Itโs the Area Under the Curve (power)
18. Loudness
โข Whatโs the Standard?
โข We Asked:
- Podcasters
- Audio Engineers
- Radio Engineers
โข Answer: There Isnโt One
โข Itโs a Hard Problem to Solve
19. Normalization
โข Peak Normalization (common)
- Maximizes Voltage, not Power
โข RMS Normalization
- Maximizes Power (=Loudness)
โข Determine a Standard Loudness Level
20. Avoid Recording to MP3!
โข MP3 is a final/release format.
โข Not designed to be decoded and re-encoded.
โข Use MP2 Instead...
โข or the highest MP3 bit rate possible.
21. AAC/M4B Files?
โข Yes, AAC is Better Than MP3
โข We Added AAC to Support iPod Bookmarks
โข Painful: Only iTunes Could Encode M4B
โข Doubled Much of Our Workflow
โข Canโt Be Easily Assembled
22. MP2: Why and When?
โข MPEG-1 Layer 2
โข Designed as an Intermediate Format
โข The Standard in Broadcast Radio
โข 128kbps per Track
โข 44.1kHz Sample Rate Preferred
24. To Summarize
โข Record at 44.1kHz Sample Rate (not in MP3!)
โข Mastering
- RMS Normalization (Pick a Standard Level)
- 80Hz Hi-Pass, 11kHz Low Pass (for voice)
- If Mono, Make L&R Digitally Identical
โข Encoding
- 64kbps when L=R
- Consider โฅ96kbps for Lโ R
- Always Use Joint Stereo