MUSICAM USA:Tech Tips

Commentary grade systems are used for transmitting voice grade audio.

Emission grade systems assume only one compression cycle will be used.

Distribution grade is used for sending audio between storage devices and allows limited cascading.

Contribution grade describes quality suitable for mastering, and allows multiple cascading.

G.722 is a very early algorithm maximized for 56 or 64 kb/s. G.722 is noted for its low delay, commentary grade bi-directional audio. Low delay makes this algorithm a good choice for remote interviewing and two-way conversations. G.722 is a low fidelity algorithm not suited for music applications. This algorithm does not cascade well and is adversely affected by post processing.

Optimized for 128 kb/s per channel, standard unenhanced MPEG Layer 2 can deliver CD quality stereo at 256 kb/s, joint stereo at 192 kb/s, and monaural at 128 kb/s. At these bit rates, standard MPEG Layer 2 can deliver distribution grade audio that is relatively immune to post processing and limited cascading. MPEG Layer 2 codecs not made by MUSICAM USA or CCS Audio Products use unenhanced Layer 2 coding, with minor variations from manufacturer to manufacturer.

MUSICAM Layer 2 encoding is an exclusive, fully compatible enhancement to the ISO/MPEG standard and yields superior audio performance at lower bit rates. MUSICAM is optimized to deliver full bandwidth, superior quality joint stereo at 112 and 128 kb/s, true stereo at 192 kb/s, and emission grade monaural at 56 and 64 kb/s. At these bit rates, MUSICAM can deliver audio relatively immune to post processing and limited cascading. Since all enhancements take place in the encoder only and are fully compatible^** with all standard MPEG Layer 2 decoders, all standard MPEG Layer 2 decoders, regardless of manufacturer, will sound better when connected to a MUSICAM codec.

Optimized for 56 and 64 kb/s per channel, MPEG Layer 3 can deliver very good sounding emission grade true stereo, with 15 kHz frequency response at bit rates as low as 112 kb/s, mono at 56 kb/s. near-CD-quality joint stereo, with 20 kHz frequency response is obtained at 112 kb/s. However, this wider bandwidth comes at a price. MPEG Layer 3 has a relatively long delay and does not cascade or post process well. In addition, some users of Layer 3 at lower bit rates have noticed audible artifacts in the higher frequencies. At higher bit rates Layer 3 delivers CD quality stereo audio, but still suffers from long delay and poor cascading performance.

Choosing an algorithm
When choosing an algorithm/bit rate/sample rate combination, you must ask yourself some questions:

What final quality do I want? In some instances, your audio will be subject to multiple encode/decode cycles as it moves through the program distribution chain. Although some algorithms cascade better than others, any algorithm, when used at low bit rates, does not cascade well.

Will there be any post-processing or equalization used? Some algorithms are negatively affected by post processing.

Is delay a factor? Encode/Decode delays can range from a few milliseconds to nearly a second. Long delays make 2-way communications difficult.

What brand codec is at the other end? Yes, this does make a difference.

Which end will be receiving the audio? If one end is not a MUSICAM USA or CCS Audio product, it does make a difference.

Is cost a factor? When using ISDN, higher bit rates require multiple B channel connections.

If you are sending stereo, another question that needs to be answered is which mode to use; stereo or joint stereo

Stereo and Dual Mono—From a coding standpoint, these two modes are identical. Half of the total bits available from the transmission line are allocated for the left channel and half the bits are allocated for the right channel. Left is always left, right is always right. No blending, no sound-stage manipulations.

Joint stereo—Unlike stereo, joint stereo employs real-time bit allocation techniques that dynamically assign bits to the channel and frequency bands that need them most. This results in wider bandwidths and better dynamic range. Joint stereo does manipulate left-right separation and sound stage; however, any spatial blending performed is in frequency bands where the human auditory system would normally blend the stereo signal. Joint stereo defaults to true stereo if enough bits are available. Note that real-world listening tests awarded higher scores to Joint Stereo than to discrete at low bit rates.

At 56 and 64 kb/s, mono
At low bit rates, best results are obtained using low sample rates.

Use G.722 only if necessary for compatibility or low delay. G.722 is not a high fidelity algorithm and the only advantages of using this algorithm are its near-universal availability and low delay.

Deciding whether to use MUSICAM Layer 2 or standardized Layer 3 is not as simple. On a single ISDN B channel, standard Layer 2 can deliver 8.2 kHz frequency response, MUSICAM Layer 2 can deliver 10 kHz and Layer 3 can deliver 15 kHz. However, frequency response should not be used as the sole criterion of which algorithm to use. In many cases, the artifacts introduced by Layer 3, especially after post processing, may make the added bandwidth unusable. Other factors to be considered are: 1) brand of far end codec and 2) who will be receiving the audio.

If both codecs are MUSICAM USA products, listen to Layer 3 and MUSICAM Layer 2. Use 24 kHz sampling for Layer 2; 24 or 32 kHz sampling for Layer 3. Your own ears should tell you which sounds best for the application. When you do your test, it is not necessary to drop the connection when changing algorithms or sample rates.

If the far end codec is not a MUSICAM USA product, use layer 3 at 32 kHz sampling if you are receiving the audio. Other manufacturers use basic, unenhanced MPEG Layer 2, and the received audio may sound weak and muddy at low bit rates. If far end users are recording your audio, experiment with MUSICAM Layer 2 and Layer 3 at low sample rates. Let their ears make the decision. Remember that it is not necessary to drop the connection when changing algorithms or sample rates.

Monaural audio at 112 or 128 kb/s is as near to perfect as possible with any algorithm, although MUSICAM Layer 2 allows more cascading (up to 5 times without audible degradation) than either Layer 3 or unenhanced Layer 2.

At 112 or 128 kb/s joint stereo, the best algorithm to use depends more on the brands of the codecs at both ends than anything else, although post processing and cascading should also be considered. Independent studies have shown that MUSICAM Layer 2 sounds best, but listeners and users say that MPEG Layer 3 sounds equally as good, with unenhanced Layer 2 a close third.

If both codecs are CCS Audio or MUSICAM USA products, use MUSICAM Layer 2. This will yield the best possible audio, with the highest immunity to post processing and cascading artifacts. You will also benefit from MUSICAM's short delay. If one end is not a MUSICAM USA or CCS product, then the algorithm choice depends on which end is receiving the audio. If the MUSICAM USA product is the receiving end, use Layer 3. If you are sending, use MUSICAM Layer 2 for the best audio.

If true stereo is required, or if different audio programs are to be sent on the left and right channels, use Layer 3.

MUSICAM USA wants you to get the most from your codec. If you have questions, comments, or require additional information, please contact us.

MUSICAM USA
Corporate Computer Systems, Inc.
670 North Beers Street, Building 4
Holmdel, NJ 07733 USA
Phone (732) 739-5600 • Fax (732) 739-1818
http://www.musicamusa.com

^*cdqPrima only, not available in RoadRunner.
^**Certified by the Institut fur Runfunktechnik (IRT).