chip@vector.UUCP (Chip Rosenthal) (07/18/88)
> From: pec@necntc.nec.com (Paul Cohen) > I would appreciate some help in understanding the CCITT spec G.721 > concerning ADPCM. There certainly was a lot of confusion with the ADPCM specs. As Paul noted, there are a couple of incompatible ADPCM techniques. As background, Adaptive differential pulse code modulation (ADPCM) is a method for digitally encoding speech. Normally, speech is encoded with pulse code modulation (PCM), which uses an 8-bit sample every 125usec. ADPCM uses a 4-bit sample every 125usec. It's advantage is that you only need 32Kbps of bandwidth for ADPCM as opposed to 64Kbps for PCM. With ADPCM coding, the next sample is predicted, and the 4-bit value is the difference between the actual value and the predicted value. The predicting technique depends upon the previous data, thus the term "adaptive". So, ADPCM allows you to double the number of conversations your line can carry with minimal degredation. In 1984, the CCITT released G.721 to specify an ADPCM coding technique. In the meantime, the ANSI T1Y1 committee was working on a standard for the USA. Fairly late in the process, the ANSI committee discovered a problem with the proposed algorithm. It did not work properly on 2400bps PSK modem signals. A modification to the algorithm was proposed, and adopted in July 1986. It is my understanding that the CCITT went back and adopted the T1Y1 recommendation to get around the same problem. There is one significant difference between the 1984 G.721 specification and the ANSI T1Y1 recommendation. The former uses a 16-bit quantizer (i.e. values range from 0000 to 1111), while the latter uses a 15-bit quantizer (values ranging from 0001 to 1111). This is to avoid long strings of zeros in the data stream. Older T1 lines step on bit seven of a channel if all eight bits are zeros, which would audibly corrupt an ADPCM signal. Thus, the two algorithms are not compatible. The AT&T M44 service gives you 44 voice channels on a T1 line using ADPCM. (A T1 line usually carries 24 channels.) The additional T1 bandwidth (2 8-bit channels) is used for bundled signaling so that "off hook" and similar status signals don't rob bits from the ADPCM channels. As a disclaimer, I'm not a totally impartial observer. Dallas Semiconductor makes speech compression IC's which support both ADPCM algorithms. On the other hand, these are my own opinions and not Dallas Semi's. --- Chip Rosenthal /// chip@vector.UUCP /// Dallas Semiconductor /// 214-450-0400 {uunet!warble,sun!texsun!rpp386,killer}!vector!chip I won't sing for politicians. Ain't singing for Spuds. This note's for you.