Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!elroy.jpl.nasa.gov!jarthur!nntp-server.caltech.edu!fjs From: fjs@nntp-server.caltech.edu (Fernando J. Selman) Newsgroups: comp.sys.next Subject: Re: The existance of a Software Modem and Fax Machine... Message-ID: <1991Mar23.023024.15154@nntp-server.caltech.edu> Date: 23 Mar 91 02:30:24 GMT References: <1991Mar19.214415.27530@oracle.com> <2cqt023R06Lj01@JUTS.ccc.amdahl.com> Distribution: na Organization: California Institute of Technology, Pasadena Lines: 58 kls30@duts.ccc.amdahl.com (Kent L Shephard) writes: >In article scott@mcs-server.gac.edu (Scott Hess) writes: >>In article <1991Mar19.214415.27530@oracle.com> wayer@oracle.uucp (William Ayer) writes: >> At a demonstration of the NeXT machine two years ago, Jobs suggested >> that the DSP chip would allow programmers to create a software modem / fax >> machine (very little hardware would be needed i.e. a cable). Was this >> suggestion wishful thinking, or could such a beast exist? >> >>Well, this is technically possible. Basically, the internals of a modem >>are a DSP chip plus some glue logic [ :-) ]. >> >>The problem is that in a modem the DSP is dedicated, while on the NeXT >>it isn't quite so dedicated. The amount of work needed to be done >>outside the DSP is greater on the NeXT than when it's off on that >>serial port. The DMA support isn't really there, etc, etc. You could >>upgrade your DSP memory (that would fix most of those types of >>problems), but in general I think it's a close one as to whether or >>not you can run a 9600 baud modem off the DSP. >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ >9600 baud is not the hardest thing in the world for a 56k DSP. Why now >do you think it is possible to have a dedicated modem with a slow DSP >and not a fast one? 9600 is "possible". To be realistic a 25Mhz 56k >should be able to handle 19200 baud. Remember you can do real time >spectral analysis with monster scope. That is a lot harder than >determining a bit from a phase modulated signal. (The person I told FSK. >I meant PM- phase modulation). >Simple A/D converter and a D/A converter with a phone line interface. >you could get away with bits for conversion because the frequency that >is modulated is not that high. Remember that phonse have almost no >frequency response above 3.5Khz. The DSP can handle sampling rates up >at least 88Khz. You can get a telephone interface from Radio Shack. >So the hardware is not difficult. It is my impression that when you say 9600 baud you mean 9600 bps. This is not a trivial difference. As you correctly stated, the phone lines have an upper cut-off of 3.5 kHz, thus they can handle 2400 bauds but not 9600. The modems that claim 9600 bps are actually 2400 baud modems using a modulation scheeme called QAM, Quadrature Amplitude Modulation, in which they can pack 4 bits [sic] in each period of the 2400 baud wave. This scheeme is the one used in all V.32 modems. But these modems also need to do echo cancellation to operate in full duplex mode, and I understand (but I might be wrong) that this have to be done in hardware, otherwise the local modem could not "understand" what the remote modem is saying because of the echoes from his own transmissions. I believe this to be the reason why Telebit modems are assymetric, that is not full duplex, they have not been able make an echo cancelation scheeme that works at the higher frequencies they use. Telebit uses an scheeme called DAMQAM, where QAM is the same as above, and the DAM part refers to a propietary scheeme used to send information through multiple channels that are constantly monitored. All this is possible because of the use of the glue chips mentioned above. This is why I think that to connect a simple black box with a CODEC chip to the DSP port and to then try to implement a high speed modem in hardware is impossible, IMHO. - Fernando >/* Kent L. Shephard : email - kls30@DUTS.ccc.amdahl.com */