Subject: Re: An idea using emulation
From: "Wolfram T. Arnold" <>
Date: Fri, 23 Apr 1999 15:34:23 -0700 (PDT)

On 23 Apr 1999, Russell Nelson wrote:

> John J. Cruz writes:
>  > I'm not sure if you're aware but it is not just the lack of a Linux driver
>  > that makes winmodems not good devices to use.     The winmodem I believe
>  > lacks a UART and this function is performed by the cpu.  This is one reason
>  > you are required to have a Pentium as a minimum.  There are other factors
>  > but this is enough reason to buy a real modem.  This trend of designing
>  > hardware out of software is spilling into printers, sound cards, etc.
> Did you even *read* the web site?  It explains
> why a winmodem driver is a good thing in spite of the problems you
> outline above.

I have to agree with Russ here.  I remember a posting from Ed (@lucent,
sorry Ed for plagiarizing you) saying that DSP-based modems put about the
same or less load on the host CPU as serial port modems.  That is because
the DSP chip does all the signal modulation and line training work and the
CPU just pumps data to it, but can do this by DMA transfer---more
efficiently than serial port IRQ servicing.

I just recently joined this mailing list and I have been impressed with
the amount of effort that Russ has put in to get a *constructive*
discussion going.  The *objective* of this list and web site is to make
winmodems work under Linux.  This is not some ideological argument but an
engineering endeavor, so if some people don't like the objective, post
your flames elsewhere.

Last not least, I think some political muscleflexing might help, to
convince some vendors to realse some technical specs for the DSP so that a
codec driver can be developed.  From some private discussions I know that
there is at least one vendor previously mentioned on this list that is
evaluating the interest and demand for Linux divers for their winmodem.

However, without a lot of engineering experience in that specific field, I
believe it might be quite difficult to write a codec driver (even if the
specs were available), because even the Windows drivers, although in the
n-th generation already, still have lots of problems with compensating for
non-linear signal distortions on phone lines (albeit this is a problem for
56k technology only) and other noise-related issues, and performance
changes randomly from release to release.  This is my personal experience
with a Lucent Winmodem on a not-so-crystal-clear phone line. 



  Wolfram T. Arnold | University of Oregon  |
  P.O. Box 3596     | Department of Physics | Materials Science Institute
  Eugene, OR 97403  | Eugene, OR 97403-1274 | Fax:    +1 (541) 346-3422
  U.S.A.            | U.S.A.                | Office: 077 Willamette Hall
  +1 (541) 988-0303 | Cell: (541) 729-2105  | Phone:  +1 (541) 346-2998