In praise of winmodems.

• Messages sorted by:[ date ][ thread ][ subject ][ author ]

At 10:55 AM 12/27/98 -0600, you wrote:
>>>>>> "Gert" == Gert Doering <gert@greenie.muc.de> writes:
>
>    Gert> Hi, On Sat, Dec 26, 1998 at 02:09:30PM -0000, Russell Nelson
>    Gert> wrote: [..]
>    >> I really have no idea how well a winmodem driver would scale.
>    >> I expect it would scale much better for fax than for 56Kbps.
>
>    Gert> That's what I would expect as well.  Half-duplex
>    Gert> transmission, with max. 14400 bps.
>
>It was my understanding that the winmodems used the host cpu for more
>than just bit-banging; they used it for DSP functions also, so the
>load on the cpu could get pretty heavy for even a single modem.
>
Actually, there are two types of winmodes: HSP and non-HSP. 
Lucent makes winmodem chipsets (like the one in my home computer) that are
of the non-HSP variety. They include a fixed-point DSP, so that while the
host computer has to do all of the control functionality for the modem, it
doesn't have to do any of the DSP work. It shouldn't actually be that hard
(assuming you could get the relevant pieces from Lucent) to write an
open-source driver for such a beast - lucent has exisiting drivers for
win95 that include the code that goes in the modem's DSP, so you wouldn't
have to write all of that. The problem is that you'd need most of the
lucent development kit in order to do the job. This sort of modem doesn't
add very much overhead to the host, so they might scale reasonably well. If
someone were able to get the development kit, I think a non-HSP winmodem
driver is quite possible.

HSP modems are much worse - they have no DSP, so the host has to do all of
that work as well. These beasts aren't going to scale worth a darn - and
it's probably not worth writing the drivers.


Michael Kohne
mhkohne@sonictech.com
"Evolution is God's version of domino rally"