how to build a modem?
Admin
Are there any publications (books, papers, standards) that might
demonstrate or instruct how to construct a traditional analog modem? Our
target goal is to reproduce the functions of analog modems in an embedded
environment. Here is a rough sketch of the flow of the end product:
---->T1---->T1 signalling interface-->modem protocol module-----
- telco - to handle call - handle v.34, etc
interface management
--->terminal server module-------->ethernet interface
- handle authorization, - physical ethernet wire
call specific protocols
ie tcp/ip, slip, ppp,
routing
- Thanks in advance!
BRad
Richard Neveau
>
> In <4afg9c$2...@s1.GANet.NET> ro...@mail.GANet.NET (Admin) writes:
>
> >Are there any publications (books, papers, standards) that might
> >demonstrate or instruct how to construct a traditional analog modem?
from International Telecommunication Union in Geneva but maybe
what you want is on the Standards CD-ROM from ???
(Infomagic or Walnut Hill, check their web sites I THINK it
is Infomagic but am only 90% sure)
The CD will SAVE you BIG BUCKS as ITU charges are helping support the
ITU organization and are unrelated to the cost of printing. Of course
they are the only source for 'current' , 'up to date' specs as
they pulled the plug on the electronic versions I think. (the CD
stuff was what was released before they caught on).
> Ah, a modem with an ethernet interface (and router).
> I have tried to convince a modem manufacturer that this would be a
> nice product, and I think they considered it for the future.
>
> Of course when you want a V.34 modem, V.34 is the name of the standard.
> However, you would normally not implement it yourself but go to one
> of the chip manufacturers (Motorola, Rockwell) that sell dedicated ICs
> for it.
>
> When you want the fastest path to the product above, do it like this:
> make a board with 2 ISA slots on it, and the processor and memory that
> you require for your software ("terminal server module"). Then just
> plug a standard PC card modem and ethernet adaptor in it.
> Advantage: you can swap network cards (token ring, 100 Mbps networks,
> arcnet, etc), and upgrade the modem card when required (V.34bis, ISDN).
>
> It would even be possible to do it without any hardware development at
> all by using a 3-slot bus board and a CPU board on an ISA card, which
> are readily available.
>
> Rob
The tricky part of rob's plan is the T1 --> 'PC World' bridge.
One of our sales guys wanted something like this (I THINK) but we
talked him out of it. For the volume you are better off
"cost & time-to-market" wise just putting in whatever combo of
(rack mount) modems and T1 Multiplexers/Channel banks and/or
terminal servers and/or routers that are needed.
It is nice to have it all in one shelf but do you REALLY
want to do something this complicated for a volume of a few
hundred sites ??
It can be done but you really need to use as many 'chip sets' as you
can. Just think of all the MAN YEARS of engineering time you
leverage by letting the Chip vendors figure out all the
problems implementing T1 && V.34( & V.34bis soon) && Ethernet.
------>>>>> If you use the chip sets you can get the vendor to
------>>>>> provide you with app notes and maybe even demo boards
------>>>>> to get you started.
Even with the chips sets it will be a big job. Better would be boards
but I am not aware of a PC board that would split T1 into 24 DS0 data
ports in a format you could muck with and feed into modem card(s).
(that would be a channel bank on a card I guess).
From there on it would be pretty much off the shelf stuff to handle
the 24 modem cards and route it to the ethernet. I think cisco
announced software to do routing on NT servers so maybe you could
see if they have a product you could 'extend' some. (or use Linux
or Free BSD software if you have time to support it yourself).
There are a few VME vendors with T1 cards that may be a better fit for
your application if you decide to develop something. I am not
sure what the names of the companys are but I heard Dialogic
(the voice card for PC people) have some 'relation-ships' in
the works for IVR on VME so maybe call them to get some pointers
or check the Dialogic Web page.
As someone who works on a product that busts T1 into multiple Data ports
I guess I would probably go crazy if we had to keep up with all the
'Modem' standards also. V.34 is NOT the end of the line. There is
the (not bis I hear) V.34+++ that goes at 33.? K and some guy at USR
or Rockwell may be able to push it even farther before ISDN takes over.
(Physics be damned, give me 3 Geeks with enough jolt cola and maybe
they can get a 'worm-hole' effect going. Although ISDN should cross
the price of Analog modems soon in the 'first world' anyway).
You may want to talk to Analog Devices / TI / Moto / Whoever bought
STAR Micro - and see if they can supply DSP code that will do
the V,34 Suite off the shelf. Most likely they can but that
is really only about 10 % or so of your effort most likely.
For all the effort it might be better to just use T1 Multiplexers or
routers or channel banks until ISDN takes over. Of course you
know the number of sites, etc. but we couldn't cost justify this sort
of a card even when the 'modems' were 9.6 and we wanted to
subrate 5 per DS0 to save WAN T1 pipes. There is just not enough
volume for the product. It was much cheaper to just get the extra
hardware and do it kind of like 'Back-to-Back' Channel Banks used
to be set up before T1 DAC's took over that function.
I just remembered that Motorola and USR have T1 cards
for their rack mount modems. Also Chase Research or Livingston (SP???)
Terminal server folks MAY have something they could adapt for you
to use. I am assuming you looked into these sorts of things.
Also surf the Boardwatch magazine back issues. Maybe something from
BBS world can help. The larger BBS's have had this problem for some
time now. They just didn't have the last --> PPP --> IP step.
I am not sure I understand your App 100% so maybe this is all nonsense.
If so it was worth what you paid for it ! I think you are just
looking to get the 24 'dialins' off a T1 trunk into PPP -> IP so
you can stick it on ethernet. If all this gear is in ONE
room it doesn' make cost sense to build it all on one board/box.
modems are too hard to get right (IMO of course).
If you wanted to remote the V.34 to digital conversion in a POP and
slip 2 'modem lines' into one DS0 it might make more cost sense
but you would have to split the AtoD/DtoA and the (de)Compression as
an uncompressed V.34 line may use more than 32KBps in bursts. Frame
relay or ATM would be work for that application off the shelf
(but granted - Multiple Shelves). It is kind of like audio gear.
To get the best stuff you want each function in it's own box.
Unlike audio, there is not enough volume to justify a 'receiver'
with everything in one box.
Dave
>Are there any publications (books, papers, standards) that might
>demonstrate or instruct how to construct a traditional analog modem? Our
>target goal is to reproduce the functions of analog modems in an embedded
>environment. Here is a rough sketch of the flow of the end product:
>---->T1---->T1 signalling interface-->modem protocol module-----
> - telco - to handle call - handle v.34, etc
> interface management
>
>--->terminal server module-------->ethernet interface
> - handle authorization, - physical ethernet wire
> call specific protocols
> ie tcp/ip, slip, ppp,
> routing
>- Thanks in advance!
>BRad
You could try
The Theory & Practice of Modem Design, by John Bingham
John Wiley, 1988. ISBN 0-471-85108-6
--
Dave
PGP fingerprint = 20 8F 95 22 96 D6 1C 0B 3D 4D C3 D4 50 A1 C4 34
Steven D. Ligett
Admin
Rob Janssen <pe1...@wab-tis.rabobank.nl> wrote:
>In <30CCDC51...@dp.tpd.dsccc.com> Richard Neveau <rne...@dp.tpd.dsccc.com> writes:
>
>>The tricky part of rob's plan is the T1 --> 'PC World' bridge.
>
>building a V.34 modem with ethernet interface (and builtin terminal
>server/router)
At any rate, discounting the T1 "bit", where you going to contribute
something to this thread?
- BRad
Admin
Richard Neveau <rne...@dp.tpd.dsccc.com> wrote:
>Rob Janssen wrote:
>>
>> In <4afg9c$2...@s1.GANet.NET> ro...@mail.GANet.NET (Admin) writes:
>>
>The ITU-T / CCITT standards/recommendations for V.??? are available
>from International Telecommunication Union in Geneva but maybe
>what you want is on the Standards CD-ROM from ???
>(Infomagic or Walnut Hill, check their web sites I THINK it
>is Infomagic but am only 90% sure)
Thanks for this, Ill check it out.
>> Ah, a modem with an ethernet interface (and router).
>> I have tried to convince a modem manufacturer that this would be a
>> nice product, and I think they considered it for the future.
This has been done but not on the scale that we'd like to see it.
>> Of course when you want a V.34 modem, V.34 is the name of the standard.
>> However, you would normally not implement it yourself but go to one
>> of the chip manufacturers (Motorola, Rockwell) that sell dedicated ICs
>> for it.
Maybe I wasn't clear enough with this. The reason that the above would
not provide a solution for us is because we want to use higher power DSPs
to handle more modems/unit of silicon than with traditional dsps found in
modem implementations. For instance, the IDT Orion can do about 175 MIPS
which could comforable from 6-8 v.34 modems assuming the overhead on a
v.34 process is 20-30 mips. Unless motorola or rockwell is willing to
license their v.34 code, they wont be able to help (BTW, rockwell modem
datapumps are too buggy for the quality we want).
>> When you want the fastest path to the product above, do it like this:
>> make a board with 2 ISA slots on it, and the processor and memory that
>> you require for your software ("terminal server module"). Then just
>> plug a standard PC card modem and ethernet adaptor in it.
>> Advantage: you can swap network cards (token ring, 100 Mbps networks,
>> arcnet, etc), and upgrade the modem card when required (V.34bis, ISDN).
>>
>> It would even be possible to do it without any hardware development at
>> all by using a 3-slot bus board and a CPU board on an ISA card, which
>> are readily available.
Actually, ISA would probably not be the backplane of choice. We have
looked at both VME and PCI solutions and found those to be much more
in-line with what we would need in a real-time application such as this.
>The tricky part of rob's plan is the T1 --> 'PC World' bridge.
>
>One of our sales guys wanted something like this (I THINK) but we
>talked him out of it. For the volume you are better off
>"cost & time-to-market" wise just putting in whatever combo of
>(rack mount) modems and T1 Multiplexers/Channel banks and/or
This costs so much more in wiring, rackspace, "silicon", manageabilty,
integration (who do you turn to if the things don't interoperate),
service, etc.
>terminal servers and/or routers that are needed.
>It is nice to have it all in one shelf but do you REALLY
>want to do something this complicated for a volume of a few
>hundred sites ??
Sure! Its going to happen eventually, we would just like to be there to
provide the solution before everyone else does.
>It can be done but you really need to use as many 'chip sets' as you
>can. Just think of all the MAN YEARS of engineering time you
>leverage by letting the Chip vendors figure out all the
>problems implementing T1 && V.34( & V.34bis soon) && Ethernet.
We would not like to do any electrical engineering. That is, all ic's
that we'd use would be assembled by vendors. For instance, there are
many T1 (single, dual, quad port, etc) interfaces for VME that have
drivers all ready for various real-time os's. Perfect for what we need.
Ethernet as well. The only thing that is not covered is the v.34...
>------>>>>> If you use the chip sets you can get the vendor to
>------>>>>> provide you with app notes and maybe even demo boards
>------>>>>> to get you started.
Right. We're working on writing up a lengthy formal proposal to submit
to various manufacturers/vendors for there ideas. BTW, if anyone would
care to see it, it should be done soon.
>Even with the chips sets it will be a big job. Better would be boards
>but I am not aware of a PC board that would split T1 into 24 DS0 data
>ports in a format you could muck with and feed into modem card(s).
>(that would be a channel bank on a card I guess).
>From there on it would be pretty much off the shelf stuff to handle
>the 24 modem cards and route it to the ethernet. I think cisco
>announced software to do routing on NT servers so maybe you could
>see if they have a product you could 'extend' some. (or use Linux
>or Free BSD software if you have time to support it yourself).
Again, PC would not be the backplane for this application. As for OS,
we'd use a real-time os such as VxWorks, pSOS, others.
>There are a few VME vendors with T1 cards that may be a better fit for
>your application if you decide to develop something. I am not
>sure what the names of the companys are but I heard Dialogic
>(the voice card for PC people) have some 'relation-ships' in
>the works for IVR on VME so maybe call them to get some pointers
>or check the Dialogic Web page.
also www.vita.com has a list of vme products for various fields.
>As someone who works on a product that busts T1 into multiple Data ports
>I guess I would probably go crazy if we had to keep up with all the
>'Modem' standards also. V.34 is NOT the end of the line. There is
>the (not bis I hear) V.34+++ that goes at 33.? K and some guy at USR
>or Rockwell may be able to push it even farther before ISDN takes over.
>(Physics be damned, give me 3 Geeks with enough jolt cola and maybe
>they can get a 'worm-hole' effect going. Although ISDN should cross
>the price of Analog modems soon in the 'first world' anyway).
We would depend on vendors for providing most of the parts. All we would
concern ourselves with would be the v.34, terminal server, integration.
>You may want to talk to Analog Devices / TI / Moto / Whoever bought
>STAR Micro - and see if they can supply DSP code that will do
>the V,34 Suite off the shelf. Most likely they can but that
>is really only about 10 % or so of your effort most likely.
I will look into this!
>For all the effort it might be better to just use T1 Multiplexers or
>routers or channel banks until ISDN takes over. Of course you
>know the number of sites, etc. but we couldn't cost justify this sort
>of a card even when the 'modems' were 9.6 and we wanted to
>subrate 5 per DS0 to save WAN T1 pipes. There is just not enough
>volume for the product. It was much cheaper to just get the extra
>hardware and do it kind of like 'Back-to-Back' Channel Banks used
>to be set up before T1 DAC's took over that function.
We are currently working with a company called Satell who subrate 2-4
14.4 or 28.8 calls per DS0 with integrated modems (just so you know, this
idea is alive and well).
>I just remembered that Motorola and USR have T1 cards
>for their rack mount modems. Also Chase Research or Livingston (SP???)
>Terminal server folks MAY have something they could adapt for you
>to use. I am assuming you looked into these sorts of things.
>Also surf the Boardwatch magazine back issues. Maybe something from
>BBS world can help. The larger BBS's have had this problem for some
>time now. They just didn't have the last --> PPP --> IP step.
Right. You dont think USR engineered their own T1 cards now do you? I
doubt it.
>I am not sure I understand your App 100% so maybe this is all nonsense.
>If so it was worth what you paid for it ! I think you are just
>looking to get the 24 'dialins' off a T1 trunk into PPP -> IP so
>you can stick it on ethernet. If all this gear is in ONE
>room it doesn' make cost sense to build it all on one board/box.
>modems are too hard to get right (IMO of course).
>
>If you wanted to remote the V.34 to digital conversion in a POP and
>slip 2 'modem lines' into one DS0 it might make more cost sense
>but you would have to split the AtoD/DtoA and the (de)Compression as
>an uncompressed V.34 line may use more than 32KBps in bursts. Frame
>relay or ATM would be work for that application off the shelf
>(but granted - Multiple Shelves). It is kind of like audio gear.
>To get the best stuff you want each function in it's own box.
>Unlike audio, there is not enough volume to justify a 'receiver'
>with everything in one box.
We are not concerned with packing more than one call over a single DS0.
We just want to integrate the funnction of channel banks, v.34 pools,
terminal servers into one box.
- Thanks for the responses!
BRad Block
dha...@imt.net
>In <30CCDC51...@dp.tpd.dsccc.com> Richard Neveau <rne...@dp.tpd.dsccc.com> writes:
>
>
>building a V.34 modem with ethernet interface (and builtin terminal
>server/router)
>
>--
>+------------------------------------+--------------------------------------+
>| Rob Janssen r...@knoware.nl | BBS: +31-302870036 (2300-0730 local) |
>| AMPRnet: r...@pe1chl.ampr.org | AX.25 BBS: PE1CHL@PI8WNO.#UTR.NLD.EU |
>+------------------------------------+--------------------------------------+
If the scope of the project is international, I'd be looking how to hook up
someone elses solutions - the approvals/certification process for phone line
interfaces are *scary* in a lot of countries!
Sorry, no technical help, just more worries! :-)
Dan
Geoffrey Levand
>Are there any publications (books, papers, standards) that might
>demonstrate or instruct how to construct a traditional analog modem? Our
>target goal is to reproduce the functions of analog modems in an embedded
>environment. Here is a rough sketch of the flow of the end product:
The book "Programmers Technical Reference: Data and Fax
Communications", Ziff-Davis Press, 1993 by Robert L. Hummel (ISBN
1-56276-077-7) describes basic modem functionality and gives
communication specifications. No programming or hardware design
examples are given though.
William L. Grenoble Jr.
: Are there any publications (books, papers, standards) that might
: demonstrate or instruct how to construct a traditional analog modem? Our
: target goal is to reproduce the functions of analog modems in an embedded
: environment. Here is a rough sketch of the flow of the end product:
: ---->T1---->T1 signalling interface-->modem protocol module-----
: - telco - to handle call - handle v.34, etc
: interface management
CSU/DSU Channel Bank 24 each external modems
:
: --->terminal server module-------->ethernet interface
: - handle authorization, - physical ethernet wire
: call specific protocols
: ie tcp/ip, slip, ppp,
: routing
Multi port serial board {PC running Linux} ethernet board
: - Thanks in advance!
: BRad
IF (if and only if) you want to roll your own, get a lobotamy (sorry,
that slipped out.) Your T1 provider will require that you buy/rent
the CSU/DSU from them or an approved source.
The channel bank will cost $5000 to $10,000 and can be built for less.
You can get away with a two sided board if you are careful about the
layout, and the chip sets take all the "fun" out of it. You will
loose money on the first or second item, but may be able to save money
after that. Dallas, Mitel, and Telco have handy chipsets. Intel and
Motorola have the codecs.
The V.34 modem would be a "Very Interesting" design task. InfoMagic
has the Standards CDROM, well worth it. If you asked this question,
you WANT the Standards CD.
The pilots of the "Looking Glass" aircraft wore a patch over one eye
so that if they were surprised by a nuclear war, they would only be
blinded in one eye and would be able to continue flying... When you
first read V.34, I suggest that you wear an eye patch for much the
same reason. If you survive V.34, go on to V.42, which you will need
to make your V.34 modem work with other modems. Either recommendation
can strike you blind, or cause permanent brain damage... :) :) :)
{The guys who thought up V.34 must of stayed up all night working
on it. I have a copy of the recommendation and I have read it. I _could_
write DSP code to do V.34, and it would be a _very_ interesting job.}
Now comes the Fun part. PCs were designed for two serial ports. Four
addresses are available, but they share interrupt lines. To service
24 serial ports, you need several of the multiport cards. You _COULD_
build your own board with many uarts and your own computer architecture,
but you will not save any money. Buy the multiport boards, put them
in a PC, run Linux, and add an ethernet board or two.
XYLogic has a box that accepts a T1 and connects the 24 channels to
modems, serial ports, and on to the ether. A few years ago it cost more
than the channel bank and modems. Ma Bell, (excuse me, Your Regional
Bell Operating Company (RBOC)) will usually provide the channel bank(s)
and the connection to the fiber lines. All you provide is power and
floor space. Many many people are buying V.34 external modems, not many
are buying the big box. Simple economics, buy what people are buying,
usually it is cheaper. Simple engineering, racking and stacking 72 each
USR Couriers is a real mess.
The problem with going direct from the T1 into the modem is that the
modem wants to control the digitizer. Over in comp.dsp there is a
constant thread about changing the sampling rate and all of the exotic
techniques to do it. Much easier to have a symbol tracking loop that
samples the symbols synchronously. IMHO, YMMV all the other disclaimers.
Bill Grenoble
--
bi...@neta.com
Brad Aisa
>ro...@mail.GANet.NET (Admin) wrote:
>>Are there any publications (books, papers, standards) that might
>>demonstrate or instruct how to construct a traditional analog modem? Our
>>target goal is to reproduce the functions of analog modems in an embedded
>>environment. Here is a rough sketch of the flow of the end product:
Contact the manufacturers of Digital Signal Processors, such as Texas
Instruments and Motorola. They have application notes with both
theoretical explanations as well as complete sample implementations of modems
for their DSP chips.
You might also contact companies like Rockwell and AT&T who make complete
chips.
--
Brad Aisa, Toronto, Ontario, Canada
ba...@hookup.net web archive: http://www.hookup.net/~baisa/
"The highest responsibility of philosophers is to serve as the
guardians and integrators of human knowledge." -- Ayn Rand
Christian Weisgerber
> Contact the manufacturers of Digital Signal Processors, such as Texas
> Instruments and Motorola. They have application notes with both
> theoretical explanations as well as complete sample implementations of modems
> for their DSP chips.
The sample implementations I recall seeing were V.21 or V.22 ones...
Motorola will probably _sell_ you V.34 etc. data pump code for their
68356 by now.
--
Christian 'naddy' Weisgerber na...@mips.pfalz.de
See another pointless homepage at http://www.rhein-neckar.de/~mips/.
-- currently reading: Richard Matheson, The Incredible Shrinking Man --