ibi wrote:
>
> I had a Phoebe 56k internal modem that was a wizz in State A, but when I
> moved to state B the connection dropped from 51 to 26! After trying
> three different ISP's I bit the bullet and bought a USR Sportster
> External modem. It works. It is recognized correctly by both W98 and L-M
> on my pc. I ran it under W95 without problems. I've also used it as a
> test modem on other configurations without incident.
Being that I was the one that recommended the Phoebe Modems, I
have to ask what Phoebe model was it?
Phoebe makes quite a few different modems. They have the TI Chip
models CM1456VQH-X and CM1456VQE-X which are functional
equivalents
of the U S Robotics. YOu can tell by the sounds they make upon
connection.
But they also make an incredible array of junk modems, including
Cirrus, and Rockwell chipset modems. They make lots of PCI
winmodems as well. It is important to purchase a Phoebe modem
by the EXACT model number because buying one by "description"
alone is not going to cut it.
As an ISP, I have hoed a tough road to get people connected here.
Let me tell you a bit of history which might enlighten you.
When I started out, I had analog lines into my building, and
the modems here were US Robotics Couriers V-everything. These
were 16 modems to the rack chassis, the MP/16. These did 33.6K
and are very good. They seem to connect well over the poorest
lines, and can be set via init string to stay with the call even
after a prolonged carrier drop. For instance, I would not drop
the call with my 30 second setting, so call-waiting tones,
noise, static etc. did not mean the user got disconnected.
But I decided to go 56k for two reasons. The first reason was
the customer demand for 56k service. The second reason, just as
powerful, was the maintenance overhead of the multiple analog
phone lines. At any given time there were some lines out of
service. With hunt groups, depending on the software used by
TELCO, if you have, say, 50 lines, and line 8 breaks, then
no call will proceed to lines 9-50 unless someone is ringing
on line 8 when you try your call. Line 8 just rings and rings
no answer and its hard to get past it.
This may sound trivial, but it is not. It is very discouraging
when customers call to tell you your phone rings and rings.
Sometimes the break is in premise wiring, and you can jumper
the line to busy at the telephone protector, but often the break
is in the telephone poles or underground telco wires. When this
happens outside business hours, you are kaput until next day.
At my request, Telco (GTE) put in software that allows the calls
to be randomly distributed over the dialup lines. So that a
call would never come in twice in a row on the same line. Thus
if that line 8 was bad, it would still affect only 2% of the
attempts. Next time the customer tried, he would randomly get
a different line and no problem. This was a tremendous help....
but it has the disadvantage that it is virtually impossible to
know which lines work or not, unless you test each one. This
is time consuming.
So.... with the 56k modems, the way it works is that I get the
service via channelized T-1 (PRI ISDN is n/a here). Over a single
T1 line, which is two pair of wires, you get 24 channels or 24
"lines". I bought a Livingston Lucent PM3 Portmaster. It was
very easy to set up and within an hour of hooking it up I was
getting my digital calls. Only two T1 lines to maintain.
This was ideal in many ways. The Telco routes the lines in a
special way, with special connectors and are isolated in the
underground vaults or telephone pole splice bags. We have never
had a single outage on a T-1.
The phone company was using a Nortel DMS-5 switch. This thing
gets a modem call from a customer on a regular analog phone line.
It converts the call to digital via a CODEC, and makes it
digital. The call is then routed to a line driver and multiplexer
and out to me on one of the T1 channels. All worked fairly well.
My problems were related to the usual mix of people with X2
modems finding out they needed to upgrade to V.90 because my
equipment only handled K56Flex/V90 and not X2/V90.
I also discovered the PM3's digital modems cannot be set for
loss of carrier time, nor can the Telco which detects this. The
time is locked at 700 milliseconds. So there was a flurry of
disconnects, and a lot of people opted to remain with the analog
phone lines, of which I still maintain about 10% of my lines with
the Robotics Couriers as before.
Alas, the DMS-5 was not Y2K compliant, and it was a large dinosaur
that GTE found increasingly hard to maintain. For instance, since
it was out of production for over 12 years, to add caller ID to
it cost them $100,000 in software package. They needed to replace
it, and they did this in April 1999. This was black ribbon day
for Nook Net.
They chose the DMS-10. This is a popular device with telcos. It
is cheap and easy to administer.... but ISP's have problems with
it.
The Analog-CODEC-digital-CODEC-Analog path to my 33.6k couriers
works perfectly and the connect speeds with them is still 33.6k
with
no problems. The Analog-CODEC-digital-DSP-Multiplex-Nook.Net path
is not. We surmise that the bandwidth on the digital path to
NookNet
has dropped from 64K per channel to about 56K. In fact, this is
what Nortel specs it out as. It simply is very marginal. We have
found timing problems and over 70% of the customer modems simply
became unusable. Modems that were previously connecting at 50,666
or so would only do 26,400 or so. Worse, dialing to the digital
pool with a customer 33.6K modem only gave 31.2K max. No amount
of work has corrected the problem at the telco end.
However, by tweaking things at the customer end ... extra init
strings, commas after the phone number, upgrading drivers etc; we
have satisfied about 90% of our 56k customer base. The biggest
source of headaches are the Rockwell chipset modems. These were
bad with the DMS-5 equipment, and virtual dogs with the DMS-10.
As an ISP, my hands are tied.
I believe that PRI ISDN at 23 channels per T1 will solve the
problem, but the Telco does not want to file for the needed Tariff
or get the equipment just for me (there is no use for ISDN
otherwise
in a small town).
My dismay has grown exponentially with the virtual takeover of the
Rockwell modems. 18 months ago, virtually all computers had
either analog modems, or a healthy mix of 56k modems of which the
Robotics X2 (standard or WinModem) was the leader. Rockwell came
out with its Host Signal Processing model, the HSP and later the
HCF
chipset. These chipsets are simply a phone interface that dump
the phone noise into the Pentium and let your software do the
work.
All modem functions are done by the CPU. The chips cost the OEM
about $3 to $4 per computer and largely account for the price drop
in computers. But it is simply a terrible device, and the
performance
and quirks of this modem depend on many things like CPU speed,
whether Office 97 is running in the background, whether you have
IE and Netscape running simultanously, etc. etc. This is not an
issue with regular modems, but it is with the HSP/HCF.
The moral of the story is: get a real modem, and expect
diffeernces
in performance and reliability if you move from one area to
another
that has a different Telco switch.
--
Ramon Gandia ============= Sysadmin ============== Nook Net
http://www.nook.net [EMAIL PROTECTED]
285 West First Avenue tel. 907-443-7575
P.O. Box 970 fax. 907-443-2487
Nome, Alaska 99762-0970 ==== Alaska Toll Free. 888-443-7525
