Your modem must connect to our modem and remain connected throughout your session. This page is for people facing problems making or holding good modem connections. It is about difficulty in connecting, about disconnects, and about low speed connections.

This page is also about persistent problems.  There are many causes of connection problems that are not covered here, such as those due to dieing modems or modems that for some reason aren't up to par.  Things covered here can cause you connection problems even with first-class modems in perfect condition.

This page is not about problems logging in or making network connections once good modem-modem communication is established. Those problems are of a different nature

Some people connect at the maximum speed their modem can work at and never get disconnected. A few have trouble connecting at all, and a great many are not getting all they could out of their Internet connections. Virtually all major problems and most minor problems can be eliminated.

As 56k modems become more common, the things causing even minor problems now will loom larger in their impact. Every user who wants to enjoy the benefits of these faster modems should become familiar with where the limitations on modem connection speed and robustness come from.

Brief Summary

Modems have to transmit precise data over a telephone network designed to transmit only a rough approximation. To send anything at all they have to encode it as tiny changes in an already small electrical signal and send that signal through a sea of interference, hoping the tiny changes are larger than the interferences picked up along the way. It is common for the signal to arrive at a home in nearly perfect condition only to undergo substantial degradation inside the home.  Much of this degradation stems from the use of the wrong kind of wire or from wiring practices not suitable for quality telephone wiring.

Modems are trying to use the telephone system for something it was not designed for.

The telephone's basic job is to receive sound at one end and reproduce it at the other. It can only reproduce an imperfect copy though, not nearly as good a reproduction as a stereo system can do, and certainly not an exact copy. The telephone is a compromise between sound quality and cost. Telephone sound transmission was designed to be low cost and "good enough" - good enough for the telephone's intended purpose, people talking to each other.

Computers need to transmit digital data to each other exactly. There's no scale of poor, fair, good, excellent for digital transmission, no "good enough". There are only correct and incorrect.

The job of a pair of modems then is to receive digital data at one end, convert the data to sound, transmit an approximation of the sound, and recover an exact copy of the original digital data from the approximation of the original sound. The modems must transmit the data perfectly in a form that is trasmitted only approximately.

The role of a telephone isn't just sending sound, it's sending the series of changes in sound that make up speech. The changes in speech are, for the most part, rather large changes, so even though the telephone does not reproduce them exactly, we can understand what is said.

Modems also use changes in sound, to transmit digital data. Very old modems used large changes to transmit data. A time where the sound changed could represent a bit that's on (a one bit) and a time when the sound didn't change would be a bit that's off (a zero bit). This kind of one bit per change only works for slow speeds; it ran out of speed with 2400 bps modems.

A 28.8Kbps modem runs 12 times faster by using subtle changes to represent 12 bits at a time. These changes are far smaller than the large changes making up most of speech. They are nuances of change at the very limit of what telephones are capable of.

To use such nuances requires a telephone connection that can reproduce these subtle changes at the remote end and one that does not introduce any extraneous changes of its own. The connection must be both very high fidelity (for a phone call) and free of noise.

To transmit sound over wires, telephones convert it to an electrical signal. Originally, modems used to generate sound they wanted to transmit, and a telephone listened to the sound and converted it to an electrical signal. Modern modems directly generate the electrical telephone signal representing the changing sound.

When modems are first establishing a connection, some of the hissing and other noises you hear are the modems testing the phone connection to find out how good it is. A lot of times they find that either the signal changes are not transmitted with enough fidelity or that additional signals corresponding to noise are present.

The problems modems find can be divided roughly into categories:

• Problems that are mild and stable cause low speed connections.
• Problems that are severe cause failure to connect.
• Problems that are unstable cause disconnects.
Mild and stable problems are usually poor phone-line fidelity or low-level noise on the phone line. Modems can adapt to these by not trying to make the phone line carry data as fast. In other words, they deliberately reduce the speed, similar to the way you adapt to a rough road by driving more slowly.

Mild, stable problems are often in the phone company's network, and whether they are rightly called "problems" is sometimes a matter of interpretation. For example, modems capable of 33.6k often will connect at only 28.8k on calls from one telephone exchange to another, because many of the trunk circuits that carry calls between telephone switches can't carry calls with the fidelity needed for 33.6k connections.  While these older trunks are digital, they only guarantee 7 out of every 8 bits.

Human hearing cannot detect this difference in fidelity, but modems trying to push the telephone network to the absolute limits can. They detect the slight loss of fidelity and back off on the speed to what the circuit can support. Since the phone network is a mixture of newer and older trunks, the digital fidelity you get may vary from call to call.

People living in some suburban housing developments can have a similar fidelity loss. Instead of having copper wires running from their homes to the telephone switch (which is often a long ways away), their lines may come from a small telephone outpost right at the housing development, with a fiber connection from that outpost to the telephone switch.  Some of these remotes only have fidelity adequate for 26.4k modems connections.  If you get solid connects but always at 26.4, ask your phone company if your phone service is via an SLC remote.

Severe or unstable phone problems, the kind that cause disconnects or failure to connect at all, are nearly always problems with the wiring right in user's homes. Many of these problems come from not knowing the special needs of telephone wiring and treating telephone wiring like electrical wiring. So let's look at why good telephone wiring takes special care you wouldn't expect.

Stand up! Well, ok, just imagine you're standing then.

Now look down at your feet, and imagine there's an ant on the floor by your foot. It's a tiny little thing compared to you, isn't it?

In fact, you are about 100 times as tall as the ant is long. That difference in size between you and the ant is what it means to be 100 times larger than something.

The electricity in your house is 110 volts. The electrical signal on your phone line is in the vicinity of 1 volt. Your electrical power compares to a telephone signal about the way you compare to that ant.

Now let's go for a walk. Bring your ant along. As you stroll across a lawn and gravel driveway, you don't even notice blades of grass and lumps of gravel, but your ant does. The lawn that's a nice smooth carpet to you is a forest of obstacles to your ant. Life is different when you're tiny.

In the same way, slight wiring defects like imperfect or loose connections that wouldn't matter a bit for electrical power wiring can be big obstacles to telephone signals.

However, the real problems with most house telephone wiring aren't with the wiring connections. They are with interference, or "pickup", and it's here where telephone wiring really differs from electrical wiring.

Because 28.8 modems encode 12 bits into a single change in signal, those changes occur in steps about 1/12th the size of the signal. This means interference as small as 1/12th the signal can change every bit. It will totally destroy the tranmission.

In fact, so will interference only half that size, because if the signal plus interference change by half a step, the remote modem can't tell if the signal changed by 1 step or didn't change at all. For the remote modem to know reliably the number of steps of change in the signal, the peak interference must be no more than about 1/8th the step size. 1/8th of 1/12th of the signal is 1/96th of the signal, so the interference must be about 100 times smaller than the signal.

So when you look down at your ant representing the signal size, imagine it looking down at an ant's ant representing the interference limit.

Now that we know just how small the interference has to be, let's look at some of the places it comes from.

Interference is just another name for an unwanted signal, and you live in a constant bath of signals ready to become interference. You know some of them are there, because you can detect them with electrical devices called radios and televisions.

They're weak enough so that when you want to detect them you have to do special things. One of those things you do is use antennas to capture them. The first antennas were called aerials and were just long wires, the longer the better so the signal could build up over a long distance.

Now go outside and look at the telephone wires.

So you have a modem that's trying to communicate using a method that's sensitive to tiny interferences, and you attach an aerial to it?

A lot of the basics of getting good connections can be summed up by saying don't attach an antenna to your modem.  Although the phone wires look like aerials, the only antenna attached to your modem is one of your own making, or at least one you can unmake.

When you looked at those phone wires, what you didn't realize is that even though they're long wires, they're not aerials or antennas. If they were, even your telephone wouldn't work.

Inside a phone company's cables, each telephone line's two wires are slowly twisted. This means that the interference picked up at one place is cancelled by that picked up one twist later. Interference can't build up down the length of the pair, and this way the phone company can carry phone signals for miles along wires and still deliver them to your home essentially free of interference buildup.

You can deliver this signal to your modem still free of interference if you maintain the proper twist on the wiring inside your home and use a little care in routing your wiring. If you had done that you probably wouldn't be reading this page.

Internal telephone wiring done by a professional telephone installer will be done using wire manufactured with the proper twist. However, if you did it yourself or had an electrician do it, chances are that it was treated like electrical wiring, as if all that mattered was electrical connection.  You used flat untwisted wire to install an antenna in your house and connect it to your modem.

Modems often come with a 6-foot cable to connect the modem to the wall jack.  These cables are flat, with untwisted side-by-side wires.  Just that single 6-foot section of untwisted wire can be the difference between connecting at 31.2 and 33.6.

Your home as an antenna array

If you've ever tried to adjust a small indoor antenna, such as a "rabbit ears" tv antenna, you know it's a maddening job, because when you walk close you affect what you're trying to adjust. That happens because you are an antenna also, and antennas near each other affect each other. (In fact, the twist of phone company wires in their cables is mostly to cancel their effects on each other. The main cables carrying many wires are shielded in metal jackets to isolate the wires from the outside interference.)

Your entire home is full of antennas - phone wires, power wires, pipes, metal parts of appliances and furniture, you, other people, animals, houseplants, toys, etc. Most of these are pretty poor antennas, and their influences are so random that they mostly cancel each other out anyway. However, if you've done your phone wiring with untwisted wire, it's about the best antenna of the lot, and if it winds its way very far through your house, it can pass near all sorts of other antennas, each adding its bit of interference as if they were all one big antenna.

Even so, most of the bath of interference in which you exist is so weak that all this still wouldn't matter if you didn't have your own interference transmitters right in your own home, in the midst of your antenna array. Just about everything electrical is a transmitter of interference: fluorescent lights, radios, televisions, stereo systems, refrigerators, water pumps, microwave ovens, even telephones and computers. Your own home is a gauntlet of sources of the interference you need to avoid picking up.

Some people actually attach interference transmitters called wireless telephones directly to their phone wires.

If you have extension phones in various rooms in your house, the wires to them are all branches on your main antenna. You are really trying to make sure you pick up every bit of interference.

In addition, branches to extensions have another effect. As your phone signal comes in or goes out, it also travels down each of these side branches. When it gets to the end, some of it is reflected back and added back into the phone signal. The signal is then the sum of what it is supposed to be plus an electrical echo of what it was a moment before.

The signal travels very fast, so the echo times are extremely short. They don't matter one whit when you're talking on the phone, but they matter to your modem. They make listening for the signal sent by the other modem similar to you listening to what someone is saying in an echo chamber or when a microphone is picking up the sound from a speaker.

There's another kind of echo influencing your phone line. Just as the end of a branch wire to an extension reflects some of the signal, so does every kink, bend, or twist in the wire. Yes, every twist of twisted wire causes a tiny echo.

With properly twisted wire the echos of the echos cancel each other out. This is very important, so don't try to twist your own wire.

If you do your own telephone wiring:

• Use real telephone wire.
• Avoid kinking it or bending it sharply.
• Try to keep it a few inches away from metal objects.
• Avoid routing it near electrical appliances.
• Forego extension phone branches if you can.

Copyright © 1997,1999 NetHeaven.