Why 56K baud rates are practically unachievable in a reliable fashion

There seems to be some confusion as to what can be causing the new modems not to reach the Golden ring of 56K baud rates. I may be able to clear some of this up. It probably won’t help fix the problem but at least we all point our finger at the same cause and effect.

Load Coils Load coils (impedance matching transformers) are placed along the length of a long Telco POTs line that exceeds the 6000ft termination spec. The twisted pair of an analog line is a “balanced pair” interface with specific impedance terminations 600 or 900 ohms. The impedance loads are initially found in the “Line Card” at the Telco switch and the telephone or modem. Impedance load is very different from loop resistance.
Those terminations are engineered for the specific 600 or 900-ohm value to match the Amplitude and Frequency characteristics of the analog signal. The proper impedance termination is vital to absorb the maximum possible power on the line. If the impedance loads are not properly matched to the signal, they will only absorb part of the signal and cause an improper signal transfer to the switches receiver circuitry. The problem starts to become evident when you wonder this – “If the energy of the signal is only partially absorbed, what happens to the stuff that’s not absorbed?”

All energy that is not absorbed by the termination load reflects back on to the copper pair and begins to interfere with the original signal. Because the reflected signal is usually out of phase from the original signal, this starts to cause “common mode rejection” or cancellation or loss of Amplitude at specific “standing wave” frequencies and their mathematical derivatives. The net effect of that is, the original signal begins to get crippled by its own reflection.
The Evil Load Coils are placed in the circuit at specific intervals (varies depending on wire gauge and cable bundling variables) to reduce the effective capacitance of the extended copper loop and therefore provide a higher level of predictability across the copper segment. Remember, impedance matching, in theory, only works if you can predict the signal characteristics of the waveform being impeded. Anything that drastically alters the characteristics of the waveform will make the circuits termination less effective in absorbing the maximum signal.
The argument is that, three discrete segments of 6,000 feet copper has less negative captive effects than a single copper segment that is perhaps 18,000 feet long. Therefore, in a long copper loop, Load coils are placed at strategic points to reduce the negative effects that capacitance will have on the signals characteristics. Capacitance is the enemy of high BAUD rate applications. Amplitude can be overcome with amplifiers; Capacitance is much harder to control cheaply.
This load coil works wonderfully for voice, however; load coils, with their purpose to limit excess capacitance, also greatly limit the frequency spectrum available to the end devices. By the way, load coils also help restrict impulse noise or interference from one copper segment to adjacent copper segment. So in effect may reduce the additional noise characteristics from loop interference. Now back to the limit of frequencies.
There are a few variables that affect this discovery either positive or negative. But in general, placing a load coil in a Telco voice line reduces the effective bandwidth by chopping off the top 25 percent of the available frequencies. If you were to take a tone generator to the lines you would see the highest quality signal at approximately 1,000 to 2,000 Hz. When the tone generator reached 2900 Hz there would be a significant “role off” of over 12dB per octave. This reduction explains why you can’t use load coils in digital circuits. As a matter of reference if you examined a T1 signal with a scope it would look like a very phase distorted 772kHz analog signal. Certainly any facility that limited the bandwidth to 2900 Hz would seriously choke a signal running at 772kHz. The same holds true for ISDN BRI, which has an effective signal rate of approximately 40kHz.
So, it becomes obvious that load coils can cripple the high-speed signal. Please note, that removing the load coils will only make the signal worse. Because of the nature and characteristics of how modems manipulate an analog signal, removing the load coils will just cause exponentially more distortion across the copper segment. If you want to remove the load coils you must completely re engineer the way the data signal is presented to the copper pair. That’s what they did with ISDN. I don’t think anybody is really up for the challenge at this point.

Bridge Taps

These little bastards are perhaps the most annoying and offensive of all the anomalies found in a Telco copper segment. They are by far the number one problem you all have getting modems to connect and stay connected at high speeds. Unfortunately they are riddled throughout most residential neighborhoods and corporate business parks.
W hen the phone company runs a cable down the street the cable may extend a mile or so passed your house. Although no other house or device is using your specific copper pair, the pair runs out the length of the cable. All an installer does is take the wires that come from your demarc, drag them out to a junction box or splice box on a poll or pedestal and “Tap” the wires coming from your house on to a spare copper pair that runs out the length of the cable. They do not cut the cable pair at the junction box just incase they have to use the same pair for one of your neighbors down the road when you move out. Also, in order not to drastically reduce the amplitude and of the signal coming from your telephone they do not to terminate the extended cable end either.
That means there may be a one-mile cable running from the central office out past your house and your telephone line is simply tapped into the middle of it. It’s sort of like having an additional half-mile antenna picking up all the garbage in the air and feeding it to your telephone equipment.

 

Assuming for a moment we can deal with the additional idle channel noise on the copper facility, which even the most unqualified lineman can test for, (but if you ask him what he’s testing for he probably can’t tell you), and tell you some story how your line is the quietest line on the street, the next hurdle beyond the noise is the reflected signal coming back off of the half-mile unterminated antenna they built just for you at no additional charge.
When an electrical signal hits the end of a wire it has to go somewhere. If there is no impedance load to absorb the signal, then the signal in its entirety gets reflected back over the entire copper segment. The signal that comes from your modem headed for the central office arrives at a specific time interval and the reflected signal coming back off the unterminated copper extension comes then just behind yours causing your signal to appear phase distorted.
When two out of phase signals are received at a certain impedance load of a cause a rejection affects and begin to cancel each other out. If the signals arrive 180 degrees out of phase your signal can be canceled out completely. The more the second reflected signal it is closer to 180 degrees the more the signal will be attenuated and phase distorted. At lower frequencies problem is not terribly dramatic as these reflections are only fractions of a waveform out of phase. But when you make the waveforms smaller as is the case with higher frequencies the problem becomes exponentially more apparent. Phase is much more an issue with smaller, shorter or higher frequency waveforms. In a nutshell, we are screwed.
Even when you can get them to, a customer’s residence and have a line tested, they DO NOT test the frequency response of the line. They perform the following two tests.
1. The first test it is a 1,000 Hz tone coming from the central office switch and measured at the customers demark for attenuation. The 1,000 Hz tone is significant because it is smack in the middle of the voice frequency range (your vocal cords cannot really make sound above 3000 Hz). Assuming they have their test equipment Setup and terminated properly they can get a dB measurement of the 1,000 Hz tone and say that is in some specified range. Please note that there are no documented specified ranges for analog voice lines using this test setup. The results of this test are merely to identify the signal level coming in. I believe somewhere between -6dbm and -20dbm is acceptable but I’m not really sure.
2. The second test they may perform it is terminating the line and doing a Cmsg noise test measuring the title channel noise on the facility with no signal provided. Ideally there are some specifications that identify how much of this noise is tolerable for a voice circuit but I’m sure the tester doesn’t know where that acceptable threshold is. The tester might as well pickup of the telephone receiver and blow in to it to see if the line passes mustard….
The problem was this noise test is that putting a signal on the line generates additional noise and interference that the modems have to contend with all the time. This noise is not present when the line is just quietly terminated at each end.
The test I would really like them to perform is a frequencies sweep between 300 and 4000 Hz and a Bridge tapped test to determine how out of phase my signal may look by the time it gets to the central office and whether it falls into specific characteristics too. If these tests are routinely performed on the TOLL grade facilities between Telco central offices they should also be performed at the customer’s request on the drop side of the switched circuit. It time consuming, it’s out of the question.
I’ve had the conversation with several Telco repair supervisors about lifting Bridge taps and performing signal-to-noise ratio tests. The conversation ends when they say, “we do not guarantee data of above 2400 Baud”. They are right too this is not a conditioned data line and I have no recourse.
Incidentally, this bridge tap problem became apparent right after 9600-baud modem’s where replaced with 14.4k modems. The problem then got escalated with 28.8 and now begins to take on monstrous effects on the new super modems.
It is an interesting argument, putting these modems on voice lines and pushing the envelope beyond their engineered usefulness. The real culprit in this equation is not the Telco’s. It is the bastard scumbag modem manufacturers pulling the wool over the eyes of the consumer and expecting us to make it work and/or absorb all of the shit when it doesn’t. Telco’s do not want you to use high speed modems on your voice lines anyway so you’ll find a very limited support from that camp.