Modem Protocols Explained
MODULATION/STANDARD CONNECTION SPEEDS (BPS) Bell 103 110, 150, 300. CCITT V.21 110, 150, 300. Bell 212A 1200. CCITT V.22 1200, 600. CCITT V.22 bis 2400. CCITT V.32 9600, 4800, 2400. CCITT V.32 bis 14400, 12000, 9600, 7200. Telebit PEP High speeds, proprietary to Telebit. US Robotics HST High speeds, proprietary to US Robotics. "V.Terbo" 20000 bps, not a standard, never will be. "V.Fast" Preliminary implementations of V.34. "V.FC" "V.Fast-Class". ITU-T V.34 28800, 26400, 24000, 21600, 19200, 16800, 14400
For consistency we use "ITU-T" to refer to CCITT standards.
If the answering modem supports the type of modulation you have selected, the connection is made immediately. Otherwise the modems attempt to "fall back" to a lower type of modulation, the highest one they have in common. For example, if a 2400-bps V.22bis modem calls a 14400-bps V.32bis modem, the two modems will communicate at 2400 bps, the highest speed they have have in common. Fallback is generally reliable among the ITU-T standard modulations:
V.34 ->. V.32bis ->. V.32 ->. V.22bis ->. V.22In the USA, most modems fall back from V.22 to Bell 103, but different (incompatible) low-speed modulation methods are used in other countries, such as ITU-T V.21 or V.23. Most modern high-speed modems have a command to choose the low-speed fallback standard; in the USA, you should set your modem to fall back to the Bell 103 standard.
Within a particular modulation (V.32 and up) it is also possible to fall back to lower speeds when a higher-speed connection can't be made, for example because of excessive noise on the telephone connection. For example, a V.32bis connection can be made at 14400 bps, 12000 bps, 9600 bps, or 7200 bps. A V.34 connection can be made at 28800, 26400, 24000, 21600, 19200, 16800, or 14400 bps.
Mixing proprietary modulations like PEP or HST into the fallback scheme generally requires special configuration settings on your modem.
It is also possible for the two modems to change speeds in the middle of a connection, perhaps because they have detected an increase in telephone-line noise; this is called "retraining" -- the connection goes "on hold" for a couple seconds, but is not broken. Retraining occurs only if both modems have this capability and have agreed with each other to use it.
CONNECT 1200which your communication software takes as a signal to change its interface speed to 1200 bps before attempting to go "online" with the remote computer or service.
Most modern modems can be configured to fix their interface speed at a given value, rather than change it according to the connection speed. This is desirable when using data compression. In this case, the CONNECTION SPEED (or MODULATION SPEED) between the two modems is different from the INTERFACE SPEED between the modem and the computer. The modem performs the speed conversion between its telephone side and its data side, and your communications software must be configured to IGNORE the speed given in the CONNECT message.
MNP Level 1, 2, 3, or 4
ITU-T V.42 = LAPM
Telebit PEP (proprietary)
US Robotics HST (proprietary)
When modems' initial error-control methods do not agree, automatic fallback is usually as follows:
V.42 ->. MNP 4 ->. MNP 3 ->. MNP 2 ->. MNP 1 ->. noneWhen PEP, HST, or other proprietary methods are involved, special configuration settings are needed on the modems to specify the fallback sequence.
Please note that no connection can ever be free of errors. The error correction technique used between the modems might be extremely effective, but it is not foolproof. More to the point, the connections between the modems and the computers are not error-corrected, nor or the data paths within the computers. Thus it is still quite common to experience data loss or corruption, even on an error-corrected modem connection. Common causes include: buffer overflows (often due to a lack of adequate flow control between the modem and the computer -- see below), interrupt conflicts, loose connectors, and malfunctioning devices.
Beware of RPI modems. They do NOT perform error correction themselves, but rely on external software to do it. Most software does not. These modems are NOT SUPPORTED at Columbia University.
Effectiveness of MNP 5 and V.42bis compression vary between 0% and 400% or higher, depending on the nature of the data. Compression fallback:
V.42bis ->. MNP 5 ->. noneWhen PEP or HST is involved, special configuration settings are needed on the modem to specify how these fit into the fallback sequence. Again, beware of RPI modems. They do not do compression themselves, but rely on external software to do it.
When using error correction or compression, or modems that are capable of retraining, it is essential to enable an effective form of flow control between each modem and the computer (or terminal, or other device) it is immediately connected to. Without effective flow control, data will be lost when one device sends data faster than the other one can receive it.
Flow control between the two modems is handled by the underlying error modem-to-modem correction protocol: MNP or V.42. If there is no underlying error-correction protocol, then there can be no flow control between the modems, and therefore no protection against data loss EVEN IF there is flow control between the modem and the computer. This applies, in particular, to RPI modems.
The following additional sources of information may or may not be useful. Since they are outside Columbia University, we have no control over their content (accuracy, timeliness, maturity, political correctness). Use at your own risk.