Generating and checking for HDLC flags and performing HDLC zero-bit stuffing and deletion The PC does everything else, such as the image rasterization and data compression. Many fax protocol functions are very timing-sensitive, which can bea problem for PCs with multitasking operating systems--especially UNIX, which is usually not good at meeting such real-time requirements. However, Class 1 fax boards are the most flexible, since almost all functions are done in software on the PC, such as the following:       Error-correcting mode (which is specified in T.30)       Adaptive answering, in which the fax board (which also has a data modem) decides whether an incoming call is a fax or data call and handles it      accordingly (this function usually can be performed only by a Class 1 fax board) In addition to the Class 1 functions, Class 2-based fax boards perform the following functions (so the PC does not need to):       Line supervision (that is, establishing and clearing the call)       Support of error-correcting mode (though this is not widely implemented) Note that not all Class 2 fax boards provide a Class 1 software interface (that is, they do the Class 1 functions, but automatically). Because of the long time between the availability of the first drafts of the Class 2 standard (SP-2388, document TR-29/89-21R8, dated March 21, 1990,and a later version released August 1990) and the final approved standard (EIA/TIA/ANSI-592, released November 1992, with first products availablein 1994), many fax boards (still) support only the draft. To distinguish these (there were changes made to the final standard), conformance with the draftstandard is indicated by "Class 2" (there are many undocumented variations of the implementation of this de facto Standard), and conformance with thefinal standard is indicated by "Class 2.0" (this is not widely supported, with the significant exception of USRobotics and ZyXEL). The ITU-T's international version of Class 2.0 is called T.32. It includes more functions, though, such as the data link functions described inTIA/EIA-602. In addition to the Class 2 functions, Class 3 boards were expected do the rasterization, based on higher-level information from the PC (such as text inASCII and graphics data in TIFF format). Initially, it was expected that offloading more functions to the fax board would result in Class 2 (and greater) fax boards making Class 1 boards obsolete.However, today's faster PCs can easily handle all the tasks necessary for faxing, even when using a "dumb" Class 1 fax board. Since the PC has thenecessary processing power anyway, why waste money building more power into the fax board? It is likely that the Class 3 standard will never beapproved. There are three incompatible specifications to support Binary File Transfer (BFT) between fax boards, so that they can send the original file (in theoriginal file format), not a rasterized version of it:       CCITT's T.434 (and the similar TIA/EIA-614)       Microsoft's At Work Fax (which also supports public key encryption, which provides authentication and encryption)       Intel's not-very-popular CAS, which includes a BFT capability (which works only with CAS) Only Class 1 (and not Class 2 or 2.0) fax boards can implement BFT or ECM. A summary of fax options is shown in the figure below. Figure 1: Fax                                                                         The Group III fax capabilities are specified in the following:       ITU-T T.4 (which is the same as EIA/TIA-465). This describes the page widths, scan resolutions, transmission times per line, and      one-dimensional compression scheme that are supported.       ITU-T T.30 (which is the same as EIA/TIA-466). This covers the negotiation method and options that are negotiated at connection time and the      protocol that is used to manage the session. A pointer page to fax information on the Internet is at the site http://www.faximum.com/FAQs/fax. A copy of the T.4 Standard is at gopher://wiretap.spies.com:70/00/Library/Techdoc/Standard/ccitt.t4. A copy of the test pages used for fax transmission speed testing is at http://www.cs.waikato.ac.nz/~singlis/ccitt.html. Some additional information is at http://www.grayfax.com/faxsminar.html (they make fax test equipment) and in the Usenet newsgroup comp.dcom.fax. See Asynchronous, AT Command Set, Authentication, CAS, Data Compression, DID, DN, Encryption, HDLC, ISDN, Modem, Paper, Pixel, RLE,Synchronous, and Usenet. ------------------------------HDLCHigh-Level Data Link ControlA bit-oriented, synchronous, link layer, data-framing, flow control, and error detection and correction protocol. Uses a header with control information and a trailing cyclic redundancy check character (which is usually 16 or 32 bits in length). Implementations are both standard subsets (see below) or vendor-specific (such as that used for the 56,000-bits/s interfaces on a vendor's remotebridge or router). IBM calls HDLC SDLC. Some standard subsets are listed in the following table.                   HDLC Subset                                                Used for 802.2        Logical link control                                                FDDI, Token Ring, and some Ethernet LANs LAP        Link Access Procedure                                                Early X.25 implementations LAPB        Link Access Procedure, Balanced                                                Current X.25 implementations LAPD        Link Access Procedure for the ISDN D channel                                                ISDN D channel and frame relay LAPM        Link Access Procedure for Modems                                                Error-correcting modems (specified as part of V.42)BISYNC is (was) an older method of synchronous data communications. See 802.2, BISYNC, CRC, DLC, Frame Relay, ISDN, MNP, SDLC, Synchronous, V.42, and XModem. ------------------------------CRCCyclic Redundancy CodeA key component in the error-detecting capabilities of many protocols. A number of bits (usually 16 or 32) generated from, and appended to the end of, a block of data to provide error detection. The message receiver alsogenerates a CRC from the block of data and compares it to the one appended to the received message. If the two match, then there is a high probabilitythat the received message has not been corrupted. There are two commonly used 16-bit CRC generator polynomials. The ITU-T (CCITT) standard 16-bit generator polynomial (used on X.25 networks,and by the Kermit, YModem, and ZModem file transfer protocols, for example) is represented by x^16 + x^12 + x^5 + 1. This is called CRC-CCITT andrepresents the binary number 10001000000100001 (note that there is a one in bit positions 16, 12, 5, and 0). IBM protocols such as SDLC use CRC-16, which is represented by x^16 + x^15 + x^2 + 1. In either case, the CRC is the remainder after binary division of the message (taken as a long string of ones and zeros, regardless of byte boundaries) bythe generator polynomial.                                                   CRC-CCITT CRCs detect:       All single- and double-bit errors       All errors of an odd number of bits       All error bursts of 16 bits or less (the length of an error burst is the number of bits between and including the first errored bit and the last errored      bit--any number of bits between may be errored)       In summary, 99.998% of all errors (that is 1/6 ) Because of these error-detection capabilities, 16-bit CRCs are usually limited to use with message of less than 4 kbytes (there are enough ways tocorrupt messages larger than 4 kbytes that catching "only" 99.998% of them is considered inadequate). 32-bit CRCs are used for messages up to 64 kbytes in length. Such CRCs detect 99.999999977% (that is, 3/2 ) of all errors. The generator polynomial forthe 32-bit CRCs used for both Ethernet and Token Ring is x^32 + x^26 + x^23 + x^22 + x^16 + x^12 + x^11 + x^10 + x^8 + x^7 + x^5 + x^4 + x^2 + x +1. See Checksum, Parity, HDLC, Kermit, MNP, V.42, and XModem. ------------------------------ModemModulator/DemodulatorAn electronic device that converts binary data (for example, the +/-12 V EIA-232 signals from a PC's COM port) to analog tones and voltages that aresuitable for transmission over standard dial-up or leased line telephone lines. Given the fixed 3,000-Hz bandwidth of typical POTS analog lines and the increasing power of Digital Signal Processing (DSP) ICs, newer modems usemore complex types of modulation to provide faster data rates. The following table summarizes the types of dial-up (in contrast to leased line) modems.  Modulation            Data Rate (bits/s)                             Comments Bell 103                         300                               Bell 212A                       1,200                             Usually compatible with V.22 V.22bis                       2,400                               V.32                       9,600                               V.32bis                      14,400                               V.terbo                      19,200                             Pre-V.34, nonstandard modulation, with fewer features than V.34 V.FC                      28,800 V.34                      28,800                             Was called V.fast during development V.34bis                    33,600[a]                             Not yet standardized Footnotes:        [a] The maximum speed has not yet been determined. Current external dial-up modems (they are separate boxes that connect to your PC's serial COM port, through an EIA-232 cable) will usually handleboth synchronous (as required by X.25 and IBM 3270 and 5250 protocols) and asynchronous (as used by standard PC serial COM ports and most bulletinboard systems) data communications. Internal modems (which are printed circuit boards that mount inside a PC and include a built-in serial COM port) are often asynchronous only. Most current modems have both built-in V.42 error correction and V.42bis data compression. Many current modems also have fax capability:       Fax machines (and PC fax modems) use different types of modems than data modems; since a fax transmission is basically one-way during a      call, half-duplex modems are used (they are less expensive).       All current fax machines and modems that use standard telephone lines are Group 3. Most have an Error Correction Mode (ECM).       Fax boards require a command interface for the PC to communicate with the fax modem. There are several such interfaces, but the most      common is Class 1, which requires the host PC (and not the fax modem's on-board processor) to do most of the work (this provides the most      flexibility and features, the lowest cost, and usually the best performance, since the PC's processor is far faster than any fax modem's dinky little      single-chip controller). At connection time, modems send tones to each other to negotiate the fastest mutually supported modulation method that will work over whateverquality line has been established for that call. If you wanted the best modem on your block, you would want one with the following:       V.34 modulation, which sends data at up to 28,800 bits/s, depending on how good a phone connection you get and whether the other end has an      equally fast modem.       V.42 error detection and correction, which specifies both LAPM and MNP 4 so that you will be compatible with whatever the modem at the other      end may support.       V.42 bis data compression and maybe MNP 5 data compression too--in case that's all the modem at the other end has.       V.17 fax modulation, which sends fax images at 14,400 bits/s (if the fax machine or modem at the other end has this modulation and the      connection is good enough) or slower (most commonly V.29 9,600 bits/s).       The fax modem or software should also implement ECM. The "universe" of dial-up modems and features is shown in the accompanying figure. Figure 1: Modem                                                                         Most modems are based on ICs made by Rockwell International. See Asynchronous, Baud, CAS, DCE, DTE, EIA/TIA-232, EIA/TIA TSB-37a, Fax, Full-duplex, MNP, PCM, POTS, Rockwell International, RPI,Synchronous, V.8 through V.120, and WAN. ------------------------------PCMPulse Code ModulationA method of digitizing audio (turning it into those ones and zeros that computers love so much). Periodically (8,000 times a second for telephone systems, 44,100 times a second per right and left channel for audio CDs) samples the input and produces(for example) an 8-bit (for telephone systems) or 16-bit (for audio CDs) value representing the amplitude of the audio input at that instant in time. Since digitizing assigns a specific (binary) number for any input amplitude, and only a given number of amplitudes are available (for example, 256 for8-bit digitizing), chances are that the assigned number will be a little bit too high or a little bit too low compared to the actual input. This error is calledquantization error and produces quantization noise at the output. Since better accuracy is required for lower signal levels (since the quantization error is more significant compared to the signal), digitization values areassigned closer together for lower signal levels. This type of nonlinear Analog-to-Digital converter (or A/D) is called a CODEC (Coder/Decoder). There are two such nonlinearities standardized: