root@zswamp.fidonet.org (Geoffrey Welsh) (11/19/90)
I recently saw a lexicon posted here, and thought that this might not be a
bad place to alpha-test my spiel on 9600+ bps dialup modems. Some of you may
benefit from it, but I'd appreciate E-Mail from anyone who wishes to correct,
contribute, or whatever. Please keep in mind that this is meant to give
vanilla 2400 bps users an idea of what others are talking about.
High-Speed Dialup Modems
PEP Before even 2400 bps dialup modems received public acceptance, a
company called Telebit Technologies was developing a 9600 bps modem
based on a multicarrier modulation scheme. The Telebit TrailBlazer
sends data at speeds from 0 to 8 baud on each of 511 carriers (spaced
every 7.8 Hz from DC to 4 kHz) using Quadrature Amplitude Modulation
(QAM).
PEP stands for the Packetized Ensemble Protocol, which the
TrailBlazer uses to correct errors, reassemble the data, and adjust
the modems' operating parameters to achieve optimum performance. This
was called Dynamically Adaptive Multicarrier QAM, or DAMQAM. Telebit
later added MNP support for non-DAMQAM connections and PEP2 (a
protocol with data compression) to what became known as the
TrailBlazer Plus.
The TrailBlazer gained great acceptance in the Unix world not only
for its speed and ability to operate reliably over a wide range of
line conditions, but also because it featured 'protocol spoofing',
which means that the modem would recognize the initiation sequence of
selected protocols and intercept them. In effect, the computer systems
would be using XMODEM, Kermit, or UUCP-g only to carry the data as far
as the modem, which would then send the data using its own fast and
efficient protocol, to the other modem... which simulated the same
protocol for the other host, giving each computer the impression that
it was indeed speaking directly to the other. This resulted in
tranbsfer rates never before seen over dialup lines or even long
serial wire connections.
Modems from Microcom and other manufacturers also featured protocol
spoofing, but they were not compatible with Telebit's multicarrier
modulation scheme, and thus never gained extensive popularity in the
market.
MNP Microcom Networking Protocol, or MNP as it is called, consists of
several layers of software protocols, often implemented within the
modem itself, which detect and correct errors and provide improved
throughput.
i. Basic link: MNP service classes 1 through 3.
MNP1 is much like XMODEM or Kermit in that the stream of characters is
broken into blocks of fixed length, and a checksum is attached to
each. After each block is sent, the transmitter waits for an
acknowledgement from the receiver that the block arrived intact before
moving on. If the receiver indicates an error, the block is
retransmitted.
MNP2 implements a 'sliding window' protocol; that is, several blocks
are sent, and the receiver identifies intact and damaged blocks by
number. Because the transmitter is permitted to send a fixed number of
blocks ahead of the oldest unacknowledged block, the wasted capability
(the time that the transmitter must wait, silent, for the signal to
continue) is drastically reduced or entirely eliminated. Throughputs
approaching the bit rate divided by ten are possible.
MNP3 uses the same sliding window arrangement as MNP2, but takes
advantage of synchronous data links, such as those in the carriers of
modems operting at 1200 bps and above. It strips the start and stop
bits of incoming characters, and sends only the 8 data bits. Thus, the
throughput may approach the bit rate divided by eight, or a 25% boost
over asynchronous framing techniques.
All MNP connections use one of these three service classes as the
foundation for their operation.
ii. Features
MNP4, called dynamic packet assembly, permits the use of data blocks
of varying size. Because the relative overhead of larger packets is
smaller than that of standard packets, MNP4 should increase
throughput. by using larger packets under good conditions and smaller
ones under noisy conditions (reducing the volume off data that must be
retransmitted). MNP4 may be used with any of MNP1 through MNP3.
MNP5 implements a form of dynamic data compression. It detects
repeated patterns in the data stream and reduces redundancy in the
transmissions. When transitting text files, this can significantly
increase throughput but, when used on files which have already been
compressed using run-length, Huffman, or Lempel-Ziv techniques, the
overhead may actually reduce the effective data throughput. It is
wiesest to disable MNP service class 5 when it is known that the
majority of information to be transmitted is in compressed form. MNP5
may be used with any of MNP1 through MNP3, with or without MNP4.
Higher levels of MNP are proprietary to Microcom and are rarely found
in other manufacturers' modems.
V.29 The CCITT developed standards for 9600 bps communication over the
V.32 dialup telephone network. The CCITT recommendation V.29 described a
half-duplex technique which required the modems to reverse the
direction of the carrier whenever it was necessary to send data in the
other direction. V.29 is used in many inexpensive 9600 bps modems,
some of which are compatible with CCITT Group III Facsimile standad.
V.32 specified that echo cancellation be used to permit
simultaneous communication in both directions. Echo cancellation with
the necessary precision is complex, and the electronics to do it
proportionately expensive. Recently, the mass production of V.32
chipsets has lowered the cost of V.32 modems significantly.
HST Because V.32 was, at the time of its introduction, very expensive
to implement, USRobotics derived from V.32 an asymmetrical modulation
technique which provided a 9600 bps 'forward channel' and a 300 bps
'back channel'. This allowed the HST to send keystrokes, ACK packets,
etc. back to the host without requiring that the high-speed carrier
direction be reversed - but was incompatible with any other 9600 bps
modulation technique. The HST used a variation of MNP for its error
correction.
The HST became popular because of its relatively low price
(compared to the complex Telebit modem and the premium Hayes product)
and because of the 'SYSOP offer' (an agressive marketing plan designed
to place many of these modems with operators of public-access systems,
forcing those who communicated with them to consider the HST when they
wanted to upgrade to 9600 bps).
V-9600 At the same time, Hayes was also developing a less expensive V.32
derivative. In stead of accepting a reduced speed backchannel, Hayes
concentrated on refining the ping-pong technique used to simulate full
duplex using a single carrier. The resulting product, the V-Series
Smartmodem 9600, was also incompatible with other 9600 bps modulation
techniques.
For error correction, Hayes offered a choice between X.25 and LAP-B
(Link Access Protocol, Balanced), an SDLC derivative. Neither of these
offered compatibility with the emerging industry standard, MNP.
Hayes also introduced a 2400 bps modem which offered X.25 or LAP-B,
and a "V-Series Modem Enhancer", which gave existing Hayes Smartmodem
2400 owners the same features.
HST 14.4 Faced with significant competition from inexpensive V.29 modems
(some of which used powerful data compression techniques to achieve
outstanding throughput), USRobotics increased their physical bit rate
from 9600 to 14400 bps. The resulting 50% boost in throughput gave
them a performance edge over most other manufacturers. This model of
HST is often called the "HST fourteen-four".
T2500 It was not long before owners of Telebit, USRobotics, and Hayes
HST DS modems became concerned that their considerable investment in high
Ultra speed modems did not always permit them to communicate at 9600 bps
with others who had also spent a lot on high speed modems. This
pressure, combined with the lowering cost of V.32 chipsets, led each
manufacturer to introduce a 'dual standard' model, which was capable
of communicating both with the manufacturer's own modems and with
others which supported V.32. These were the Telebit T2500, The
USRobotics Courier HST Dual Standard, and the Hayes V-Series Ultra
Smartmodem 9600.
V.42 In the mean time, the CCITT had been meeting to develop a standard
for error correction between dialup modems. The result, recommendation
V.42, included two protocols: LAP-M (the Link Access Protcol for
Modems, derived from SDLC) and MNP. The
latter, up to service class 4, was included as 'Annex A' to maintain
compatibility with the large pool of existing MNP-equipped modems.
V.42bis Aware also that MNP5 and other data compression techniques had
become popular, the CCITT released recommendation V.42bis, which
specified a data compression technique which would provide dramatic
performance improvements where possible, but which would not become a
burden when the information being sent was already well compressed.
V.32bis Currently in the discussion phase, V.32bis may be released by
February of 1991. It will describe a full-duplex carrier technique
that will provide 14400 bps physical bit rate, effectively 'catching
up' to the HST 14.4. Some manufacturers have already begun advertising
V.32bis modems, but there is no gusrantee that they will be compatible
with the final specification.
--
UUCP: watmath!xenitec!zswamp!root | 602-66 Mooregate Crescent
Internet: root@zswamp.fidonet.org | Kitchener, Ontario
FidoNet: SYSOP, 1:221/171 | N2M 5E6 CANADA
Data: (519) 742-8939 | (519) 741-9553
MC Hammer, n. Device used to ensure firm seating of MicroChannel boards
Try our new Bud 'C' compiler... it specializes in 'case' statements!jparnas@larouch.uucp (Jacob Parnas) (11/22/90)
In article <4726.27476730@zswamp.fidonet.org>, root@zswamp.fidonet.org (Geoffrey Welsh) writes: |> |> V.32bis Currently in the discussion phase, V.32bis may be released by |> February of 1991. It will describe a full-duplex carrier technique |> that will provide 14400 bps physical bit rate, effectively 'catching |> up' to the HST 14.4. Some manufacturers have already begun advertising |> V.32bis modems, but there is no gusrantee that they will be compatible |> with the final specification. I think V.32bis will more than catch up to the HST because it is full duplex 14400 in each direction, not fast in one direction and slow in the other. ------------------------------------------------------------------------------ | Jacob M. Parnas | DISCLAIMER: The above message is from | | IBM Thomas J. Watson Research Ctr. | me and is not from my employer. IBM | | Arpanet: jparnas@ibm.com | might completely disagree with me. | | Bitnet: jparnas@yktvmx.bitnet \---------------------------------------| | Home: ..!uunet!bywater!acheron!larouch!jparnas | Phone: (914) 945-1635 | ------------------------------------------------------------------------------