Talk:Modem

From Wikipedia, the free encyclopedia

"Bus" network topology This article is part of WikiProject Computer networking, an attempt to build a comprehensive and detailed guide to Computer networking on Wikipedia. If you would like to participate, you can edit the article attached to this page, or visit the project page, where you can join the project and/or contribute to the discussion.
Computer networking Portal
B This article has been rated as B-Class on the assessment scale.
Mid This article is on a subject of mid priority within Computer networking for inclusion in Wikipedia 1.0.
Computer networking To-do:
WP:TEL This article is within the scope of WikiProject Telecommunications, an attempt to build a comprehensive and detailed guide to telecommunications on Wikipedia. If you would like to participate, you can edit the article attached to this page, or visit the project page, where you can join the project as a "full time member" and/or contribute to the discussion.

Contents

[edit] ridiculous statement

from the article: "Without the discovery and eventual application of trellis modulation, maximum telephone rates would have been limited to 3429 baud * 4 bits/symbol == approximately 14 kilobits per second using traditional QAM".

This is statement is speculative, and on top of that, incorrect. There are many other ways to improve the modulation - for example, OFDM would give the same spectral effiency, as witnessed in ADSL protocols (which use OFDM). please remove statement.

[edit] UTC does improve speeds?

europrobe 07:15, 2005 May 19 (UTCdoes improve speeds by using a compression algorithm optimized for internet content (html etc). 'appears to be' should be rewritten with a technical sourced explanation.

The reason why we're having this conversation is robbed bit signaling. This "steals" the least-significant bit in every 6th byte transmitted in each DS0. (That is: less than one control bit per byte.) On average, only half of these bits will have the "wrong" value from the point of view of the bearer channel. The actual theoretical bitrate on channels subject to robbed bit signalling is therefore 64 - (8 / 6 / 2) kbit/s = 63.333 kbit/s. This assumes either only one T-1 circuit subject to robbed-bit signalling being in the end-to-end communications channel, or superframe synchronisation between all concatenated T-1 circuits. For this bitrate to be available to modems, these modems would have to be able to achieve superframe synchronisation themselves (i.e. not just bit synchronisation with the communications channel). A very tall order indeed. JanCeuleers 19:38, 24 July 2006 (UTC)

[edit] FCC Speed Myth

'Current FCC regulations (Part 68) limit the top speed of these modems to 53Kbps.' There is no such regulation on speed limits. The limitation is on power output which limited the old X2 to 53.3K. Newer modems can acheive 54K and under perfect conditions 56K. Someone please rewrite and source. see http://www.google.com/search?hl=en&lr=&q=fcc+modem+speed+myth+56k

[edit] AT&T history

I have not found much information on AT&T modems which preceded the 202. Modem designers were active in the 1950s and did publish several papers which I have not examined. SAGE is covered by several papers. http://www.att.com/history/milestone_1958.html provides a different version of the modem shown in the wiki article. The device is labelled RECORDED CARRIER SUBSET. Another device (not in wiki image) is labelled DIGITAL SUBSET.

[edit] AT&T documentation

Most of my claims regarding dating of AT&T 202, 201 and 103 came from Bell Telephone Magazine issues in the 1960 to 1963 era. BTM Autumn 1963 suggests 103A is a 1963 development but other sources claim 1962. I will stick with 62 until evidence is available.

[edit] 1200bps history

Almost all of this information came from: Data Communications A User's Handbook published by Racal-Vadic sometime after 1978. This sentence: "According to Vadic, this made the 212 protocol incompatible with acoustic coupling." is a personal recollection.

[edit] Modulation technique history

Quite a bit could be written about abandoned modulation schemes. There were some problems with CCITT's recommendation of V.29 which was covered by Codex patents. There was a belief that V.29 modulation with multiple circles in constellation provided better protection against phase hits than did the Bell 209 scheme with a square constellation. Perhaps I should create and import and some drawings. Rdmoore6 05:38, 19 January 2006 (UTC)

See the entry on modulation first (unless you wrote it :)
I think some corrections are also called for. To my knowledge, 24 64Kbps lines make a T-1 since forever (DS-0 was standardized around 1957, AT&T ESS-1A computerized phone switches began deployment around 1969-1973 I think), and since about 8kbps gets used for control, the theoretical limit for a US analog phone line has _always_ been 56kbps (though many people didn't think so before modulation schemes past 2400bps were thought of). Digital switches helped reach that max by having a cleaner signal path, but it was always at least theoretically possible.
(In the "older tech had certain advantages" department, if you made a local call on a step-switch phone system, you basically had a clean piece of copper from one handset to the other and could probably run a megabit at least (with modern ICs, which of course didn't exist at the same time as step-switches were in public use)).
On a related note, I never heard of echo cancellation as being a motivation for telebit's multiple carriers. though I'm not saying it wasn't. They were promoted as being useful because a very common form of signal loss on phone lines was loss within a narrow frequency band; interference might nuke 1580-1700 Hz, for example, and that would totally hose a modulation scheme that needed the whole bandwidth relatively clean. Plus, the multiple carrier scheme allowed for gracefull fallback, whereas earlier schemes would lose a minimum 25% capacity (9600 to 7200).
The inaccuracies noted above make me have concern for the accuracy of other parts of the article; I don't remember Hayes being the only folks to allow a computer to control answering and calling out, they were just the ones who managed to become the defacto standard, especially in the microcomputer market. DEC, IBM, and AT&T/Teletype all made non-Hayes compatibles, but I don't know if any of them were introduced before 1981.

[edit] SAGE Modems

I'm also not sure SAGE units would be the first modems. Other than being connected to "computers", is there much difference between the boxes that hooked a SAGE unit to a copper line and what wire services and teleprinters used?
Akb4 22:31, 11 May 2006 (UTC)

The argument for SAGE "Data Terminals" being the first modern modems is based on three papers presented at the AIEE Fall General Meeting in Pittsburgh October 26-31, 1958. These were subsequently reprinted in the January 1959 issue of AIEE Transactions, vol 77, pt I, 1958 (January 1959 issue). The article entitled "SAGE Data Terminals" by R.O. Soffel and E.G. Spack is of particular interest.

A SAGE DDT was effectively a simple modem which transmitted at either 1200 bps or 1600 bps. A 2 kHz carrier was used. "Since some of the channels which carry the line signal have narrow bandwidths, vestigial side-band transmission is employed; the usable sideband is limited to approximately half the bit rate." Data clock was supplied by the data source and passed to the data sink. Unlike RS-232 modems the data source also supplied word marks which were passed to the sink. Unlike RS-232 the data source could be located about a mile from the modulator (DDT).

An important aspect of the SAGE DDT/DDR scheme is the modulation technique. Prior to WWII FSK was used by telegraph companies. Reichspost had a particular efficient FDM system for slow telex channels. I suspect the significance of SAGE system was that it was the first instance of synchronous data transmission.--Rdmoore6 02:56, 7 August 2006 (UTC)

[edit] Um, what?

I find the following:

Because signals travelling through the air must be analog (digital signals cannot be transmitted without a guided transmission medium such as wire)

to be very suspect. Can anyone explain to me what this lines is trying to say? Because I'm pretty sure what it actually says isn't true... User:Glenn Willen (Talk) 18:04, 28 March 2006 (UTC)


This can get a bit philosophical, but you could say "digital signals" don't even really exist. Signals are analog, though they can represent digital values. Note that even a signal which consists only of two levels of 0V and 1V "looks binary", but IS analog (with an ASK modulation). That's why on reception one has to deal with the degraded levels and decide wether a, say, 0.4V signal was 0V or 1V in origin.
Digital signals are an abstraction. They have a finite number of levels, which are not measured in Volts and can't degrade from one level into another. You would have a 0 level and a 1 lever, for example, with nothing in between.
Even IF the original sentence meant digital modulations instead of digital signals, it would still be wrong. Digital modulations can of course be transmitted by radio - look at WiFi or GSM, which are digital systems using digital modulations over the air.
--Ermija 10:43, 23 April 2006 (UTC)
I'd say a "signal" could be digital. But to move that signal around via an analog medium, such as a radio wave or a piece of wire, you would have to encode it via an analog means, such as modulation. At the very least whoever wrote the original phrase was unclear; it's true that digital signals travelling thru the air must do it by piggybacking on something analog, but that's true of signals on wires, too! Anyway, I'm rewriting it. Akb4 23:11, 11 May 2006 (UTC)
oh heck, out of time. the whole radio modem section needs lots of work. I'd recommend making it clear that radio modems are not really different than other kinds. Akb4 23:20, 11 May 2006 (UTC)
If anyone does rework this, then please be careful about the terminology. For example, encoding is different from modulation, and modulation is not a form of encoding (nor vice versa). In general, a data communications transmission device can contain both an encoder and a modulator (and the receiver obviously a demodulator and decoder). Recommended reading: Bernard Sklar's excellent book on digital communications. JanCeuleers 12:50, 25 July 2006 (UTC)
It's not just the radio modem section that needs work: the introduction says that voiceband modems "converts those sounds back into 1s and 0s"!
Unless you want to talk about encoding schemes, or explain the difference between a 'digital' and a 'binary' signal, those signals on the analog lines are 1s and 0s all the way. The digital signal is represented by an electrical analog on both sides of the modem. A voiceband modem converts signals between a modulated signal and a baseband signal, or between a coded signal and an uncoded signal, not between 'analog' and 'digital'.
A 'modem' is not an 'analog to digital converter'. A modem would have some kind of 'analog to digital converter' on both sides. If you were looking at a simple RS232 interface, the adc would be a single bit adc, perhaps formed by a clock recovery circuit and a comparator - just like the single bit ADC on the 'analog' side of a 300 baud modem.
Anyway, those 'analog' phone lines are called that because they carry analog voice conversations. When you connect a modem to them, they are just as 'digital' as -the most digital thing you can think of -. Using the analog phone line as a definition of a 'modem', an ADSL modem is not a modem because it does not connect to an 'analog' phone line. The problem is, that leads to the difficulty explaining what a radio modem is: clearly it doesn't connect to an analog phone line, but it is a modulator/demodulator, so what is it, and why?
Perhaps it's time to go through the whole article and delete references to 'analog' (and analogue). (david) 218.214.148.10 07:08, 20 November 2006 (UTC)
Our anonymous correspondent makes valid criticisms but poor recommendations. Yes, the distinctions among "Digital", "Discrete" and "Baseband" are not make clear in the article. They could be either discussed or referenced.
On the other hand, the ability radio lacks is baseband, not digital. If a DSL modem is not wired into an analog line, to what kind does it connect? And most modems have a ADC and DAC inside them. Otherwise the DSP couldn't do its job. Jim.henderson 14:30, 20 November 2006 (UTC)
The data is digital on both sides. Implying that the data is not digital on the 'analog' side is a source of confusion. Can the (obvious) confusion caused by the misuse of these terms be corrected?
BTW, a DSL modem connects to a Digital Subscriber Line :~) (david) 218.214.148.10


All things are analog. Even inside a basic computer device, a 3.3 volt level (for example) represents a '1', and 0 volt level represents a '0' state. But it doesn't have to be exactly those values.... a '1' state is actually a range from 2.0 upto 3.7. Hence the signal is truly an analog signal, and when we high-speed engineers model our circuits via simulation, we use analog to catch any flaws in the design (like indeterminate states or bad timing).
Back to the original statement: Whether you are using radiowaves, or wires, or two cans with a string tied between them, makes no difference. They are ALL analog mediums. They all have a wide range of varying voltages, currents, or frequencies. What makes a wire "digital", or a radio station "Hybrid Digital", is not the medium. The medium is still analog. What changes is the conscious decision to define 2.0-to-3.7 volts (for example) as '1' and values below 1.0 volts as '0'.
It's not the medium that changed; it's our interpretation of the incoming voltages that has changed.
theaveng 12:10, 20 July 2007 (UTC)



Earlier this month I wired two DSL modems to phone lines. Plain old telephone lines. The phones still ring; all analog jobs are still done. Yeah, the same wires are carrying voiceband signals, also called baseband and analog, while they carry complex, frequency divided partial response highband signals. And yes, those complex signals are Discrete signals and Digital signals. Perhaps the article should go into detail on how these words, as well as Analog, apply to modems, but it's already a long article with a very long history of equipment no longer used. Is it time to split off a "Modem History" article and add more theoretical material to the "Modem" article? Jim.henderson 02:10, 28 November 2006 (UTC)

[edit] Echo cancellation

The discussion about side tone does not seem correct to me. It is true to say that telephone sets create some feedback from the microphone to the earpiece, and if a modem were acoustically coupled to a telephone set then this would be a source of echo. The real problem, it seems to me, is that the outgoing and incoming signals are transmitted and received on the same physical twisted pair. They are combined in the modem by a 2/4-wire transformer circuit, and again on the analog line circuit in the central office, usually embedded on a hybrid circuit (i.e. the H in BORSCHT). Each one of these conversions causes reflections, which means that the echo canceller needs multiple taps. JanCeuleers 15:07, 6 June 2006 (UTC)

Yes, there is an Echo cancellation article, but it is sparse and dry, and that's where a more detailed discussion belongs. The link should replace, not supplement, any mention of "sidetone." Jim.henderson 02:10, 28 November 2006 (UTC)

[edit] Questions regarding the v.90 section

Reading the v.90 section brought several question and concerns to mind regarding accuracy and omission of important data. The following statement may be inaccurate in some ways: "Given that v.90 only works when the circuit is digital end-to-end, there is little reason to take the digital data from the computer, convert it to analog sounds, and then have the phone system convert it back to digital data so the process can be repeated at the far end of the phone line." It is my understanding that home modem users could benefit v.90 if the following criteria where met:

  • The ISP used v.90 modems connected the telephone network digitally.
  • The digital signal from the ISP could only be converted to analog once at the telephone central office (or central switch) before going to the ISP customer's house. Thus the connection does not need to be digital end-to-end, only between the ISP and the caller's CO. Multiple analog to digital or digital to analog conversions would limit speeds to 33.6kbps. This also limits direct connection between two home users of v.90 modems to 33.6kbps since they both have analog connections their CO's.

The os also two issues with v.90 this section seems to omit. The first is the v.90 allows only higher speeds for downloading. Uploading was still limited to 33.6kbs. The second issue (at least in the U.S.) was that while full 56kbps was theoretically possible, FCC regs limit the speed 54kbps. I am indeed remembering things correctly? If so, then this section needs to be rewritten to reflect this info. --Cab88 23:10, 13 June 2006 (UTC)

I found the answers to my questions myself using the book Upgrading and Repairing PCs (16th Edition) by Scott Mueller. The book backed ups my understanding and as such I decided to rewrite the section to conform to info in the book. --Cab88 11:03, 16 June 2006 (UTC)

[edit] Just putting in my two cents

Don't forget, that a carrier wave is not produced with analog modems on speeds lower than 1200 baud.

Um, yes, there is a carrier wave. (A baseband carrier wave). It is modulated at the signal frequency (the baud rate). If you turn the speaker on, you can hear the carrier wave, and the modulation. For V21 (asynchronous transmission at rates up to 300 baud), the carriers were 1080 Hz and 1750 Hz. Bell 103 used a similar system, but it was normally quoted giving the min/max frequencies, rather than the carrier frequency: conceptually, you could think of it as modulating 4 frequencies, rather than 2. 218.214.148.10

[edit] Analog modem

This phrase is used in the article to refer to a voiceband modem. All modems are analog modems; that's implicit in its job description. Some, such as microwave modems and DSL modems are wideband; that is they exploit bandwidth beyond the 3.6 Khz dial-up telephone voiceband, but the majority of 20th Century modems were voiceband dial-up modems. The only literal sense in which some modems are "analog" and some are not is, all modern ones use DSPs inside, while some 20th Century ones used only LC filters, op amps and other analog parts inside to generate and discriminate signals. Should "voiceband" be 1) explained, and 2) used in all parts of the article that now say "analog"? Jim.henderson 14:33, 29 October 2006 (UTC)

Cab88: DSL, cable and satellite modems are actually real modems, as they are not baseband tranceivers. All three convert digital data into analog signals for transmission with a (one or more) modulated carrier(s).

A modem is not an Analog to Digital Converter. Cable Modems do not 'covert digital data into analog signals'. A Cable Modem has analog signals on both sides, and has digital data on both sides. The digital data is carried by an electrical analog. A modem modulates and demodulates. DSL, cable and satallite modems are actually real modems, not baseband tranceivers: but the analog side of a voice modem is called 'analog' for historical reasons, because it was used to carry an electrical analog of an air pressure wave, not because one side of the modem was 'digital' and the other 'analog' 218.214.148.10
So, if a voiceband modem is called "analog" because it is voiceband, why not simply cut to the chase and tell it like it is? That is: Voiceband modem. History is lovely but not when it gets in the way of straight talk.
Jim.henderson 01:40, 28 November 2006 (UTC)

[edit] What OSI layer are modems working at? Why is PPP over modem not tunneling?

The article should discuss the OSI layers of the modems.

Have I understood it correctly?

The oldest modems were only working at the physical layer, since their function was pure modulation and demodulation over a circuit switched channel. Data were either transmitted asynchronously, using start and stop bits, or synchronously at constant bit-rate, using self-synchronizing line code (?).

In 1981 the Hayes smartmodem supported computer controlled number dialling and hang-up, which could be considered as Medium Access Control adressing. (???)

Modern modems also include Logical Link Control, since they support error correction and deal with frames.

Modern modems also include data compression. In the OSI model, this is part of the presentation layer. However the presentation layer is a end-to-end layer, while modems are point-to-point links, so in this case it should be considered as data link protocol.

From higher layer protocols point of view, a modem behaves as if it were a circuit switched physical link (an asynchronous serial link), since the interface to a computer typically is an RS232 asynchronous serial link.

Question: Why is PPP over a POTS or ISDN modems not considered as tunneling, while PPPoE, PPPoA and [[[PPPoX]]] belongs to Category:Tunneling protocols? To my understading, PPP over modem is an LLC protocol encapsulated in another. Mange01 15:03, 22 November 2006 (UTC)

[edit] Shannon and Nyquist limit in PCM systems? Why different uplink and downlink speed in V.92 and V.90 modems?

User:134.58.253.130 added some discussion on the Shannon limit for modems. I have a few questions related to this:

  • You claim that the SNR is 50dB. Today almost all telephone switches in the world are digital, using PCM. Can the maximum possible SNR be calculated for PCM, if there is no noise? (I have made an attempt below, by relating it to the Nyquist capacity.)
  • A V.92 downlink can handle 56kbps in the PCM based downlink but 48kbps in the uplink from an analog user. Why? Can this difference be explained by different Shannon limits in the PCM case as in the analog case. Is it impossible to increase the uplink rate because of this Shannon limit? Perhaps the inter-symbol-interference may be higher in the uplink, since the modem symbols are not synchronized with the PCM sample instants? Perhaps it is not possible to use all 256 levels, since it may be hard to identify the maximum possible amplitude?

I think it is more appropriate to use the Nyquist capacity limit instead of the Shannon limit in the PCM case. The Shannon capacity is the maximum net bitrate if an ideal error correction code is used (which is not possible in the real world). Nyquist gives the gross bitrate if we now the maximum number of levels. We may assume that that we could use all N=256 levels, resulting in a gross bit rate capacity of 2*B*log2256 = 16*B bit/s, where B is the analog bandwidth in Hertz. If we assume that the bandwidth is B=4,000 Hz (i.e. ideal filtering), the modulation would result in according to Nyquist we could get 16*4,000 = 64,000 bit/s. If we instead assume only 3,400-300 = 3,100 Hertz Bandwidth, we would get 49,600 bit/s. In practice a frequencies outside the passband may also be utilized. The V.92 maximum downlink speed corresponds to 56,000/16 = 3500 Hertz Bandwidth.

N = 256 levels, without any noise, gives the same bit rate as the Shannon limit would give, if the SNR was 20 log10 256 = 48 decibel. However, note that Shannon gives the ideal net bit rate, and Nyquist the gross bit rate

Mange01 14:49, 30 November 2006 (UTC)

[edit] How is spectrum shaping achieved in V.92 PCM downlink?

How is the POTS channel filter (with passband of at least300 to 3,400 Hertz, in practice a little bit wider) compensated for in the 56kbps V.90/V.92 downlink? Somehow the spectrum must be shaped to avoid energy outside the passband.

My guess:

  • Alternative 1: The modem generates 8,000 symbols per second, but only utize 7 bit per sample (128 levels), resulting in 56000 bit/s. One bit per sample is used to generate a signal that compensates for the spectrum content that is outside the passband, resulting in very low spectral energy in that area. The probllem is that the energy outside the passband can not be completely ellminated in this way. I am not clear on which bit that is optimal to use.
  • Alternative 2: The modem generates 7,000 symbols per second, but utilize 8 bit per sample (256 levels), resulting in 56000 bit/s. The sample rate is than changed to 8,000 samples per second using upsampling with a factor of 8 (inserting 7 zero value samples between each symbol), digital interpolation filtering and downsampling with a factor of 7 (only keeping every 7:th sample)? The problem is that this upsampling process may cause quantization error.
  • Alternative 3: Only 14 of 16 samples are utilized for information. 256 levels per sample are utilized. The last two samples are utlized to compensate for energy outside the passband. No quantization errors are added, since filtering is not added to the useful samples. The spectrum of the compensation signal (the last two samples in every block of 16 samples) would mainly affect the spectrum below 250 Hertz and above 3750 Hertz).

Mange01 16:58, 30 November 2006 (UTC)

[edit] Bonding modems

I started a section on bonding modems (aka inverse multiplexing modem). Their is more info I did not add in the link I provided in the references section if anyone wants to add to or improve my wording in the section. --Cab88 13:34, 6 March 2007 (UTC)

[edit] V.92

This section states that bandwidth seems to be limited to 56k for voiceband modems, but I remember reading somewhere that there the limit has been formally proven.


[edit] DSL "modems"

Um, DSL connection devices operate on a different principle and everything. Shouldn't that immediately be made clear in the subsection intro, rather than buried in the body text of the subsection? <looks confused>

--Kim Bruning 14:29, 4 June 2007 (UTC)

How different principle? They modulate; they transmit; they receive; they demodulate. They adapt between the baseband signals of the computer and the frequency dependent characteristics of a longer transmission line. The differences are outlined in the DSL modem article but none of them are differences between a real modem and something that isn't really a modem. Jim.henderson 14:36, 4 June 2007 (UTC)

[edit] Merge

Not at all. Modem is already very large and covers all kinds, voiceband, broadband, interplanetary, antique, still under manufacture, etc. 56k modem is about the only kind of voiceband modem still made, and should be the article people use when they want to find out how the modem in their home computer works. It should have all the details and none of the ancient historical background. Alas, as it stands, 56k modem is a crummy article and should be somewhat expanded and greatly rewritten, but it should remain a separate article. Jim.henderson 03:14, 13 July 2007 (UTC)

[edit] Rewrite

This article is hard to read and certain sections repeat eg 'V90' and 'Using digital lines and PCM'. The article gives greater coverage to voiceband modems than other modem types such as DSL modems, due to the many developments over the last 40 years. The descriptions of the variants of voiceband modems is now largely of historic interest. The article does not descibe the relationship between modems and networks, such as todays internet. For example a voiceband modem network typically used to be two identical modems over an analogue voice circuit but nowadays is typically a wide range of different 56k modems connecting over a digital PSTN to a shared central modem server. The section on 'Popularity' refers to using 56k voiceband modems to access the internet, obviously DSL modems are gaining in popularity. A better structure for this article is needed.John a s (talk) 23:23, 12 January 2008 (UTC)


I agree the article is repetitive. I have thought about eliminating most of the first section and merging it with the next section (about breaking the 9.6k barrier, data compression, et cetera). The only reason I did not do it sooner was because of time; maybe I will tackle it this weekend. ----- As for newer modems like DSL, it's probably worth adding a section immediately following the 56k section, but I would not eliminate the earlier stuff. POTS modems dominated for over 30 years; DSL is still just a baby in comparison. (And POTS modems still have uses; like in my hotel. A 56k is my only option.)
"For example a voiceband modem network typically used to be two identical modems over an analogue voice circuit but nowadays is typically a wide range of different 56k modems connecting over a digital PSTN" <---- The article does discuss this in regards to the transition from 33.8k to 56k. —Preceding unsigned comment added by Theaveng (talk • contribs) 20:43, 14 January 2008 (UTC)