Welcome! to the Modems section

Don't Blame the Modem (Improving Connection Speeds)

This section is probably of most interest to owners of V.90, K56Flex & X2 modems who are tearing their hair out because they are the proud owners of 56k modems that simply will not connect at 56,000 bit/s. Once again, there is much below that will be of interest to all modem users.

Contents

• A Little History
  • Brief explanation
  • The situation now
  • V.90 The one-eyed man's view
• What Can You Do to get better connection speeds?
  • DUN shows 57,600 or 115,200 bit/s
  • DUN takes a long time to connect
    • DUN Server Types [+ screen shot]
  • Check it out practical measures to improve connection speed
    • How long is yours? Speed impact of the modem line
    • And how many of them? Speed impact of other telephones
    • And are they unusual? Speed impact of other devices
    • And is it corroded? Speed impact of noise
    • Or too thin? Speed impact of telephone gain / AGC
    • Or out of your control? Speed impact of SNR
  • NETCOM x2 Line Tester
  • Modem Speed Test Pages & other help

In 1998 - if you believe what each company says - Rockwell Semiconductor Systems and US Robotics both won the commercial war that they had been waging against each other in the previous year for the new standard in modem telecommunications. The reality, of course, is that modem users lost & are still suffering the collateral damage now, more than a year later. Brief explanation: In February 1998 the ITU introduced V.90, a new internationally-agreed standard for modem-to-modem communications that promised 56,000 bits-per-second [bit/s] downstream-rates and 33,600 bit/s upstream rates; this standard was ratified in September of the same year. Now, what you have just read is already untrue and part of the hype that always surrounds these things. 56k bit/s and 33.6k bit/s are theoretical maximum downstream/upstream rates and almost never achieved in practice - indeed there are people that will never ever achieve even the maximum upstream rate in either direction, but more of this later. Before V.90 was introduced the various chipset manufacturers had attempted to use their commercial muscle to enforce mutually-incompatible and proprietary standards developed by themselves - see boxes at right. This latter situation had led to ISP's having to have multiple dial-up lines - one for K56Flex™, one for X2™, one for ISDN and another for previous standards. V.90 now promised to do away with all of that nonsense... eventually.

A Little History   V.90 An open connection protocol established by the ITU, with speeds from 28,000 bit/s to 60,000 bit/s in steps of 1,333, used by the following chipset manufacturers: All K56Flex™ A proprietary connection protocol, with speeds from 32,000 bit/s to 60,000 bit/s in steps of 2,000, developed and/or used by the following chipset manufacturers: Cirrus Logic Lucent Motorola Rockwell - now Conexant X2™ A proprietary connection protocol, with speeds from 32,000 bit/s to 61,333 bit/s in steps of 1,333, developed and/or used by the following chipset manufacturers: Texas Instruments US Robotics - now 3Com

This is V.90 At the ISP's server Headend is the digital equivalent of a modem This communicates at 56k upstream and 33k downstream - the reverse of the V.90 modem in your system. These are made by different manufacturers Freeserve's Digital Modems are provided, as an example, by Cisco, and when first installed were declared to be a Beta version of the V.90 standard. Digital Modem manufacturers - just as with personal modem manufacturers - have managed a better-or-worse conformance to the standard. V.90 was ratified in Sep 98 and your modem chipset - let alone the modem - had possibly already left the manufacturing plant by then. Most of the chipset manufacturers do not produce the modem, only the chipset. Software takes twice as long to produce as the person writing it thinks... ...and is then full of bugs [the polite word for errors]. Firmware is software in silicon. Windows98 came out before V.90 This means that even a Windows98 standard modem driver will not connect at V.90 speeds.

The situation now: Modem chipset manufacturers and software writers are desperately trying to produce silicon and software drivers that conform to the standard, and succeeding, more or less. The situation is compounded by many things. Any of the sections detailed at left would make life difficult; put them together, & it is a miracle that any modem connects at all. As one example, your modem - which is more-or-less compliant with a standard - has to connect with the Headend Digital Modem - which is more-or-less compliant with the same standard. Get the picture?

V.90 As long as you do not mind being led by a one-eyed man, on the right is as untechnical an explanation of V.90 as I can manage. As a non-telephone engineer I'm grateful to Alan Fowler for his input on this. A more technical description from Alan is here. V.90 performs it's magic by assuming that the network is digital on the downstream side and voice-band on the upstream side. Now, I know, I know - why not assume that it is digital both ways? but that's how it is, so try to live with it or get an ISDN link. The ISP has a direct digital link to the main exchange and thus one of the conversions is absent, reducing potential errors and increasing the maximum potential speed at which communication can take place. Notice, however, that word 'potential' - other factors on the line can reduce the actual practically-achievable speed, & this leads nicely into the next section:

No, this is V.90 The telecommunication system, which our mothers use to ask why we haven't phoned them recently, & we use to connect to the internet, was originally designed to allow one human to talk to another human. This technically means turning sound waves into electrical energy at one end, pushing that electrical energy down a copper wire connecting the two ends, and then turning electrical energy back into complaints/laughter/love or whatever at the other end. Now, since all things including conversation are a two-way exchange two copper wires are needed - the famous twisted pair. This is simple enough once a microphone and a speaker has been invented & each party has one of each and there are only two people - you possibly know the story of when Alexander Graham Bell invented a third telephone and joined it to the first pair of telephones. He tried to dial one of the first two phones & found the line was engaged. Another issue comes with distance, and distance is always going to be an issue, as why use a phone if you are close together? A simple twisted-pair wire will work up to a distance of a number of miles, but eventually the sound at the other end will get fainter & fainter until inaudible. This is due to the resistance of the copper wire to the electrical signal (if an alternating signal - such as sound waves translated into electrical energy - is used, the term impedance is used instead). This translates into the 'gain' on our telephone line. A telephone which is close to the exchange needs very much less gain than one which is far from the exchange. Telephone gain is one of the aspects which will be talked of below in terms of getting a better communication speed from the modem. Modems are designed to allow a computer - which talks in 'bits' of information - to communicate along a system which was not designed with them in mind. The modem does this by mod-ulating [convert computer bits into noise] and dem-odulating [convert the noise back into computer-speak] - hence mod-dem. The problem comes when the other person sounds very tinny. This would happen a lot with early telephone exchanges or international calls, and still does now with cellular phones - so much for progress. This is due to both the restricted bandwidth of the telephone system [meaning that the telephone attenuates both the bass & treble of the human voice into extinction] and also frequency shift or distortion. Humans are much cleverer than computers and, normally, much less pedantic. If mother sounds like she has her head in a tin can, or is accompanied by shoosh-ing noises, or is very faint, or is insisting on playing her Val Doonican records at full volume in the background again, we can still make out the words. A computer would process this ever so much more quickly than us and then throw up a dialog saying 'fatal exception' and shut down. No contest. A computer, however, can communicate along a twisted-pair copper wire at high speed if allowed to do so in a way that suits its own, pedantic self - a network card, as an example, will do this at 10,000,000 bit/s, no problem - but even with a clear line the practical maximum speed of voice-based communication is close to the V34bis standard of 33,600 bit/s. Enter now the digital telephone network. Telco's [Tel-ecommunication Co-mpanies - technology produces acronyms like frogs produce spawn] across the world have been replacing their old electro-mechanical exchanges with brand spanking-new digital exchanges for the last two decades now. This white-heat of technological innovation has led Britain to the point where even parts of Scotland & Wales are telecommunatically digitalised. Of course, other parts are not (such as parts of central London!), and folks served by these older exchanges can neither release their caller-id info nor communicate faster than V34bis with a modem. Such advances have also been avoided throughout Russia, which means that if you decide to visit our Slavic cousins you had better take an old 14,400 bit/s modem with you should you desire to keep in touch with the 'civilised' world. Elsewhere these changes now mean that instead of having a pretty switchboard girl put us on hold it can be done by a piece of silicon instead, and we can listen to some music whilst we wait. If you are at all unsure as to which type of exchange your telephone is serviced by, here is a simple way of finding out: the next time you make a telephone call [try the speaking clock, it doesn't mind] press any button on the telephone. If you hear a tone the exchange is digital. If it is a series of clicks it may not be, & I suggest you ask your Telco to find out. The problem for a digital telephone network is that it still had to connect & work with the old, analogue (voice-band) exchanges. As an example, I've still got a BT speaker-telephone that I bought in the late-70's for £100 - equivalent to approx. £1,000 now. This is a pulse-dial phone rather than tone-dial, yet I can still use it successfully to dial a number. If the network had gone completely digital then everyone would have had to buy digital telephones - which is the situation with ISDN now - and not talk to their neighbours who were still connected to an old voice-band exchange. Instead the network compromised by being digital internally and voice-band externally. Voice calls are carried internally across a 64,000 bit/s channel and, since the trunk-lines connecting exchanges are far wider than this, many voice-calls can be multiplexed across the same line. For most domestic subscribers only the final part of the line - the copper wires from exchange to house (the 'local loop') - are voice-band, and some commercial concerns install a digital link direct to their premises. This means that a typical domestic voice call gets at least two conversions - first is analogue-to-digital, then digital-to-analogue - although curiously there can be other conversions along the way. These conversions inevitably introduce errors and give a ceiling to the maximum speed at which a data-call could be made of 34,822 bit/s [Shannon 1948] in a channel capable of 64,000 bit/s.

Getting better connection speeds First, a quick bucket of cold water (called reality) for owners of 56k modems - you will never achieve connection speeds of 56,000 bit/s, not ever, and any speed over 33,600 is a 56k connection.

What can you do...?

Another initial question, of course, is 'What speed is my system connected at?' The easiest way for Windows' users to find the starting connection speed with a dial-up line is to hover the cursor over the little icon sat in the system tray after connection (see panel right). This icon looks like two computer screens connected by a line and belongs to Dial-Up Networking [DUN]. As with all things in life, DUN can lie.

The System Tray This is the part of the Task Bar that contains the time; the Task Bar is the grey-strip normally at the bottom of the screen that contains the Windows 'Start' button

Accepted Wisdom for Showing DTE Speed Chipset Extra Settings Motorola atq=1\v=4 Rockwell PCI at+mr=2 Other Rockwell w2 USR &a3 All at&f1 Speed Fix for USR WinModems These modems are very prone to this problem. Nigel has made a fix available for them - go here. mdu.exe This was developed by & is copyright BVRP Software, & can also be sourced at modemhelp.com. It principally sucks info from the Registry but will also print out the ati responses if the com port is selected. It is very useful if you do not know where in the registry the modem is located. The Registry Editor is loaded by entering "REGEDIT" (no quotes) in a 'Run' box. Beware, as wrong entries in the registry can bring your computer grinding permanently to a halt, & in this you are on your own.

DUN shows 57,600 or 115,200 bit/s then the modem's .inf file is either the wrong one, or inaccurate, & DUN is reporting the DTE speed rather than the DCE speed - go here for more information. If the .inf file is the correct one then go to Extra Settings and enter one of the strings as at left. Even "correct" INF files are sometimes incorrect The last string in the list at left sets the modem to factory-defaults. Unfortunately even this desperate measure will not work for some modems, & one reason may be that the Responses section of the .inf file is missing such as the following (from my own modem's .inf file):- [MfgAddReg] ... ; Add the following rates to support V.90 ... HKR, Responses, "<cr><lf>CONNECT 45333<cr><lf>", 1, 02, 00, 15,b1,00,00, 00,00,00,00 HKR, Responses, "<cr><lf>CONNECT 46666<cr><lf>", 1, 02, 00, 4a,b6,00,00, 00,00,00,00 Now, whilst K56Flex is in steps of 2,000 starting at 28,000 & finishing at 60,000 bit/s, V.90 is in steps of 1,333, leading to some very strange connect-speeds indeed. If the .inf file is missing any of these speeds then the computer may default to show the DTE speed. If you are determined to correct this terrible error then it means either hacking the .inf file & re-installing the modem and/or hacking the Registry direct. With my modem this is:- HKEY_LOCAL_MACHINE System CurrentControlSet Services Class Modem 0000 Responses You can find your own by using mdu.exe - see also box at left.

DUN takes a long time to connect Whilst this is the speed of connection rather than the connection speed it seems to be very common. What follows is specific to Freeserve but may have relevance for other ISP's, particularly if you also have AOL. The typical picture is of dialling out with long, long periods of hand-shaking but no connection with an eventual time-out resulting in the whole process being repeated over again, perhaps many times, until connection is eventually made [or not]. Whilst there is excellent help on the Freeserve support pages here is a quick troubleshooter for this one:- Check the Network & DUN settings A screen shot of the minimum Network entries needed is here. Unless your setup requires other network-facilities any others should be removed. Do not change settings in entries here unless you know exactly what you are doing - changes here over-ride those in DUN with no warning. Indeed it may be worthwhile to follow instructions on the same page as the screen-shot to remove & re-install DUN and thus the dial-up adaptor & retrieve your Freeserve account as to obtain those defaults.

The DUN settings that are important here are on the Server Types page of DUN Properties. It is reached: open My Computer open Dial-Up Networking highlight the Freeserve connectoid & go File | Properties click on Server Types page If yours looks any different than the screen shot at right then a probable answer has been found for DUN taking a long time to connect. For the curious, NetBEUI is a Windows-for-Workgroups peer-to-peer protocol and IPX/SPX is a Netware protocol. Neither are needed for standard dialup lines. A tick in the box 'Log on to network' indicates: a missing or corrupt Dial-up adaptor in Network (possibly) an account with AOL you have created the connectoid manually If this does not solve the problem then the best course of action is to start the modem troubleshooter here. If that doesn't find the problem & your system is OK then the likelihood is that the ISP is having software and/or hardware problems of their own. Something to experiment with is to put a tick into Enable Software Compression; this has speeded up downloads for some people with some files. True hardware (i.e. not soft) modems implement compression in silicon - this is the LAP-M, MNP & V42Bis standards - so ticking this option should not make a difference. However, it has helped some & you may wish to experiment.

Check your house Having sorted out the computer & the modem as potential culprits attention now moves along the Line of Communication to the house. One first check - is the power supply to the computer earthed? Non-earthed power supplies have been reported to halve modem connection speed.

How long is yours? Shorten the modem line-cable. Many people have a line extension from the phone socket to the modem which is many metres long, and this modem line is responsible for far more speed loss than most people realise. Indeed, someone reported a 10,000 bit/s increase in speed just by replacing the modem cable for one from another manufacturer, yet of the same length. The ideal is to have as short a length of phone cable between the telephone exchange and the modem as possible and, whilst this does lead to visions of plugging the modem into a socket positioned on the outside of the exchange, there are many practical steps which everyone can take inside of their own house to shorten their internal high-loss lines. If you are contemplating replacing your own modem line-cable do remember that there is no such thing as a standard modem cable, and that you will need to obtain the correct replacement - take the old one along for comparison. Indeed, one computer manufacturer recently supplied cables which caused every telephone in the house to 'screech' on connection.

or, lengthen it! 17 Jan 00 As an odd addendum to this section, Ray reported in the freeserve.help.modems newsgroup improving his connection from 9,600 bit/s to 46,666 bit/s by increasing his cable by 40 feet! Due to a window replacement, his 30+ year old telephone wiring was replaced with brand-new BT-approved 4-wire cable. It was a shorter run with fewer connections. And his connections dropped from approx 46,000 to 14,400 or less!!! Reverting back to the former wiring restored the previous connection speeds. The second feature was that his telephone had now become a radio receiver, playing Radio 5. The local radio shop suggested adding extra wire, & thus altering the tuning of the wire. Hence the extra 40 feet, which both stopped Radio 5 & restored his connection speeds. Ray's home page

And how many of them? Detach every other device. Whilst a telephone in every room may seem convenient it will eventually cause problems. This is usually considered in terms of Ring Equivalence Number [REN] & means that if you have too many devices on the telephone line one or more of them will not ring if someone phones you, but it also impacts the connection speed that the modem can establish. The issue here is one of the 'gain' on the line, & it is mentioned above and below. Each device will suck some of the juice from the line & may not leave enough for the modem. Testing this out is easy enough - detach every device [be it telephone, fax, answering machine, or whatever] - and see if it makes a difference. If it does then re-attach them one-by-one to see which one(s) make a big difference.

BT Relate phones These have been reported to reduce internet connection speeds. Try unplugging before net connections. 12 Jun 99 Thanks to JOVINBLAK for reporting that his BT model 'Relate 300' was responsible for severely reducing his net speeds.

And are they unusual? Various devices can get attached to the telephone line by various authorities. The effect of most of them seems to be to reduce the modem connection speed. Get a cable line... It is an odd situation. As mentioned above, the same twisted-pair wire on an internal network will allow 10 million bit/s, American experience with digital subscriber line technology has shown the existing infrastructure to be capable of 1.5 million bit/s or more, and even BT will drive it at 64,000 bit/s, yet when 2 lines are multiplexed (DACS boxes - see right) it becomes capable of no more than 14,400 bit/s. It has to be said that those in the UK with the best connection speeds usually seem to be on cable lines. 'Nuff said. ..for data connection. A [British] Telewest customer discovered that the box in the street had the line to each customer designated as either a voice or data customer. Persuading the engineer to switch his line to data doubled his connection speeds.

Sky Digiboxes These are reported to reduce connection speed. Please be aware of any contract details on minimum length of connection to the line. DACS boxes (Digital Access Carrier System) If you or a neighbour have had a second or new line installed by BT & it has been achieved with a DACS box then you will never achieve connection speeds above 33,600 bit/s. This is particularly aggravating for people that installed the second line specifically for internet connection! DACS is a means of multiplexing two voice-band telephone numbers across one physical telephone line & will be betrayed by the presence of a box on the outside of the house. It is obviously cheaper for the Telco to do this than to install another physical twisted-pair line, but the trade-off is in reduced bandwidth for at least one connection. BT's Terms of Trading state that it is required to provide a line capable of carrying faxes (i.e. 14,400 bit/s) and therefore installing a DACS line is well within this contract. It is possible to stipulate before installation that any additional lines are achieved with a Direct Exchange Line [DEL] but too late afterwards. One ray of hope may be that, usually, only one of the DACS lines is of highly-reduced bandwidth & a call to 151 may be able to get the modem put onto the high-bandwidth line. More info is here.

And is it corroded? Try the quiet line test. The Bible tells us that the moth and rust doth corrupt, and this most certainly is true at the telephone junction box. Typical signs are a noisy line. BT maintains an automated quiet line test (amongst other facilities) at 17070. My own experience with BT from many years back indicates that their twisted-pair line is actually a twisted-threesome of which only 2 lines are used. The engineers can switch this at the exchange without physical intervention and possibly bypass a poor connection on the third wire (worked for me)*. Other points are unshielded wires passing close to electrical motors/machinery introducing interference/hum onto the line. The biggest culprit here, of course, is probably the computer. Line cards within the local or main exchange are also capable of generating noise. *thanks to Peter Hurst - ex-BT - for pointing out that my information was nonsense! He explains that all underground wires are twisted-pairs, and that BT was probably using ‘split legs’ - “..the engineer will use a faulty pair and find one good wire in that pair to swap with yours”.

Or too thin? Pump up the volume. The final aspect in being able to affect the speed of connection concerns the 'Gain' on the line (see also here & here). This means, quite simply, is the line too faint or too loud? To use, once again, BT as the example, it normally subjects the line to Automatic Gain Control (AGC). This usually works well for a voice call but is not necessarily most effective for modem communication as the modem usually contains AGC within it's own codec, & the two can conflict. Many users have reported dramatic speed improvements after asking for the gain to be increased/decreased. Just in case you are now confused by the fact that the gain may need to be put up or down, it is quite simple. 'Gain' talks of how much voltage the Telco is using to drive the telephone signal down the line. They do not want to melt the cables so the voltage is quite low, but a voltage it is nonetheless. If your house is close to the exchange this voltage can be quite low, whereas if far from the exchange it will need to be much higher. Also, across time corrosion at the various junctions will increase the resistance on the line & the voltage will need to be a little higher.

Changing the Gain - BT Lines Dialling 151 & asking for the gain to be changed will automatically cause AGC to be switched off. If you explain that it is to allow a better internet connection the engineers are usually very helpful. Cable line-voltages Cable connections in the UK usually drive the line at a different voltage to BT connections. This has been reported to cause problems for certain modems. You will need to contact your modem manufacturer for a fix if this affects you (sorry, no details).

Or out of your control? There is one more item that impacts dramatically on connection speed - but there is nothing that you can do about it - and this is the signal-to-noise power ratio (SNR) on the entire line. Now, this is a topic for real Techno-Sluts & take the box at right under advisement as I'm a little rusty on these matters.	SNR - signal-to-noise power ratio SNR is usually expressed in positive decibels [dB] and the original measure of logarithmic power ratio was, of course, the Bell (after Alexander Graham Bell - who else). dB is used because you need to shout ten times harder to be heard twice as loud. Or something such. The US telephone network has a theoretical noise floor of 39.5 (expressed in positive dB) and realistically achieves 35 to 36 dB (here more is better - I have no info on equivalent figures for the UK network). 56k connection requires 44.4 dB or better.
Modems make a much more intimate examination of the line than you might expect when they connect, & line quality is only one of the factors. Some modems also are able to give a report following connection which includes line quality: Use HyperTerminal to connect to your favourite ISP. At the point where the modem is still connected on the line to the ISP server it can report on many things to do with the connection. The command strings vary modem-to-modem:
Checking Signal-to-Noise Power Ratio [SNR] during Connection Conexant Chipsets (Rockwell Chipsets) at&v1 at%l for line signal at%q for line quality All in -dB. A line quality under 25 is OK. Lucent LT WinModem Chipsets ati11 8 Estimated Noise Level [No base reference - more is worse] 3Com Chipsets (USR Chipsets) atl6 ati11 aty11 All in +dB. A line quality over 40 is good.
Don't forget to enter ath0 to hang up

NETCOM x2 Line Tester Netcom [in the UK] used to run a very nifty line-tester that produced multiple results, and amongst which was the SNR [shown below in dark blue] plus a graphical display of the line frequency-response. If the line can support x2 it can also support V.90. Unfortunately, sometime before 21 May 99 it stopped working (thanks to JohnFHH for reporting this) - the modem connects but nothing further happens, & this situation remains the same as at 8 Feb 00. This is a damn shame, as there is no better test for a modem. Zoltrix maintains a Phone Line Test (details here), but unless Netcom change their mind, what follows is purely a history lesson. Use HyperTerminal & enter the lines as shown below. The connection is standard V34bis, so don't get the collywobbles at what appears to be a slow connection. A sample from my own modem follows (the colour graphical output is much prettier, by the way, but shown here in text for speed of display).
+++ate1 OK atdt0845 079 8022 CONNECT 31200/LAPM Connected to ARCHOST v1.05 03-04-99 05:39:00 Caller number: 13380 Do you want ANSI/PC color graphics? (Y/N) >N Please wait, testing in progress... +-----------------------------------------------------------------------+ | This connection will support x2! | +-----------------------------------------------------------------------+ Press any key to continue... +-----------------------+-----------------+-----------------------------+ | x2-STATS Version 1.01 | 03-04-99 05:39 | Caller 13380 | +-----------------------+-----------+-----+-----------------------------+ | Elapsed Time 00:00:12 | Modulation V.34+ | | Blocks Received 1 | Speed 31200/31200 | | Blers 0 | Symbol Rate 3200/3200 | | Blocks sent 22 | Carrier Frequency 1829/1829 | | Link Naks 0 | Trellis Code 64S-4D/16S-4D | | Blocks resent 0 | Nonlinear Encoding ON/ON | | Link Timeouts 0 | Precoding OFF/ON | | Chars sent 0 | Shaping ON/ON | | Octets sent 158 | Preemphasis 6/0 | | Chars lost 0 | Rx Lev/TX Lev/SNR 20.0/18.2/44.2 | | Chars Received 0 | Echo Loss Near Far | | Octets Received 1 | Roundtrip Delay 18 | | Protocol LAPM | Retrains Request/Grant 0/0 | | Block Size 128 | Fallback Enabled | | Window Size 15 | HST Line Reversals 0 | | Compression NONE | HST Equalization Long | | Dictionary Size | SV: 07/26/96 DSP: 06/14/96 | | String Length | Reason: Online | +-----------------------------------+-----------------------------------+ Press any key to continue... +---------------------------------------------------------------+ | -18 | . x x . . . . . . . . . . . . . . . . . . . . . . | 0 | | -20 | . X X X X x . . . . . . . . . . . . . . . . . . . | 2 | | -22 | . X X X X X X X x . . . . . . . . . . . . . . . . | 4 | | -24 | X X X X X X X X X X X x x . . . . . . . . . . . . | 6 | | -26 | X X X X X X X X X X X X X X X x x x . . . . . . . | 8 | | -28 | X X X X X X X X X X X X X X X X X X X x . . . . . | 10 | | -30 | X X X X X X X X X X X X X X X X X X X X X . . . . | 12 | | -32 | X X X X X X X X X X X X X X X X X X X X X X . . . | 14 | | -34 | X X X X X X X X X X X X X X X X X X X X X X x . . | 16 | | -36 | X X X X X X X X X X X X X X X X X X X X X X X . . | 18 | | -38 | X X X X X X X X X X X X X X X X X X X X X X X x . | 20 | | -40 | X X X X X X X X X X X X X X X X X X X X X X X X . | 22 | | -42 | X X X X X X X X X X X X X X X X X X X X X X X X . | 24 | | -44 |=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=X=x=| 26 | |Level+---------------------------------------------------+Atten| | 0 0 0 0 0 0 1 1 1 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3 | | 1 3 4 6 7 9 0 2 3 5 6 8 9 1 2 4 5 7 8 0 1 3 4 6 7 | | 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 | | 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 | +---------------------------------------------------------------+ Press any key to continue... Thank you for using the NETCOM x2 Line Tester. We hope you found what you were looking for! The NETCOM Operations Team Press any key to continue... Caller number: 13380 Thank you for calling. Disconnecting now. NO CARRIER +++ath0 OK
My Motorola soft modem regularly achieves 49,333 bit/s on this (very poor) line.

Modem Speed Test Pages & other help:

	Modem Speed Test Page. Uses Java to objectively test the modem speed. Very useful.
	Modem Speed Test Pages at Toast Net. Four locations in USA to objectively test the modem speed on both text & graphics d/l. Mine was good on text, poor on graphics.
	Bandwidth Speed Test Page at MSN. My lowly little Motorola scored brilliantly! If you have problems with the main link, goto the bandwidth page & click on the 'Bandwidth Speed Test' link. Thanks to Michelle for this link.
	Optimizing Internet Connections. This is a page for Gaming fanatics [there's redundancy for you] & therefore concentrates on good ping values. An excellent listing of Windows 95 update files.
	Modem Tweaking by Tweak3D. This is on a site for Gaming enthusiasts & concentrates on increasing the speed & stability of the modem. In part it discusses using iSpeed from High Mountain Software.
lagless.com	Primarily aimed at anyone who plays online games with a modem

In addition, this site's Downloads & Links page has a whole slew of sites offering various form's of help.