This page was last updated at 03:58:26 CDT on Fri Sep 5, 1997. Comments on these results should be directed to William Garfield, wdg@hal-pc.org. Academ Consulting Services is providing space for these results as a public service and neither endorses or provides funding for this testing.
Expecting 56K or 33,600 or 28,800 bps? Your results will vary. If you find that you're unable to achieve full speed connections and/or your new high speed modem sometimes behaves erratically, the reason is usually a phone line impairment. Most often the cause is insufficient bandwidth, poor signal to noise ratio, or an imbalance in the phone lines, or a combination thereof... either yours, the phone lines of the system you are calling, or in the lines and telephone switch equipment -anywhere- along the way. Here are the minimums: 33,600 bps V.34 requires 3429 Hz bandwidth, from 244 - 3674 Hz. 31,200 bps V.34 requires 3200 Hz bandwidth, from 359 - 3559 Hz. 28,800 bps V.FC/V.34 requires 3200 Hz bandwidth, from 320 - 3520 Hz. 26,400 bps V.FC/V.34 requires 3000 Hz bandwidth, from 375 - 3375 Hz. 24,000 bps V.FC/V.34 requires 2800 Hz bandwidth, from 467 - 3267 Hz. 56k: (see "Expecting 56k?" toward end of article, below) I should qualify the above by stating that at 28.8 speeds and below, V.34 can shift the whole baseband down a bit further to compensate for a poor top end but I'm sure everyone gets the basic idea. By contrast... a 14.4k bps (V.32bis) link requires a usable bandwidth of only 2400 Hz, from 600 - 3000 Hz. Alas, while most U.S. domestic phone lines can easily support the requirements of 14400/V.32 bis, many fall short of meeting the minimum technical specifications necessary for V.FC and V.34 to function at _FULL_ speed. This, in a nutshell, is what you (and several others) are experiencing. Consider this: Compared to 14.4k V.32 bis, 33,600 requires 43% more available bandwidth, 28,800 bps requires 33% more bandwidth, 26,400 bps requires 25% more bandwidth and 24,000 bps needs 17% more bandwidth. This additional bandwidth _MUST_ be there from end to end, all the way from one modem to the other. Either you have it or you don't, and your modem is as much as telling you via its performance. Now, before you go screaming at your phone company, beware that their ancient tariffs on file with the FCC and state Public Utility Commissions specify a voice grade bandwidth of only 300 - 3000 Hz. Therefore, if 14.4/v.32bis worked ok and you're now achieving at least that level of performance or better with your V.34/V.FC modem, then you can rest assured that your line at least -meets- the minimums. The phone company is under no obligation to provide anything beyond 300-3000, though most of us do enjoy considerably better than minimum spec lines. Here are a few things you can TRY for improving your high speed modeming: Go throughout the house and disconnect -ALL- telephonic devices attached to the phone line. This includes extension phones, answering machines, fax machines, caller-id boxes, line-in-use indicators, cordless phone base units, demon dialers, and voltage spike protectors or line filters like those commonly found in PC Desktop master-switch power directors and power line conditioning units. Don't forget remote utility meter reading devices and burglar alarms which may also be attached to the phone line. If you find that any of this helps, then start plugging things back in one by one, until the culprit is identified. It could even be a combination of things. At the office you may discover that your modem will perform better when provisioned with an outside line, ie, one which _doesn't_ go through the office pbx or a multi-line key system. If your telephone wiring is a rat's nest, and/or you've strung some extension lines yourself and not used genuine twisted-pair telephone wiring, consider having a professional replace your haywired additions. Your telephone wiring should also be well _away_ from the A/C power wiring in walls, ceilings & floors. AC power wiring should be crossed only at right angles. The old (non-twisted) 4-conductor "cloverleaf" wire (red/green/yellow/black) should not be used for 2-line service since the parallel conductors will cause crosstalk. If you have to make a vertical drop down a wall that also has power wires in it, a good rule of thumb is to allow at least 24-inches horizontal separation. Don't waste your money putting in "Category-5" data grade cable - it isn't required, not even for ISDN lines. However, twisted-pair phone wiring is sometimes hard to find whereas 'Cat-5' wire is often plentiful, so use it if you can't find anything else. If your phone service arrives in your area via a subscriber loop concentrator or "remote terminal" or other "pair gain" device such as the ubiquitous SLC-96 ("SLICK"), then all bets are off and you may have to settle for 24,000 and below on most calls, with 26,400 being a rare treat, indeed. I haven't heard of too many folks regularly achieving 28,800 through a non-integrated slick. If you're using a USR modem, be SURE you have your serial port speed set to operate at 38,400 bps **MINIMUM**. The new USR's will not permit the -data LINK- to operate at speeds any faster than the serial port... this has already bitten several unsuspecting users. ***** EXPECTING 56K? (X2/K56Flex) As of this writing (Jul '97) the few vendors offering these radically new (and somewhat controversial) modems all have one thing in common; they too often don't live up to all the advertising hype or meet the customer's performance expectations. Unless you *read closely* or *listened closely*, the advertising hype would have us believing that the new X2/K56Flex modems will operate at twice the normal speed of your 28.8/V34 modem. The fact is these new modems often won't achieve anything close to those implied speeds. If you did read the ad or press release closely, you may have noticed that those statements were all couched in very vague and ambiguous terms. No one "promised" anything. Words like "nearly" and "almost" and a vague reference to "standard modem speeds" are used to lead you into expecting something that the new modems simply did not promise and for many users fall substantially short of delivering. In actual performance testing, the new "56k" modems can, DEPENDING ON PHONE LINE CONDITIONS, operate at speeds that, on average, yield about 42-48k. True there are some -rare- exceptions where users report 52-53k, but I need to emphasize that those are _highly unusual_ exceptions where the user's dumb luck affords them access to a "near lab-grade" telephone line. These are often too the instances where both the user and the ISP are served out of the same telco switching office and the user is located very, very near to his local telco office, typically less than a mile or so. One modem manufacturer, during a nationally televised interview on CNBC, suggested that up to 90% of all US Domestic phone lines could be used with this new technology. If we take into account that well in excess of 90% of all those phone lines are located in high-density commercial business areas (and thus near the telco central office), then his statement is factually correct. However, if we interpret his statement to imply that 90% of everyone will be able to use the new technology, then his statement is blatantly false. In suburban residential areas around the large metropolitan cities and in many rural areas the telephone companies make frequent use of 'subscriber-loop-concentrators' or slicks as they are known. By the manufacturer's own acknowledgement, the new 56k modems won't work in this situation. The result then is in actuality, only about 50% of all residential users can utilize the magical new 56k technology, and only then if they're reasonably "close" to their telco switching office (usually less than 18,000 cable feet). This is not a good thing. The modem manufacturers have established some long distance dial-up 'test numbers' for you to call in and check your lines, but these tests are anecdotal at best and absolutely meaningless unless the test equipment is located at *YOUR* local service provider and you're placing a *LOCAL* call to reach it. The reason here is that toll-free and long distance calls are completed over 100% full digital telephone circuits. By contrast, a local call from one side of town to the other may *not* be carried over full digital facilities or may encounter a technique known as "robbed-bit-signalling" to indicate status of the call or digital pads. Therefore, the manufacturer's long distance or toll-free test numbers may yield false or misleading results that OFTEN AREN'T ACHIEVABLE over a local call. One local provider here in the Houston, TX area *has* put up one of these test numbers (713-693-0399) and I think the results speak for themselves. It is not uncommon for a local Houston caller to fail the local test but pass the manufacturer's "long-distance" test. The local test number isn't a more stringent test, merely it represents a more realistic "real conditions" test. My 'toungue-in-cheek' advice to anyone contemplating purchase of a "56k" modem is to first *WAIT* until your local service provider offers 56k connectivity. Then buy (with a credit card) and immediately try the new modem. If it works, you're a winner! If it doesn't work or it's performance is marginal or unacceptable, don't delay, return it today. *CONNECTIVITY ISSUES WITH 1996-VINTAGE USR SPORTSTER MODEMS* (and certain TI-chipset based Sportster clones) If you're having trouble just getting a connection (at any speed) and/or you find that you're frequently losing that connection and have to call back, look at your modem. These are both MODEM PROBLEMS that the author has found to frequently follow certain USR Sportster (and Sportster clone) modems which were sold over an approximate 14-month period between OCTOBER 1995 and DECEMBER 1996. Sportsters purchased in this time frame for the most part represent, in the author's opinion, "transitional period" modems that were manufactured during the time the product was undergoing some very radical design changes and steep price cuts. They sure were cheap though. It appears that the product finally began to stabilize in manufactured units sometime during the 3rd quarter of 1996, but those were brand new stock and so pretty scarce on dealer shelves until during the 1996 Christmas holidays. Even then the shelf stock at CompUSA and Circuit City stores looked to me to be 'salted' with a high percentage of the prior models. If you have a Sportster that was bought last year ('96) and have been experiencing a high incidence of trouble connecting or staying connected to your internet provider, my suggestion would be to consider trying a different modem. The manufacturer may also have a later firmware revision that might also help, but that still leaves you with the "Edsel". The most effective cure seems to be a replacement. By contrast, there's some good news. Recent production Sportsters (those appearing in the new red, white & blue "56k" packages) seem to be **much** improved over last year's models, though alas, prices are back up again. Seems as always you get what you pay for.
Many Thanks to:
www@academ.com