Target 56kHow realistic are the claims of the 56kbit/s modem makers? Are we being built up to a download heaven only to find data crawling out of the Net? Steve Bush investigates You sit there, crying in frustration, as the image you are downloading dribbles on to the screen at 1.4kbit/s. The answer has to be a new modem, one of those nice new 56kbit/s gadgets, that should speed the whole thing up - or will it? 56k modems seem to be
the way to get ISDN-like (64kbit/s) performance without paying for a special connection to the network, but the situation is not as simple as some would claim. First, there is the question of which 56k standard to plump for. Two were initially proposed and implemented. At the beginning of 1997, K56flex and X2 were slugging it out like Betamax and VHS - although not necessarily in that order. This was preventing potential users from buying 56k modems for fear that they would tie-up
with the loosing protocol. The situation even caused a mini-boom in the sales of older V.34 28.8kbit/s modems. Luckily for everyone, the International Telecommunications Union has stepped in, got hold of both proto-standards, and proposed a compromise called V.90. Being neither K56flex or X2, but related to both, modem makers have all had to do some work, but not a prohibitive amount. The result is saleable production 56k modems which either conform to either K56flex or X2 initially, but will be upgradeable to V.90 on-line. The upgrades should happen soon, now that the ITUhas ratified V.90. The individual modem makers just need to finalised their V.90 implementation code. Straight V.90 modems are expected to appear on the shelves this March. Line quality is the other big factor determining how much data can be transmitted between terminals. "Modems can achieve 56kbit/s as long as there is only one D/A converter in the communication chain," said Annick Evenepoel of PC Card modem maker Xircom. In reality, this means that only data from a fully digital Internet service provider or similar organisation has a chance of achieving the full rate. This must then stay in digital format until the subscriber's local street box, where the one D to A conversion is allowed before the final analogue line to the user. Data in the reverse direction has to use the local analogue line and be A/D converted, precluding 56kbit/s traffic. The back channel is therefore limited to more conventional V.34 coding and 33.6kbit/s. In practice this is not too much of a problem as short requests for data are usually all that is travelling back. The chance of getting 56kbit/s through the forward channel, assuming that it is all digital, is further diluted by other factors. Devices with characteristics similar to analogue to digital converters, including different types of echo cancelling equipment as well as line boosters or repeaters are sited by Xircom as reducing the possible data rate. Satellite links are another problem, they are echo-cancelled to an extreme level and cannot be used at 56k. With all this against them, international modem transfers using 56kbit/s modems are still claimed to be possible providing they don't break the ‘only on D/A converter' rule. Some countries have political reasons for preventing digital trunks crossing their borders, so there is no chance of full speed transmission to them. Taking all these factors together, and adding in the congestion that occurs on networks, it should be possible to get some figures on real and expected data transfer rates. However, network operators seem singularly unwilling to discuss the performance that modems can achieve across their networks for fear of unfavourable comparisons with other companies. BT, in a statement to EW, while stressing that it believes this performance matches that of other telecommunication companies for data transfer over the public network: "A recent test of modem performance showed that at least nine out of ten calls on BT's network can be handled at speeds of 24kbit/s or greater. These tests are ongoing. Historically, performance guarantees on modem calls have not been made because of the complexity and variety of modems themselves and the network equipment which modem calls will pass through." Signals & Software is a UK-based software modem company that is currently developing a 56kbit/s modems for infrastructure and subscriber use. MD Dave Morley, said: "Despite what the marketing departments claim, I don't believe that anyone has got up to 56kbit/s yet. As of a couple of weeks ago, the best was 48kbit/s and people had been stuck between 45 and 46kbit/s for quite a while. The biggest debilitating effect is bit-stealing by the networks, especially when they are busy." Even when 56kbit/s is not achievable, modems suited for it are claimed to be better than those that can only handle 33.6kbit/s maximum. "Something else to remember about 56k modems is that they make better 33.6k modems," claims Xircom's Evenepoel.
How to get 56kbit/s Getting 56kbit/s through a telephone line requires a fundamental shift in technique from moving data at 33.6kbit/s. "33.6kbit/s modems use a modified form of quadrature amplitude modulation (QAM),"said Dave Morley, managing director of software modem company Signals & Software, "This technique won't stretch to 56k." Instead, developers assumed that data will be introduced to the network digitally and stay that way all the way to the pulse code modulation (PCM) D to A codec at the subscribers local street box. "On the network, data is transmitted at in 8-bit words at 8kHz,"said Morley, "This gives a maximum data rate of 64kbit/s. On some systems, notably in America, one bits is stolen for the network clock and the rate comes down to 56kbit/s." The subscribers modem therefore has the job of deducing what 8-bit word was presented to the codec by measuring the voltage that it produced.
As the voltage has been mangled by the subsequent analogue line between street box and user, the modem's first job is to remove phase and amplitude distortions. This it does by developing an inverse filter during an initial training sequence. It then recovers a timing clock: "The complicated part," according to Morley, and finally regenerate the data stream.